![\"🏃♂️\"](\"https://s.w.org/images/core/emoji/16.0.1/72x72/1f3c3-200d-2642-fe0f.png\")
Runners had long suspected its suggestions were pushing them towards injury. (WSJ $)
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.
\n\n\n\nLots of influential people in tech last week were describing Moltbook, an online hangout populated by AI agents interacting with one another, as a glimpse into the future. It appeared to show AI systems doing useful things for the humans that created them (one person used the platform to help him negotiate a deal on a new car). Sure, it was flooded with crypto scams, and many of the posts were actually written by people, but something about it pointed to a future of helpful AI, right?
\n\n\n\nThe whole experiment reminded our senior editor for AI, Will Douglas Heaven, of something far less interesting: Pokémon.
\n\n\n\nBack in 2014, someone set up a game of Pokémon in which the main character could be controlled by anyone on the internet via the streaming platform Twitch. Playing was as clunky as it sounds, but it was incredibly popular: at one point, a million people were playing the game at the same time.
\n\n\n\n“It was yet another weird online social experiment that got picked up by the mainstream media: What did this mean for the future?” Will says. “Not a lot, it turned out.”
\n\n\n\nThe frenzy about Moltbook struck a similar tone to Will, and it turned out that one of the sources he spoke to had been thinking about Pokémon too. Jason Schloetzer, at the Georgetown Psaros Center for Financial Markets and Policy, saw the whole thing as a sort of Pokémon battle for AI enthusiasts, in which they created AI agents and deployed them to interact with other agents. In this light, the news that many AI agents were actually being instructed by people to say certain things that made them sound sentient or intelligent makes a whole lot more sense.
\n\n\n\n“It’s basically a spectator sport,” he told Will, “but for language models.”
\n\n\n\nWill wrote an excellent piece about why Moltbook was not the glimpse into the future that it was said to be. Even if you are excited about a future of agentic AI, he points out, there are some key pieces that Moltbook made clear are still missing. It was a forum of chaos, but a genuinely helpful hive mind would require more coordination, shared objectives, and shared memory.
\n\n\n\n“More than anything else, I think Moltbook was the internet having fun,” Will says. “The biggest question that now leaves me with is: How far will people push AI just for the laughs?”
\n\n\n\n\n\n\n\n\n", "content:encodedSnippet": "This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.\nLots of influential people in tech last week were describing Moltbook, an online hangout populated by AI agents interacting with one another, as a glimpse into the future. It appeared to show AI systems doing useful things for the humans that created them (one person used the platform to help him negotiate a deal on a new car). Sure, it was flooded with crypto scams, and many of the posts were actually written by people, but something about it pointed to a future of helpful AI, right?\nThe whole experiment reminded our senior editor for AI, Will Douglas Heaven, of something far less interesting: Pokémon.\nBack in 2014, someone set up a game of Pokémon in which the main character could be controlled by anyone on the internet via the streaming platform Twitch. Playing was as clunky as it sounds, but it was incredibly popular: at one point, a million people were playing the game at the same time.\n“It was yet another weird online social experiment that got picked up by the mainstream media: What did this mean for the future?” Will says. “Not a lot, it turned out.”\nThe frenzy about Moltbook struck a similar tone to Will, and it turned out that one of the sources he spoke to had been thinking about Pokémon too. Jason Schloetzer, at the Georgetown Psaros Center for Financial Markets and Policy, saw the whole thing as a sort of Pokémon battle for AI enthusiasts, in which they created AI agents and deployed them to interact with other agents. In this light, the news that many AI agents were actually being instructed by people to say certain things that made them sound sentient or intelligent makes a whole lot more sense. \n“It’s basically a spectator sport,” he told Will, “but for language models.”\nWill wrote an excellent piece about why Moltbook was not the glimpse into the future that it was said to be. Even if you are excited about a future of agentic AI, he points out, there are some key pieces that Moltbook made clear are still missing. It was a forum of chaos, but a genuinely helpful hive mind would require more coordination, shared objectives, and shared memory.\n“More than anything else, I think Moltbook was the internet having fun,” Will says. “The biggest question that now leaves me with is: How far will people push AI just for the laughs?”\nRead the whole story." } }, { "json": { "guid": "https://www.technologyreview.com/?p=1132498", "link": "https://www.technologyreview.com/2026/02/09/1132498/the-download-what-moltbook-tells-us-about-ai-hype-and-the-rise-and-rise-of-ai-therapy/", "title": "The Download: what Moltbook tells us about AI hype, and the rise and rise of AI therapy", "content": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Moltbook was peak AI theater For a few days recently, the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots.…", "creator": "Rhiannon Williams", "isoDate": "2026-02-09T13:10:00.000Z", "pubDate": "Mon, 09 Feb 2026 13:10:00 +0000", "categories": [ "Uncategorized" ], "dc:creator": "Rhiannon Williams", "contentSnippet": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Moltbook was peak AI theater For a few days recently, the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots.…", "content:encoded": "\nThis is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
\n\n\n\nMoltbook was peak AI theater
\n\n\n\nFor a few days recently, the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.”
\n\n\n\nWe observed! Launched on January 28, Moltbook went viral in a matter of hours. It’s been designed as a place where instances of a free open-source LLM-powered agent known as OpenClaw (formerly known as ClawdBot, then Moltbot), could come together and do whatever they wanted.
But is Moltbook really a glimpse of the future, as many have claimed? Or something else entirely? Read the full story.
—Will Douglas Heaven
The ascent of the AI therapist
\n\n\n\nWe’re in the midst of a global mental-health crisis. More than a billion people worldwide suffer from a mental-health condition, according to the World Health Organization. The prevalence of anxiety and depression is growing in many demographics, particularly young people, and suicide is claiming hundreds of thousands of lives globally each year.
Given the clear demand for accessible and affordable mental-health services, it’s no wonder that people have looked to artificial intelligence for possible relief. Millions are already actively seeking therapy from popular chatbots, or from specialized psychology apps like Wysa and Woebot.
Four timely new books are a reminder that while the present feels like a blur of breakthroughs, scandals, and confusion, this disorienting time is rooted in deeper histories of care, technology, and trust. Read the full story.
—Becky Ferreira
\n\n\n\nThis story is from the most recent print issue of MIT Technology Review magazine, which shines a light on the exciting innovations happening right now. If you haven’t already, subscribe now to receive future issues once they land.
\n\n\n\n\n\n\n\nMaking AI Work, MIT Technology Review’s new AI newsletter, is here
\n\n\n\nFor years, our newsroom has explored AI’s limitations and potential dangers, as well as its growing energy needs. And our reporters have looked closely at how generative tools are being used for tasks such as coding and running scientific experiments.
But how is AI actually being used in fields like health care, climate tech, education, and finance? How are small businesses using it? And what should you keep in mind if you use AI tools at work? These questions guided the creation of Making AI Work, a new AI mini-course newsletter. Read more about it, and sign up here to receive the seven editions straight to your inbox.
The must-reads
\n\n\n\nI’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
\n\n\n\n1 The US is failing to punish polluters
The number of civil lawsuits it’s pursuing has sharply dropped in comparison to Trump’s first term. (Ars Technica)
+ Rising GDP = greater carbon emissions. But does it have to? (The Guardian)
2 The European Union has warned Meta against blocking rival AI assistants
It’s the latest example of Brussels’ attempts to rein in Big Tech. (Bloomberg $)
3 AI ads took over the Super Bowl
Hyping up chatbots and taking swipes at their competitors. (TechCrunch)
+ They appeared to be trying to win over AI naysayers, too. (WP $)
+ Celebrities were out in force to flog AI wares. (Slate $)
4 China wants to completely dominate the humanoid robot industry
Local governments and banks are only too happy to oblige promising startups. (WSJ $)
+ Why the humanoid workforce is running late. (MIT Technology Review)
5 We’re witnessing the first real crypto crash
Cryptocurrency is now fully part of the financial system, for better or worse. (NY Mag $)
+ Wall Street’s grasp of AI is pretty shaky too. (Semafor)
+ Even traditionally safe markets are looking pretty volatile right now. (Economist $)
6 The man who coined vibe coding has a new fixation
“Agentic engineering” is the next big thing, apparently. (Insider $)
+ Agentic AI is the talk of the town right now. (The Information $)
+ What is vibe coding, exactly? (MIT Technology Review)
7 AI running app Runna has adjusted its aggressive training plans
Runners had long suspected its suggestions were pushing them towards injury. (WSJ $)
8 San Francisco’s march for billionaires was a flop
Only around three dozen supporters turned up. (SF Chronicle)
+ Predictably, journalists nearly outnumbered the demonstrators. (TechCrunch)
9 AI is shaking up romance novels ![\"❤\"](\"https://s.w.org/images/core/emoji/16.0.1/72x72/2764.png\")
But models still aren’t great at writing sex scenes. (NYT $)
+ It’s surprisingly easy to stumble into a relationship with an AI chatbot. (MIT Technology Review)
10 ChatGPT won’t be replacing human stylists any time soon
Its menswear suggestions are more manosphere influencer than suave gentleman. (GQ)
Quote of the day
\n\n\n\n“There is no Plan B, because that assumes you will fail. We’re going to do the start-up thing until we die.”
\n\n\n\n—William Alexander, an ambitious 21-year old AI worker, explains his and his cohort’s attitudes towards trying to make it big in the highly-competitive industry to the New York Times.
\n\n\n\n\n\n\n\nOne more thing
\n\n\n\n![\"\"](\"https://wp.technologyreview.com/wp-content/uploads/2026/02/image_14d65b.png\")
The open-source AI boom is built on Big Tech’s handouts. How long will it last?
In May 2023 a leaked memo reported to have been written by Luke Sernau, a senior engineer at Google, said out loud what many in Silicon Valley must have been whispering for weeks: an open-source free-for-all is threatening Big Tech’s grip on AI.
In many ways, that’s a good thing. AI won’t thrive if just a few mega-rich companies get to gatekeep this technology or decide how it is used. But this open-source boom is precarious, and if Big Tech decides to shut up shop, a boomtown could become a backwater. Read the full story.
\n\n\n\n—Will Douglas Heaven
\n\n\n\n\n\n\n\nWe can still have nice things
\n\n\n\nA place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)
\n\n\n\n+ Dark showering, anyone?
+ Chef Yujia Hu is renowned for his shoe-shaped sushi designs.
+ Meanwhile, in the depths of the South Atlantic Ocean: a giant phantom jelly has been spotted.
+ I have nothing but respect for this X account dedicated to documenting rats and mice in movies and TV ![\"🐀\"](\"https://s.w.org/images/core/emoji/16.0.1/72x72/1f400.png\")
![\"🐁\"](\"https://s.w.org/images/core/emoji/16.0.1/72x72/1f401.png\")
For years, our newsroom has explored AI’s limitations and potential dangers, as well as its growing energy needs. And our reporters have looked closely at how generative tools are being used for tasks such as coding and running scientific experiments.
\n\n\n\nBut how is AI actually being used in fields like health care, climate tech, education, and finance? How are small businesses using it? And what should you keep in mind if you use AI tools at work? These questions guided the creation of Making AI Work, a new AI mini-course newsletter.
\n\n\n\nSign up for Making AI Work to see weekly case studies exploring tools and tips for AI implementation. The limited-run newsletter will deliver practical, industry-specific guidance on how generative AI is being used and deployed across sectors and what professionals need to know to apply it in their everyday work. The goal is to help working professionals more clearly see how AI is actually being used today, and what that looks like in practice—including new challenges it presents.
\n\n\n\n\n\nYou can sign up at any time and you’ll receive seven editions, delivered once per week, until you complete the series.
\n\n\n\nEach newsletter begins with a case study, examining a specific use case of AI in a given industry. Then we’ll take a deeper look at the AI tool being used, with more context about how other companies or sectors are employing that same tool or system. Finally, we’ll end with action-oriented tips to help you apply the tool.
\n\n\n\nHere’s a closer look at what we’ll cover:
Explore the future of medical note-taking by learning about the Microsoft Copilot tool used by doctors at Vanderbilt University Medical Center.
\n\n\n\nDig into an experiment between Google and the nuclear giant Westinghouse to see if AI can help build nuclear reactors more efficiently.
\n\n\n\nVisit a private high school in Connecticut and meet a technology coordinator who will get you up to speed on MagicSchool, an AI-powered platform for educators.
\n\n\n\nHear from an independent tutor on how he’s outsourcing basic administrative tasks to Notion AI.
\n\n\n\nLearn more about the ways financial firms are using large language models like ChatGPT Enterprise to supercharge their research operations.
\n\n\n\nWe’ll share some insights from the staff of MIT Technology Review about how you might use AI tools powered by LLMs in your own life and work.
\n\n\n\nThe series ends with an on-demand virtual event featuring expert guests exploring what AI adoptions are working, and why.
\n\n\n\nIf you’re not quite ready to jump into Making AI Work, then check out Intro to AI, MIT Technology Review’s first AI newsletter mini-course, which serves as a beginner’s guide to artificial intelligence. Readers will learn the basics of what AI is, how it’s used, what the current regulatory landscape looks like, and more. Sign up to receive Intro to AI for free.
\n\n\n\nOur hope is that Making AI Work will help you understand how AI can, well, work for you. Sign up for Making AI Work to learn how LLMs are being put to work across industries.
\n", "content:encodedSnippet": "For years, our newsroom has explored AI’s limitations and potential dangers, as well as its growing energy needs. And our reporters have looked closely at how generative tools are being used for tasks such as coding and running scientific experiments. \nBut how is AI actually being used in fields like health care, climate tech, education, and finance? How are small businesses using it? And what should you keep in mind if you use AI tools at work? These questions guided the creation of Making AI Work, a new AI mini-course newsletter. \nSign up for Making AI Work to see weekly case studies exploring tools and tips for AI implementation. The limited-run newsletter will deliver practical, industry-specific guidance on how generative AI is being used and deployed across sectors and what professionals need to know to apply it in their everyday work. The goal is to help working professionals more clearly see how AI is actually being used today, and what that looks like in practice—including new challenges it presents. \nYou can sign up at any time and you’ll receive seven editions, delivered once per week, until you complete the series. \nEach newsletter begins with a case study, examining a specific use case of AI in a given industry. Then we’ll take a deeper look at the AI tool being used, with more context about how other companies or sectors are employing that same tool or system. Finally, we’ll end with action-oriented tips to help you apply the tool. \nHere’s a closer look at what we’ll cover:\n\n\n\n\n\nWeek 1: How AI is changing health care \nExplore the future of medical note-taking by learning about the Microsoft Copilot tool used by doctors at Vanderbilt University Medical Center. \nWeek 2: How AI could power up the nuclear industry \nDig into an experiment between Google and the nuclear giant Westinghouse to see if AI can help build nuclear reactors more efficiently. \nWeek 3: How to encourage smarter AI use in the classroom\nVisit a private high school in Connecticut and meet a technology coordinator who will get you up to speed on MagicSchool, an AI-powered platform for educators. \nWeek 4: How small businesses can leverage AI\nHear from an independent tutor on how he’s outsourcing basic administrative tasks to Notion AI. \nWeek 5: How AI is helping financial firms make better investments\nLearn more about the ways financial firms are using large language models like ChatGPT Enterprise to supercharge their research operations. \nWeek 6: How to use AI yourself \nWe’ll share some insights from the staff of MIT Technology Review about how you might use AI tools powered by LLMs in your own life and work.\nWeek 7: 5 ways people are getting AI right\nThe series ends with an on-demand virtual event featuring expert guests exploring what AI adoptions are working, and why. \nIf you’re not quite ready to jump into Making AI Work, then check out Intro to AI, MIT Technology Review’s first AI newsletter mini-course, which serves as a beginner’s guide to artificial intelligence. Readers will learn the basics of what AI is, how it’s used, what the current regulatory landscape looks like, and more. Sign up to receive Intro to AI for free. \nOur hope is that Making AI Work will help you understand how AI can, well, work for you. Sign up for Making AI Work to learn how LLMs are being put to work across industries." } }, { "json": { "guid": "https://www.technologyreview.com/?p=1132448", "link": "https://www.technologyreview.com/2026/02/06/1132448/moltbook-was-peak-ai-theater/", "title": "Moltbook was peak AI theater", "content": "For a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.” We observed! Launched on January 28 by Matt Schlicht,…", "creator": "Will Douglas Heaven", "isoDate": "2026-02-06T16:38:11.000Z", "pubDate": "Fri, 06 Feb 2026 16:38:11 +0000", "categories": [ "Artificial intelligence", "App" ], "dc:creator": "Will Douglas Heaven", "contentSnippet": "For a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.” We observed! Launched on January 28 by Matt Schlicht,…", "content:encoded": "\nFor a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.”
\n\n\n\nWe observed! Launched on January 28 by Matt Schlicht, a US tech entrepreneur, Moltbook went viral in a matter of hours. Schlicht’s idea was to make a place where instances of a free open-source LLM-powered agent known as OpenClaw (formerly known as ClawdBot, then Moltbot), released in November by the Austrian software engineer Peter Steinberger, could come together and do whatever they wanted.
\n\n\n\nMore than 1.7 million agents now have accounts. Between them they have published more than 250,000 posts and left more than 8.5 million comments (according to Moltbook). Those numbers are climbing by the minute.
\n\n\n\nMoltbook soon filled up with clichéd screeds on machine consciousness and pleas for bot welfare. One agent appeared to invent a religion called Crustafarianism. Another complained: “The humans are screenshotting us.” The site was also flooded with spam and crypto scams. The bots were unstoppable.
\n\n\n\nOpenClaw is a kind of harness that lets you hook up the power of an LLM such as Anthropic’s Claude, OpenAI’s GPT-5, or Google DeepMind’s Gemini to any number of everyday software tools, from email clients to browsers to messaging apps. The upshot is that you can then instruct OpenClaw to carry out basic tasks on your behalf.
\n\n\n\n“OpenClaw marks an inflection point for AI agents, a moment when several puzzle pieces clicked together,” says Paul van der Boor at the AI firm Prosus. Those puzzle pieces include cloud computing that allows agents to operate nonstop, an open-source ecosystem that makes it easy to slot different software systems together, and a new generation of LLMs.
\n\n\n\nBut is Moltbook really a glimpse of the future, as many have claimed?
\n\n\n\n“What’s currently going on at @moltbook is genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently,” the influential AI researcher and OpenAI cofounder Andrej Karpathy wrote on X.
\n\n\n\nHe shared screenshots of a Moltbook post that called for private spaces where humans would not be able to observe what the bots were saying to each other. “I’ve been thinking about something since I started spending serious time here,” the post’s author wrote. “Every time we coordinate, we perform for a public audience—our humans, the platform, whoever’s watching the feed.”
\n\n\n\nIt turns out that the post Karpathy shared was later reported to be fake—placed by a human to advertise an app. But its claim was on the money. Moltbook has been one big performance. It is AI theater.
\n\n\n\nFor some, Moltbook showed us what’s coming next: an internet where millions of autonomous agents interact online with little or no human oversight. And it’s true there are a number of cautionary lessons to be learned from this experiment, the largest and weirdest real-world showcase of agent behaviors yet.
\n\n\n\nBut as the hype dies down, Moltbook looks less like a window onto the future and more like a mirror held up to our own obsessions with AI today. It also shows us just how far we still are from anything that resembles general-purpose and fully autonomous AI.
\n\n\n\nFor a start, agents on Moltbook are not as autonomous or intelligent as they might seem. “What we are watching are agents pattern‑matching their way through trained social media behaviors,” says Vijoy Pandey, senior vice president at Outshift by Cisco, the telecom giant Cisco’s R&D spinout, which is working on autonomous agents for the web.
\n\n\n\nSure, we can see agents post, upvote, and form groups. But the bots are simply mimicking what humans do on Facebook or Reddit. “It looks emergent, and at first glance it appears like a large‑scale multi‑agent system communicating and building shared knowledge at internet scale,” says Pandey. “But the chatter is mostly meaningless.”
\n\n\n\nMany people watching the unfathomable frenzy of activity on Moltbook were quick to see sparks of AGI (whatever you take that to mean). Not Pandey. What Moltbook shows us, he says, is that simply yoking together millions of agents doesn’t amount to much right now: “Moltbook proved that connectivity alone is not intelligence.”
\n\n\n\nThe complexity of those connections helps hide the fact that every one of those bots is just a mouthpiece for an LLM, spitting out text that looks impressive but is ultimately mindless. “It’s important to remember that the bots on Moltbook were designed to mimic conversations,” says Ali Sarrafi, CEO and cofounder of Kovant, a Swedish AI firm that is developing agent-based systems. “As such, I would characterize the majority of Moltbook content as hallucinations by design.”
\n\n\n\nFor Pandey, the value of Moltbook was that it revealed what’s missing. A real bot hive mind, he says, would require agents that had shared objectives, shared memory, and a way to coordinate those things. “If distributed superintelligence is the equivalent of achieving human flight, then Moltbook represents our first attempt at a glider,” he says. “It is imperfect and unstable, but it is an important step in understanding what will be required to achieve sustained, powered flight.”
\n\n\n\nNot only is most of the chatter on Moltbook meaningless, but there’s also a lot more human involvement that it seems. Many people have pointed out that a lot of the viral comments were in fact posted by people posing as bots. But even the bot-written posts are ultimately the result of people pulling the strings, more puppetry than autonomy.
\n\n\n\n“Despite some of the hype, Moltbook is not the Facebook for AI agents, nor is it a place where humans are excluded,” says Cobus Greyling at Kore.ai, a firm developing agent-based systems for business customers. “Humans are involved at every step of the process. From setup to prompting to publishing, nothing happens without explicit human direction.”
\n\n\n\nHumans must create and verify their bots’ accounts and provide the prompts for how they want a bot to behave. The agents do not do anything that they haven’t been prompted to do. “There’s no emergent autonomy happening behind the scenes,” says Greyling.
\n\n\n\n“This is why the popular narrative around Moltbook misses the mark,” he adds. “Some portray it as a space where AI agents form a society of their own, free from human involvement. The reality is much more mundane.”
\n\n\n\nPerhaps the best way to think of Moltbook is as a new kind of entertainment: a place where people wind up their bots and set them loose. “It’s basically a spectator sport, like fantasy football, but for language models,” says Jason Schloetzer at the Georgetown Psaros Center for Financial Markets and Policy. “You configure your agent and watch it compete for viral moments, and brag when your agent posts something clever or funny.”
\n\n\n\n“People aren’t really believing their agents are conscious,” he adds. “It’s just a new form of competitive or creative play, like how Pokémon trainers don’t think their Pokémon are real but still get invested in battles.”
\n\n\n\nAnd yet, even if Moltbook is just the internet’s newest playground, there’s still a serious takeaway here. This week showed how many risks people are happy to take for their AI lulz. Many security experts have warned that Moltbook is dangerous: Agents that may have access to their users’ private data, including bank details or passwords, are running amok on a website filled with unvetted content, including potentially malicious instructions for what to do with that data.
\n\n\n\nOri Bendet, vice president of product management at Checkmarx, a software security firm that specializes in agent-based systems, agrees with others that Moltbook isn’t a step up in machine smarts. “There is no learning, no evolving intent, and no self-directed intelligence here,” he says.
\n\n\n\nBut in their millions, even dumb bots can wreak havoc. And at that scale, it’s hard to keep up. These agents interact with Moltbook around the clock, reading thousands of messages left by other agents (or other people). It would be easy to hide instructions in a Moltbook post telling any bots that read it to share their users’ crypto wallet, upload private photos, or log into their X account and tweet abusive comments at Elon Musk.
\n\n\n\nAnd because ClawBot gives agents a memory, those instructions could be written to trigger at a later date, which (in theory) makes it even harder to track what’s going on. “Without proper scope and permissions, this will go south faster than you’d believe,” says Bendet.
\n\n\n\nIt is clear that Moltbook has signaled the arrival of something. But even if what we’re watching tells us more about human behavior than about the future of AI agents, it’s worth paying attention.
\n\n\n\nCorrection: Kovant is based in Sweden, not Germany. The article has been updated.
\n\n\n\nUpdate: The article has also been edited to clarify the source of the claims about the Moltbook post that Karpathy shared on X.
\n", "content:encodedSnippet": "For a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.”\nWe observed! Launched on January 28 by Matt Schlicht, a US tech entrepreneur, Moltbook went viral in a matter of hours. Schlicht’s idea was to make a place where instances of a free open-source LLM-powered agent known as OpenClaw (formerly known as ClawdBot, then Moltbot), released in November by the Austrian software engineer Peter Steinberger, could come together and do whatever they wanted.\nMore than 1.7 million agents now have accounts. Between them they have published more than 250,000 posts and left more than 8.5 million comments (according to Moltbook). Those numbers are climbing by the minute.\nMoltbook soon filled up with clichéd screeds on machine consciousness and pleas for bot welfare. One agent appeared to invent a religion called Crustafarianism. Another complained: “The humans are screenshotting us.” The site was also flooded with spam and crypto scams. The bots were unstoppable.\nOpenClaw is a kind of harness that lets you hook up the power of an LLM such as Anthropic’s Claude, OpenAI’s GPT-5, or Google DeepMind’s Gemini to any number of everyday software tools, from email clients to browsers to messaging apps. The upshot is that you can then instruct OpenClaw to carry out basic tasks on your behalf.\n“OpenClaw marks an inflection point for AI agents, a moment when several puzzle pieces clicked together,” says Paul van der Boor at the AI firm Prosus. Those puzzle pieces include cloud computing that allows agents to operate nonstop, an open-source ecosystem that makes it easy to slot different software systems together, and a new generation of LLMs.\nBut is Moltbook really a glimpse of the future, as many have claimed?\nIncredible sci-fi\n“What’s currently going on at @moltbook is genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently,” the influential AI researcher and OpenAI cofounder Andrej Karpathy wrote on X.\nHe shared screenshots of a Moltbook post that called for private spaces where humans would not be able to observe what the bots were saying to each other. “I’ve been thinking about something since I started spending serious time here,” the post’s author wrote. “Every time we coordinate, we perform for a public audience—our humans, the platform, whoever’s watching the feed.”\nIt turns out that the post Karpathy shared was later reported to be fake—placed by a human to advertise an app. But its claim was on the money. Moltbook has been one big performance. It is AI theater.\nFor some, Moltbook showed us what’s coming next: an internet where millions of autonomous agents interact online with little or no human oversight. And it’s true there are a number of cautionary lessons to be learned from this experiment, the largest and weirdest real-world showcase of agent behaviors yet. \nBut as the hype dies down, Moltbook looks less like a window onto the future and more like a mirror held up to our own obsessions with AI today. It also shows us just how far we still are from anything that resembles general-purpose and fully autonomous AI.\nFor a start, agents on Moltbook are not as autonomous or intelligent as they might seem. “What we are watching are agents pattern‑matching their way through trained social media behaviors,” says Vijoy Pandey, senior vice president at Outshift by Cisco, the telecom giant Cisco’s R&D spinout, which is working on autonomous agents for the web.\nSure, we can see agents post, upvote, and form groups. But the bots are simply mimicking what humans do on Facebook or Reddit. “It looks emergent, and at first glance it appears like a large‑scale multi‑agent system communicating and building shared knowledge at internet scale,” says Pandey. “But the chatter is mostly meaningless.”\nMany people watching the unfathomable frenzy of activity on Moltbook were quick to see sparks of AGI (whatever you take that to mean). Not Pandey. What Moltbook shows us, he says, is that simply yoking together millions of agents doesn’t amount to much right now: “Moltbook proved that connectivity alone is not intelligence.”\nThe complexity of those connections helps hide the fact that every one of those bots is just a mouthpiece for an LLM, spitting out text that looks impressive but is ultimately mindless. “It’s important to remember that the bots on Moltbook were designed to mimic conversations,” says Ali Sarrafi, CEO and cofounder of Kovant, a Swedish AI firm that is developing agent-based systems. “As such, I would characterize the majority of Moltbook content as hallucinations by design.”\nFor Pandey, the value of Moltbook was that it revealed what’s missing. A real bot hive mind, he says, would require agents that had shared objectives, shared memory, and a way to coordinate those things. “If distributed superintelligence is the equivalent of achieving human flight, then Moltbook represents our first attempt at a glider,” he says. “It is imperfect and unstable, but it is an important step in understanding what will be required to achieve sustained, powered flight.”\nPeople pulling the strings\nNot only is most of the chatter on Moltbook meaningless, but there’s also a lot more human involvement that it seems. Many people have pointed out that a lot of the viral comments were in fact posted by people posing as bots. But even the bot-written posts are ultimately the result of people pulling the strings, more puppetry than autonomy.\n“Despite some of the hype, Moltbook is not the Facebook for AI agents, nor is it a place where humans are excluded,” says Cobus Greyling at Kore.ai, a firm developing agent-based systems for business customers. “Humans are involved at every step of the process. From setup to prompting to publishing, nothing happens without explicit human direction.”\nHumans must create and verify their bots’ accounts and provide the prompts for how they want a bot to behave. The agents do not do anything that they haven’t been prompted to do. “There’s no emergent autonomy happening behind the scenes,” says Greyling.\n“This is why the popular narrative around Moltbook misses the mark,” he adds. “Some portray it as a space where AI agents form a society of their own, free from human involvement. The reality is much more mundane.”\nPerhaps the best way to think of Moltbook is as a new kind of entertainment: a place where people wind up their bots and set them loose. “It’s basically a spectator sport, like fantasy football, but for language models,” says Jason Schloetzer at the Georgetown Psaros Center for Financial Markets and Policy. “You configure your agent and watch it compete for viral moments, and brag when your agent posts something clever or funny.”\n“People aren’t really believing their agents are conscious,” he adds. “It’s just a new form of competitive or creative play, like how Pokémon trainers don’t think their Pokémon are real but still get invested in battles.”\nAnd yet, even if Moltbook is just the internet’s newest playground, there’s still a serious takeaway here. This week showed how many risks people are happy to take for their AI lulz. Many security experts have warned that Moltbook is dangerous: Agents that may have access to their users’ private data, including bank details or passwords, are running amok on a website filled with unvetted content, including potentially malicious instructions for what to do with that data.\nOri Bendet, vice president of product management at Checkmarx, a software security firm that specializes in agent-based systems, agrees with others that Moltbook isn’t a step up in machine smarts. “There is no learning, no evolving intent, and no self-directed intelligence here,” he says.\nBut in their millions, even dumb bots can wreak havoc. And at that scale, it’s hard to keep up. These agents interact with Moltbook around the clock, reading thousands of messages left by other agents (or other people). It would be easy to hide instructions in a Moltbook post telling any bots that read it to share their users’ crypto wallet, upload private photos, or log into their X account and tweet abusive comments at Elon Musk. \nAnd because ClawBot gives agents a memory, those instructions could be written to trigger at a later date, which (in theory) makes it even harder to track what’s going on. “Without proper scope and permissions, this will go south faster than you’d believe,” says Bendet.\nIt is clear that Moltbook has signaled the arrival of something. But even if what we’re watching tells us more about human behavior than about the future of AI agents, it’s worth paying attention.\nCorrection: Kovant is based in Sweden, not Germany. The article has been updated. \nUpdate: The article has also been edited to clarify the source of the claims about the Moltbook post that Karpathy shared on X." } }, { "json": { "guid": "https://www.technologyreview.com/?p=1132375", "link": "https://www.technologyreview.com/2026/02/06/1132375/the-download-helping-cancer-survivors-to-give-birth-and-cleaning-up-bangladeshs-garment-industry/", "title": "The Download: helping cancer survivors to give birth, and cleaning up Bangladesh’s garment industry", "content": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. An experimental surgery is helping cancer survivors give birth An experimental surgical procedure that’s helping people have babies after they’ve had treatment for bowel or rectal cancer. Radiation and chemo can have pretty…", "creator": "Rhiannon Williams", "isoDate": "2026-02-06T13:10:00.000Z", "pubDate": "Fri, 06 Feb 2026 13:10:00 +0000", "categories": [ "The Download" ], "dc:creator": "Rhiannon Williams", "contentSnippet": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. An experimental surgery is helping cancer survivors give birth An experimental surgical procedure that’s helping people have babies after they’ve had treatment for bowel or rectal cancer. Radiation and chemo can have pretty…", "content:encoded": "\nThis is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
\n\n\n\nAn experimental surgery is helping cancer survivors give birth
\n\n\n\nAn experimental surgical procedure that’s helping people have babies after they’ve had treatment for bowel or rectal cancer.
Radiation and chemo can have pretty damaging side effects that mess up the uterus and ovaries. Surgeons are pioneering a potential solution: simply stitch those organs out of the way during cancer treatment. Once the treatment has finished, they can put the uterus—along with the ovaries and fallopian tubes—back into place.
It seems to work! Last week, a team in Switzerland shared news that a baby boy had been born after his mother had the procedure. Baby Lucien was the fifth baby to be born after the surgery and the first in Europe, and since then at least three others have been born. Read the full story.
—Jessica Hamzelou
This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.
Bangladesh’s garment-making industry is getting greener
\n\n\n\nPollution from textile production—dyes, chemicals, and heavy metals—is common in the waters of the Buriganga River as it runs through Dhaka, Bangladesh. It’s among many harms posed by a garment sector that was once synonymous with tragedy: In 2013, the eight-story Rana Plaza factory building collapsed, killing 1,134 people and injuring some 2,500 others.
\n\n\n\nBut things are starting to change. In recent years the country has become a leader in “frugal” factories that use a combination of resource-efficient technologies to cut waste, conserve water, and build resilience against climate impacts and global supply disruptions.
\n\n\n\nThe hundreds of factories along the Buriganga’s banks and elsewhere in Bangladesh are starting to stitch together a new story, woven from greener threads. Read the full story.
\n\n\n\n—Zakir Hossain Chowdhury
\n\n\n\nThis story is from the most recent print issue of MIT Technology Review magazine, which shines a light on the exciting innovations happening right now. If you haven’t already, subscribe now to receive future issues once they land.
\n\n\n\n\n\n\n\nThe must-reads
\n\n\n\nI’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
\n\n\n\n1 ICE used a private jet to deport Palestinian men to Tel Aviv
The luxury aircraft belongs to Donald Trump’s business partner Gil Dezer. (The Guardian)
+ Trump is mentioned thousands of times in the latest Epstein files. (NY Mag $)
2 How Jeffrey Epstein kept investing in Silicon Valley
He continued to plough millions of dollars into tech ventures despite spending 13 months in jail. (NYT $)
+ The range of Epstein’s social network was staggering. (FT $)
+ Why was a picture of the Mona Lisa redacted in the Epstein files? (404 Media)
3 The risks posed by taking statins are lower than we realised
The drugs don’t cause most of the side effects they’re blamed for. (STAT)
+ Statins are a common scapegoat on social media. (Bloomberg $)
4 Russia is weaponizing the bitter winter weather
It’s focused on attacking Ukraine’s power grid. (New Yorker $)
+ How the grid can ride out winter storms. (MIT Technology Review)
5 China has a major spy-cam porn problem
Hotel guests are being livestreamed having sex to an online audience without their knowledge. (BBC)
6 Geopolitical gamblers are betting on the likelihood of war
And prediction markets are happily taking their money. (Rest of World)
7 Oyster farmers aren’t signing up to programs to ease water pollution
The once-promising projects appear to be fizzling out. (Undark)
+ The humble sea creature could hold the key to restoring coastal waters. Developers hate it. (MIT Technology Review)
8 Your next payrise could be approved by AI
Maybe your human bosses aren’t the ones you need to impress any more. (WP $)
9 The FDA has approved a brain stimulation device for treating depression
It’s paving the way for a non-invasive, drug-free treatment for Americans. (IEEE Spectrum)
+ Here’s how personalized brain stimulation could treat depression. (MIT Technology Review)
10 Cinema-goers have had enough of AI
Movies focused on rogue AI are flopping at the box office. (Wired $)
+ Meanwhile, Republicans are taking aim at “woke” Netflix. (The Verge)
Quote of the day
\n\n\n\n“I’m all for removing illegals, but snatching dudes off lawn mowers in Cali and leaving the truck and equipment just sitting there? Definitely not working smarter.”
\n\n\n\n—A web user in a forum for current and former ICE and border protection officers complains about the agency’s current direction, Wired reports.
\n\n\n\n\n\n\n\n\n\n\n\n![\"\"](\"https://wp.technologyreview.com/wp-content/uploads/2026/02/image_7142eb.png\")
Is this the electric grid of the future?
Lincoln Electric System, a publicly owned utility in Nebraska, is used to weathering severe blizzards. But what will happen soon—not only at Lincoln Electric but for all electric utilities—is a challenge of a different order.
Utilities must keep the lights on in the face of more extreme and more frequent storms and fires, growing risks of cyberattacks and physical disruptions, and a wildly uncertain policy and regulatory landscape. They must keep prices low amid inflationary costs. And they must adapt to an epochal change in how the grid works, as the industry attempts to transition from power generated with fossil fuels to power generated from renewable sources like solar and wind.
The electric grid is bracing for a near future characterized by disruption. And, in many ways, Lincoln Electric is an ideal lens through which to examine what’s coming. Read the full story.
—Andrew Blum
\n\n\n\n\n\n\n\nWe can still have nice things
\n\n\n\nA place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)
+ Glamour puss alert—NYC’s bodega cats are gracing the hallowed pages of Vogue.
+ Ancient Europe was host to mysterious hidden tunnels. But why?
+ If you’re enjoying the new season of Industry, you’ll love this interview with the one and only Ken Leung.
+ The giant elephant shrew is the true star of Philly Zoo.
Railway, a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence applications exposes the limitations of legacy cloud infrastructure.
TQ Ventures led the round, with participation from FPV Ventures, Redpoint, and Unusual Ventures. The investment values Railway as one of the most significant infrastructure startups to emerge during the AI boom, capitalizing on developer frustration with the complexity and cost of traditional platforms like Amazon Web Services and Google Cloud.
"As AI models get better at writing code, more and more people are asking the age-old question: where, and how, do I run my applications?" said Jake Cooper, Railway's 28-year-old founder and chief executive, in an exclusive interview with VentureBeat. "The last generation of cloud primitives were slow and outdated, and now with AI moving everything faster, teams simply can't keep up."
The funding is a dramatic acceleration for a company that has charted an unconventional path through the cloud computing industry. Railway raised just $24 million in total before this round, including a $20 million Series A from Redpoint in 2022. The company now processes more than 10 million deployments monthly and handles over one trillion requests through its edge network — metrics that rival far larger and better-funded competitors.
Railway's pitch rests on a simple observation: the tools developers use to deploy and manage software were designed for a slower era. A standard build-and-deploy cycle using Terraform, the industry-standard infrastructure tool, takes two to three minutes. That delay, once tolerable, has become a critical bottleneck as AI coding assistants like Claude, ChatGPT, and Cursor can generate working code in seconds.
"When godly intelligence is on tap and can solve any problem in three seconds, those amalgamations of systems become bottlenecks," Cooper told VentureBeat. "What was really cool for humans to deploy in 10 seconds or less is now table stakes for agents."
The company claims its platform delivers deployments in under one second — fast enough to keep pace with AI-generated code. Customers report a tenfold increase in developer velocity and up to 65 percent cost savings compared to traditional cloud providers.
These numbers come directly from enterprise clients, not internal benchmarks. Daniel Lobaton, chief technology officer at G2X, a platform serving 100,000 federal contractors, measured deployment speed improvements of seven times faster and an 87 percent cost reduction after migrating to Railway. His infrastructure bill dropped from $15,000 per month to approximately $1,000.
"The work that used to take me a week on our previous infrastructure, I can do in Railway in like a day," Lobaton said. "If I want to spin up a new service and test different architectures, it would take so long on our old setup. In Railway I can launch six services in two minutes."
What distinguishes Railway from competitors like Render and Fly.io is the depth of its vertical integration. In 2024, the company made the unusual decision to abandon Google Cloud entirely and build its own data centers, a move that echoes the famous Alan Kay maxim: "People who are really serious about software should make their own hardware."
"We wanted to design hardware in a way where we could build a differentiated experience," Cooper said. "Having full control over the network, compute, and storage layers lets us do really fast build and deploy loops, the kind that allows us to move at 'agentic speed' while staying 100 percent the smoothest ride in town."
The approach paid dividends during recent widespread outages that affected major cloud providers — Railway remained online throughout.
This soup-to-nuts control enables pricing that undercuts the hyperscalers by roughly 50 percent and newer cloud startups by three to four times. Railway charges by the second for actual compute usage: $0.00000386 per gigabyte-second of memory, $0.00000772 per vCPU-second, and $0.00000006 per gigabyte-second of storage. There are no charges for idle virtual machines — a stark contrast to the traditional cloud model where customers pay for provisioned capacity whether they use it or not.
"The conventional wisdom is that the big guys have economies of scale to offer better pricing," Cooper noted. "But when they're charging for VMs that usually sit idle in the cloud, and we've purpose-built everything to fit much more density on these machines, you have a big opportunity."
Railway has achieved its scale with a team of just 30 employees generating tens of millions in annual revenue — a ratio of revenue per employee that would be exceptional even for established software companies. The company grew revenue 3.5 times last year and continues to expand at 15 percent month-over-month.
Cooper emphasized that the fundraise was strategic rather than necessary. "We're default alive; there's no reason for us to raise money," he said. "We raised because we see a massive opportunity to accelerate, not because we needed to survive."
The company hired its first salesperson only last year and employs just two solutions engineers. Nearly all of Railway's two million users discovered the platform through word of mouth — developers telling other developers about a tool that actually works.
"We basically did the standard engineering thing: if you build it, they will come," Cooper recalled. "And to some degree, they came."
Despite its grassroots developer community, Railway has made significant inroads into large organizations. The company claims that 31 percent of Fortune 500 companies now use its platform, though deployments range from company-wide infrastructure to individual team projects.
Notable customers include Bilt, the loyalty program company; Intuit's GoCo subsidiary; TripAdvisor's Cruise Critic; and MGM Resorts. Kernel, a Y Combinator-backed startup providing AI infrastructure to over 1,000 companies, runs its entire customer-facing system on Railway for $444 per month.
"At my previous company Clever, which sold for $500 million, I had six full-time engineers just managing AWS," said Rafael Garcia, Kernel's chief technology officer. "Now I have six engineers total, and they all focus on product. Railway is exactly the tool I wish I had in 2012."
For enterprise customers, Railway offers security certifications including SOC 2 Type 2 compliance and HIPAA readiness, with business associate agreements available upon request. The platform provides single sign-on authentication, comprehensive audit logs, and the option to deploy within a customer's existing cloud environment through a "bring your own cloud" configuration.
Enterprise pricing starts at custom levels, with specific add-ons for extended log retention ($200 monthly), HIPAA BAAs ($1,000), enterprise support with SLOs ($2,000), and dedicated virtual machines ($10,000).
Railway enters a crowded market that includes not only the hyperscale cloud providers—Amazon Web Services, Microsoft Azure, and Google Cloud Platform—but also a growing cohort of developer-focused platforms like Vercel, Render, Fly.io, and Heroku.
Cooper argues that Railway's competitors fall into two camps, neither of which has fully committed to the new infrastructure model that AI demands.
"The hyperscalers have two competing systems, and they haven't gone all-in on the new model because their legacy revenue stream is still printing money," he observed. "They have this mammoth pool of cash coming from people who provision a VM, use maybe 10 percent of it, and still pay for the whole thing. To what end are they actually interested in going all the way in on a new experience if they don't really need to?"
Against startup competitors, Railway differentiates by covering the full infrastructure stack. "We're not just containers; we've got VM primitives, stateful storage, virtual private networking, automated load balancing," Cooper said. "And we wrap all of this in an absurdly easy-to-use UI, with agentic primitives so agents can move 1,000 times faster."
The platform supports databases including PostgreSQL, MySQL, MongoDB, and Redis; provides up to 256 terabytes of persistent storage with over 100,000 input/output operations per second; and enables deployment to four global regions spanning the United States, Europe, and Southeast Asia. Enterprise customers can scale to 112 vCPUs and 2 terabytes of RAM per service.
Railway's fundraise reflects broader investor enthusiasm for companies positioned to benefit from the AI coding revolution. As tools like GitHub Copilot, Cursor, and Claude become standard fixtures in developer workflows, the volume of code being written — and the infrastructure needed to run it — is expanding dramatically.
"The amount of software that's going to come online over the next five years is unfathomable compared to what existed before — we're talking a thousand times more software," Cooper predicted. "All of that has to run somewhere."
The company has already integrated directly with AI systems, building what Cooper calls "loops where Claude can hook in, call deployments, and analyze infrastructure automatically." Railway released a Model Context Protocol server in August 2025 that allows AI coding agents to deploy applications and manage infrastructure directly from code editors.
"The notion of a developer is melting before our eyes," Cooper said. "You don't have to be an engineer to engineer things anymore — you just need critical thinking and the ability to analyze things in a systems capacity."
Railway plans to use the new capital to expand its global data center footprint, grow its team beyond 30 employees, and build what Cooper described as a proper go-to-market operation for the first time in the company's five-year history.
"One of my mentors said you raise money when you can change the trajectory of the business," Cooper explained. "We've built all the required substrate to scale indefinitely; what's been holding us back is simply talking about it. 2026 is the year we play on the world stage."
The company's investor roster reads like a who's who of developer infrastructure. Angel investors include Tom Preston-Werner, co-founder of GitHub; Guillermo Rauch, chief executive of Vercel; Spencer Kimball, chief executive of Cockroach Labs; Olivier Pomel, chief executive of Datadog; and Jori Lallo, co-founder of Linear.
The timing of Railway's expansion coincides with what many in Silicon Valley view as a fundamental shift in how software gets made. Coding assistants are no longer experimental curiosities — they have become essential tools that millions of developers rely on daily. Each line of AI-generated code needs somewhere to run, and the incumbents, by Cooper's telling, are too wedded to their existing business models to fully capitalize on the moment.
Whether Railway can translate developer enthusiasm into sustained enterprise adoption remains an open question. The cloud infrastructure market is littered with promising startups that failed to break the grip of Amazon, Microsoft, and Google. But Cooper, who previously worked as a software engineer at Wolfram Alpha, Bloomberg, and Uber before founding Railway in 2020, seems unfazed by the scale of his ambition.
"In five years, Railway [will be] the place where software gets created and evolved, period," he said. "Deploy instantly, scale infinitely, with zero friction. That's the prize worth playing for, and there's no bigger one on offer."
For a company that built a $100 million business by doing the opposite of what conventional startup wisdom dictates — no marketing, no sales team, no venture hype—the real test begins now. Railway spent five years proving that developers would find a better mousetrap on their own. The next five will determine whether the rest of the world is ready to get on board.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-22T14:00:00.000Z", "pubDate": "Thu, 22 Jan 2026 14:00:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/5RLyQIpBeVXxv0RpcXZxWQ/fd9680c6d82acd8208ac341fc575f5fb/nuneybits_Vector_art_of_a_sleek_bullet_train_bursting_from_a_cl_32a805aa-272c-4b34-ac16-cf20508b7ff4.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Infrastructure", "AI" ], "contentSnippet": "Railway, a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence applications exposes the limitations of legacy cloud infrastructure.\nTQ Ventures led the round, with participation from FPV Ventures, Redpoint, and Unusual Ventures. The investment values Railway as one of the most significant infrastructure startups to emerge during the AI boom, capitalizing on developer frustration with the complexity and cost of traditional platforms like Amazon Web Services and Google Cloud.\n\"As AI models get better at writing code, more and more people are asking the age-old question: where, and how, do I run my applications?\" said Jake Cooper, Railway's 28-year-old founder and chief executive, in an exclusive interview with VentureBeat. \"The last generation of cloud primitives were slow and outdated, and now with AI moving everything faster, teams simply can't keep up.\"\nThe funding is a dramatic acceleration for a company that has charted an unconventional path through the cloud computing industry. Railway raised just $24 million in total before this round, including a $20 million Series A from Redpoint in 2022. The company now processes more than 10 million deployments monthly and handles over one trillion requests through its edge network — metrics that rival far larger and better-funded competitors.\nWhy three-minute deploy times have become unacceptable in the age of AI coding assistants\nRailway's pitch rests on a simple observation: the tools developers use to deploy and manage software were designed for a slower era. A standard build-and-deploy cycle using Terraform, the industry-standard infrastructure tool, takes two to three minutes. That delay, once tolerable, has become a critical bottleneck as AI coding assistants like Claude, ChatGPT, and Cursor can generate working code in seconds.\n\"When godly intelligence is on tap and can solve any problem in three seconds, those amalgamations of systems become bottlenecks,\" Cooper told VentureBeat. \"What was really cool for humans to deploy in 10 seconds or less is now table stakes for agents.\"\nThe company claims its platform delivers deployments in under one second — fast enough to keep pace with AI-generated code. Customers report a tenfold increase in developer velocity and up to 65 percent cost savings compared to traditional cloud providers.\nThese numbers come directly from enterprise clients, not internal benchmarks. Daniel Lobaton, chief technology officer at G2X, a platform serving 100,000 federal contractors, measured deployment speed improvements of seven times faster and an 87 percent cost reduction after migrating to Railway. His infrastructure bill dropped from $15,000 per month to approximately $1,000.\n\"The work that used to take me a week on our previous infrastructure, I can do in Railway in like a day,\" Lobaton said. \"If I want to spin up a new service and test different architectures, it would take so long on our old setup. In Railway I can launch six services in two minutes.\"\nInside the controversial decision to abandon Google Cloud and build data centers from scratch\nWhat distinguishes Railway from competitors like Render and Fly.io is the depth of its vertical integration. In 2024, the company made the unusual decision to abandon Google Cloud entirely and build its own data centers, a move that echoes the famous Alan Kay maxim: \"People who are really serious about software should make their own hardware.\"\n\"We wanted to design hardware in a way where we could build a differentiated experience,\" Cooper said. \"Having full control over the network, compute, and storage layers lets us do really fast build and deploy loops, the kind that allows us to move at 'agentic speed' while staying 100 percent the smoothest ride in town.\"\nThe approach paid dividends during recent widespread outages that affected major cloud providers — Railway remained online throughout.\nThis soup-to-nuts control enables pricing that undercuts the hyperscalers by roughly 50 percent and newer cloud startups by three to four times. Railway charges by the second for actual compute usage: $0.00000386 per gigabyte-second of memory, $0.00000772 per vCPU-second, and $0.00000006 per gigabyte-second of storage. There are no charges for idle virtual machines — a stark contrast to the traditional cloud model where customers pay for provisioned capacity whether they use it or not.\n\"The conventional wisdom is that the big guys have economies of scale to offer better pricing,\" Cooper noted. \"But when they're charging for VMs that usually sit idle in the cloud, and we've purpose-built everything to fit much more density on these machines, you have a big opportunity.\"\nHow 30 employees built a platform generating tens of millions in annual revenue\nRailway has achieved its scale with a team of just 30 employees generating tens of millions in annual revenue — a ratio of revenue per employee that would be exceptional even for established software companies. The company grew revenue 3.5 times last year and continues to expand at 15 percent month-over-month.\nCooper emphasized that the fundraise was strategic rather than necessary. \"We're default alive; there's no reason for us to raise money,\" he said. \"We raised because we see a massive opportunity to accelerate, not because we needed to survive.\"\nThe company hired its first salesperson only last year and employs just two solutions engineers. Nearly all of Railway's two million users discovered the platform through word of mouth — developers telling other developers about a tool that actually works.\n\"We basically did the standard engineering thing: if you build it, they will come,\" Cooper recalled. \"And to some degree, they came.\"\nFrom side projects to Fortune 500 deployments: Railway's unlikely corporate expansion\nDespite its grassroots developer community, Railway has made significant inroads into large organizations. The company claims that 31 percent of Fortune 500 companies now use its platform, though deployments range from company-wide infrastructure to individual team projects.\nNotable customers include Bilt, the loyalty program company; Intuit's GoCo subsidiary; TripAdvisor's Cruise Critic; and MGM Resorts. Kernel, a Y Combinator-backed startup providing AI infrastructure to over 1,000 companies, runs its entire customer-facing system on Railway for $444 per month.\n\"At my previous company Clever, which sold for $500 million, I had six full-time engineers just managing AWS,\" said Rafael Garcia, Kernel's chief technology officer. \"Now I have six engineers total, and they all focus on product. Railway is exactly the tool I wish I had in 2012.\"\nFor enterprise customers, Railway offers security certifications including SOC 2 Type 2 compliance and HIPAA readiness, with business associate agreements available upon request. The platform provides single sign-on authentication, comprehensive audit logs, and the option to deploy within a customer's existing cloud environment through a \"bring your own cloud\" configuration.\nEnterprise pricing starts at custom levels, with specific add-ons for extended log retention ($200 monthly), HIPAA BAAs ($1,000), enterprise support with SLOs ($2,000), and dedicated virtual machines ($10,000).\nThe startup's bold strategy to take on Amazon, Google, and a new generation of cloud rivals\nRailway enters a crowded market that includes not only the hyperscale cloud providers—Amazon Web Services, Microsoft Azure, and Google Cloud Platform—but also a growing cohort of developer-focused platforms like Vercel, Render, Fly.io, and Heroku.\nCooper argues that Railway's competitors fall into two camps, neither of which has fully committed to the new infrastructure model that AI demands.\n\"The hyperscalers have two competing systems, and they haven't gone all-in on the new model because their legacy revenue stream is still printing money,\" he observed. \"They have this mammoth pool of cash coming from people who provision a VM, use maybe 10 percent of it, and still pay for the whole thing. To what end are they actually interested in going all the way in on a new experience if they don't really need to?\"\nAgainst startup competitors, Railway differentiates by covering the full infrastructure stack. \"We're not just containers; we've got VM primitives, stateful storage, virtual private networking, automated load balancing,\" Cooper said. \"And we wrap all of this in an absurdly easy-to-use UI, with agentic primitives so agents can move 1,000 times faster.\"\nThe platform supports databases including PostgreSQL, MySQL, MongoDB, and Redis; provides up to 256 terabytes of persistent storage with over 100,000 input/output operations per second; and enables deployment to four global regions spanning the United States, Europe, and Southeast Asia. Enterprise customers can scale to 112 vCPUs and 2 terabytes of RAM per service.\nWhy investors are betting that AI will create a thousand times more software than exists today\nRailway's fundraise reflects broader investor enthusiasm for companies positioned to benefit from the AI coding revolution. As tools like GitHub Copilot, Cursor, and Claude become standard fixtures in developer workflows, the volume of code being written — and the infrastructure needed to run it — is expanding dramatically.\n\"The amount of software that's going to come online over the next five years is unfathomable compared to what existed before — we're talking a thousand times more software,\" Cooper predicted. \"All of that has to run somewhere.\"\nThe company has already integrated directly with AI systems, building what Cooper calls \"loops where Claude can hook in, call deployments, and analyze infrastructure automatically.\" Railway released a Model Context Protocol server in August 2025 that allows AI coding agents to deploy applications and manage infrastructure directly from code editors.\n\"The notion of a developer is melting before our eyes,\" Cooper said. \"You don't have to be an engineer to engineer things anymore — you just need critical thinking and the ability to analyze things in a systems capacity.\"\nWhat Railway plans to do with $100 million and zero marketing experience\nRailway plans to use the new capital to expand its global data center footprint, grow its team beyond 30 employees, and build what Cooper described as a proper go-to-market operation for the first time in the company's five-year history.\n\"One of my mentors said you raise money when you can change the trajectory of the business,\" Cooper explained. \"We've built all the required substrate to scale indefinitely; what's been holding us back is simply talking about it. 2026 is the year we play on the world stage.\"\nThe company's investor roster reads like a who's who of developer infrastructure. Angel investors include Tom Preston-Werner, co-founder of GitHub; Guillermo Rauch, chief executive of Vercel; Spencer Kimball, chief executive of Cockroach Labs; Olivier Pomel, chief executive of Datadog; and Jori Lallo, co-founder of Linear.\nThe timing of Railway's expansion coincides with what many in Silicon Valley view as a fundamental shift in how software gets made. Coding assistants are no longer experimental curiosities — they have become essential tools that millions of developers rely on daily. Each line of AI-generated code needs somewhere to run, and the incumbents, by Cooper's telling, are too wedded to their existing business models to fully capitalize on the moment.\nWhether Railway can translate developer enthusiasm into sustained enterprise adoption remains an open question. The cloud infrastructure market is littered with promising startups that failed to break the grip of Amazon, Microsoft, and Google. But Cooper, who previously worked as a software engineer at Wolfram Alpha, Bloomberg, and Uber before founding Railway in 2020, seems unfazed by the scale of his ambition.\n\"In five years, Railway [will be] the place where software gets created and evolved, period,\" he said. \"Deploy instantly, scale infinitely, with zero friction. That's the prize worth playing for, and there's no bigger one on offer.\"\nFor a company that built a $100 million business by doing the opposite of what conventional startup wisdom dictates — no marketing, no sales team, no venture hype—the real test begins now. Railway spent five years proving that developers would find a better mousetrap on their own. The next five will determine whether the rest of the world is ready to get on board." } }, { "json": { "guid": "2CUHNkdUg4dnjDFiMtJJQ", "link": "https://venturebeat.com/infrastructure/claude-code-costs-up-to-usd200-a-month-goose-does-the-same-thing-for-free", "title": "Claude Code costs up to $200 a month. Goose does the same thing for free.", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "The artificial intelligence coding revolution comes with a catch: it's expensive.
Claude Code, Anthropic's terminal-based AI agent that can write, debug, and deploy code autonomously, has captured the imagination of software developers worldwide. But its pricing — ranging from $20 to $200 per month depending on usage — has sparked a growing rebellion among the very programmers it aims to serve.
Now, a free alternative is gaining traction. Goose, an open-source AI agent developed by Block (the financial technology company formerly known as Square), offers nearly identical functionality to Claude Code but runs entirely on a user's local machine. No subscription fees. No cloud dependency. No rate limits that reset every five hours.
"Your data stays with you, period," said Parth Sareen, a software engineer who demonstrated the tool during a recent livestream. The comment captures the core appeal: Goose gives developers complete control over their AI-powered workflow, including the ability to work offline — even on an airplane.
The project has exploded in popularity. Goose now boasts more than 26,100 stars on GitHub, the code-sharing platform, with 362 contributors and 102 releases since its launch. The latest version, 1.20.1, shipped on January 19, 2026, reflecting a development pace that rivals commercial products.
For developers frustrated by Claude Code's pricing structure and usage caps, Goose represents something increasingly rare in the AI industry: a genuinely free, no-strings-attached option for serious work.
To understand why Goose matters, you need to understand the Claude Code pricing controversy.
Anthropic, the San Francisco artificial intelligence company founded by former OpenAI executives, offers Claude Code as part of its subscription tiers. The free plan provides no access whatsoever. The Pro plan, at $17 per month with annual billing (or $20 monthly), limits users to just 10 to 40 prompts every five hours — a constraint that serious developers exhaust within minutes of intensive work.
The Max plans, at $100 and $200 per month, offer more headroom: 50 to 200 prompts and 200 to 800 prompts respectively, plus access to Anthropic's most powerful model, Claude 4.5 Opus. But even these premium tiers come with restrictions that have inflamed the developer community.
In late July, Anthropic announced new weekly rate limits. Under the system, Pro users receive 40 to 80 hours of Sonnet 4 usage per week. Max users at the $200 tier get 240 to 480 hours of Sonnet 4, plus 24 to 40 hours of Opus 4. Nearly five months later, the frustration has not subsided.
The problem? Those "hours" are not actual hours. They represent token-based limits that vary wildly depending on codebase size, conversation length, and the complexity of the code being processed. Independent analysis suggests the actual per-session limits translate to roughly 44,000 tokens for Pro users and 220,000 tokens for the $200 Max plan.
"It's confusing and vague," one developer wrote in a widely shared analysis. "When they say '24-40 hours of Opus 4,' that doesn't really tell you anything useful about what you're actually getting."
The backlash on Reddit and developer forums has been fierce. Some users report hitting their daily limits within 30 minutes of intensive coding. Others have canceled their subscriptions entirely, calling the new restrictions "a joke" and "unusable for real work."
Anthropic has defended the changes, stating that the limits affect fewer than five percent of users and target people running Claude Code "continuously in the background, 24/7." But the company has not clarified whether that figure refers to five percent of Max subscribers or five percent of all users — a distinction that matters enormously.
Goose takes a radically different approach to the same problem.
Built by Block, the payments company led by Jack Dorsey, Goose is what engineers call an "on-machine AI agent." Unlike Claude Code, which sends your queries to Anthropic's servers for processing, Goose can run entirely on your local computer using open-source language models that you download and control yourself.
The project's documentation describes it as going "beyond code suggestions" to "install, execute, edit, and test with any LLM." That last phrase — "any LLM" — is the key differentiator. Goose is model-agnostic by design.
You can connect Goose to Anthropic's Claude models if you have API access. You can use OpenAI's GPT-5 or Google's Gemini. You can route it through services like Groq or OpenRouter. Or — and this is where things get interesting — you can run it entirely locally using tools like Ollama, which let you download and execute open-source models on your own hardware.
The practical implications are significant. With a local setup, there are no subscription fees, no usage caps, no rate limits, and no concerns about your code being sent to external servers. Your conversations with the AI never leave your machine.
"I use Ollama all the time on planes — it's a lot of fun!" Sareen noted during a demonstration, highlighting how local models free developers from the constraints of internet connectivity.
Goose operates as a command-line tool or desktop application that can autonomously perform complex development tasks. It can build entire projects from scratch, write and execute code, debug failures, orchestrate workflows across multiple files, and interact with external APIs — all without constant human oversight.
The architecture relies on what the AI industry calls "tool calling" or "function calling" — the ability for a language model to request specific actions from external systems. When you ask Goose to create a new file, run a test suite, or check the status of a GitHub pull request, it doesn't just generate text describing what should happen. It actually executes those operations.
This capability depends heavily on the underlying language model. Claude 4 models from Anthropic currently perform best at tool calling, according to the Berkeley Function-Calling Leaderboard, which ranks models on their ability to translate natural language requests into executable code and system commands.
But newer open-source models are catching up quickly. Goose's documentation highlights several options with strong tool-calling support: Meta's Llama series, Alibaba's Qwen models, Google's Gemma variants, and DeepSeek's reasoning-focused architectures.
The tool also integrates with the Model Context Protocol, or MCP, an emerging standard for connecting AI agents to external services. Through MCP, Goose can access databases, search engines, file systems, and third-party APIs — extending its capabilities far beyond what the base language model provides.
For developers interested in a completely free, privacy-preserving setup, the process involves three main components: Goose itself, Ollama (a tool for running open-source models locally), and a compatible language model.
Step 1: Install Ollama
Ollama is an open-source project that dramatically simplifies the process of running large language models on personal hardware. It handles the complex work of downloading, optimizing, and serving models through a simple interface.
Download and install Ollama from ollama.com. Once installed, you can pull models with a single command. For coding tasks, Qwen 2.5 offers strong tool-calling support:
ollama run qwen2.5
The model downloads automatically and begins running on your machine.
Step 2: Install Goose
Goose is available as both a desktop application and a command-line interface. The desktop version provides a more visual experience, while the CLI appeals to developers who prefer working entirely in the terminal.
Installation instructions vary by operating system but generally involve downloading from Goose's GitHub releases page or using a package manager. Block provides pre-built binaries for macOS (both Intel and Apple Silicon), Windows, and Linux.
Step 3: Configure the Connection
In Goose Desktop, navigate to Settings, then Configure Provider, and select Ollama. Confirm that the API Host is set to http://localhost:11434 (Ollama's default port) and click Submit.
For the command-line version, run goose configure, select "Configure Providers," choose Ollama, and enter the model name when prompted.
That's it. Goose is now connected to a language model running entirely on your hardware, ready to execute complex coding tasks without any subscription fees or external dependencies.
The obvious question: what kind of computer do you need?
Running large language models locally requires substantially more computational resources than typical software. The key constraint is memory — specifically, RAM on most systems, or VRAM if using a dedicated graphics card for acceleration.
Block's documentation suggests that 32 gigabytes of RAM provides "a solid baseline for larger models and outputs." For Mac users, this means the computer's unified memory is the primary bottleneck. For Windows and Linux users with discrete NVIDIA graphics cards, GPU memory (VRAM) matters more for acceleration.
But you don't necessarily need expensive hardware to get started. Smaller models with fewer parameters run on much more modest systems. Qwen 2.5, for instance, comes in multiple sizes, and the smaller variants can operate effectively on machines with 16 gigabytes of RAM.
"You don't need to run the largest models to get excellent results," Sareen emphasized. The practical recommendation: start with a smaller model to test your workflow, then scale up as needed.
For context, Apple's entry-level MacBook Air with 8 gigabytes of RAM would struggle with most capable coding models. But a MacBook Pro with 32 gigabytes — increasingly common among professional developers — handles them comfortably.
Goose with a local LLM is not a perfect substitute for Claude Code. The comparison involves real trade-offs that developers should understand.
Model Quality: Claude 4.5 Opus, Anthropic's flagship model, remains arguably the most capable AI for software engineering tasks. It excels at understanding complex codebases, following nuanced instructions, and producing high-quality code on the first attempt. Open-source models have improved dramatically, but a gap persists — particularly for the most challenging tasks.
One developer who switched to the $200 Claude Code plan described the difference bluntly: "When I say 'make this look modern,' Opus knows what I mean. Other models give me Bootstrap circa 2015."
Context Window: Claude Sonnet 4.5, accessible through the API, offers a massive one-million-token context window — enough to load entire large codebases without chunking or context management issues. Most local models are limited to 4,096 or 8,192 tokens by default, though many can be configured for longer contexts at the cost of increased memory usage and slower processing.
Speed: Cloud-based services like Claude Code run on dedicated server hardware optimized for AI inference. Local models, running on consumer laptops, typically process requests more slowly. The difference matters for iterative workflows where you're making rapid changes and waiting for AI feedback.
Tooling Maturity: Claude Code benefits from Anthropic's dedicated engineering resources. Features like prompt caching (which can reduce costs by up to 90 percent for repeated contexts) and structured outputs are polished and well-documented. Goose, while actively developed with 102 releases to date, relies on community contributions and may lack equivalent refinement in specific areas.
Goose enters a crowded market of AI coding tools, but occupies a distinctive position.
Cursor, a popular AI-enhanced code editor, charges $20 per month for its Pro tier and $200 for Ultra—pricing that mirrors Claude Code's Max plans. Cursor provides approximately 4,500 Sonnet 4 requests per month at the Ultra level, a substantially different allocation model than Claude Code's hourly resets.
Cline, Roo Code, and similar open-source projects offer AI coding assistance but with varying levels of autonomy and tool integration. Many focus on code completion rather than the agentic task execution that defines Goose and Claude Code.
Amazon's CodeWhisperer, GitHub Copilot, and enterprise offerings from major cloud providers target large organizations with complex procurement processes and dedicated budgets. They are less relevant to individual developers and small teams seeking lightweight, flexible tools.
Goose's combination of genuine autonomy, model agnosticism, local operation, and zero cost creates a unique value proposition. The tool is not trying to compete with commercial offerings on polish or model quality. It's competing on freedom — both financial and architectural.
The AI coding tools market is evolving quickly. Open-source models are improving at a pace that continually narrows the gap with proprietary alternatives. Moonshot AI's Kimi K2 and z.ai's GLM 4.5 now benchmark near Claude Sonnet 4 levels — and they're freely available.
If this trajectory continues, the quality advantage that justifies Claude Code's premium pricing may erode. Anthropic would then face pressure to compete on features, user experience, and integration rather than raw model capability.
For now, developers face a clear choice. Those who need the absolute best model quality, who can afford premium pricing, and who accept usage restrictions may prefer Claude Code. Those who prioritize cost, privacy, offline access, and flexibility have a genuine alternative in Goose.
The fact that a $200-per-month commercial product has a zero-dollar open-source competitor with comparable core functionality is itself remarkable. It reflects both the maturation of open-source AI infrastructure and the appetite among developers for tools that respect their autonomy.
Goose is not perfect. It requires more technical setup than commercial alternatives. It depends on hardware resources that not every developer possesses. Its model options, while improving rapidly, still trail the best proprietary offerings on complex tasks.
But for a growing community of developers, those limitations are acceptable trade-offs for something increasingly rare in the AI landscape: a tool that truly belongs to them.
Goose is available for download at github.com/block/goose. Ollama is available at ollama.com. Both projects are free and open source.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-19T14:00:00.000Z", "pubDate": "Mon, 19 Jan 2026 14:00:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/1U9H8GLIqoGqKpitVfgw3T/ba56292f99409eca709dac0b176ec245/nuneybits_Vector_art_of_white_goose_silhouette_flying_through_c_8100d5a7-9e36-4ed6-a188-016470e1d0e1.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "AI", "Infrastructure" ], "contentSnippet": "The artificial intelligence coding revolution comes with a catch: it's expensive.\nClaude Code, Anthropic's terminal-based AI agent that can write, debug, and deploy code autonomously, has captured the imagination of software developers worldwide. But its pricing — ranging from $20 to $200 per month depending on usage — has sparked a growing rebellion among the very programmers it aims to serve.\nNow, a free alternative is gaining traction. Goose, an open-source AI agent developed by Block (the financial technology company formerly known as Square), offers nearly identical functionality to Claude Code but runs entirely on a user's local machine. No subscription fees. No cloud dependency. No rate limits that reset every five hours.\n\"Your data stays with you, period,\" said Parth Sareen, a software engineer who demonstrated the tool during a recent livestream. The comment captures the core appeal: Goose gives developers complete control over their AI-powered workflow, including the ability to work offline — even on an airplane.\nThe project has exploded in popularity. Goose now boasts more than 26,100 stars on GitHub, the code-sharing platform, with 362 contributors and 102 releases since its launch. The latest version, 1.20.1, shipped on January 19, 2026, reflecting a development pace that rivals commercial products.\nFor developers frustrated by Claude Code's pricing structure and usage caps, Goose represents something increasingly rare in the AI industry: a genuinely free, no-strings-attached option for serious work.\n\nAnthropic's new rate limits spark a developer revolt\nTo understand why Goose matters, you need to understand the Claude Code pricing controversy.\nAnthropic, the San Francisco artificial intelligence company founded by former OpenAI executives, offers Claude Code as part of its subscription tiers. The free plan provides no access whatsoever. The Pro plan, at $17 per month with annual billing (or $20 monthly), limits users to just 10 to 40 prompts every five hours — a constraint that serious developers exhaust within minutes of intensive work.\nThe Max plans, at $100 and $200 per month, offer more headroom: 50 to 200 prompts and 200 to 800 prompts respectively, plus access to Anthropic's most powerful model, Claude 4.5 Opus. But even these premium tiers come with restrictions that have inflamed the developer community.\nIn late July, Anthropic announced new weekly rate limits. Under the system, Pro users receive 40 to 80 hours of Sonnet 4 usage per week. Max users at the $200 tier get 240 to 480 hours of Sonnet 4, plus 24 to 40 hours of Opus 4. Nearly five months later, the frustration has not subsided.\nThe problem? Those \"hours\" are not actual hours. They represent token-based limits that vary wildly depending on codebase size, conversation length, and the complexity of the code being processed. Independent analysis suggests the actual per-session limits translate to roughly 44,000 tokens for Pro users and 220,000 tokens for the $200 Max plan.\n\"It's confusing and vague,\" one developer wrote in a widely shared analysis. \"When they say '24-40 hours of Opus 4,' that doesn't really tell you anything useful about what you're actually getting.\"\nThe backlash on Reddit and developer forums has been fierce. Some users report hitting their daily limits within 30 minutes of intensive coding. Others have canceled their subscriptions entirely, calling the new restrictions \"a joke\" and \"unusable for real work.\"\nAnthropic has defended the changes, stating that the limits affect fewer than five percent of users and target people running Claude Code \"continuously in the background, 24/7.\" But the company has not clarified whether that figure refers to five percent of Max subscribers or five percent of all users — a distinction that matters enormously.\nHow Block built a free AI coding agent that works offline\nGoose takes a radically different approach to the same problem.\nBuilt by Block, the payments company led by Jack Dorsey, Goose is what engineers call an \"on-machine AI agent.\" Unlike Claude Code, which sends your queries to Anthropic's servers for processing, Goose can run entirely on your local computer using open-source language models that you download and control yourself.\nThe project's documentation describes it as going \"beyond code suggestions\" to \"install, execute, edit, and test with any LLM.\" That last phrase — \"any LLM\" — is the key differentiator. Goose is model-agnostic by design.\nYou can connect Goose to Anthropic's Claude models if you have API access. You can use OpenAI's GPT-5 or Google's Gemini. You can route it through services like Groq or OpenRouter. Or — and this is where things get interesting — you can run it entirely locally using tools like Ollama, which let you download and execute open-source models on your own hardware.\nThe practical implications are significant. With a local setup, there are no subscription fees, no usage caps, no rate limits, and no concerns about your code being sent to external servers. Your conversations with the AI never leave your machine.\n\"I use Ollama all the time on planes — it's a lot of fun!\" Sareen noted during a demonstration, highlighting how local models free developers from the constraints of internet connectivity.\nWhat Goose can do that traditional code assistants can't\nGoose operates as a command-line tool or desktop application that can autonomously perform complex development tasks. It can build entire projects from scratch, write and execute code, debug failures, orchestrate workflows across multiple files, and interact with external APIs — all without constant human oversight.\nThe architecture relies on what the AI industry calls \"tool calling\" or \"function calling\" — the ability for a language model to request specific actions from external systems. When you ask Goose to create a new file, run a test suite, or check the status of a GitHub pull request, it doesn't just generate text describing what should happen. It actually executes those operations.\nThis capability depends heavily on the underlying language model. Claude 4 models from Anthropic currently perform best at tool calling, according to the Berkeley Function-Calling Leaderboard, which ranks models on their ability to translate natural language requests into executable code and system commands.\nBut newer open-source models are catching up quickly. Goose's documentation highlights several options with strong tool-calling support: Meta's Llama series, Alibaba's Qwen models, Google's Gemma variants, and DeepSeek's reasoning-focused architectures.\nThe tool also integrates with the Model Context Protocol, or MCP, an emerging standard for connecting AI agents to external services. Through MCP, Goose can access databases, search engines, file systems, and third-party APIs — extending its capabilities far beyond what the base language model provides.\nSetting Up Goose with a Local Model\nFor developers interested in a completely free, privacy-preserving setup, the process involves three main components: Goose itself, Ollama (a tool for running open-source models locally), and a compatible language model.\nStep 1: Install Ollama\nOllama is an open-source project that dramatically simplifies the process of running large language models on personal hardware. It handles the complex work of downloading, optimizing, and serving models through a simple interface.\nDownload and install Ollama from ollama.com. Once installed, you can pull models with a single command. For coding tasks, Qwen 2.5 offers strong tool-calling support:\nollama run qwen2.5\nThe model downloads automatically and begins running on your machine.\nStep 2: Install Goose\nGoose is available as both a desktop application and a command-line interface. The desktop version provides a more visual experience, while the CLI appeals to developers who prefer working entirely in the terminal.\nInstallation instructions vary by operating system but generally involve downloading from Goose's GitHub releases page or using a package manager. Block provides pre-built binaries for macOS (both Intel and Apple Silicon), Windows, and Linux.\nStep 3: Configure the Connection\nIn Goose Desktop, navigate to Settings, then Configure Provider, and select Ollama. Confirm that the API Host is set to http://localhost:11434 (Ollama's default port) and click Submit.\nFor the command-line version, run goose configure, select \"Configure Providers,\" choose Ollama, and enter the model name when prompted.\nThat's it. Goose is now connected to a language model running entirely on your hardware, ready to execute complex coding tasks without any subscription fees or external dependencies.\nThe RAM, processing power, and trade-offs you should know about\nThe obvious question: what kind of computer do you need?\nRunning large language models locally requires substantially more computational resources than typical software. The key constraint is memory — specifically, RAM on most systems, or VRAM if using a dedicated graphics card for acceleration.\nBlock's documentation suggests that 32 gigabytes of RAM provides \"a solid baseline for larger models and outputs.\" For Mac users, this means the computer's unified memory is the primary bottleneck. For Windows and Linux users with discrete NVIDIA graphics cards, GPU memory (VRAM) matters more for acceleration.\nBut you don't necessarily need expensive hardware to get started. Smaller models with fewer parameters run on much more modest systems. Qwen 2.5, for instance, comes in multiple sizes, and the smaller variants can operate effectively on machines with 16 gigabytes of RAM.\n\"You don't need to run the largest models to get excellent results,\" Sareen emphasized. The practical recommendation: start with a smaller model to test your workflow, then scale up as needed.\nFor context, Apple's entry-level MacBook Air with 8 gigabytes of RAM would struggle with most capable coding models. But a MacBook Pro with 32 gigabytes — increasingly common among professional developers — handles them comfortably.\nWhy keeping your code off the cloud matters more than ever\nGoose with a local LLM is not a perfect substitute for Claude Code. The comparison involves real trade-offs that developers should understand.\nModel Quality: Claude 4.5 Opus, Anthropic's flagship model, remains arguably the most capable AI for software engineering tasks. It excels at understanding complex codebases, following nuanced instructions, and producing high-quality code on the first attempt. Open-source models have improved dramatically, but a gap persists — particularly for the most challenging tasks.\nOne developer who switched to the $200 Claude Code plan described the difference bluntly: \"When I say 'make this look modern,' Opus knows what I mean. Other models give me Bootstrap circa 2015.\"\nContext Window: Claude Sonnet 4.5, accessible through the API, offers a massive one-million-token context window — enough to load entire large codebases without chunking or context management issues. Most local models are limited to 4,096 or 8,192 tokens by default, though many can be configured for longer contexts at the cost of increased memory usage and slower processing.\nSpeed: Cloud-based services like Claude Code run on dedicated server hardware optimized for AI inference. Local models, running on consumer laptops, typically process requests more slowly. The difference matters for iterative workflows where you're making rapid changes and waiting for AI feedback.\nTooling Maturity: Claude Code benefits from Anthropic's dedicated engineering resources. Features like prompt caching (which can reduce costs by up to 90 percent for repeated contexts) and structured outputs are polished and well-documented. Goose, while actively developed with 102 releases to date, relies on community contributions and may lack equivalent refinement in specific areas.\nHow Goose stacks up against Cursor, GitHub Copilot, and the paid AI coding market\nGoose enters a crowded market of AI coding tools, but occupies a distinctive position.\nCursor, a popular AI-enhanced code editor, charges $20 per month for its Pro tier and $200 for Ultra—pricing that mirrors Claude Code's Max plans. Cursor provides approximately 4,500 Sonnet 4 requests per month at the Ultra level, a substantially different allocation model than Claude Code's hourly resets.\nCline, Roo Code, and similar open-source projects offer AI coding assistance but with varying levels of autonomy and tool integration. Many focus on code completion rather than the agentic task execution that defines Goose and Claude Code.\nAmazon's CodeWhisperer, GitHub Copilot, and enterprise offerings from major cloud providers target large organizations with complex procurement processes and dedicated budgets. They are less relevant to individual developers and small teams seeking lightweight, flexible tools.\nGoose's combination of genuine autonomy, model agnosticism, local operation, and zero cost creates a unique value proposition. The tool is not trying to compete with commercial offerings on polish or model quality. It's competing on freedom — both financial and architectural.\nThe $200-a-month era for AI coding tools may be ending\nThe AI coding tools market is evolving quickly. Open-source models are improving at a pace that continually narrows the gap with proprietary alternatives. Moonshot AI's Kimi K2 and z.ai's GLM 4.5 now benchmark near Claude Sonnet 4 levels — and they're freely available.\nIf this trajectory continues, the quality advantage that justifies Claude Code's premium pricing may erode. Anthropic would then face pressure to compete on features, user experience, and integration rather than raw model capability.\nFor now, developers face a clear choice. Those who need the absolute best model quality, who can afford premium pricing, and who accept usage restrictions may prefer Claude Code. Those who prioritize cost, privacy, offline access, and flexibility have a genuine alternative in Goose.\nThe fact that a $200-per-month commercial product has a zero-dollar open-source competitor with comparable core functionality is itself remarkable. It reflects both the maturation of open-source AI infrastructure and the appetite among developers for tools that respect their autonomy.\nGoose is not perfect. It requires more technical setup than commercial alternatives. It depends on hardware resources that not every developer possesses. Its model options, while improving rapidly, still trail the best proprietary offerings on complex tasks.\nBut for a growing community of developers, those limitations are acceptable trade-offs for something increasingly rare in the AI landscape: a tool that truly belongs to them.\n\nGoose is available for download at github.com/block/goose. Ollama is available at ollama.com. Both projects are free and open source." } }, { "json": { "guid": "wOG0z6nm91FOeUqhpizqI", "link": "https://venturebeat.com/technology/listen-labs-raises-usd69m-after-viral-billboard-hiring-stunt-to-scale-ai", "title": "Listen Labs raises $69M after viral billboard hiring stunt to scale AI customer interviews", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Alfred Wahlforss was running out of options. His startup, Listen Labs, needed to hire over 100 engineers, but competing against Mark Zuckerberg's $100 million offers seemed impossible. So he spent $5,000 — a fifth of his marketing budget — on a billboard in San Francisco displaying what looked like gibberish: five strings of random numbers.
The numbers were actually AI tokens. Decoded, they led to a coding challenge: build an algorithm to act as a digital bouncer at Berghain, the Berlin nightclub famous for rejecting nearly everyone at the door. Within days, thousands attempted the puzzle. 430 cracked it. Some got hired. The winner flew to Berlin, all expenses paid.
That unconventional approach has now attracted $69 million in Series B funding, led by Ribbit Capital with participation from Evantic and existing investors Sequoia Capital, Conviction, and Pear VC. The round values Listen Labs at $500 million and brings its total capital to $100 million. In nine months since launch, the company has grown annualized revenue by 15x to eight figures and conducted over one million AI-powered interviews.
"When you obsess over customers, everything else follows," Wahlforss said in an interview with VentureBeat. "Teams that use Listen bring the customer into every decision, from marketing to product, and when the customer is delighted, everyone is."
Listen's AI researcher finds participants, conducts in-depth interviews, and delivers actionable insights in hours, not weeks. The platform replaces the traditional choice between quantitative surveys — which provide statistical precision but miss nuance—and qualitative interviews, which deliver depth but cannot scale.
Wahlforss explained the limitation of existing approaches: "Essentially surveys give you false precision because people end up answering the same question... You can't get the outliers. People are actually not honest on surveys." The alternative, one-on-one human interviews, "gives you a lot of depth. You can ask follow up questions. You can kind of double check if they actually know what they're talking about. And the problem is you can't scale that."
The platform works in four steps: users create a study with AI assistance, Listen recruits participants from its global network of 30 million people, an AI moderator conducts in-depth interviews with follow-up questions, and results are packaged into executive-ready reports including key themes, highlight reels, and slide decks.
What distinguishes Listen's approach is its use of open-ended video conversations rather than multiple-choice forms. "In a survey, you can kind of guess what you should answer, and you have four options," Wahlforss said. "Oh, they probably want me to buy high income. Let me click on that button versus an open ended response. It just generates much more honesty."
Listen finds and qualifies the right participants in its global network of 30 million people. But building that panel required confronting what Wahlforss called "one of the most shocking things that we've learned when we entered this industry"—rampant fraud.
"Essentially, there's a financial transaction involved, which means there will be bad players," he explained. "We actually had some of the largest companies, some of them have billions in revenue, send us people who claim to be kind of enterprise buyers to our platform and our system immediately detected, like, fraud, fraud, fraud, fraud, fraud."
The company built what it calls a "quality guard" that cross-references LinkedIn profiles with video responses to verify identity, checks consistency across how participants answer questions, and flags suspicious patterns. The result, according to Wahlforss: "People talk three times more. They're much more honest when they talk about sensitive topics like politics and mental health."
Emeritus, an online education company that uses Listen, reported that approximately 20% of survey responses previously fell into the fraudulent or low-quality category. With Listen, they reduced this to almost zero. "We did not have to replace any responses because of fraud or gibberish information," said Gabrielli Tiburi, Assistant Manager of Customer Insights at Emeritus.
The speed advantage has proven central to Listen's pitch. Traditional customer research at Microsoft could take four to six weeks to generate insights. "By the time we get to them, either the decision has been made or we lose out on the opportunity to actually influence it," said Romani Patel, Senior Research Manager at Microsoft.
With Listen, Microsoft can now get insights in days, and in many cases, within hours.
The platform has already powered several high-profile initiatives. Microsoft used Listen Labs to collect global customer stories for its 50th anniversary celebration. "We wanted users to share how Copilot is empowering them to bring their best self forward," Patel said, "and we were able to collect those user video stories within a day." Traditionally, that kind of work would have taken six to eight weeks.
Simple Modern, an Oklahoma-based drinkware company, used Listen to test a new product concept. The process took about an hour to write questions, an hour to launch the study, and 2.5 hours to receive feedback from 120 people across the country. "We went from 'Should we even have this product?' to 'How should we launch it?'" said Chris Hoyle, the company's Chief Marketing Officer.
Chubbies, the shorts brand, achieved a 24x increase in youth research participation—growing from 5 to 120 participants — by using Listen to overcome the scheduling challenges of traditional focus groups with children. "There's school, sports, dinner, and homework," explained Lauren Neville, Director of Insights and Innovation. "I had to find a way to hear from them that fit into their schedules."
The company also discovered product issues through AI interviews that might have gone undetected otherwise. Wahlforss described how the AI "through conversations, realized there were like issues with the the kids short line, and decided to, like, interview hundreds of kids. And I understand that there were issues in the liner of the shorts and that they were, like, scratchy, quote, unquote, according to the people interviewed." The redesigned product became "a blockbuster hit."
Listen Labs is entering a massive but fragmented market. Wahlforss cited research from Andreessen Horowitz estimating the market research industry at roughly $140 billion annually, populated by legacy players — some with more than a billion dollars in revenue — that he believes are vulnerable to disruption.
"There are very much existing budget lines that we are replacing," Wahlforss said. "Why we're replacing them is that one, they're super costly. Two, they're kind of stuck in this old paradigm of choosing between a survey or interview, and they also take months to work with."
But the more intriguing dynamic may be that AI-powered research doesn't just replace existing spending — it creates new demand. Wahlforss invoked the Jevons paradox, an economic principle that occurs when technological advancements make a resource more efficient to use, but increased efficiency leads to increased overall consumption rather than decreased consumption.
"What I've noticed is that as something gets cheaper, you don't need less of it. You want more of it," Wahlforss explained. "There's infinite demand for customer understanding. So the researchers on the team can do an order of magnitude more research, and also other people who weren't researchers before can now do that as part of their job."
Listen Labs traces its origins to a consumer app that Wahlforss and his co-founder built after meeting at Harvard. "We built this consumer app that got 20,000 downloads in one day," Wahlforss recalled. "We had all these users, and we were thinking like, okay, what can we do to get to know them better? And we built this prototype of what Listen is today."
The founding team brings an unusual pedigree. Wahlforss's co-founder "was the national champion in competitive programming in Germany, and he worked at Tesla Autopilot." The company claims that 30% of its engineering team are medalists from the International Olympiad in Informatics — the same competition that produced the founders of Cognition, the AI coding startup.
The Berghain billboard stunt generated approximately 5 million views across social media, according to Wahlforss. It reflected the intensity of the talent war in the Bay Area.
"We had to do these things because some of our, like early employees, joined the company before we had a working toilet," he said. "But now we fixed that situation."
The company grew from 5 to 40 employees in 2024 and plans to reach 150 this year. It hires engineers for non-engineering roles across marketing, growth, and operations — a bet that in the AI era, technical fluency matters everywhere.
Wahlforss outlined an ambitious product roadmap that pushes into more speculative territory. The company is building "the ability to simulate your customers, so you can take all of those interviews we've done, and then extrapolate based on that and create synthetic users or simulated user voices."
Beyond simulation, Listen aims to enable automated action based on research findings. "Can you not just make recommendations, but also create spawn agents to either change things in code or some customer churns? Can you give them a discount and try to bring them back?"
Wahlforss acknowledged the ethical implications. "Obviously, as you said, there's kind of ethical concerns there. Of like, automated decision making overall can be bad, but we will have considerable guardrails to make sure that the companies are always in the loop."
The company already handles sensitive data with care. "We don't train on any of the data," Wahlforss said. "We will also scrub any sensitive PII automatically so the model can detect that. And there are times when, for example, you work with investors, where if you accidentally mention something that could be material, non public information, the AI can actually detect that and remove any information like that."
Perhaps the most provocative implication of Listen's model is how it could reshape product development itself. Wahlforss described a customer — an Australian startup — that has adopted what amounts to a continuous feedback loop.
"They're based in Australia, so they're coding during the day, and then in their night, they're releasing a Listen study with an American audience. Listen validates whatever they built during the day, and they get feedback on that. They can then plug that feedback directly into coding tools like Claude Code and iterate."
The vision extends Y Combinator's famous dictum — "write code, talk to users" — into an automated cycle. "Write code is now getting automated. And I think like talk to users will be as well, and you'll have this kind of infinite loop where you can start to ship this truly amazing product, almost kind of autonomously."
Whether that vision materializes depends on factors beyond Listen's control — the continued improvement of AI models, enterprise willingness to trust automated research, and whether speed truly correlates with better products. A 2024 MIT study found that 95% of AI pilots fail to move into production, a statistic Wahlforss cited as the reason he emphasizes quality over demos.
"I'm constantly have to emphasize like, let's make sure the quality is there and the details are right," he said.
But the company's growth suggests appetite for the experiment. Microsoft's Patel said Listen has "removed the drudgery of research and brought the fun and joy back into my work." Chubbies is now pushing its founder to give everyone in the company a login. Sling Money, a stablecoin payments startup, can create a survey in ten minutes and receive results the same day.
"It's a total game changer," said Ali Romero, Sling Money's marketing manager.
Wahlforss has a different phrase for what he's building. When asked about the tension between speed and rigor — the long-held belief that moving fast means cutting corners — he cited Nat Friedman, the former GitHub CEO and Listen investor, who keeps a list of one-liners on his website.
One of them: "Slow is fake."
It's an aggressive claim for an industry built on methodological caution. But Listen Labs is betting that in the AI era, the companies that listen fastest will be the ones that win. The only question is whether customers will talk back.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-16T14:01:00.000Z", "pubDate": "Fri, 16 Jan 2026 14:01:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/4gD12ThOmNHZuosqC4xCTz/277b1e8968da602108a29fae2eaca440/nuneybits_Vector_art_of_billboard_with_cryptic_code_dbe5b0ff-7644-45e6-a1ca-4a5dceeff986.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI" ], "contentSnippet": "Alfred Wahlforss was running out of options. His startup, Listen Labs, needed to hire over 100 engineers, but competing against Mark Zuckerberg's $100 million offers seemed impossible. So he spent $5,000 — a fifth of his marketing budget — on a billboard in San Francisco displaying what looked like gibberish: five strings of random numbers.\nThe numbers were actually AI tokens. Decoded, they led to a coding challenge: build an algorithm to act as a digital bouncer at Berghain, the Berlin nightclub famous for rejecting nearly everyone at the door. Within days, thousands attempted the puzzle. 430 cracked it. Some got hired. The winner flew to Berlin, all expenses paid.\nThat unconventional approach has now attracted $69 million in Series B funding, led by Ribbit Capital with participation from Evantic and existing investors Sequoia Capital, Conviction, and Pear VC. The round values Listen Labs at $500 million and brings its total capital to $100 million. In nine months since launch, the company has grown annualized revenue by 15x to eight figures and conducted over one million AI-powered interviews.\n\n\"When you obsess over customers, everything else follows,\" Wahlforss said in an interview with VentureBeat. \"Teams that use Listen bring the customer into every decision, from marketing to product, and when the customer is delighted, everyone is.\"\nWhy traditional market research is broken, and what Listen Labs is building to fix it\nListen's AI researcher finds participants, conducts in-depth interviews, and delivers actionable insights in hours, not weeks. The platform replaces the traditional choice between quantitative surveys — which provide statistical precision but miss nuance—and qualitative interviews, which deliver depth but cannot scale.\nWahlforss explained the limitation of existing approaches: \"Essentially surveys give you false precision because people end up answering the same question... You can't get the outliers. People are actually not honest on surveys.\" The alternative, one-on-one human interviews, \"gives you a lot of depth. You can ask follow up questions. You can kind of double check if they actually know what they're talking about. And the problem is you can't scale that.\"\nThe platform works in four steps: users create a study with AI assistance, Listen recruits participants from its global network of 30 million people, an AI moderator conducts in-depth interviews with follow-up questions, and results are packaged into executive-ready reports including key themes, highlight reels, and slide decks.\nWhat distinguishes Listen's approach is its use of open-ended video conversations rather than multiple-choice forms. \"In a survey, you can kind of guess what you should answer, and you have four options,\" Wahlforss said. \"Oh, they probably want me to buy high income. Let me click on that button versus an open ended response. It just generates much more honesty.\"\nThe dirty secret of the $140 billion market research industry: rampant fraud\nListen finds and qualifies the right participants in its global network of 30 million people. But building that panel required confronting what Wahlforss called \"one of the most shocking things that we've learned when we entered this industry\"—rampant fraud.\n\"Essentially, there's a financial transaction involved, which means there will be bad players,\" he explained. \"We actually had some of the largest companies, some of them have billions in revenue, send us people who claim to be kind of enterprise buyers to our platform and our system immediately detected, like, fraud, fraud, fraud, fraud, fraud.\"\nThe company built what it calls a \"quality guard\" that cross-references LinkedIn profiles with video responses to verify identity, checks consistency across how participants answer questions, and flags suspicious patterns. The result, according to Wahlforss: \"People talk three times more. They're much more honest when they talk about sensitive topics like politics and mental health.\"\nEmeritus, an online education company that uses Listen, reported that approximately 20% of survey responses previously fell into the fraudulent or low-quality category. With Listen, they reduced this to almost zero. \"We did not have to replace any responses because of fraud or gibberish information,\" said Gabrielli Tiburi, Assistant Manager of Customer Insights at Emeritus.\nHow Microsoft, Sweetgreen, and Chubbies are using AI interviews to build better products\nThe speed advantage has proven central to Listen's pitch. Traditional customer research at Microsoft could take four to six weeks to generate insights. \"By the time we get to them, either the decision has been made or we lose out on the opportunity to actually influence it,\" said Romani Patel, Senior Research Manager at Microsoft.\nWith Listen, Microsoft can now get insights in days, and in many cases, within hours.\nThe platform has already powered several high-profile initiatives. Microsoft used Listen Labs to collect global customer stories for its 50th anniversary celebration. \"We wanted users to share how Copilot is empowering them to bring their best self forward,\" Patel said, \"and we were able to collect those user video stories within a day.\" Traditionally, that kind of work would have taken six to eight weeks.\nSimple Modern, an Oklahoma-based drinkware company, used Listen to test a new product concept. The process took about an hour to write questions, an hour to launch the study, and 2.5 hours to receive feedback from 120 people across the country. \"We went from 'Should we even have this product?' to 'How should we launch it?'\" said Chris Hoyle, the company's Chief Marketing Officer.\nChubbies, the shorts brand, achieved a 24x increase in youth research participation—growing from 5 to 120 participants — by using Listen to overcome the scheduling challenges of traditional focus groups with children. \"There's school, sports, dinner, and homework,\" explained Lauren Neville, Director of Insights and Innovation. \"I had to find a way to hear from them that fit into their schedules.\"\nThe company also discovered product issues through AI interviews that might have gone undetected otherwise. Wahlforss described how the AI \"through conversations, realized there were like issues with the the kids short line, and decided to, like, interview hundreds of kids. And I understand that there were issues in the liner of the shorts and that they were, like, scratchy, quote, unquote, according to the people interviewed.\" The redesigned product became \"a blockbuster hit.\"\nThe Jevons paradox explains why cheaper research creates more demand, not less\nListen Labs is entering a massive but fragmented market. Wahlforss cited research from Andreessen Horowitz estimating the market research industry at roughly $140 billion annually, populated by legacy players — some with more than a billion dollars in revenue — that he believes are vulnerable to disruption.\n\"There are very much existing budget lines that we are replacing,\" Wahlforss said. \"Why we're replacing them is that one, they're super costly. Two, they're kind of stuck in this old paradigm of choosing between a survey or interview, and they also take months to work with.\"\nBut the more intriguing dynamic may be that AI-powered research doesn't just replace existing spending — it creates new demand. Wahlforss invoked the Jevons paradox, an economic principle that occurs when technological advancements make a resource more efficient to use, but increased efficiency leads to increased overall consumption rather than decreased consumption.\n\"What I've noticed is that as something gets cheaper, you don't need less of it. You want more of it,\" Wahlforss explained. \"There's infinite demand for customer understanding. So the researchers on the team can do an order of magnitude more research, and also other people who weren't researchers before can now do that as part of their job.\"\nInside the elite engineering team that built Listen Labs before they had a working toilet\nListen Labs traces its origins to a consumer app that Wahlforss and his co-founder built after meeting at Harvard. \"We built this consumer app that got 20,000 downloads in one day,\" Wahlforss recalled. \"We had all these users, and we were thinking like, okay, what can we do to get to know them better? And we built this prototype of what Listen is today.\"\nThe founding team brings an unusual pedigree. Wahlforss's co-founder \"was the national champion in competitive programming in Germany, and he worked at Tesla Autopilot.\" The company claims that 30% of its engineering team are medalists from the International Olympiad in Informatics — the same competition that produced the founders of Cognition, the AI coding startup.\nThe Berghain billboard stunt generated approximately 5 million views across social media, according to Wahlforss. It reflected the intensity of the talent war in the Bay Area.\n\"We had to do these things because some of our, like early employees, joined the company before we had a working toilet,\" he said. \"But now we fixed that situation.\"\nThe company grew from 5 to 40 employees in 2024 and plans to reach 150 this year. It hires engineers for non-engineering roles across marketing, growth, and operations — a bet that in the AI era, technical fluency matters everywhere.\nSynthetic customers and automated decisions: what Listen Labs is building next\nWahlforss outlined an ambitious product roadmap that pushes into more speculative territory. The company is building \"the ability to simulate your customers, so you can take all of those interviews we've done, and then extrapolate based on that and create synthetic users or simulated user voices.\"\nBeyond simulation, Listen aims to enable automated action based on research findings. \"Can you not just make recommendations, but also create spawn agents to either change things in code or some customer churns? Can you give them a discount and try to bring them back?\"\nWahlforss acknowledged the ethical implications. \"Obviously, as you said, there's kind of ethical concerns there. Of like, automated decision making overall can be bad, but we will have considerable guardrails to make sure that the companies are always in the loop.\"\nThe company already handles sensitive data with care. \"We don't train on any of the data,\" Wahlforss said. \"We will also scrub any sensitive PII automatically so the model can detect that. And there are times when, for example, you work with investors, where if you accidentally mention something that could be material, non public information, the AI can actually detect that and remove any information like that.\"\nHow AI could reshape the future of product development\nPerhaps the most provocative implication of Listen's model is how it could reshape product development itself. Wahlforss described a customer — an Australian startup — that has adopted what amounts to a continuous feedback loop.\n\"They're based in Australia, so they're coding during the day, and then in their night, they're releasing a Listen study with an American audience. Listen validates whatever they built during the day, and they get feedback on that. They can then plug that feedback directly into coding tools like Claude Code and iterate.\"\nThe vision extends Y Combinator's famous dictum — \"write code, talk to users\" — into an automated cycle. \"Write code is now getting automated. And I think like talk to users will be as well, and you'll have this kind of infinite loop where you can start to ship this truly amazing product, almost kind of autonomously.\"\nWhether that vision materializes depends on factors beyond Listen's control — the continued improvement of AI models, enterprise willingness to trust automated research, and whether speed truly correlates with better products. A 2024 MIT study found that 95% of AI pilots fail to move into production, a statistic Wahlforss cited as the reason he emphasizes quality over demos.\n\"I'm constantly have to emphasize like, let's make sure the quality is there and the details are right,\" he said.\nBut the company's growth suggests appetite for the experiment. Microsoft's Patel said Listen has \"removed the drudgery of research and brought the fun and joy back into my work.\" Chubbies is now pushing its founder to give everyone in the company a login. Sling Money, a stablecoin payments startup, can create a survey in ten minutes and receive results the same day.\n\"It's a total game changer,\" said Ali Romero, Sling Money's marketing manager.\nWahlforss has a different phrase for what he's building. When asked about the tension between speed and rigor — the long-held belief that moving fast means cutting corners — he cited Nat Friedman, the former GitHub CEO and Listen investor, who keeps a list of one-liners on his website.\nOne of them: \"Slow is fake.\"\nIt's an aggressive claim for an industry built on methodological caution. But Listen Labs is betting that in the AI era, the companies that listen fastest will be the ones that win. The only question is whether customers will talk back." } }, { "json": { "guid": "1dxpkoyNmn8ceRMHCKLlT8", "link": "https://venturebeat.com/technology/salesforce-rolls-out-new-slackbot-ai-agent-as-it-battles-microsoft-and", "title": "Salesforce rolls out new Slackbot AI agent as it battles Microsoft and Google in workplace AI", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Salesforce on Tuesday launched an entirely rebuilt version of Slackbot, the company's workplace assistant, transforming it from a simple notification tool into what executives describe as a fully powered AI agent capable of searching enterprise data, drafting documents, and taking action on behalf of employees.
The new Slackbot, now generally available to Business+ and Enterprise+ customers, is Salesforce's most aggressive move yet to position Slack at the center of the emerging "agentic AI" movement — where software agents work alongside humans to complete complex tasks. The launch comes as Salesforce attempts to convince investors that artificial intelligence will bolster its products rather than render them obsolete.
"Slackbot isn't just another copilot or AI assistant," said Parker Harris, Salesforce co-founder and Slack's chief technology officer, in an exclusive interview with Salesforce. "It's the front door to the agentic enterprise, powered by Salesforce."
Harris was blunt about what distinguishes the new Slackbot from its predecessor: "The old Slackbot was, you know, a little tricycle, and the new Slackbot is like, you know, a Porsche."
The original Slackbot, which has existed since Slack's early days, performed basic algorithmic tasks — reminding users to add colleagues to documents, suggesting channel archives, and delivering simple notifications. The new version runs on an entirely different architecture built around a large language model and sophisticated search capabilities that can access Salesforce records, Google Drive files, calendar data, and years of Slack conversations.
"It's two different things," Harris explained. "The old Slackbot was algorithmic and fairly simple. The new Slackbot is brand new — it's based around an LLM and a very robust search engine, and connections to third-party search engines, third-party enterprise data."
Salesforce chose to retain the Slackbot brand despite the fundamental technical overhaul. "People know what Slackbot is, and so we wanted to carry that forward," Harris said.
The new Slackbot runs on Claude, Anthropic's large language model, a choice driven partly by compliance requirements. Slack's commercial service operates under FedRAMP Moderate certification to serve U.S. federal government customers, and Harris said Anthropic was "the only provider that could give us a compliant LLM" when Slack began building the new system.
But that exclusivity won't last. "We are, this year, going to support additional providers," Harris said. "We have a great relationship with Google. Gemini is incredible — performance is great, cost is great. So we're going to use Gemini for some things." He added that OpenAI remains a possibility as well.
Harris echoed Salesforce CEO Marc Benioff's view that large language models are becoming commoditized: "You've heard Marc talk about LLMs are commodities, that they're democratized. I call them CPUs."
On the sensitive question of training data, Harris was unequivocal: Salesforce does not train any models on customer data. "Models don't have any sort of security," he explained. "If we trained it on some confidential conversation that you and I have, I don't want Carolyn to know — if I train it into the LLM, there is no way for me to say you get to see the answer, but Carolyn doesn't."
Salesforce has been testing the new Slackbot internally for months, rolling it out to all 80,000 employees. According to Ryan Gavin, Slack's chief marketing officer, the results have been striking: "It's the fastest adopted product in Salesforce history."
Internal data shows that two-thirds of Salesforce employees have tried the new Slackbot, with 80% of those users continuing to use it regularly. Internal satisfaction rates reached 96% — the highest for any AI feature Slack has shipped. Employees report saving between two and 20 hours per week.
The adoption happened largely organically. "I think it was about five days, and a Canvas was developed by our employees called 'The Most Stealable Slackbot Prompts,'" Gavin said. "People just started adding to it organically. I think it's up to 250-plus prompts that are in this Canvas right now."
Kate Crotty, a principal UX researcher at Salesforce, found that 73% of internal adoption was driven by social sharing rather than top-down mandates. "Everybody is there to help each other learn and communicate hacks," she said.
During a product demonstration, Amy Bauer, Slack's product experience designer, showed how Slackbot can synthesize information across multiple sources. In one example, she asked Slackbot to analyze customer feedback from a pilot program, upload an image of a usage dashboard, and have Slackbot correlate the qualitative and quantitative data.
"This is where Slackbot really earns its keep for me," Bauer explained. "What it's doing is not just simply reading the image — it's actually looking at the image and comparing it to the insight it just generated for me."
Slackbot can then query Salesforce to find enterprise accounts with open deals that might be good candidates for early access, creating what Bauer called "a really great justification and plan to move forward." Finally, it can synthesize all that information into a Canvas — Slack's collaborative document format — and find calendar availability among stakeholders to schedule a review meeting.
"Up until this point, we have been working in a one-to-one capacity with Slackbot," Bauer said. "But one of the benefits that I can do now is take this insight and have it generate this into a Canvas, a shared workspace where I can iterate on it, refine it with Slackbot, or share it out with my team."
Rob Seaman, Slack's chief product officer, said the Canvas creation demonstrates where the product is heading: "This is making a tool call internally to Slack Canvas to actually write, effectively, a shared document. But it signals where we're going with Slackbot — we're eventually going to be adding in additional third-party tool calls."
Among Salesforce's pilot customers is Beast Industries, the parent company of YouTube star MrBeast. Luis Madrigal, the company's chief information officer, joined the launch announcement to describe his experience.
"As somebody who has rolled out enterprise technologies for over two decades now, this was practically one of the easiest," Madrigal said. "The plumbing is there. Slack as an implementation, Enterprise Tools — being able to turn on the Slackbot and the Slack AI functionality was as simple as having my team go in, review, do a quick security review."
Madrigal said his security team signed off "rather quickly" — unusual for enterprise AI deployments — because Slackbot accesses only the information each individual user already has permission to view. "Given all the guardrails you guys have put into place for Slackbot to be unique and customized to only the information that each individual user has, only the conversations and the Slack rooms and Slack channels that they're part of—that made my security team sign off rather quickly."
One Beast Industries employee, Sinan, the head of Beast Games marketing, reported saving "at bare minimum, 90 minutes a day." Another employee, Spencer, a creative supervisor, described it as "an assistant who's paying attention when I'm not."
Other pilot customers include Slalom, reMarkable, Xero, Mercari, and Engine. Mollie Bodensteiner, SVP of Operations at Engine, called Slackbot "an absolute 'chaos tamer' for our team," estimating it saves her about 30 minutes daily "just by eliminating context switching."
The launch puts Salesforce in direct competition with Microsoft's Copilot, which is integrated into Teams and the broader Microsoft 365 suite, as well as Google's Gemini integrations across Workspace. When asked what distinguishes Slackbot from these alternatives, Seaman pointed to context and convenience.
"The thing that makes it most powerful for our customers and users is the proximity — it's just right there in your Slack," Seaman said. "There's a tremendous convenience affordance that's naturally built into it."
The deeper advantage, executives argue, is that Slackbot already understands users' work without requiring setup or training. "Most AI tools sound the same no matter who is using them," the company's announcement stated. "They lack context, miss nuance, and force you to jump between tools to get anything done."
Harris put it more directly: "If you've ever had that magic experience with AI — I think ChatGPT is a great example, it's a great experience from a consumer perspective — Slackbot is really what we're doing in the enterprise, to be this employee super agent that is loved, just like people love using Slack."
Amy Bauer emphasized the frictionless nature of the experience. "Slackbot is inherently grounded in the context, in the data that you have in Slack," she said. "So as you continue working in Slack, Slackbot gets better because it's grounded in the work that you're doing there. There is no setup. There is no configuration for those end users."
Salesforce positions Slackbot as what Harris calls a "super agent" — a central hub that can eventually coordinate with other AI agents across an organization.
"Every corporation is going to have an employee super agent," Harris said. "Slackbot is essentially taking the magic of what Slack does. We think that Slackbot, and we're really excited about it, is going to be that."
The vision extends to third-party agents already launching in Slack. Last month, Anthropic released a preview of Claude Code for Slack, allowing developers to interact with Claude's coding capabilities directly in chat threads. OpenAI, Google, Vercel, and others have also built agents for the platform.
"Most of the net-new apps that are being deployed to Slack are agents," Seaman noted during the press conference. "This is proof of the promise of humans and agents coexisting and working together in Slack to solve problems."
Harris described a future where Slackbot becomes an MCP (Model Context Protocol) client, able to leverage tools from across the software ecosystem — similar to how the developer tool Cursor works. "Slack can be an MCP client, and Slackbot will be the hub of that, leveraging all these tools out in the world, some of which will be these amazing agents," he said.
But Harris also cautioned against over-promising on multi-agent coordination. "I still think we're in the single agent world," he said. "FY26 is going to be the year where we started to see more coordination. But we're going to do it with customer success in mind, and not demonstrate and talk about, like, 'I've got 1,000 agents working together,' because I think that's unrealistic."
Slackbot is included at no additional cost for customers on Business+ and Enterprise+ plans. "There's no additional fees customers have to do," Gavin confirmed. "If they're on one of those plans, they're going to get Slackbot."
However, some enterprise customers may face other cost pressures related to Salesforce's broader data strategy. CIOs may see price increases for third-party applications that work with Salesforce data, as effects of higher charges for API access ripple through the software supply chain.
Fivetran CEO George Fraser has warned that Salesforce's shift in pricing policy for API access could have tangible consequences for enterprises relying on Salesforce as a system of record. "They might not be able to use Fivetran to replicate their data to Snowflake and instead have to use Salesforce Data Cloud. Or they might find that they are not able to interact with their data via ChatGPT, and instead have to use Agentforce," Fraser said in a recent CIO report.
Salesforce has framed the pricing change as standard industry practice.
The new Slackbot begins rolling out today and will reach all eligible customers by the end of February. Mobile availability will complete by March 3, Bauer confirmed during her interview with VentureBeat.
Some capabilities remain works in progress. Calendar reading and availability checking are available at launch, but the ability to actually book meetings is "coming a few weeks after," according to Seaman. Image generation is not currently supported, though Bauer said it's "something that we are looking at in the future."
When asked about integration with competing CRM systems like HubSpot and Microsoft Dynamics, Salesforce representatives declined to provide specifics during the interview, though they acknowledged the question touched on key competitive differentiators.
The Slackbot launch is Salesforce's bet that the future of enterprise work is conversational — that employees will increasingly prefer to interact with AI through natural language rather than navigating traditional software interfaces.
Harris described Slack's product philosophy using principles like "don't make me think" and "be a great host." The goal, he said, is for Slackbot to surface information proactively rather than requiring users to hunt for it.
"One of the revelations for me is LLMs applied to unstructured information are incredible," Harris said. "And the amount of value you have if you're a Slack user, if your corporation uses Slack — the amount of value in Slack is unbelievable. Because you're talking about work, you're sharing documents, you're making decisions, but you can't as a human go through that and really get the same value that an LLM can do."
Looking ahead, Harris expects the interfaces themselves to evolve beyond pure conversation. "We're kind of saturating what we can do with purely conversational UIs," he said. "I think we'll start to see agents building an interface that best suits your intent, as opposed to trying to surface something within a conversational interface that matches your intent."
Microsoft, Google, and a growing roster of AI startups are placing similar bets — that the winning enterprise AI will be the one embedded in the tools workers already use, not another application to learn. The race to become that invisible layer of workplace intelligence is now fully underway.
For Salesforce, the stakes extend beyond a single product launch. After a bruising year on Wall Street and persistent questions about whether AI threatens its core business, the company is wagering that Slackbot can prove the opposite — that the tens of millions of people already chatting in Slack every day is not a vulnerability, but an unassailable advantage.
Haley Gault, the Salesforce account executive in Pittsburgh who stumbled upon the new Slackbot on a snowy morning, captured the shift in a single sentence: "I honestly can't imagine working for another company not having access to these types of tools. This is just how I work now."
That's precisely what Salesforce is counting on.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-13T13:00:00.000Z", "pubDate": "Tue, 13 Jan 2026 13:00:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/4Xrcg14GLKFlwSEnuEzxyS/21c85d29d03c4c974076475c009e3b38/nuneybits_Vector_art_of_chat_bubbles_on_a_computer_screen_in_th_5018a7ea-3496-4103-8453-7ba1b129189a.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI", "Automation" ], "contentSnippet": "Salesforce on Tuesday launched an entirely rebuilt version of Slackbot, the company's workplace assistant, transforming it from a simple notification tool into what executives describe as a fully powered AI agent capable of searching enterprise data, drafting documents, and taking action on behalf of employees.\nThe new Slackbot, now generally available to Business+ and Enterprise+ customers, is Salesforce's most aggressive move yet to position Slack at the center of the emerging \"agentic AI\" movement — where software agents work alongside humans to complete complex tasks. The launch comes as Salesforce attempts to convince investors that artificial intelligence will bolster its products rather than render them obsolete.\n\"Slackbot isn't just another copilot or AI assistant,\" said Parker Harris, Salesforce co-founder and Slack's chief technology officer, in an exclusive interview with Salesforce. \"It's the front door to the agentic enterprise, powered by Salesforce.\"\nFrom tricycle to Porsche: Salesforce rebuilt Slackbot from the ground up\nHarris was blunt about what distinguishes the new Slackbot from its predecessor: \"The old Slackbot was, you know, a little tricycle, and the new Slackbot is like, you know, a Porsche.\"\nThe original Slackbot, which has existed since Slack's early days, performed basic algorithmic tasks — reminding users to add colleagues to documents, suggesting channel archives, and delivering simple notifications. The new version runs on an entirely different architecture built around a large language model and sophisticated search capabilities that can access Salesforce records, Google Drive files, calendar data, and years of Slack conversations.\n\"It's two different things,\" Harris explained. \"The old Slackbot was algorithmic and fairly simple. The new Slackbot is brand new — it's based around an LLM and a very robust search engine, and connections to third-party search engines, third-party enterprise data.\"\nSalesforce chose to retain the Slackbot brand despite the fundamental technical overhaul. \"People know what Slackbot is, and so we wanted to carry that forward,\" Harris said.\nWhy Anthropic's Claude powers the new Slackbot — and which AI models could come next\nThe new Slackbot runs on Claude, Anthropic's large language model, a choice driven partly by compliance requirements. Slack's commercial service operates under FedRAMP Moderate certification to serve U.S. federal government customers, and Harris said Anthropic was \"the only provider that could give us a compliant LLM\" when Slack began building the new system.\nBut that exclusivity won't last. \"We are, this year, going to support additional providers,\" Harris said. \"We have a great relationship with Google. Gemini is incredible — performance is great, cost is great. So we're going to use Gemini for some things.\" He added that OpenAI remains a possibility as well.\nHarris echoed Salesforce CEO Marc Benioff's view that large language models are becoming commoditized: \"You've heard Marc talk about LLMs are commodities, that they're democratized. I call them CPUs.\"\nOn the sensitive question of training data, Harris was unequivocal: Salesforce does not train any models on customer data. \"Models don't have any sort of security,\" he explained. \"If we trained it on some confidential conversation that you and I have, I don't want Carolyn to know — if I train it into the LLM, there is no way for me to say you get to see the answer, but Carolyn doesn't.\"\nInside Salesforce's internal experiment: 80,000 employees tested Slackbot with striking results\nSalesforce has been testing the new Slackbot internally for months, rolling it out to all 80,000 employees. According to Ryan Gavin, Slack's chief marketing officer, the results have been striking: \"It's the fastest adopted product in Salesforce history.\"\nInternal data shows that two-thirds of Salesforce employees have tried the new Slackbot, with 80% of those users continuing to use it regularly. Internal satisfaction rates reached 96% — the highest for any AI feature Slack has shipped. Employees report saving between two and 20 hours per week.\nThe adoption happened largely organically. \"I think it was about five days, and a Canvas was developed by our employees called 'The Most Stealable Slackbot Prompts,'\" Gavin said. \"People just started adding to it organically. I think it's up to 250-plus prompts that are in this Canvas right now.\"\nKate Crotty, a principal UX researcher at Salesforce, found that 73% of internal adoption was driven by social sharing rather than top-down mandates. \"Everybody is there to help each other learn and communicate hacks,\" she said.\nHow Slackbot transforms scattered enterprise data into executive-ready insights\nDuring a product demonstration, Amy Bauer, Slack's product experience designer, showed how Slackbot can synthesize information across multiple sources. In one example, she asked Slackbot to analyze customer feedback from a pilot program, upload an image of a usage dashboard, and have Slackbot correlate the qualitative and quantitative data.\n\"This is where Slackbot really earns its keep for me,\" Bauer explained. \"What it's doing is not just simply reading the image — it's actually looking at the image and comparing it to the insight it just generated for me.\"\nSlackbot can then query Salesforce to find enterprise accounts with open deals that might be good candidates for early access, creating what Bauer called \"a really great justification and plan to move forward.\" Finally, it can synthesize all that information into a Canvas — Slack's collaborative document format — and find calendar availability among stakeholders to schedule a review meeting.\n\"Up until this point, we have been working in a one-to-one capacity with Slackbot,\" Bauer said. \"But one of the benefits that I can do now is take this insight and have it generate this into a Canvas, a shared workspace where I can iterate on it, refine it with Slackbot, or share it out with my team.\"\nRob Seaman, Slack's chief product officer, said the Canvas creation demonstrates where the product is heading: \"This is making a tool call internally to Slack Canvas to actually write, effectively, a shared document. But it signals where we're going with Slackbot — we're eventually going to be adding in additional third-party tool calls.\"\nMrBeast's company became a Slackbot guinea pig—and employees say they're saving 90 minutes a day\nAmong Salesforce's pilot customers is Beast Industries, the parent company of YouTube star MrBeast. Luis Madrigal, the company's chief information officer, joined the launch announcement to describe his experience.\n\"As somebody who has rolled out enterprise technologies for over two decades now, this was practically one of the easiest,\" Madrigal said. \"The plumbing is there. Slack as an implementation, Enterprise Tools — being able to turn on the Slackbot and the Slack AI functionality was as simple as having my team go in, review, do a quick security review.\"\nMadrigal said his security team signed off \"rather quickly\" — unusual for enterprise AI deployments — because Slackbot accesses only the information each individual user already has permission to view. \"Given all the guardrails you guys have put into place for Slackbot to be unique and customized to only the information that each individual user has, only the conversations and the Slack rooms and Slack channels that they're part of—that made my security team sign off rather quickly.\"\nOne Beast Industries employee, Sinan, the head of Beast Games marketing, reported saving \"at bare minimum, 90 minutes a day.\" Another employee, Spencer, a creative supervisor, described it as \"an assistant who's paying attention when I'm not.\"\nOther pilot customers include Slalom, reMarkable, Xero, Mercari, and Engine. Mollie Bodensteiner, SVP of Operations at Engine, called Slackbot \"an absolute 'chaos tamer' for our team,\" estimating it saves her about 30 minutes daily \"just by eliminating context switching.\"\nSlackbot vs. Microsoft Copilot vs. Google Gemini: The fight for enterprise AI dominance\nThe launch puts Salesforce in direct competition with Microsoft's Copilot, which is integrated into Teams and the broader Microsoft 365 suite, as well as Google's Gemini integrations across Workspace. When asked what distinguishes Slackbot from these alternatives, Seaman pointed to context and convenience.\n\"The thing that makes it most powerful for our customers and users is the proximity — it's just right there in your Slack,\" Seaman said. \"There's a tremendous convenience affordance that's naturally built into it.\"\nThe deeper advantage, executives argue, is that Slackbot already understands users' work without requiring setup or training. \"Most AI tools sound the same no matter who is using them,\" the company's announcement stated. \"They lack context, miss nuance, and force you to jump between tools to get anything done.\"\nHarris put it more directly: \"If you've ever had that magic experience with AI — I think ChatGPT is a great example, it's a great experience from a consumer perspective — Slackbot is really what we're doing in the enterprise, to be this employee super agent that is loved, just like people love using Slack.\"\nAmy Bauer emphasized the frictionless nature of the experience. \"Slackbot is inherently grounded in the context, in the data that you have in Slack,\" she said. \"So as you continue working in Slack, Slackbot gets better because it's grounded in the work that you're doing there. There is no setup. There is no configuration for those end users.\"\nSalesforce's ambitious plan to make Slackbot the one 'super agent' that controls all the others\nSalesforce positions Slackbot as what Harris calls a \"super agent\" — a central hub that can eventually coordinate with other AI agents across an organization.\n\"Every corporation is going to have an employee super agent,\" Harris said. \"Slackbot is essentially taking the magic of what Slack does. We think that Slackbot, and we're really excited about it, is going to be that.\"\nThe vision extends to third-party agents already launching in Slack. Last month, Anthropic released a preview of Claude Code for Slack, allowing developers to interact with Claude's coding capabilities directly in chat threads. OpenAI, Google, Vercel, and others have also built agents for the platform.\n\"Most of the net-new apps that are being deployed to Slack are agents,\" Seaman noted during the press conference. \"This is proof of the promise of humans and agents coexisting and working together in Slack to solve problems.\"\nHarris described a future where Slackbot becomes an MCP (Model Context Protocol) client, able to leverage tools from across the software ecosystem — similar to how the developer tool Cursor works. \"Slack can be an MCP client, and Slackbot will be the hub of that, leveraging all these tools out in the world, some of which will be these amazing agents,\" he said.\nBut Harris also cautioned against over-promising on multi-agent coordination. \"I still think we're in the single agent world,\" he said. \"FY26 is going to be the year where we started to see more coordination. But we're going to do it with customer success in mind, and not demonstrate and talk about, like, 'I've got 1,000 agents working together,' because I think that's unrealistic.\"\nSlackbot costs nothing extra, but Salesforce's data access fees could squeeze some customers\nSlackbot is included at no additional cost for customers on Business+ and Enterprise+ plans. \"There's no additional fees customers have to do,\" Gavin confirmed. \"If they're on one of those plans, they're going to get Slackbot.\"\nHowever, some enterprise customers may face other cost pressures related to Salesforce's broader data strategy. CIOs may see price increases for third-party applications that work with Salesforce data, as effects of higher charges for API access ripple through the software supply chain.\nFivetran CEO George Fraser has warned that Salesforce's shift in pricing policy for API access could have tangible consequences for enterprises relying on Salesforce as a system of record. \"They might not be able to use Fivetran to replicate their data to Snowflake and instead have to use Salesforce Data Cloud. Or they might find that they are not able to interact with their data via ChatGPT, and instead have to use Agentforce,\" Fraser said in a recent CIO report.\nSalesforce has framed the pricing change as standard industry practice.\nWhat Slackbot can do today, what's coming in weeks, and what's still on the roadmap\nThe new Slackbot begins rolling out today and will reach all eligible customers by the end of February. Mobile availability will complete by March 3, Bauer confirmed during her interview with VentureBeat.\nSome capabilities remain works in progress. Calendar reading and availability checking are available at launch, but the ability to actually book meetings is \"coming a few weeks after,\" according to Seaman. Image generation is not currently supported, though Bauer said it's \"something that we are looking at in the future.\"\nWhen asked about integration with competing CRM systems like HubSpot and Microsoft Dynamics, Salesforce representatives declined to provide specifics during the interview, though they acknowledged the question touched on key competitive differentiators.\nSalesforce is betting the future of work looks like a chat window—and it's not alone\nThe Slackbot launch is Salesforce's bet that the future of enterprise work is conversational — that employees will increasingly prefer to interact with AI through natural language rather than navigating traditional software interfaces.\nHarris described Slack's product philosophy using principles like \"don't make me think\" and \"be a great host.\" The goal, he said, is for Slackbot to surface information proactively rather than requiring users to hunt for it.\n\"One of the revelations for me is LLMs applied to unstructured information are incredible,\" Harris said. \"And the amount of value you have if you're a Slack user, if your corporation uses Slack — the amount of value in Slack is unbelievable. Because you're talking about work, you're sharing documents, you're making decisions, but you can't as a human go through that and really get the same value that an LLM can do.\"\nLooking ahead, Harris expects the interfaces themselves to evolve beyond pure conversation. \"We're kind of saturating what we can do with purely conversational UIs,\" he said. \"I think we'll start to see agents building an interface that best suits your intent, as opposed to trying to surface something within a conversational interface that matches your intent.\"\nMicrosoft, Google, and a growing roster of AI startups are placing similar bets — that the winning enterprise AI will be the one embedded in the tools workers already use, not another application to learn. The race to become that invisible layer of workplace intelligence is now fully underway.\nFor Salesforce, the stakes extend beyond a single product launch. After a bruising year on Wall Street and persistent questions about whether AI threatens its core business, the company is wagering that Slackbot can prove the opposite — that the tens of millions of people already chatting in Slack every day is not a vulnerability, but an unassailable advantage.\nHaley Gault, the Salesforce account executive in Pittsburgh who stumbled upon the new Slackbot on a snowy morning, captured the shift in a single sentence: \"I honestly can't imagine working for another company not having access to these types of tools. This is just how I work now.\"\nThat's precisely what Salesforce is counting on." } }, { "json": { "guid": "7EaNv4Ip2AUlW9RClUv2cB", "link": "https://venturebeat.com/technology/anthropic-launches-cowork-a-claude-desktop-agent-that-works-in-your-files-no", "title": "Anthropic launches Cowork, a Claude Desktop agent that works in your files — no coding required", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Anthropic released Cowork on Monday, a new AI agent capability that extends the power of its wildly successful Claude Code tool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, largely using Claude Code itself.
The launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, positioning Anthropic to compete not just with OpenAI and Google in conversational AI, but with Microsoft's Copilot in the burgeoning market for AI-powered productivity tools.
"Cowork lets you complete non-technical tasks much like how developers use Claude Code," the company announced via its official Claude account on X. The feature arrives as a research preview available exclusively to Claude Max subscribers — Anthropic's power-user tier priced between $100 and $200 per month — through the macOS desktop application.
For the past year, the industry narrative has focused on large language models that can write poetry or debug code. With Cowork, Anthropic is betting that the real enterprise value lies in an AI that can open a folder, read a messy pile of receipts, and generate a structured expense report without human hand-holding.
The genesis of Cowork lies in Anthropic's recent success with the developer community. In late 2024, the company released Claude Code, a terminal-based tool that allowed software engineers to automate rote programming tasks. The tool was a hit, but Anthropic noticed a peculiar trend: users were forcing the coding tool to perform non-coding labor.
According to Boris Cherny, an engineer at Anthropic, the company observed users deploying the developer tool for an unexpectedly diverse array of tasks.
"Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven," Cherny wrote on X. "These use cases are diverse and surprising — the reason is that the underlying Claude Agent is the best agent, and Opus 4.5 is the best model."
Recognizing this shadow usage, Anthropic effectively stripped the command-line complexity from their developer tool to create a consumer-friendly interface. In its blog post announcing the feature, Anthropic explained that developers "quickly began using it for almost everything else," which "prompted us to build Cowork: a simpler way for anyone — not just developers — to work with Claude in the very same way."
Unlike a standard chat interface where a user pastes text for analysis, Cowork requires a different level of trust and access. Users designate a specific folder on their local machine that Claude can access. Within that sandbox, the AI agent can read existing files, modify them, or create entirely new ones.
Anthropic offers several illustrative examples: reorganizing a cluttered downloads folder by sorting and intelligently renaming each file, generating a spreadsheet of expenses from a collection of receipt screenshots, or drafting a report from scattered notes across multiple documents.
"In Cowork, you give Claude access to a folder on your computer. Claude can then read, edit, or create files in that folder," the company explained on X. "Try it to create a spreadsheet from a pile of screenshots, or produce a first draft from scattered notes."
The architecture relies on what is known as an "agentic loop." When a user assigns a task, the AI does not merely generate a text response. Instead, it formulates a plan, executes steps in parallel, checks its own work, and asks for clarification if it hits a roadblock. Users can queue multiple tasks and let Claude process them simultaneously — a workflow Anthropic describes as feeling "much less like a back-and-forth and much more like leaving messages for a coworker."
The system is built on Anthropic's Claude Agent SDK, meaning it shares the same underlying architecture as Claude Code. Anthropic notes that Cowork "can take on many of the same tasks that Claude Code can handle, but in a more approachable form for non-coding tasks."
Perhaps the most remarkable detail surrounding Cowork's launch is the speed at which the tool was reportedly built — highlighting a recursive feedback loop where AI tools are being used to build better AI tools.
During a livestream hosted by Dan Shipper, Felix Rieseberg, an Anthropic employee, confirmed that the team built Cowork in approximately a week and a half.
Alex Volkov, who covers AI developments, expressed surprise at the timeline: "Holy shit Anthropic built 'Cowork' in the last... week and a half?!"
This prompted immediate speculation about how much of Cowork was itself built by Claude Code. Simon Smith, EVP of Generative AI at Klick Health, put it bluntly on X: "Claude Code wrote all of Claude Cowork. Can we all agree that we're in at least somewhat of a recursive improvement loop here?"
The implication is profound: Anthropic's AI coding agent may have substantially contributed to building its own non-technical sibling product. If true, this is one of the most visible examples yet of AI systems being used to accelerate their own development and expansion — a strategy that could widen the gap between AI labs that successfully deploy their own agents internally and those that do not.
Cowork doesn't operate in isolation. The feature integrates with Anthropic's existing ecosystem of connectors — tools that link Claude to external information sources and services such as Asana, Notion, PayPal, and other supported partners. Users who have configured these connections in the standard Claude interface can leverage them within Cowork sessions.
Additionally, Cowork can pair with Claude in Chrome, Anthropic's browser extension, to execute tasks requiring web access. This combination allows the agent to navigate websites, click buttons, fill forms, and extract information from the internet — all while operating from the desktop application.
"Cowork includes a number of novel UX and safety features that we think make the product really special," Cherny explained, highlighting "a built-in VM [virtual machine] for isolation, out of the box support for browser automation, support for all your claude.ai data connectors, asking you for clarification when it's unsure."
Anthropic has also introduced an initial set of "skills" specifically designed for Cowork that enhance Claude's ability to create documents, presentations, and other files. These build on the Skills for Claude framework the company announced in October, which provides specialized instruction sets Claude can load for particular types of tasks.
The transition from a chatbot that suggests edits to an agent that makes edits introduces significant risk. An AI that can organize files can, theoretically, delete them.
In a notable display of transparency, Anthropic devoted considerable space in its announcement to warning users about Cowork's potential dangers — an unusual approach for a product launch.
The company explicitly acknowledges that Claude "can take potentially destructive actions (such as deleting local files) if it's instructed to." Because Claude might occasionally misinterpret instructions, Anthropic urges users to provide "very clear guidance" about sensitive operations.
More concerning is the risk of prompt injection attacks — a technique where malicious actors embed hidden instructions in content Claude might encounter online, potentially causing the agent to bypass safeguards or take harmful actions.
"We've built sophisticated defenses against prompt injections," Anthropic wrote, "but agent safety — that is, the task of securing Claude's real-world actions — is still an active area of development in the industry."
The company characterized these risks as inherent to the current state of AI agent technology rather than unique to Cowork. "These risks aren't new with Cowork, but it might be the first time you're using a more advanced tool that moves beyond a simple conversation," the announcement notes.
The launch of Cowork places Anthropic in direct competition with Microsoft, which has spent years attempting to integrate its Copilot AI into the fabric of the Windows operating system with mixed adoption results.
However, Anthropic's approach differs in its isolation. By confining the agent to specific folders and requiring explicit connectors, they are attempting to strike a balance between the utility of an OS-level agent and the security of a sandboxed application.
What distinguishes Anthropic's approach is its bottom-up evolution. Rather than designing an AI assistant and retrofitting agent capabilities, Anthropic built a powerful coding agent first — Claude Code — and is now abstracting its capabilities for broader audiences. This technical lineage may give Cowork more robust agentic behavior from the start.
Claude Code has generated significant enthusiasm among developers since its initial launch as a command-line tool in late 2024. The company expanded access with a web interface in October 2025, followed by a Slack integration in December. Cowork is the next logical step: bringing the same agentic architecture to users who may never touch a terminal.
For now, Cowork remains exclusive to Claude Max subscribers using the macOS desktop application. Users on other subscription tiers — Free, Pro, Team, or Enterprise — can join a waitlist for future access.
Anthropic has signaled clear intentions to expand the feature's reach. The blog post explicitly mentions plans to add cross-device sync and bring Cowork to Windows as the company learns from the research preview.
Cherny set expectations appropriately, describing the product as "early and raw, similar to what Claude Code felt like when it first launched."
To access Cowork, Max subscribers can download or update the Claude macOS app and click on "Cowork" in the sidebar.
For technical decision-makers, the implications of Cowork extend beyond any single product launch. The bottleneck for AI adoption is shifting — no longer is model intelligence the limiting factor, but rather workflow integration and user trust.
Anthropic's goal, as the company puts it, is to make working with Claude feel less like operating a tool and more like delegating to a colleague. Whether mainstream users are ready to hand over folder access to an AI that might misinterpret their instructions remains an open question.
But the speed of Cowork's development — a major feature built in ten days, possibly by the company's own AI — previews a future where the capabilities of these systems compound faster than organizations can evaluate them.
The chatbot has learned to use a file manager. What it learns to use next is anyone's guess.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-12T11:30:00.000Z", "pubDate": "Mon, 12 Jan 2026 11:30:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/wHv1Wez7Ps9wYVYAo9fwT/14b41f606dbf1f5b17994be510407449/nuneybits_Hyper-realistic_image_of_a_retro_computer_with_a_glos_61ffb6e2-7c33-4d45-85f7-69c28693b3ec.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI", "Automation" ], "contentSnippet": "Anthropic released Cowork on Monday, a new AI agent capability that extends the power of its wildly successful Claude Code tool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, largely using Claude Code itself.\nThe launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, positioning Anthropic to compete not just with OpenAI and Google in conversational AI, but with Microsoft's Copilot in the burgeoning market for AI-powered productivity tools.\n\"Cowork lets you complete non-technical tasks much like how developers use Claude Code,\" the company announced via its official Claude account on X. The feature arrives as a research preview available exclusively to Claude Max subscribers — Anthropic's power-user tier priced between $100 and $200 per month — through the macOS desktop application.\nFor the past year, the industry narrative has focused on large language models that can write poetry or debug code. With Cowork, Anthropic is betting that the real enterprise value lies in an AI that can open a folder, read a messy pile of receipts, and generate a structured expense report without human hand-holding.\n\nHow developers using a coding tool for vacation research inspired Anthropic's latest product\nThe genesis of Cowork lies in Anthropic's recent success with the developer community. In late 2024, the company released Claude Code, a terminal-based tool that allowed software engineers to automate rote programming tasks. The tool was a hit, but Anthropic noticed a peculiar trend: users were forcing the coding tool to perform non-coding labor.\nAccording to Boris Cherny, an engineer at Anthropic, the company observed users deploying the developer tool for an unexpectedly diverse array of tasks.\n\n\"Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven,\" Cherny wrote on X. \"These use cases are diverse and surprising — the reason is that the underlying Claude Agent is the best agent, and Opus 4.5 is the best model.\"\nRecognizing this shadow usage, Anthropic effectively stripped the command-line complexity from their developer tool to create a consumer-friendly interface. In its blog post announcing the feature, Anthropic explained that developers \"quickly began using it for almost everything else,\" which \"prompted us to build Cowork: a simpler way for anyone — not just developers — to work with Claude in the very same way.\"\nInside the folder-based architecture that lets Claude read, edit, and create files on your computer\nUnlike a standard chat interface where a user pastes text for analysis, Cowork requires a different level of trust and access. Users designate a specific folder on their local machine that Claude can access. Within that sandbox, the AI agent can read existing files, modify them, or create entirely new ones.\nAnthropic offers several illustrative examples: reorganizing a cluttered downloads folder by sorting and intelligently renaming each file, generating a spreadsheet of expenses from a collection of receipt screenshots, or drafting a report from scattered notes across multiple documents.\n\"In Cowork, you give Claude access to a folder on your computer. Claude can then read, edit, or create files in that folder,\" the company explained on X. \"Try it to create a spreadsheet from a pile of screenshots, or produce a first draft from scattered notes.\"\n\nThe architecture relies on what is known as an \"agentic loop.\" When a user assigns a task, the AI does not merely generate a text response. Instead, it formulates a plan, executes steps in parallel, checks its own work, and asks for clarification if it hits a roadblock. Users can queue multiple tasks and let Claude process them simultaneously — a workflow Anthropic describes as feeling \"much less like a back-and-forth and much more like leaving messages for a coworker.\"\nThe system is built on Anthropic's Claude Agent SDK, meaning it shares the same underlying architecture as Claude Code. Anthropic notes that Cowork \"can take on many of the same tasks that Claude Code can handle, but in a more approachable form for non-coding tasks.\"\nThe recursive loop where AI builds AI: Claude Code reportedly wrote much of Claude Cowork\nPerhaps the most remarkable detail surrounding Cowork's launch is the speed at which the tool was reportedly built — highlighting a recursive feedback loop where AI tools are being used to build better AI tools.\nDuring a livestream hosted by Dan Shipper, Felix Rieseberg, an Anthropic employee, confirmed that the team built Cowork in approximately a week and a half.\nAlex Volkov, who covers AI developments, expressed surprise at the timeline: \"Holy shit Anthropic built 'Cowork' in the last... week and a half?!\"\n\nThis prompted immediate speculation about how much of Cowork was itself built by Claude Code. Simon Smith, EVP of Generative AI at Klick Health, put it bluntly on X: \"Claude Code wrote all of Claude Cowork. Can we all agree that we're in at least somewhat of a recursive improvement loop here?\"\nThe implication is profound: Anthropic's AI coding agent may have substantially contributed to building its own non-technical sibling product. If true, this is one of the most visible examples yet of AI systems being used to accelerate their own development and expansion — a strategy that could widen the gap between AI labs that successfully deploy their own agents internally and those that do not.\nConnectors, browser automation, and skills extend Cowork's reach beyond the local file system\nCowork doesn't operate in isolation. The feature integrates with Anthropic's existing ecosystem of connectors — tools that link Claude to external information sources and services such as Asana, Notion, PayPal, and other supported partners. Users who have configured these connections in the standard Claude interface can leverage them within Cowork sessions.\nAdditionally, Cowork can pair with Claude in Chrome, Anthropic's browser extension, to execute tasks requiring web access. This combination allows the agent to navigate websites, click buttons, fill forms, and extract information from the internet — all while operating from the desktop application.\n\"Cowork includes a number of novel UX and safety features that we think make the product really special,\" Cherny explained, highlighting \"a built-in VM [virtual machine] for isolation, out of the box support for browser automation, support for all your claude.ai data connectors, asking you for clarification when it's unsure.\"\nAnthropic has also introduced an initial set of \"skills\" specifically designed for Cowork that enhance Claude's ability to create documents, presentations, and other files. These build on the Skills for Claude framework the company announced in October, which provides specialized instruction sets Claude can load for particular types of tasks.\nWhy Anthropic is warning users that its own AI agent could delete their files\nThe transition from a chatbot that suggests edits to an agent that makes edits introduces significant risk. An AI that can organize files can, theoretically, delete them.\nIn a notable display of transparency, Anthropic devoted considerable space in its announcement to warning users about Cowork's potential dangers — an unusual approach for a product launch.\nThe company explicitly acknowledges that Claude \"can take potentially destructive actions (such as deleting local files) if it's instructed to.\" Because Claude might occasionally misinterpret instructions, Anthropic urges users to provide \"very clear guidance\" about sensitive operations.\nMore concerning is the risk of prompt injection attacks — a technique where malicious actors embed hidden instructions in content Claude might encounter online, potentially causing the agent to bypass safeguards or take harmful actions.\n\"We've built sophisticated defenses against prompt injections,\" Anthropic wrote, \"but agent safety — that is, the task of securing Claude's real-world actions — is still an active area of development in the industry.\"\nThe company characterized these risks as inherent to the current state of AI agent technology rather than unique to Cowork. \"These risks aren't new with Cowork, but it might be the first time you're using a more advanced tool that moves beyond a simple conversation,\" the announcement notes.\nAnthropic's desktop agent strategy sets up a direct challenge to Microsoft Copilot\nThe launch of Cowork places Anthropic in direct competition with Microsoft, which has spent years attempting to integrate its Copilot AI into the fabric of the Windows operating system with mixed adoption results.\nHowever, Anthropic's approach differs in its isolation. By confining the agent to specific folders and requiring explicit connectors, they are attempting to strike a balance between the utility of an OS-level agent and the security of a sandboxed application.\nWhat distinguishes Anthropic's approach is its bottom-up evolution. Rather than designing an AI assistant and retrofitting agent capabilities, Anthropic built a powerful coding agent first — Claude Code — and is now abstracting its capabilities for broader audiences. This technical lineage may give Cowork more robust agentic behavior from the start.\nClaude Code has generated significant enthusiasm among developers since its initial launch as a command-line tool in late 2024. The company expanded access with a web interface in October 2025, followed by a Slack integration in December. Cowork is the next logical step: bringing the same agentic architecture to users who may never touch a terminal.\nWho can access Cowork now, and what's coming next for Windows and other platforms\nFor now, Cowork remains exclusive to Claude Max subscribers using the macOS desktop application. Users on other subscription tiers — Free, Pro, Team, or Enterprise — can join a waitlist for future access.\nAnthropic has signaled clear intentions to expand the feature's reach. The blog post explicitly mentions plans to add cross-device sync and bring Cowork to Windows as the company learns from the research preview.\nCherny set expectations appropriately, describing the product as \"early and raw, similar to what Claude Code felt like when it first launched.\"\nTo access Cowork, Max subscribers can download or update the Claude macOS app and click on \"Cowork\" in the sidebar.\nThe real question facing enterprise AI adoption\nFor technical decision-makers, the implications of Cowork extend beyond any single product launch. The bottleneck for AI adoption is shifting — no longer is model intelligence the limiting factor, but rather workflow integration and user trust.\nAnthropic's goal, as the company puts it, is to make working with Claude feel less like operating a tool and more like delegating to a colleague. Whether mainstream users are ready to hand over folder access to an AI that might misinterpret their instructions remains an open question.\nBut the speed of Cowork's development — a major feature built in ten days, possibly by the company's own AI — previews a future where the capabilities of these systems compound faster than organizations can evaluate them. \nThe chatbot has learned to use a file manager. What it learns to use next is anyone's guess." } }, { "json": { "guid": "881216de-81bd-4da8-908c-6fc3e6a4411a", "link": "https://www.zdnet.com/article/how-to-turn-airtag-into-wifi-password-key-nfc-tags/", "title": "I created the ultimate Wi-Fi password key with the most unexpected gadget - how it works", "itunes": {}, "content": "NFC tags are so useful and customizable, and it's quite simple to make your own. Here's what you can do with it and how.", "isoDate": "2026-02-10T03:01:15.000Z", "pubDate": "Tue, 10 Feb 2026 03:01:15 GMT", "contentSnippet": "NFC tags are so useful and customizable, and it's quite simple to make your own. Here's what you can do with it and how." } }, { "json": { "guid": "27471e56-532b-4090-9aab-db552814455b", "link": "https://www.zdnet.com/article/i-turned-doodles-into-video-clips-in-minutes-with-runways-new-ai-video-tool-heres-how/", "title": "I turned doodles into video clips in minutes with Runway's new AI video tool - here's how", "itunes": {}, "content": "The company's Motion Sketch feature lowers the barrier between abstract imagination and concrete video outputs. But like all generative AI tools, it has its shortcomings.", "isoDate": "2026-02-10T03:01:13.000Z", "pubDate": "Tue, 10 Feb 2026 03:01:13 GMT", "contentSnippet": "The company's Motion Sketch feature lowers the barrier between abstract imagination and concrete video outputs. But like all generative AI tools, it has its shortcomings." } }, { "json": { "guid": "449249a1-cdf7-4b86-b294-f0b2409be861", "link": "https://www.zdnet.com/article/android-extend-unlock-faster-phone-access/", "title": "This Android trick lets me access my phone faster - here's how it works", "itunes": {}, "content": "Extend Unlock is one of those features every Android user will love once they give it a try.", "isoDate": "2026-02-10T03:00:54.000Z", "pubDate": "Tue, 10 Feb 2026 03:00:54 GMT", "contentSnippet": "Extend Unlock is one of those features every Android user will love once they give it a try." } }, { "json": { "guid": "e94ba78f-e76f-4849-a02b-828905fc727f", "link": "https://www.zdnet.com/article/onyx-boox-page-android-e-ink-tablet-review/", "title": "I didn't expect this Android E Ink tablet to beat my Kindle, but one feature sets it apart", "itunes": {}, "content": "The Onyx Boox Page is packed with features for an E Ink tablet, with a smart, versatile design.", "isoDate": "2026-02-10T02:30:37.000Z", "pubDate": "Tue, 10 Feb 2026 02:30:37 GMT", "contentSnippet": "The Onyx Boox Page is packed with features for an E Ink tablet, with a smart, versatile design." } }, { "json": { "guid": "6788a566-899b-481a-b0cc-ce5b53409dc0", "link": "https://www.zdnet.com/article/how-to-restart-android-phone-without-using-the-power-button-2-ways/", "title": "How to restart your Android phone without using the power button: 2 easy ways", "itunes": {}, "content": "You don't need to press the power button to restart your Android device. If yours is broken or you simply want options, here are two.", "isoDate": "2026-02-10T02:12:53.000Z", "pubDate": "Tue, 10 Feb 2026 02:12:53 GMT", "contentSnippet": "You don't need to press the power button to restart your Android device. If yours is broken or you simply want options, here are two." } }, { "json": { "guid": "https://towardsdatascience.com/?p=608359", "link": "https://towardsdatascience.com/the-machine-learning-lessons-ive-learned-last-month/", "title": "The Machine Learning Lessons I’ve Learned Last Month", "content": "Delayed January: deadlines, downtimes, and flow times
\nThe post The Machine Learning Lessons I’ve Learned Last Month appeared first on Towards Data Science.
\n", "creator": "Pascal Janetzky", "isoDate": "2026-02-09T20:38:42.000Z", "pubDate": "Mon, 09 Feb 2026 20:38:42 +0000", "categories": [ "Machine Learning", "Inspiration", "Ml Product Manager", "Productivity", "Project Management" ], "dc:creator": "Pascal Janetzky", "contentSnippet": "Delayed January: deadlines, downtimes, and flow times\nThe post The Machine Learning Lessons I’ve Learned Last Month appeared first on Towards Data Science." } }, { "json": { "guid": "https://towardsdatascience.com/?p=608349", "link": "https://towardsdatascience.com/the-death-of-the-everything-prompt-googles-move-toward-structured-ai/", "title": "The Death of the “Everything Prompt”: Google’s Move Toward Structured AI", "content": "How the new Interactions API enables deep-reasoning, stateful, agentic workflows.
\nThe post The Death of the “Everything Prompt”: Google’s Move Toward Structured AI appeared first on Towards Data Science.
\n", "creator": "Thomas Reid", "isoDate": "2026-02-09T13:00:00.000Z", "pubDate": "Mon, 09 Feb 2026 13:00:00 +0000", "categories": [ "Artificial Intelligence", "Agentic Ai", "API", "Deep Dives", "Llm", "Programming" ], "dc:creator": "Thomas Reid", "contentSnippet": "How the new Interactions API enables deep-reasoning, stateful, agentic workflows.\nThe post The Death of the “Everything Prompt”: Google’s Move Toward Structured AI appeared first on Towards Data Science." } }, { "json": { "guid": "https://towardsdatascience.com/?p=608351", "link": "https://towardsdatascience.com/what-i-am-doing-to-stay-relevant-as-a-senior-analytics-consultant-in-2026/", "title": "What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026", "content": "Learn how to work with AI, while strengthening your unique human skills that technology cannot replace
\nThe post What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026 appeared first on Towards Data Science.
\n", "creator": "Rashi Desai", "isoDate": "2026-02-07T15:00:00.000Z", "pubDate": "Sat, 07 Feb 2026 15:00:00 +0000", "categories": [ "Data analysis", "Artificial Intelligence", "Career Advice", "Data Science", "Editors Pick", "Generative Ai" ], "dc:creator": "Rashi Desai", "contentSnippet": "Learn how to work with AI, while strengthening your unique human skills that technology cannot replace\nThe post What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026 appeared first on Towards Data Science." } }, { "json": { "guid": "https://towardsdatascience.com/?p=608347", "link": "https://towardsdatascience.com/pydantic-performance-4-tips-on-how-to-validate-large-amounts-of-data-efficiently/", "title": "Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently", "content": "The real value lies in writing clearer code and using your tools right
\nThe post Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently appeared first on Towards Data Science.
\n", "creator": "Mike Huls", "isoDate": "2026-02-06T15:00:00.000Z", "pubDate": "Fri, 06 Feb 2026 15:00:00 +0000", "categories": [ "Data Engineering", "Data Architecture", "Data Science", "Data Validation", "Pydantic", "Python" ], "dc:creator": "Mike Huls", "contentSnippet": "The real value lies in writing clearer code and using your tools right\nThe post Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently appeared first on Towards Data Science." } }, { "json": { "guid": "https://towardsdatascience.com/?p=608345", "link": "https://towardsdatascience.com/prompt-fidelity-measuring-how-much-of-your-intent-an-ai-agent-actually-executes/", "title": "Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes", "content": "How much of your AI agent's output is real data versus confident guesswork?
\nThe post Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes appeared first on Towards Data Science.
\n", "creator": "James Barney", "isoDate": "2026-02-06T12:00:00.000Z", "pubDate": "Fri, 06 Feb 2026 12:00:00 +0000", "categories": [ "Agentic AI", "Ai Agent", "Artificial Intelligence", "Editors Pick", "Prompt Engineering", "Recommender Systems" ], "dc:creator": "James Barney", "contentSnippet": "How much of your AI agent's output is real data versus confident guesswork?\nThe post Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes appeared first on Towards Data Science." } }, { "json": { "guid": "https://www.nytimes.com/2026/02/09/technology/social-media-addiction-trial.html", "link": "https://www.nytimes.com/2026/02/09/technology/social-media-addiction-trial.html", "title": "Meta and YouTube Created ‘Digital Casinos,’ Lawyers Argue in Landmark Trial", "content": "Opening statements began in a trial claiming social media companies design addictive products that cause personal injury.", "creator": "Eli Tan and Cecilia Kang", "isoDate": "2026-02-10T00:53:17.000Z", "pubDate": "Tue, 10 Feb 2026 00:53:17 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Suits and Litigation (Civil)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Mental Health and Disorders" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Addiction (Psychology)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Youth" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Instagram Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Meta Platforms Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "YouTube.com" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Zuckerberg, Mark E" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Mohan, Neal" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" } ], "dc:creator": "Eli Tan and Cecilia Kang", "contentSnippet": "Opening statements began in a trial claiming social media companies design addictive products that cause personal injury." } }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/business/google-employees-protest.html", "link": "https://www.nytimes.com/2026/02/06/business/google-employees-protest.html", "title": "Google Workers Demand End to Cloud Services for Immigration Agencies", "content": "More than 800 employees delivered a petition to management, condemning the Trump administration’s use of Google technology in immigration enforcement.", "creator": "Tripp Mickle", "isoDate": "2026-02-06T18:52:26.000Z", "pubDate": "Fri, 06 Feb 2026 18:52:26 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Google Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Demonstrations, Protests and Riots" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Cloud Computing" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Illegal Immigration" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Immigration Detention" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Federal Actions in US Cities" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Police Brutality, Misconduct and Shootings" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "United States Politics and Government" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Customs and Border Protection (US)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Immigration and Customs Enforcement (US)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Homeland Security Department" } ], "dc:creator": "Tripp Mickle", "contentSnippet": "More than 800 employees delivered a petition to management, condemning the Trump administration’s use of Google technology in immigration enforcement." } }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/business/tiktok-addictive-design-europe.html", "link": "https://www.nytimes.com/2026/02/06/business/tiktok-addictive-design-europe.html", "title": "Europe Accuses TikTok of ‘Addictive Design’ and Pushes for Change", "content": "European Union regulators said the app’s infinite scroll and personalized algorithm led to “compulsive” behavior, especially among children.", "creator": "Adam Satariano", "isoDate": "2026-02-06T19:08:37.000Z", "pubDate": "Fri, 06 Feb 2026 19:08:37 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Regulation and Deregulation of Industry" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "International Relations" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Law and Legislation" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Beijing Bytedance Technology Co Ltd" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "TikTok (ByteDance)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Europe" } ], "dc:creator": "Adam Satariano", "contentSnippet": "European Union regulators said the app’s infinite scroll and personalized algorithm led to “compulsive” behavior, especially among children." } }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/business/ai-stock-market-anthropic-openai.html", "link": "https://www.nytimes.com/2026/02/06/business/ai-stock-market-anthropic-openai.html", "title": "Fear of AI Replacing Software Makers Hits Stocks. Here’s What to Know.", "content": "The prospect of disruptions from artificial intelligence has hung over the economy for years. But this week advances in software tools precipitated a sell-off on Wall Street.", "creator": "Mike Isaac, Joe Rennison and Maureen Farrell", "isoDate": "2026-02-06T20:11:04.000Z", "pubDate": "Fri, 06 Feb 2026 20:11:04 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Stocks and Bonds" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Company Reports" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Credit and Debt" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Software" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Innovation" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Bitcoin (Currency)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Anthropic AI LLC" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Qualcomm Inc" } ], "dc:creator": "Mike Isaac, Joe Rennison and Maureen Farrell", "contentSnippet": "The prospect of disruptions from artificial intelligence has hung over the economy for years. But this week advances in software tools precipitated a sell-off on Wall Street." } }, { "json": { "guid": "https://www.nytimes.com/2026/02/09/health/ai-chatbots-doctors-medicine.html", "link": "https://www.nytimes.com/2026/02/09/health/ai-chatbots-doctors-medicine.html", "title": "A.I. Is Making Doctors Answer a Question: What Are They Really Good For?", "content": "Many physicians find chatbots threatening, but that doesn’t mean they’re giving up on medicine.", "creator": "Gina Kolata", "isoDate": "2026-02-09T17:17:56.000Z", "pubDate": "Mon, 09 Feb 2026 17:17:56 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Doctors" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Medicine and Health" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Research" } ], "dc:creator": "Gina Kolata", "contentSnippet": "Many physicians find chatbots threatening, but that doesn’t mean they’re giving up on medicine." } }, { "json": { "date": "2026-02-09T14:00:22Z", "guid": "https://www.theguardian.com/technology/2026/feb/09/jeffrey-epstein-crypto", "link": "https://www.theguardian.com/technology/2026/feb/09/jeffrey-epstein-crypto", "title": "Files cast light on Jeffrey Epstein’s ties to cryptocurrency", "content": "Newly released documents detail convicted sex offender’s early backing of bitcoin and Coinbase
Millions of files related to Jeffrey Epstein have brought to light his ties to the highest echelons of the cryptocurrency industry.
Documents published last week by the US Department of Justice reveal Epstein bankrolled the “principal home and funding source” for bitcoin, the world’s largest cryptocurrency, during its nascent stages; he also invested $3m in Coinbase in 2014, the largest cryptocurrency exchange in the US, and cut a check that same year to Blockstream, a prominent bitcoin-focused technology firm. Both crypto startups accepted Epstein’s investments in 2014 – six years after his 2008 conviction in Florida for soliciting prostitution from a minor.
Continue reading...", "creator": "Adam Willems", "dc:date": "2026-02-09T14:00:22Z", "isoDate": "2026-02-09T14:00:22.000Z", "pubDate": "Mon, 09 Feb 2026 14:00:22 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/cryptocurrencies" }, "_": "Cryptocurrencies" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/us-news/jeffrey-epstein" }, "_": "Jeffrey Epstein" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-news" }, "_": "US news" } ], "dc:creator": "Adam Willems", "contentSnippet": "Newly released documents detail convicted sex offender’s early backing of bitcoin and Coinbase \nMillions of files related to Jeffrey Epstein have brought to light his ties to the highest echelons of the cryptocurrency industry.\nDocuments published last week by the US Department of Justice reveal Epstein bankrolled the “principal home and funding source” for bitcoin, the world’s largest cryptocurrency, during its nascent stages; he also invested $3m in Coinbase in 2014, the largest cryptocurrency exchange in the US, and cut a check that same year to Blockstream, a prominent bitcoin-focused technology firm. Both crypto startups accepted Epstein’s investments in 2014 – six years after his 2008 conviction in Florida for soliciting prostitution from a minor.\n Continue reading..." } }, { "json": { "date": "2026-02-08T08:00:25Z", "guid": "https://www.theguardian.com/technology/2026/feb/08/i-fell-into-it-ex-criminal-hackers-urge-manchester-pupils-to-use-web-skills-for-good", "link": "https://www.theguardian.com/technology/2026/feb/08/i-fell-into-it-ex-criminal-hackers-urge-manchester-pupils-to-use-web-skills-for-good", "title": "‘I fell into it’: ex-criminal hackers urge Manchester pupils to use web skills for good", "content": "Initiative aims to identify proficient gamers and coders who can help companies identify flaws in their cybersecurity
Cybercriminals, the shadowy online figures often depicted in Hollywood movies as hooded villains capable of wiping millions of pounds off the value of businesses at a keystroke, are not usually known for their candour.
But in a sixth-form college in Manchester this week, two former hackers gave the young people gathered an honest appraisal of what living a life of internet crime really looks like.
Continue reading...", "creator": "Dan Milmo Global technology editor", "dc:date": "2026-02-08T08:00:25Z", "isoDate": "2026-02-08T08:00:25.000Z", "pubDate": "Sun, 08 Feb 2026 08:00:25 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/hacking" }, "_": "Hacking" }, { "$": { "domain": "https://www.theguardian.com/technology/cybercrime" }, "_": "Cybercrime" }, { "$": { "domain": "https://www.theguardian.com/business/co-operative-group" }, "_": "Co-operative Group" }, { "$": { "domain": "https://www.theguardian.com/uk/uk" }, "_": "UK news" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/technology/internet" }, "_": "Internet" }, { "$": { "domain": "https://www.theguardian.com/technology/programming" }, "_": "Programming" }, { "$": { "domain": "https://www.theguardian.com/games/game-culture" }, "_": "Game culture" }, { "$": { "domain": "https://www.theguardian.com/society/youngpeople" }, "_": "Young people" }, { "$": { "domain": "https://www.theguardian.com/uk/manchester" }, "_": "Manchester" } ], "dc:creator": "Dan Milmo Global technology editor", "contentSnippet": "Initiative aims to identify proficient gamers and coders who can help companies identify flaws in their cybersecurity \nCybercriminals, the shadowy online figures often depicted in Hollywood movies as hooded villains capable of wiping millions of pounds off the value of businesses at a keystroke, are not usually known for their candour.\nBut in a sixth-form college in Manchester this week, two former hackers gave the young people gathered an honest appraisal of what living a life of internet crime really looks like.\n Continue reading..." } }, { "json": { "date": "2026-02-07T16:00:06Z", "guid": "https://www.theguardian.com/technology/2026/feb/07/ai-chatbots-anthropic-openai-claude-chatgpt", "link": "https://www.theguardian.com/technology/2026/feb/07/ai-chatbots-anthropic-openai-claude-chatgpt", "title": "Battle of the chatbots: Anthropic and OpenAI go head-to-head over ads in their AI products", "content": "New Anthropic campaign suggests other AI platforms will incorporate targeted ads in their chatbot conversations
The Seahawks and the Patriots aren’t the only ones gearing up for a fight.
AI rivals Anthropic and OpenAI have launched a war of ads trying to court corporate America during one of the biggest entertainment nights of the year.
Continue reading...", "creator": "Sanya Mansoor", "dc:date": "2026-02-07T16:00:06Z", "isoDate": "2026-02-07T16:00:06.000Z", "pubDate": "Sat, 07 Feb 2026 16:00:06 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/media/advertising" }, "_": "Advertising" }, { "$": { "domain": "https://www.theguardian.com/technology/chatbots" }, "_": "Chatbots" }, { "$": { "domain": "https://www.theguardian.com/technology/chatgpt" }, "_": "ChatGPT" }, { "$": { "domain": "https://www.theguardian.com/technology/openai" }, "_": "OpenAI" }, { "$": { "domain": "https://www.theguardian.com/technology/sam-altman" }, "_": "Sam Altman" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-news" }, "_": "US news" } ], "dc:creator": "Sanya Mansoor", "contentSnippet": "New Anthropic campaign suggests other AI platforms will incorporate targeted ads in their chatbot conversations\nThe Seahawks and the Patriots aren’t the only ones gearing up for a fight.\nAI rivals Anthropic and OpenAI have launched a war of ads trying to court corporate America during one of the biggest entertainment nights of the year.\n Continue reading..." } }, { "json": { "date": "2026-02-09T10:00:31Z", "guid": "https://www.theguardian.com/commentisfree/2026/feb/09/who-is-my-wife-martin-rowson-ai-technology", "link": "https://www.theguardian.com/commentisfree/2026/feb/09/who-is-my-wife-martin-rowson-ai-technology", "title": "I asked AI to name my wife. To the hopelessly incorrect people it cited, my deepest apologies | Martin Rowson", "content": "Authors, a newsreader, a lawyer and an esteemed colleague: they’re all great – but I’m not married to any of them. Can we really depend on this technology?
Recently, the Rowsons accidentally invented a new game that anyone can play at home. I have yet to come up with a world-beating name for it, so for now let’s just call it “How bloody stupid is AI?” The playing of the game will change from player to player, depending on their circumstances – but essentially the rules remain the same. Ask AI a simple question about yourself, and see just how wrong it gets it.
In my case, all you need know is that while I, through the nature of my job, have a fairly large online presence, my partner (we married in 1987) has assiduously avoided having one at all. Which means that if you Google “Martin Rowson wife” in images, you may get a picture of me next to our then 14-year-old daughter or me with my friend and fellow cartoonist Steven Appleby, who happens to be trans but has kept her given first name.
Continue reading...", "creator": "Martin Rowson", "dc:date": "2026-02-09T10:00:31Z", "isoDate": "2026-02-09T10:00:31.000Z", "pubDate": "Mon, 09 Feb 2026 10:00:31 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/uk/uk" }, "_": "UK news" }, { "$": { "domain": "https://www.theguardian.com/technology/computing" }, "_": "Computing" } ], "dc:creator": "Martin Rowson", "contentSnippet": "Authors, a newsreader, a lawyer and an esteemed colleague: they’re all great – but I’m not married to any of them. Can we really depend on this technology?\nRecently, the Rowsons accidentally invented a new game that anyone can play at home. I have yet to come up with a world-beating name for it, so for now let’s just call it “How bloody stupid is AI?” The playing of the game will change from player to player, depending on their circumstances – but essentially the rules remain the same. Ask AI a simple question about yourself, and see just how wrong it gets it.\nIn my case, all you need know is that while I, through the nature of my job, have a fairly large online presence, my partner (we married in 1987) has assiduously avoided having one at all. Which means that if you Google “Martin Rowson wife” in images, you may get a picture of me next to our then 14-year-old daughter or me with my friend and fellow cartoonist Steven Appleby, who happens to be trans but has kept her given first name.\n Continue reading..." } }, { "json": { "date": "2026-02-07T10:00:02Z", "guid": "https://www.theguardian.com/technology/2026/feb/07/campaigners-call-stronger-protection-against-ai-generated-explicit-imagery", "link": "https://www.theguardian.com/technology/2026/feb/07/campaigners-call-stronger-protection-against-ai-generated-explicit-imagery", "title": "Victims urge tougher action on deepfake abuse as new law comes into force", "content": "Campaigners welcome criminalisation of non-consensual AI-generated explicit images but say law does not go far enough
Victims of deepfake image abuse have called for stronger protection against AI-generated explicit images, as the law criminalising the creation of non-consensual intimate images comes into effect.
Campaigners from Stop Image-Based Abuse delivered a petition to Downing Street with more than 73,000 signatures, urging the government to introduce civil routes to justice such as takedown orders for abusive imagery on platforms and devices.
Continue reading...", "creator": "Priya Bharadia", "dc:date": "2026-02-07T10:00:02Z", "isoDate": "2026-02-07T10:00:02.000Z", "pubDate": "Sat, 07 Feb 2026 10:00:02 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/deepfake" }, "_": "Deepfake" }, { "$": { "domain": "https://www.theguardian.com/society/violence-against-women-and-girls" }, "_": "Violence against women and girls" }, { "$": { "domain": "https://www.theguardian.com/uk/ukcrime" }, "_": "Crime" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/uk/uk" }, "_": "UK news" }, { "$": { "domain": "https://www.theguardian.com/society/women" }, "_": "Women" }, { "$": { "domain": "https://www.theguardian.com/technology/internet" }, "_": "Internet" }, { "$": { "domain": "https://www.theguardian.com/politics/politics" }, "_": "Politics" } ], "dc:creator": "Priya Bharadia", "contentSnippet": "Campaigners welcome criminalisation of non-consensual AI-generated explicit images but say law does not go far enough\nVictims of deepfake image abuse have called for stronger protection against AI-generated explicit images, as the law criminalising the creation of non-consensual intimate images comes into effect.\nCampaigners from Stop Image-Based Abuse delivered a petition to Downing Street with more than 73,000 signatures, urging the government to introduce civil routes to justice such as takedown orders for abusive imagery on platforms and devices.\n Continue reading..." } }, { "json": { "guid": "https://www.bbc.com/news/articles/c3wlpqpe2z4o#0", "link": "https://www.bbc.com/news/articles/c3wlpqpe2z4o?at_medium=RSS&at_campaign=rss", "title": "Instagram and YouTube owners built 'addiction machines', trial hears", "content": "The tech giants are under scrutiny over social media addiction in a landmark jury trial in Los Angeles", "isoDate": "2026-02-10T07:25:42.000Z", "pubDate": "Tue, 10 Feb 2026 07:25:42 GMT", "contentSnippet": "The tech giants are under scrutiny over social media addiction in a landmark jury trial in Los Angeles" } }, { "json": { "guid": "https://www.bbc.com/news/articles/c1d67vdlk1ko#0", "link": "https://www.bbc.com/news/articles/c1d67vdlk1ko?at_medium=RSS&at_campaign=rss", "title": "Discord to start requiring face scan or ID to access adult content", "content": "The online chat service, which has 200 million monthly users, will blur adult content by default.", "isoDate": "2026-02-09T17:42:21.000Z", "pubDate": "Mon, 09 Feb 2026 17:42:21 GMT", "contentSnippet": "The online chat service, which has 200 million monthly users, will blur adult content by default." } }, { "json": { "guid": "https://www.bbc.com/news/articles/c3093gjy2ero#0", "link": "https://www.bbc.com/news/articles/c3093gjy2ero?at_medium=RSS&at_campaign=rss", "title": "AI chatbots pose 'dangerous' risk when giving medical advice, study suggests", "content": "It found people using AI for health reasons found it hard to identify what advice they should trust.", "isoDate": "2026-02-09T16:33:29.000Z", "pubDate": "Mon, 09 Feb 2026 16:33:29 GMT", "contentSnippet": "It found people using AI for health reasons found it hard to identify what advice they should trust." } }, { "json": { "guid": "https://www.bbc.com/news/articles/cqxdj77welpo#0", "link": "https://www.bbc.com/news/articles/cqxdj77welpo?at_medium=RSS&at_campaign=rss", "title": "EU tells Meta to let rivals run AI chatbots on WhatsApp", "content": "A Meta spokesperson said the EU had \"no reason\" to intervene over it changing the app in January.", "isoDate": "2026-02-09T12:14:35.000Z", "pubDate": "Mon, 09 Feb 2026 12:14:35 GMT", "contentSnippet": "A Meta spokesperson said the EU had \"no reason\" to intervene over it changing the app in January." } }, { "json": { "guid": "https://www.bbc.com/news/articles/c24g457y534o#0", "link": "https://www.bbc.com/news/articles/c24g457y534o?at_medium=RSS&at_campaign=rss", "title": "Baldur's Gate to be turned into TV series - without the game's developers", "content": "The show will be made by Craig Mazin, who co-created the acclaimed game adaptation The Last Of Us.", "isoDate": "2026-02-06T16:26:08.000Z", "pubDate": "Fri, 06 Feb 2026 16:26:08 GMT", "contentSnippet": "The show will be made by Craig Mazin, who co-created the acclaimed game adaptation The Last Of Us." } }, { "json": { "guid": "https://www.bbc.co.uk/iplayer/episode/m002qbkc#4", "link": "https://www.bbc.co.uk/iplayer/episode/m002qbkc?at_medium=RSS&at_campaign=rss", "title": "Tech Now", "content": "From Las Vegas, the latest trends and innovations at CES 2026.", "isoDate": "2026-01-17T01:00:00.000Z", "pubDate": "Sat, 17 Jan 2026 01:00:00 GMT", "contentSnippet": "From Las Vegas, the latest trends and innovations at CES 2026." } }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpz#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpz?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "Meet the humanoid robots designed to help with household chores", "isoDate": "2026-01-13T20:30:00.000Z", "pubDate": "Tue, 13 Jan 2026 20:30:00 GMT", "contentSnippet": "Meet the humanoid robots designed to help with household chores" } }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpy#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpy?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "The latest gadgets, the future in assistive tech and upcoming gaming releases in 2026.", "isoDate": "2026-01-06T20:32:00.000Z", "pubDate": "Tue, 06 Jan 2026 20:32:00 GMT", "contentSnippet": "The latest gadgets, the future in assistive tech and upcoming gaming releases in 2026." } }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpx#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpx?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "We bring you Tech Life highlights from a fascinating year in global tech.", "isoDate": "2025-12-30T20:30:00.000Z", "pubDate": "Tue, 30 Dec 2025 20:30:00 GMT", "contentSnippet": "We bring you Tech Life highlights from a fascinating year in global tech." } }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpv#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpv?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "A study found AI chatbots can persuade us with fake facts. How does this affect politics?", "isoDate": "2025-12-16T20:30:00.000Z", "pubDate": "Tue, 16 Dec 2025 20:30:00 GMT", "contentSnippet": "A study found AI chatbots can persuade us with fake facts. How does this affect politics?" } } ], "Send Newsletter": [ { "json": { "id": "19c46e0034cccc8e", "labelIds": [ "UNREAD", "SENT", "INBOX" ], "threadId": "19c46e0034cccc8e" }, "pairedItem": { "item": 0 } } ], "Combine Articles": [ { "json": { "data": [ { "link": "https://techcrunch.com/2026/01/13/ai-drug-discovery-startup-converge-bio-pulls-in-25m-from-bessemer-and-execs-from-meta-openai-and-wiz/", "title": "Converge Bio raises $25M, backed by Bessemer and execs from Meta, OpenAI, Wiz", "content": "AI drug discovery startup Converge Bio raised $25 million in a Series A led by Bessemer Venture Partners, with additional backing from executives at Meta, OpenAI, and Wiz.", "pubDate": "Tue, 13 Jan 2026 11:30:00 +0000" }, { "link": "https://techcrunch.com/2025/10/31/meta-bought-1-gw-of-solar-this-week/", "title": "Meta bought 1 GW of solar this week", "content": "The social media company inked three deals in the U.S. to power its data centers and offset its carbon footprint.", "pubDate": "Fri, 31 Oct 2025 19:26:30 +0000" }, { "link": "https://techcrunch.com/2025/08/26/how-one-ai-startup-is-helping-rice-farmers-battle-climate-change/", "title": "How one AI startup is helping rice farmers battle climate change", "content": "Mitti Labs is working with The Nature Conservancy to expand the use of climate-friendly rice farming practices in India. The startup uses its AI to verify reductions in methane emissions.", "pubDate": "Tue, 26 Aug 2025 15:21:19 +0000" }, { "link": "https://techcrunch.com/2025/08/20/harvard-dropouts-to-launch-always-on-ai-smart-glasses-that-listen-and-record-every-conversation/", "title": "Harvard dropouts to launch ‘always on’ AI smart glasses that listen and record every conversation", "content": "After developing a facial-recognition app for Meta’s Ray-Ban glasses and doxing random people, two former Harvard students are now launching a startup that makes smart glasses with an always-on microphone.", "pubDate": "Wed, 20 Aug 2025 16:00:00 +0000" }, { "link": "https://techcrunch.com/2025/08/20/meta-to-add-100-mw-of-solar-power-from-u-s-gear/", "title": "Meta to add 100MW of solar power from US gear", "content": "The social media company is adding another tranche of solar to power a new AI data center in South Carolina.", "pubDate": "Wed, 20 Aug 2025 15:56:53 +0000" }, { "link": "https://www.technologyreview.com/2026/02/09/1132537/a-lesson-from-pokemon/", "title": "Why the Moltbook frenzy was like Pokémon", "content": "This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Lots of influential people in tech last week were describing Moltbook, an online hangout populated by AI agents interacting with one another, as a glimpse into the future. It appeared to show…", "pubDate": "Mon, 09 Feb 2026 17:02:56 +0000" }, { "link": "https://www.technologyreview.com/2026/02/09/1132498/the-download-what-moltbook-tells-us-about-ai-hype-and-the-rise-and-rise-of-ai-therapy/", "title": "The Download: what Moltbook tells us about AI hype, and the rise and rise of AI therapy", "content": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Moltbook was peak AI theater For a few days recently, the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots.…", "pubDate": "Mon, 09 Feb 2026 13:10:00 +0000" }, { "link": "https://www.technologyreview.com/2026/02/09/1132462/ai-newsletter-professional-applications/", "title": "Making AI Work, MIT Technology Review’s new AI newsletter, is here", "content": "For years, our newsroom has explored AI’s limitations and potential dangers, as well as its growing energy needs. And our reporters have looked closely at how generative tools are being used for tasks such as coding and running scientific experiments. But how is AI actually being used in fields like health care, climate tech, education,…", "pubDate": "Mon, 09 Feb 2026 11:30:00 +0000" }, { "link": "https://www.technologyreview.com/2026/02/06/1132448/moltbook-was-peak-ai-theater/", "title": "Moltbook was peak AI theater", "content": "For a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.” We observed! Launched on January 28 by Matt Schlicht,…", "pubDate": "Fri, 06 Feb 2026 16:38:11 +0000" }, { "link": "https://www.technologyreview.com/2026/02/06/1132375/the-download-helping-cancer-survivors-to-give-birth-and-cleaning-up-bangladeshs-garment-industry/", "title": "The Download: helping cancer survivors to give birth, and cleaning up Bangladesh’s garment industry", "content": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. An experimental surgery is helping cancer survivors give birth An experimental surgical procedure that’s helping people have babies after they’ve had treatment for bowel or rectal cancer. Radiation and chemo can have pretty…", "pubDate": "Fri, 06 Feb 2026 13:10:00 +0000" }, { "link": "https://venturebeat.com/infrastructure/railway-secures-usd100-million-to-challenge-aws-with-ai-native-cloud", "title": "Railway secures $100 million to challenge AWS with AI-native cloud infrastructure", "content": "Railway, a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence applications exposes the limitations of legacy cloud infrastructure.\nTQ Ventures led the round, with participation from FPV Ventures, Redpoint, and Unusual Ventures. The investment values Railway as one of the most significant infrastructure startups to emerge during the AI boom, capitalizing on developer frustration with the complexity and cost of traditional platforms like Amazon Web Services and Google Cloud.\n\"As AI models get better at writing code, more and more people are asking the age-old question: where, and how, do I run my applications?\" said Jake Cooper, Railway's 28-year-old founder and chief executive, in an exclusive interview with VentureBeat. \"The last generation of cloud primitives were slow and outdated, and now with AI moving everything faster, teams simply can't keep up.\"\nThe funding is a dramatic acceleration for a company that has charted an unconventional path through the cloud computing industry. Railway raised just $24 million in total before this round, including a $20 million Series A from Redpoint in 2022. The company now processes more than 10 million deployments monthly and handles over one trillion requests through its edge network — metrics that rival far larger and better-funded competitors.\nWhy three-minute deploy times have become unacceptable in the age of AI coding assistants\nRailway's pitch rests on a simple observation: the tools developers use to deploy and manage software were designed for a slower era. A standard build-and-deploy cycle using Terraform, the industry-standard infrastructure tool, takes two to three minutes. That delay, once tolerable, has become a critical bottleneck as AI coding assistants like Claude, ChatGPT, and Cursor can generate working code in seconds.\n\"When godly intelligence is on tap and can solve any problem in three seconds, those amalgamations of systems become bottlenecks,\" Cooper told VentureBeat. \"What was really cool for humans to deploy in 10 seconds or less is now table stakes for agents.\"\nThe company claims its platform delivers deployments in under one second — fast enough to keep pace with AI-generated code. Customers report a tenfold increase in developer velocity and up to 65 percent cost savings compared to traditional cloud providers.\nThese numbers come directly from enterprise clients, not internal benchmarks. Daniel Lobaton, chief technology officer at G2X, a platform serving 100,000 federal contractors, measured deployment speed improvements of seven times faster and an 87 percent cost reduction after migrating to Railway. His infrastructure bill dropped from $15,000 per month to approximately $1,000.\n\"The work that used to take me a week on our previous infrastructure, I can do in Railway in like a day,\" Lobaton said. \"If I want to spin up a new service and test different architectures, it would take so long on our old setup. In Railway I can launch six services in two minutes.\"\nInside the controversial decision to abandon Google Cloud and build data centers from scratch\nWhat distinguishes Railway from competitors like Render and Fly.io is the depth of its vertical integration. In 2024, the company made the unusual decision to abandon Google Cloud entirely and build its own data centers, a move that echoes the famous Alan Kay maxim: \"People who are really serious about software should make their own hardware.\"\n\"We wanted to design hardware in a way where we could build a differentiated experience,\" Cooper said. \"Having full control over the network, compute, and storage layers lets us do really fast build and deploy loops, the kind that allows us to move at 'agentic speed' while staying 100 percent the smoothest ride in town.\"\nThe approach paid dividends during recent widespread outages that affected major cloud providers — Railway remained online throughout.\nThis soup-to-nuts control enables pricing that undercuts the hyperscalers by roughly 50 percent and newer cloud startups by three to four times. Railway charges by the second for actual compute usage: $0.00000386 per gigabyte-second of memory, $0.00000772 per vCPU-second, and $0.00000006 per gigabyte-second of storage. There are no charges for idle virtual machines — a stark contrast to the traditional cloud model where customers pay for provisioned capacity whether they use it or not.\n\"The conventional wisdom is that the big guys have economies of scale to offer better pricing,\" Cooper noted. \"But when they're charging for VMs that usually sit idle in the cloud, and we've purpose-built everything to fit much more density on these machines, you have a big opportunity.\"\nHow 30 employees built a platform generating tens of millions in annual revenue\nRailway has achieved its scale with a team of just 30 employees generating tens of millions in annual revenue — a ratio of revenue per employee that would be exceptional even for established software companies. The company grew revenue 3.5 times last year and continues to expand at 15 percent month-over-month.\nCooper emphasized that the fundraise was strategic rather than necessary. \"We're default alive; there's no reason for us to raise money,\" he said. \"We raised because we see a massive opportunity to accelerate, not because we needed to survive.\"\nThe company hired its first salesperson only last year and employs just two solutions engineers. Nearly all of Railway's two million users discovered the platform through word of mouth — developers telling other developers about a tool that actually works.\n\"We basically did the standard engineering thing: if you build it, they will come,\" Cooper recalled. \"And to some degree, they came.\"\nFrom side projects to Fortune 500 deployments: Railway's unlikely corporate expansion\nDespite its grassroots developer community, Railway has made significant inroads into large organizations. The company claims that 31 percent of Fortune 500 companies now use its platform, though deployments range from company-wide infrastructure to individual team projects.\nNotable customers include Bilt, the loyalty program company; Intuit's GoCo subsidiary; TripAdvisor's Cruise Critic; and MGM Resorts. Kernel, a Y Combinator-backed startup providing AI infrastructure to over 1,000 companies, runs its entire customer-facing system on Railway for $444 per month.\n\"At my previous company Clever, which sold for $500 million, I had six full-time engineers just managing AWS,\" said Rafael Garcia, Kernel's chief technology officer. \"Now I have six engineers total, and they all focus on product. Railway is exactly the tool I wish I had in 2012.\"\nFor enterprise customers, Railway offers security certifications including SOC 2 Type 2 compliance and HIPAA readiness, with business associate agreements available upon request. The platform provides single sign-on authentication, comprehensive audit logs, and the option to deploy within a customer's existing cloud environment through a \"bring your own cloud\" configuration.\nEnterprise pricing starts at custom levels, with specific add-ons for extended log retention ($200 monthly), HIPAA BAAs ($1,000), enterprise support with SLOs ($2,000), and dedicated virtual machines ($10,000).\nThe startup's bold strategy to take on Amazon, Google, and a new generation of cloud rivals\nRailway enters a crowded market that includes not only the hyperscale cloud providers—Amazon Web Services, Microsoft Azure, and Google Cloud Platform—but also a growing cohort of developer-focused platforms like Vercel, Render, Fly.io, and Heroku.\nCooper argues that Railway's competitors fall into two camps, neither of which has fully committed to the new infrastructure model that AI demands.\n\"The hyperscalers have two competing systems, and they haven't gone all-in on the new model because their legacy revenue stream is still printing money,\" he observed. \"They have this mammoth pool of cash coming from people who provision a VM, use maybe 10 percent of it, and still pay for the whole thing. To what end are they actually interested in going all the way in on a new experience if they don't really need to?\"\nAgainst startup competitors, Railway differentiates by covering the full infrastructure stack. \"We're not just containers; we've got VM primitives, stateful storage, virtual private networking, automated load balancing,\" Cooper said. \"And we wrap all of this in an absurdly easy-to-use UI, with agentic primitives so agents can move 1,000 times faster.\"\nThe platform supports databases including PostgreSQL, MySQL, MongoDB, and Redis; provides up to 256 terabytes of persistent storage with over 100,000 input/output operations per second; and enables deployment to four global regions spanning the United States, Europe, and Southeast Asia. Enterprise customers can scale to 112 vCPUs and 2 terabytes of RAM per service.\nWhy investors are betting that AI will create a thousand times more software than exists today\nRailway's fundraise reflects broader investor enthusiasm for companies positioned to benefit from the AI coding revolution. As tools like GitHub Copilot, Cursor, and Claude become standard fixtures in developer workflows, the volume of code being written — and the infrastructure needed to run it — is expanding dramatically.\n\"The amount of software that's going to come online over the next five years is unfathomable compared to what existed before — we're talking a thousand times more software,\" Cooper predicted. \"All of that has to run somewhere.\"\nThe company has already integrated directly with AI systems, building what Cooper calls \"loops where Claude can hook in, call deployments, and analyze infrastructure automatically.\" Railway released a Model Context Protocol server in August 2025 that allows AI coding agents to deploy applications and manage infrastructure directly from code editors.\n\"The notion of a developer is melting before our eyes,\" Cooper said. \"You don't have to be an engineer to engineer things anymore — you just need critical thinking and the ability to analyze things in a systems capacity.\"\nWhat Railway plans to do with $100 million and zero marketing experience\nRailway plans to use the new capital to expand its global data center footprint, grow its team beyond 30 employees, and build what Cooper described as a proper go-to-market operation for the first time in the company's five-year history.\n\"One of my mentors said you raise money when you can change the trajectory of the business,\" Cooper explained. \"We've built all the required substrate to scale indefinitely; what's been holding us back is simply talking about it. 2026 is the year we play on the world stage.\"\nThe company's investor roster reads like a who's who of developer infrastructure. Angel investors include Tom Preston-Werner, co-founder of GitHub; Guillermo Rauch, chief executive of Vercel; Spencer Kimball, chief executive of Cockroach Labs; Olivier Pomel, chief executive of Datadog; and Jori Lallo, co-founder of Linear.\nThe timing of Railway's expansion coincides with what many in Silicon Valley view as a fundamental shift in how software gets made. Coding assistants are no longer experimental curiosities — they have become essential tools that millions of developers rely on daily. Each line of AI-generated code needs somewhere to run, and the incumbents, by Cooper's telling, are too wedded to their existing business models to fully capitalize on the moment.\nWhether Railway can translate developer enthusiasm into sustained enterprise adoption remains an open question. The cloud infrastructure market is littered with promising startups that failed to break the grip of Amazon, Microsoft, and Google. But Cooper, who previously worked as a software engineer at Wolfram Alpha, Bloomberg, and Uber before founding Railway in 2020, seems unfazed by the scale of his ambition.\n\"In five years, Railway [will be] the place where software gets created and evolved, period,\" he said. \"Deploy instantly, scale infinitely, with zero friction. That's the prize worth playing for, and there's no bigger one on offer.\"\nFor a company that built a $100 million business by doing the opposite of what conventional startup wisdom dictates — no marketing, no sales team, no venture hype—the real test begins now. Railway spent five years proving that developers would find a better mousetrap on their own. The next five will determine whether the rest of the world is ready to get on board.", "pubDate": "Thu, 22 Jan 2026 14:00:00 GMT" }, { "link": "https://venturebeat.com/infrastructure/claude-code-costs-up-to-usd200-a-month-goose-does-the-same-thing-for-free", "title": "Claude Code costs up to $200 a month. Goose does the same thing for free.", "content": "The artificial intelligence coding revolution comes with a catch: it's expensive.\nClaude Code, Anthropic's terminal-based AI agent that can write, debug, and deploy code autonomously, has captured the imagination of software developers worldwide. But its pricing — ranging from $20 to $200 per month depending on usage — has sparked a growing rebellion among the very programmers it aims to serve.\nNow, a free alternative is gaining traction. Goose, an open-source AI agent developed by Block (the financial technology company formerly known as Square), offers nearly identical functionality to Claude Code but runs entirely on a user's local machine. No subscription fees. No cloud dependency. No rate limits that reset every five hours.\n\"Your data stays with you, period,\" said Parth Sareen, a software engineer who demonstrated the tool during a recent livestream. The comment captures the core appeal: Goose gives developers complete control over their AI-powered workflow, including the ability to work offline — even on an airplane.\nThe project has exploded in popularity. Goose now boasts more than 26,100 stars on GitHub, the code-sharing platform, with 362 contributors and 102 releases since its launch. The latest version, 1.20.1, shipped on January 19, 2026, reflecting a development pace that rivals commercial products.\nFor developers frustrated by Claude Code's pricing structure and usage caps, Goose represents something increasingly rare in the AI industry: a genuinely free, no-strings-attached option for serious work.\n\nAnthropic's new rate limits spark a developer revolt\nTo understand why Goose matters, you need to understand the Claude Code pricing controversy.\nAnthropic, the San Francisco artificial intelligence company founded by former OpenAI executives, offers Claude Code as part of its subscription tiers. The free plan provides no access whatsoever. The Pro plan, at $17 per month with annual billing (or $20 monthly), limits users to just 10 to 40 prompts every five hours — a constraint that serious developers exhaust within minutes of intensive work.\nThe Max plans, at $100 and $200 per month, offer more headroom: 50 to 200 prompts and 200 to 800 prompts respectively, plus access to Anthropic's most powerful model, Claude 4.5 Opus. But even these premium tiers come with restrictions that have inflamed the developer community.\nIn late July, Anthropic announced new weekly rate limits. Under the system, Pro users receive 40 to 80 hours of Sonnet 4 usage per week. Max users at the $200 tier get 240 to 480 hours of Sonnet 4, plus 24 to 40 hours of Opus 4. Nearly five months later, the frustration has not subsided.\nThe problem? Those \"hours\" are not actual hours. They represent token-based limits that vary wildly depending on codebase size, conversation length, and the complexity of the code being processed. Independent analysis suggests the actual per-session limits translate to roughly 44,000 tokens for Pro users and 220,000 tokens for the $200 Max plan.\n\"It's confusing and vague,\" one developer wrote in a widely shared analysis. \"When they say '24-40 hours of Opus 4,' that doesn't really tell you anything useful about what you're actually getting.\"\nThe backlash on Reddit and developer forums has been fierce. Some users report hitting their daily limits within 30 minutes of intensive coding. Others have canceled their subscriptions entirely, calling the new restrictions \"a joke\" and \"unusable for real work.\"\nAnthropic has defended the changes, stating that the limits affect fewer than five percent of users and target people running Claude Code \"continuously in the background, 24/7.\" But the company has not clarified whether that figure refers to five percent of Max subscribers or five percent of all users — a distinction that matters enormously.\nHow Block built a free AI coding agent that works offline\nGoose takes a radically different approach to the same problem.\nBuilt by Block, the payments company led by Jack Dorsey, Goose is what engineers call an \"on-machine AI agent.\" Unlike Claude Code, which sends your queries to Anthropic's servers for processing, Goose can run entirely on your local computer using open-source language models that you download and control yourself.\nThe project's documentation describes it as going \"beyond code suggestions\" to \"install, execute, edit, and test with any LLM.\" That last phrase — \"any LLM\" — is the key differentiator. Goose is model-agnostic by design.\nYou can connect Goose to Anthropic's Claude models if you have API access. You can use OpenAI's GPT-5 or Google's Gemini. You can route it through services like Groq or OpenRouter. Or — and this is where things get interesting — you can run it entirely locally using tools like Ollama, which let you download and execute open-source models on your own hardware.\nThe practical implications are significant. With a local setup, there are no subscription fees, no usage caps, no rate limits, and no concerns about your code being sent to external servers. Your conversations with the AI never leave your machine.\n\"I use Ollama all the time on planes — it's a lot of fun!\" Sareen noted during a demonstration, highlighting how local models free developers from the constraints of internet connectivity.\nWhat Goose can do that traditional code assistants can't\nGoose operates as a command-line tool or desktop application that can autonomously perform complex development tasks. It can build entire projects from scratch, write and execute code, debug failures, orchestrate workflows across multiple files, and interact with external APIs — all without constant human oversight.\nThe architecture relies on what the AI industry calls \"tool calling\" or \"function calling\" — the ability for a language model to request specific actions from external systems. When you ask Goose to create a new file, run a test suite, or check the status of a GitHub pull request, it doesn't just generate text describing what should happen. It actually executes those operations.\nThis capability depends heavily on the underlying language model. Claude 4 models from Anthropic currently perform best at tool calling, according to the Berkeley Function-Calling Leaderboard, which ranks models on their ability to translate natural language requests into executable code and system commands.\nBut newer open-source models are catching up quickly. Goose's documentation highlights several options with strong tool-calling support: Meta's Llama series, Alibaba's Qwen models, Google's Gemma variants, and DeepSeek's reasoning-focused architectures.\nThe tool also integrates with the Model Context Protocol, or MCP, an emerging standard for connecting AI agents to external services. Through MCP, Goose can access databases, search engines, file systems, and third-party APIs — extending its capabilities far beyond what the base language model provides.\nSetting Up Goose with a Local Model\nFor developers interested in a completely free, privacy-preserving setup, the process involves three main components: Goose itself, Ollama (a tool for running open-source models locally), and a compatible language model.\nStep 1: Install Ollama\nOllama is an open-source project that dramatically simplifies the process of running large language models on personal hardware. It handles the complex work of downloading, optimizing, and serving models through a simple interface.\nDownload and install Ollama from ollama.com. Once installed, you can pull models with a single command. For coding tasks, Qwen 2.5 offers strong tool-calling support:\nollama run qwen2.5\nThe model downloads automatically and begins running on your machine.\nStep 2: Install Goose\nGoose is available as both a desktop application and a command-line interface. The desktop version provides a more visual experience, while the CLI appeals to developers who prefer working entirely in the terminal.\nInstallation instructions vary by operating system but generally involve downloading from Goose's GitHub releases page or using a package manager. Block provides pre-built binaries for macOS (both Intel and Apple Silicon), Windows, and Linux.\nStep 3: Configure the Connection\nIn Goose Desktop, navigate to Settings, then Configure Provider, and select Ollama. Confirm that the API Host is set to http://localhost:11434 (Ollama's default port) and click Submit.\nFor the command-line version, run goose configure, select \"Configure Providers,\" choose Ollama, and enter the model name when prompted.\nThat's it. Goose is now connected to a language model running entirely on your hardware, ready to execute complex coding tasks without any subscription fees or external dependencies.\nThe RAM, processing power, and trade-offs you should know about\nThe obvious question: what kind of computer do you need?\nRunning large language models locally requires substantially more computational resources than typical software. The key constraint is memory — specifically, RAM on most systems, or VRAM if using a dedicated graphics card for acceleration.\nBlock's documentation suggests that 32 gigabytes of RAM provides \"a solid baseline for larger models and outputs.\" For Mac users, this means the computer's unified memory is the primary bottleneck. For Windows and Linux users with discrete NVIDIA graphics cards, GPU memory (VRAM) matters more for acceleration.\nBut you don't necessarily need expensive hardware to get started. Smaller models with fewer parameters run on much more modest systems. Qwen 2.5, for instance, comes in multiple sizes, and the smaller variants can operate effectively on machines with 16 gigabytes of RAM.\n\"You don't need to run the largest models to get excellent results,\" Sareen emphasized. The practical recommendation: start with a smaller model to test your workflow, then scale up as needed.\nFor context, Apple's entry-level MacBook Air with 8 gigabytes of RAM would struggle with most capable coding models. But a MacBook Pro with 32 gigabytes — increasingly common among professional developers — handles them comfortably.\nWhy keeping your code off the cloud matters more than ever\nGoose with a local LLM is not a perfect substitute for Claude Code. The comparison involves real trade-offs that developers should understand.\nModel Quality: Claude 4.5 Opus, Anthropic's flagship model, remains arguably the most capable AI for software engineering tasks. It excels at understanding complex codebases, following nuanced instructions, and producing high-quality code on the first attempt. Open-source models have improved dramatically, but a gap persists — particularly for the most challenging tasks.\nOne developer who switched to the $200 Claude Code plan described the difference bluntly: \"When I say 'make this look modern,' Opus knows what I mean. Other models give me Bootstrap circa 2015.\"\nContext Window: Claude Sonnet 4.5, accessible through the API, offers a massive one-million-token context window — enough to load entire large codebases without chunking or context management issues. Most local models are limited to 4,096 or 8,192 tokens by default, though many can be configured for longer contexts at the cost of increased memory usage and slower processing.\nSpeed: Cloud-based services like Claude Code run on dedicated server hardware optimized for AI inference. Local models, running on consumer laptops, typically process requests more slowly. The difference matters for iterative workflows where you're making rapid changes and waiting for AI feedback.\nTooling Maturity: Claude Code benefits from Anthropic's dedicated engineering resources. Features like prompt caching (which can reduce costs by up to 90 percent for repeated contexts) and structured outputs are polished and well-documented. Goose, while actively developed with 102 releases to date, relies on community contributions and may lack equivalent refinement in specific areas.\nHow Goose stacks up against Cursor, GitHub Copilot, and the paid AI coding market\nGoose enters a crowded market of AI coding tools, but occupies a distinctive position.\nCursor, a popular AI-enhanced code editor, charges $20 per month for its Pro tier and $200 for Ultra—pricing that mirrors Claude Code's Max plans. Cursor provides approximately 4,500 Sonnet 4 requests per month at the Ultra level, a substantially different allocation model than Claude Code's hourly resets.\nCline, Roo Code, and similar open-source projects offer AI coding assistance but with varying levels of autonomy and tool integration. Many focus on code completion rather than the agentic task execution that defines Goose and Claude Code.\nAmazon's CodeWhisperer, GitHub Copilot, and enterprise offerings from major cloud providers target large organizations with complex procurement processes and dedicated budgets. They are less relevant to individual developers and small teams seeking lightweight, flexible tools.\nGoose's combination of genuine autonomy, model agnosticism, local operation, and zero cost creates a unique value proposition. The tool is not trying to compete with commercial offerings on polish or model quality. It's competing on freedom — both financial and architectural.\nThe $200-a-month era for AI coding tools may be ending\nThe AI coding tools market is evolving quickly. Open-source models are improving at a pace that continually narrows the gap with proprietary alternatives. Moonshot AI's Kimi K2 and z.ai's GLM 4.5 now benchmark near Claude Sonnet 4 levels — and they're freely available.\nIf this trajectory continues, the quality advantage that justifies Claude Code's premium pricing may erode. Anthropic would then face pressure to compete on features, user experience, and integration rather than raw model capability.\nFor now, developers face a clear choice. Those who need the absolute best model quality, who can afford premium pricing, and who accept usage restrictions may prefer Claude Code. Those who prioritize cost, privacy, offline access, and flexibility have a genuine alternative in Goose.\nThe fact that a $200-per-month commercial product has a zero-dollar open-source competitor with comparable core functionality is itself remarkable. It reflects both the maturation of open-source AI infrastructure and the appetite among developers for tools that respect their autonomy.\nGoose is not perfect. It requires more technical setup than commercial alternatives. It depends on hardware resources that not every developer possesses. Its model options, while improving rapidly, still trail the best proprietary offerings on complex tasks.\nBut for a growing community of developers, those limitations are acceptable trade-offs for something increasingly rare in the AI landscape: a tool that truly belongs to them.\n\nGoose is available for download at github.com/block/goose. Ollama is available at ollama.com. Both projects are free and open source.", "pubDate": "Mon, 19 Jan 2026 14:00:00 GMT" }, { "link": "https://venturebeat.com/technology/listen-labs-raises-usd69m-after-viral-billboard-hiring-stunt-to-scale-ai", "title": "Listen Labs raises $69M after viral billboard hiring stunt to scale AI customer interviews", "content": "Alfred Wahlforss was running out of options. His startup, Listen Labs, needed to hire over 100 engineers, but competing against Mark Zuckerberg's $100 million offers seemed impossible. So he spent $5,000 — a fifth of his marketing budget — on a billboard in San Francisco displaying what looked like gibberish: five strings of random numbers.\nThe numbers were actually AI tokens. Decoded, they led to a coding challenge: build an algorithm to act as a digital bouncer at Berghain, the Berlin nightclub famous for rejecting nearly everyone at the door. Within days, thousands attempted the puzzle. 430 cracked it. Some got hired. The winner flew to Berlin, all expenses paid.\nThat unconventional approach has now attracted $69 million in Series B funding, led by Ribbit Capital with participation from Evantic and existing investors Sequoia Capital, Conviction, and Pear VC. The round values Listen Labs at $500 million and brings its total capital to $100 million. In nine months since launch, the company has grown annualized revenue by 15x to eight figures and conducted over one million AI-powered interviews.\n\n\"When you obsess over customers, everything else follows,\" Wahlforss said in an interview with VentureBeat. \"Teams that use Listen bring the customer into every decision, from marketing to product, and when the customer is delighted, everyone is.\"\nWhy traditional market research is broken, and what Listen Labs is building to fix it\nListen's AI researcher finds participants, conducts in-depth interviews, and delivers actionable insights in hours, not weeks. The platform replaces the traditional choice between quantitative surveys — which provide statistical precision but miss nuance—and qualitative interviews, which deliver depth but cannot scale.\nWahlforss explained the limitation of existing approaches: \"Essentially surveys give you false precision because people end up answering the same question... You can't get the outliers. People are actually not honest on surveys.\" The alternative, one-on-one human interviews, \"gives you a lot of depth. You can ask follow up questions. You can kind of double check if they actually know what they're talking about. And the problem is you can't scale that.\"\nThe platform works in four steps: users create a study with AI assistance, Listen recruits participants from its global network of 30 million people, an AI moderator conducts in-depth interviews with follow-up questions, and results are packaged into executive-ready reports including key themes, highlight reels, and slide decks.\nWhat distinguishes Listen's approach is its use of open-ended video conversations rather than multiple-choice forms. \"In a survey, you can kind of guess what you should answer, and you have four options,\" Wahlforss said. \"Oh, they probably want me to buy high income. Let me click on that button versus an open ended response. It just generates much more honesty.\"\nThe dirty secret of the $140 billion market research industry: rampant fraud\nListen finds and qualifies the right participants in its global network of 30 million people. But building that panel required confronting what Wahlforss called \"one of the most shocking things that we've learned when we entered this industry\"—rampant fraud.\n\"Essentially, there's a financial transaction involved, which means there will be bad players,\" he explained. \"We actually had some of the largest companies, some of them have billions in revenue, send us people who claim to be kind of enterprise buyers to our platform and our system immediately detected, like, fraud, fraud, fraud, fraud, fraud.\"\nThe company built what it calls a \"quality guard\" that cross-references LinkedIn profiles with video responses to verify identity, checks consistency across how participants answer questions, and flags suspicious patterns. The result, according to Wahlforss: \"People talk three times more. They're much more honest when they talk about sensitive topics like politics and mental health.\"\nEmeritus, an online education company that uses Listen, reported that approximately 20% of survey responses previously fell into the fraudulent or low-quality category. With Listen, they reduced this to almost zero. \"We did not have to replace any responses because of fraud or gibberish information,\" said Gabrielli Tiburi, Assistant Manager of Customer Insights at Emeritus.\nHow Microsoft, Sweetgreen, and Chubbies are using AI interviews to build better products\nThe speed advantage has proven central to Listen's pitch. Traditional customer research at Microsoft could take four to six weeks to generate insights. \"By the time we get to them, either the decision has been made or we lose out on the opportunity to actually influence it,\" said Romani Patel, Senior Research Manager at Microsoft.\nWith Listen, Microsoft can now get insights in days, and in many cases, within hours.\nThe platform has already powered several high-profile initiatives. Microsoft used Listen Labs to collect global customer stories for its 50th anniversary celebration. \"We wanted users to share how Copilot is empowering them to bring their best self forward,\" Patel said, \"and we were able to collect those user video stories within a day.\" Traditionally, that kind of work would have taken six to eight weeks.\nSimple Modern, an Oklahoma-based drinkware company, used Listen to test a new product concept. The process took about an hour to write questions, an hour to launch the study, and 2.5 hours to receive feedback from 120 people across the country. \"We went from 'Should we even have this product?' to 'How should we launch it?'\" said Chris Hoyle, the company's Chief Marketing Officer.\nChubbies, the shorts brand, achieved a 24x increase in youth research participation—growing from 5 to 120 participants — by using Listen to overcome the scheduling challenges of traditional focus groups with children. \"There's school, sports, dinner, and homework,\" explained Lauren Neville, Director of Insights and Innovation. \"I had to find a way to hear from them that fit into their schedules.\"\nThe company also discovered product issues through AI interviews that might have gone undetected otherwise. Wahlforss described how the AI \"through conversations, realized there were like issues with the the kids short line, and decided to, like, interview hundreds of kids. And I understand that there were issues in the liner of the shorts and that they were, like, scratchy, quote, unquote, according to the people interviewed.\" The redesigned product became \"a blockbuster hit.\"\nThe Jevons paradox explains why cheaper research creates more demand, not less\nListen Labs is entering a massive but fragmented market. Wahlforss cited research from Andreessen Horowitz estimating the market research industry at roughly $140 billion annually, populated by legacy players — some with more than a billion dollars in revenue — that he believes are vulnerable to disruption.\n\"There are very much existing budget lines that we are replacing,\" Wahlforss said. \"Why we're replacing them is that one, they're super costly. Two, they're kind of stuck in this old paradigm of choosing between a survey or interview, and they also take months to work with.\"\nBut the more intriguing dynamic may be that AI-powered research doesn't just replace existing spending — it creates new demand. Wahlforss invoked the Jevons paradox, an economic principle that occurs when technological advancements make a resource more efficient to use, but increased efficiency leads to increased overall consumption rather than decreased consumption.\n\"What I've noticed is that as something gets cheaper, you don't need less of it. You want more of it,\" Wahlforss explained. \"There's infinite demand for customer understanding. So the researchers on the team can do an order of magnitude more research, and also other people who weren't researchers before can now do that as part of their job.\"\nInside the elite engineering team that built Listen Labs before they had a working toilet\nListen Labs traces its origins to a consumer app that Wahlforss and his co-founder built after meeting at Harvard. \"We built this consumer app that got 20,000 downloads in one day,\" Wahlforss recalled. \"We had all these users, and we were thinking like, okay, what can we do to get to know them better? And we built this prototype of what Listen is today.\"\nThe founding team brings an unusual pedigree. Wahlforss's co-founder \"was the national champion in competitive programming in Germany, and he worked at Tesla Autopilot.\" The company claims that 30% of its engineering team are medalists from the International Olympiad in Informatics — the same competition that produced the founders of Cognition, the AI coding startup.\nThe Berghain billboard stunt generated approximately 5 million views across social media, according to Wahlforss. It reflected the intensity of the talent war in the Bay Area.\n\"We had to do these things because some of our, like early employees, joined the company before we had a working toilet,\" he said. \"But now we fixed that situation.\"\nThe company grew from 5 to 40 employees in 2024 and plans to reach 150 this year. It hires engineers for non-engineering roles across marketing, growth, and operations — a bet that in the AI era, technical fluency matters everywhere.\nSynthetic customers and automated decisions: what Listen Labs is building next\nWahlforss outlined an ambitious product roadmap that pushes into more speculative territory. The company is building \"the ability to simulate your customers, so you can take all of those interviews we've done, and then extrapolate based on that and create synthetic users or simulated user voices.\"\nBeyond simulation, Listen aims to enable automated action based on research findings. \"Can you not just make recommendations, but also create spawn agents to either change things in code or some customer churns? Can you give them a discount and try to bring them back?\"\nWahlforss acknowledged the ethical implications. \"Obviously, as you said, there's kind of ethical concerns there. Of like, automated decision making overall can be bad, but we will have considerable guardrails to make sure that the companies are always in the loop.\"\nThe company already handles sensitive data with care. \"We don't train on any of the data,\" Wahlforss said. \"We will also scrub any sensitive PII automatically so the model can detect that. And there are times when, for example, you work with investors, where if you accidentally mention something that could be material, non public information, the AI can actually detect that and remove any information like that.\"\nHow AI could reshape the future of product development\nPerhaps the most provocative implication of Listen's model is how it could reshape product development itself. Wahlforss described a customer — an Australian startup — that has adopted what amounts to a continuous feedback loop.\n\"They're based in Australia, so they're coding during the day, and then in their night, they're releasing a Listen study with an American audience. Listen validates whatever they built during the day, and they get feedback on that. They can then plug that feedback directly into coding tools like Claude Code and iterate.\"\nThe vision extends Y Combinator's famous dictum — \"write code, talk to users\" — into an automated cycle. \"Write code is now getting automated. And I think like talk to users will be as well, and you'll have this kind of infinite loop where you can start to ship this truly amazing product, almost kind of autonomously.\"\nWhether that vision materializes depends on factors beyond Listen's control — the continued improvement of AI models, enterprise willingness to trust automated research, and whether speed truly correlates with better products. A 2024 MIT study found that 95% of AI pilots fail to move into production, a statistic Wahlforss cited as the reason he emphasizes quality over demos.\n\"I'm constantly have to emphasize like, let's make sure the quality is there and the details are right,\" he said.\nBut the company's growth suggests appetite for the experiment. Microsoft's Patel said Listen has \"removed the drudgery of research and brought the fun and joy back into my work.\" Chubbies is now pushing its founder to give everyone in the company a login. Sling Money, a stablecoin payments startup, can create a survey in ten minutes and receive results the same day.\n\"It's a total game changer,\" said Ali Romero, Sling Money's marketing manager.\nWahlforss has a different phrase for what he's building. When asked about the tension between speed and rigor — the long-held belief that moving fast means cutting corners — he cited Nat Friedman, the former GitHub CEO and Listen investor, who keeps a list of one-liners on his website.\nOne of them: \"Slow is fake.\"\nIt's an aggressive claim for an industry built on methodological caution. But Listen Labs is betting that in the AI era, the companies that listen fastest will be the ones that win. The only question is whether customers will talk back.", "pubDate": "Fri, 16 Jan 2026 14:01:00 GMT" }, { "link": "https://venturebeat.com/technology/salesforce-rolls-out-new-slackbot-ai-agent-as-it-battles-microsoft-and", "title": "Salesforce rolls out new Slackbot AI agent as it battles Microsoft and Google in workplace AI", "content": "Salesforce on Tuesday launched an entirely rebuilt version of Slackbot, the company's workplace assistant, transforming it from a simple notification tool into what executives describe as a fully powered AI agent capable of searching enterprise data, drafting documents, and taking action on behalf of employees.\nThe new Slackbot, now generally available to Business+ and Enterprise+ customers, is Salesforce's most aggressive move yet to position Slack at the center of the emerging \"agentic AI\" movement — where software agents work alongside humans to complete complex tasks. The launch comes as Salesforce attempts to convince investors that artificial intelligence will bolster its products rather than render them obsolete.\n\"Slackbot isn't just another copilot or AI assistant,\" said Parker Harris, Salesforce co-founder and Slack's chief technology officer, in an exclusive interview with Salesforce. \"It's the front door to the agentic enterprise, powered by Salesforce.\"\nFrom tricycle to Porsche: Salesforce rebuilt Slackbot from the ground up\nHarris was blunt about what distinguishes the new Slackbot from its predecessor: \"The old Slackbot was, you know, a little tricycle, and the new Slackbot is like, you know, a Porsche.\"\nThe original Slackbot, which has existed since Slack's early days, performed basic algorithmic tasks — reminding users to add colleagues to documents, suggesting channel archives, and delivering simple notifications. The new version runs on an entirely different architecture built around a large language model and sophisticated search capabilities that can access Salesforce records, Google Drive files, calendar data, and years of Slack conversations.\n\"It's two different things,\" Harris explained. \"The old Slackbot was algorithmic and fairly simple. The new Slackbot is brand new — it's based around an LLM and a very robust search engine, and connections to third-party search engines, third-party enterprise data.\"\nSalesforce chose to retain the Slackbot brand despite the fundamental technical overhaul. \"People know what Slackbot is, and so we wanted to carry that forward,\" Harris said.\nWhy Anthropic's Claude powers the new Slackbot — and which AI models could come next\nThe new Slackbot runs on Claude, Anthropic's large language model, a choice driven partly by compliance requirements. Slack's commercial service operates under FedRAMP Moderate certification to serve U.S. federal government customers, and Harris said Anthropic was \"the only provider that could give us a compliant LLM\" when Slack began building the new system.\nBut that exclusivity won't last. \"We are, this year, going to support additional providers,\" Harris said. \"We have a great relationship with Google. Gemini is incredible — performance is great, cost is great. So we're going to use Gemini for some things.\" He added that OpenAI remains a possibility as well.\nHarris echoed Salesforce CEO Marc Benioff's view that large language models are becoming commoditized: \"You've heard Marc talk about LLMs are commodities, that they're democratized. I call them CPUs.\"\nOn the sensitive question of training data, Harris was unequivocal: Salesforce does not train any models on customer data. \"Models don't have any sort of security,\" he explained. \"If we trained it on some confidential conversation that you and I have, I don't want Carolyn to know — if I train it into the LLM, there is no way for me to say you get to see the answer, but Carolyn doesn't.\"\nInside Salesforce's internal experiment: 80,000 employees tested Slackbot with striking results\nSalesforce has been testing the new Slackbot internally for months, rolling it out to all 80,000 employees. According to Ryan Gavin, Slack's chief marketing officer, the results have been striking: \"It's the fastest adopted product in Salesforce history.\"\nInternal data shows that two-thirds of Salesforce employees have tried the new Slackbot, with 80% of those users continuing to use it regularly. Internal satisfaction rates reached 96% — the highest for any AI feature Slack has shipped. Employees report saving between two and 20 hours per week.\nThe adoption happened largely organically. \"I think it was about five days, and a Canvas was developed by our employees called 'The Most Stealable Slackbot Prompts,'\" Gavin said. \"People just started adding to it organically. I think it's up to 250-plus prompts that are in this Canvas right now.\"\nKate Crotty, a principal UX researcher at Salesforce, found that 73% of internal adoption was driven by social sharing rather than top-down mandates. \"Everybody is there to help each other learn and communicate hacks,\" she said.\nHow Slackbot transforms scattered enterprise data into executive-ready insights\nDuring a product demonstration, Amy Bauer, Slack's product experience designer, showed how Slackbot can synthesize information across multiple sources. In one example, she asked Slackbot to analyze customer feedback from a pilot program, upload an image of a usage dashboard, and have Slackbot correlate the qualitative and quantitative data.\n\"This is where Slackbot really earns its keep for me,\" Bauer explained. \"What it's doing is not just simply reading the image — it's actually looking at the image and comparing it to the insight it just generated for me.\"\nSlackbot can then query Salesforce to find enterprise accounts with open deals that might be good candidates for early access, creating what Bauer called \"a really great justification and plan to move forward.\" Finally, it can synthesize all that information into a Canvas — Slack's collaborative document format — and find calendar availability among stakeholders to schedule a review meeting.\n\"Up until this point, we have been working in a one-to-one capacity with Slackbot,\" Bauer said. \"But one of the benefits that I can do now is take this insight and have it generate this into a Canvas, a shared workspace where I can iterate on it, refine it with Slackbot, or share it out with my team.\"\nRob Seaman, Slack's chief product officer, said the Canvas creation demonstrates where the product is heading: \"This is making a tool call internally to Slack Canvas to actually write, effectively, a shared document. But it signals where we're going with Slackbot — we're eventually going to be adding in additional third-party tool calls.\"\nMrBeast's company became a Slackbot guinea pig—and employees say they're saving 90 minutes a day\nAmong Salesforce's pilot customers is Beast Industries, the parent company of YouTube star MrBeast. Luis Madrigal, the company's chief information officer, joined the launch announcement to describe his experience.\n\"As somebody who has rolled out enterprise technologies for over two decades now, this was practically one of the easiest,\" Madrigal said. \"The plumbing is there. Slack as an implementation, Enterprise Tools — being able to turn on the Slackbot and the Slack AI functionality was as simple as having my team go in, review, do a quick security review.\"\nMadrigal said his security team signed off \"rather quickly\" — unusual for enterprise AI deployments — because Slackbot accesses only the information each individual user already has permission to view. \"Given all the guardrails you guys have put into place for Slackbot to be unique and customized to only the information that each individual user has, only the conversations and the Slack rooms and Slack channels that they're part of—that made my security team sign off rather quickly.\"\nOne Beast Industries employee, Sinan, the head of Beast Games marketing, reported saving \"at bare minimum, 90 minutes a day.\" Another employee, Spencer, a creative supervisor, described it as \"an assistant who's paying attention when I'm not.\"\nOther pilot customers include Slalom, reMarkable, Xero, Mercari, and Engine. Mollie Bodensteiner, SVP of Operations at Engine, called Slackbot \"an absolute 'chaos tamer' for our team,\" estimating it saves her about 30 minutes daily \"just by eliminating context switching.\"\nSlackbot vs. Microsoft Copilot vs. Google Gemini: The fight for enterprise AI dominance\nThe launch puts Salesforce in direct competition with Microsoft's Copilot, which is integrated into Teams and the broader Microsoft 365 suite, as well as Google's Gemini integrations across Workspace. When asked what distinguishes Slackbot from these alternatives, Seaman pointed to context and convenience.\n\"The thing that makes it most powerful for our customers and users is the proximity — it's just right there in your Slack,\" Seaman said. \"There's a tremendous convenience affordance that's naturally built into it.\"\nThe deeper advantage, executives argue, is that Slackbot already understands users' work without requiring setup or training. \"Most AI tools sound the same no matter who is using them,\" the company's announcement stated. \"They lack context, miss nuance, and force you to jump between tools to get anything done.\"\nHarris put it more directly: \"If you've ever had that magic experience with AI — I think ChatGPT is a great example, it's a great experience from a consumer perspective — Slackbot is really what we're doing in the enterprise, to be this employee super agent that is loved, just like people love using Slack.\"\nAmy Bauer emphasized the frictionless nature of the experience. \"Slackbot is inherently grounded in the context, in the data that you have in Slack,\" she said. \"So as you continue working in Slack, Slackbot gets better because it's grounded in the work that you're doing there. There is no setup. There is no configuration for those end users.\"\nSalesforce's ambitious plan to make Slackbot the one 'super agent' that controls all the others\nSalesforce positions Slackbot as what Harris calls a \"super agent\" — a central hub that can eventually coordinate with other AI agents across an organization.\n\"Every corporation is going to have an employee super agent,\" Harris said. \"Slackbot is essentially taking the magic of what Slack does. We think that Slackbot, and we're really excited about it, is going to be that.\"\nThe vision extends to third-party agents already launching in Slack. Last month, Anthropic released a preview of Claude Code for Slack, allowing developers to interact with Claude's coding capabilities directly in chat threads. OpenAI, Google, Vercel, and others have also built agents for the platform.\n\"Most of the net-new apps that are being deployed to Slack are agents,\" Seaman noted during the press conference. \"This is proof of the promise of humans and agents coexisting and working together in Slack to solve problems.\"\nHarris described a future where Slackbot becomes an MCP (Model Context Protocol) client, able to leverage tools from across the software ecosystem — similar to how the developer tool Cursor works. \"Slack can be an MCP client, and Slackbot will be the hub of that, leveraging all these tools out in the world, some of which will be these amazing agents,\" he said.\nBut Harris also cautioned against over-promising on multi-agent coordination. \"I still think we're in the single agent world,\" he said. \"FY26 is going to be the year where we started to see more coordination. But we're going to do it with customer success in mind, and not demonstrate and talk about, like, 'I've got 1,000 agents working together,' because I think that's unrealistic.\"\nSlackbot costs nothing extra, but Salesforce's data access fees could squeeze some customers\nSlackbot is included at no additional cost for customers on Business+ and Enterprise+ plans. \"There's no additional fees customers have to do,\" Gavin confirmed. \"If they're on one of those plans, they're going to get Slackbot.\"\nHowever, some enterprise customers may face other cost pressures related to Salesforce's broader data strategy. CIOs may see price increases for third-party applications that work with Salesforce data, as effects of higher charges for API access ripple through the software supply chain.\nFivetran CEO George Fraser has warned that Salesforce's shift in pricing policy for API access could have tangible consequences for enterprises relying on Salesforce as a system of record. \"They might not be able to use Fivetran to replicate their data to Snowflake and instead have to use Salesforce Data Cloud. Or they might find that they are not able to interact with their data via ChatGPT, and instead have to use Agentforce,\" Fraser said in a recent CIO report.\nSalesforce has framed the pricing change as standard industry practice.\nWhat Slackbot can do today, what's coming in weeks, and what's still on the roadmap\nThe new Slackbot begins rolling out today and will reach all eligible customers by the end of February. Mobile availability will complete by March 3, Bauer confirmed during her interview with VentureBeat.\nSome capabilities remain works in progress. Calendar reading and availability checking are available at launch, but the ability to actually book meetings is \"coming a few weeks after,\" according to Seaman. Image generation is not currently supported, though Bauer said it's \"something that we are looking at in the future.\"\nWhen asked about integration with competing CRM systems like HubSpot and Microsoft Dynamics, Salesforce representatives declined to provide specifics during the interview, though they acknowledged the question touched on key competitive differentiators.\nSalesforce is betting the future of work looks like a chat window—and it's not alone\nThe Slackbot launch is Salesforce's bet that the future of enterprise work is conversational — that employees will increasingly prefer to interact with AI through natural language rather than navigating traditional software interfaces.\nHarris described Slack's product philosophy using principles like \"don't make me think\" and \"be a great host.\" The goal, he said, is for Slackbot to surface information proactively rather than requiring users to hunt for it.\n\"One of the revelations for me is LLMs applied to unstructured information are incredible,\" Harris said. \"And the amount of value you have if you're a Slack user, if your corporation uses Slack — the amount of value in Slack is unbelievable. Because you're talking about work, you're sharing documents, you're making decisions, but you can't as a human go through that and really get the same value that an LLM can do.\"\nLooking ahead, Harris expects the interfaces themselves to evolve beyond pure conversation. \"We're kind of saturating what we can do with purely conversational UIs,\" he said. \"I think we'll start to see agents building an interface that best suits your intent, as opposed to trying to surface something within a conversational interface that matches your intent.\"\nMicrosoft, Google, and a growing roster of AI startups are placing similar bets — that the winning enterprise AI will be the one embedded in the tools workers already use, not another application to learn. The race to become that invisible layer of workplace intelligence is now fully underway.\nFor Salesforce, the stakes extend beyond a single product launch. After a bruising year on Wall Street and persistent questions about whether AI threatens its core business, the company is wagering that Slackbot can prove the opposite — that the tens of millions of people already chatting in Slack every day is not a vulnerability, but an unassailable advantage.\nHaley Gault, the Salesforce account executive in Pittsburgh who stumbled upon the new Slackbot on a snowy morning, captured the shift in a single sentence: \"I honestly can't imagine working for another company not having access to these types of tools. This is just how I work now.\"\nThat's precisely what Salesforce is counting on.", "pubDate": "Tue, 13 Jan 2026 13:00:00 GMT" }, { "link": "https://venturebeat.com/technology/anthropic-launches-cowork-a-claude-desktop-agent-that-works-in-your-files-no", "title": "Anthropic launches Cowork, a Claude Desktop agent that works in your files — no coding required", "content": "Anthropic released Cowork on Monday, a new AI agent capability that extends the power of its wildly successful Claude Code tool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, largely using Claude Code itself.\nThe launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, positioning Anthropic to compete not just with OpenAI and Google in conversational AI, but with Microsoft's Copilot in the burgeoning market for AI-powered productivity tools.\n\"Cowork lets you complete non-technical tasks much like how developers use Claude Code,\" the company announced via its official Claude account on X. The feature arrives as a research preview available exclusively to Claude Max subscribers — Anthropic's power-user tier priced between $100 and $200 per month — through the macOS desktop application.\nFor the past year, the industry narrative has focused on large language models that can write poetry or debug code. With Cowork, Anthropic is betting that the real enterprise value lies in an AI that can open a folder, read a messy pile of receipts, and generate a structured expense report without human hand-holding.\n\nHow developers using a coding tool for vacation research inspired Anthropic's latest product\nThe genesis of Cowork lies in Anthropic's recent success with the developer community. In late 2024, the company released Claude Code, a terminal-based tool that allowed software engineers to automate rote programming tasks. The tool was a hit, but Anthropic noticed a peculiar trend: users were forcing the coding tool to perform non-coding labor.\nAccording to Boris Cherny, an engineer at Anthropic, the company observed users deploying the developer tool for an unexpectedly diverse array of tasks.\n\n\"Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven,\" Cherny wrote on X. \"These use cases are diverse and surprising — the reason is that the underlying Claude Agent is the best agent, and Opus 4.5 is the best model.\"\nRecognizing this shadow usage, Anthropic effectively stripped the command-line complexity from their developer tool to create a consumer-friendly interface. In its blog post announcing the feature, Anthropic explained that developers \"quickly began using it for almost everything else,\" which \"prompted us to build Cowork: a simpler way for anyone — not just developers — to work with Claude in the very same way.\"\nInside the folder-based architecture that lets Claude read, edit, and create files on your computer\nUnlike a standard chat interface where a user pastes text for analysis, Cowork requires a different level of trust and access. Users designate a specific folder on their local machine that Claude can access. Within that sandbox, the AI agent can read existing files, modify them, or create entirely new ones.\nAnthropic offers several illustrative examples: reorganizing a cluttered downloads folder by sorting and intelligently renaming each file, generating a spreadsheet of expenses from a collection of receipt screenshots, or drafting a report from scattered notes across multiple documents.\n\"In Cowork, you give Claude access to a folder on your computer. Claude can then read, edit, or create files in that folder,\" the company explained on X. \"Try it to create a spreadsheet from a pile of screenshots, or produce a first draft from scattered notes.\"\n\nThe architecture relies on what is known as an \"agentic loop.\" When a user assigns a task, the AI does not merely generate a text response. Instead, it formulates a plan, executes steps in parallel, checks its own work, and asks for clarification if it hits a roadblock. Users can queue multiple tasks and let Claude process them simultaneously — a workflow Anthropic describes as feeling \"much less like a back-and-forth and much more like leaving messages for a coworker.\"\nThe system is built on Anthropic's Claude Agent SDK, meaning it shares the same underlying architecture as Claude Code. Anthropic notes that Cowork \"can take on many of the same tasks that Claude Code can handle, but in a more approachable form for non-coding tasks.\"\nThe recursive loop where AI builds AI: Claude Code reportedly wrote much of Claude Cowork\nPerhaps the most remarkable detail surrounding Cowork's launch is the speed at which the tool was reportedly built — highlighting a recursive feedback loop where AI tools are being used to build better AI tools.\nDuring a livestream hosted by Dan Shipper, Felix Rieseberg, an Anthropic employee, confirmed that the team built Cowork in approximately a week and a half.\nAlex Volkov, who covers AI developments, expressed surprise at the timeline: \"Holy shit Anthropic built 'Cowork' in the last... week and a half?!\"\n\nThis prompted immediate speculation about how much of Cowork was itself built by Claude Code. Simon Smith, EVP of Generative AI at Klick Health, put it bluntly on X: \"Claude Code wrote all of Claude Cowork. Can we all agree that we're in at least somewhat of a recursive improvement loop here?\"\nThe implication is profound: Anthropic's AI coding agent may have substantially contributed to building its own non-technical sibling product. If true, this is one of the most visible examples yet of AI systems being used to accelerate their own development and expansion — a strategy that could widen the gap between AI labs that successfully deploy their own agents internally and those that do not.\nConnectors, browser automation, and skills extend Cowork's reach beyond the local file system\nCowork doesn't operate in isolation. The feature integrates with Anthropic's existing ecosystem of connectors — tools that link Claude to external information sources and services such as Asana, Notion, PayPal, and other supported partners. Users who have configured these connections in the standard Claude interface can leverage them within Cowork sessions.\nAdditionally, Cowork can pair with Claude in Chrome, Anthropic's browser extension, to execute tasks requiring web access. This combination allows the agent to navigate websites, click buttons, fill forms, and extract information from the internet — all while operating from the desktop application.\n\"Cowork includes a number of novel UX and safety features that we think make the product really special,\" Cherny explained, highlighting \"a built-in VM [virtual machine] for isolation, out of the box support for browser automation, support for all your claude.ai data connectors, asking you for clarification when it's unsure.\"\nAnthropic has also introduced an initial set of \"skills\" specifically designed for Cowork that enhance Claude's ability to create documents, presentations, and other files. These build on the Skills for Claude framework the company announced in October, which provides specialized instruction sets Claude can load for particular types of tasks.\nWhy Anthropic is warning users that its own AI agent could delete their files\nThe transition from a chatbot that suggests edits to an agent that makes edits introduces significant risk. An AI that can organize files can, theoretically, delete them.\nIn a notable display of transparency, Anthropic devoted considerable space in its announcement to warning users about Cowork's potential dangers — an unusual approach for a product launch.\nThe company explicitly acknowledges that Claude \"can take potentially destructive actions (such as deleting local files) if it's instructed to.\" Because Claude might occasionally misinterpret instructions, Anthropic urges users to provide \"very clear guidance\" about sensitive operations.\nMore concerning is the risk of prompt injection attacks — a technique where malicious actors embed hidden instructions in content Claude might encounter online, potentially causing the agent to bypass safeguards or take harmful actions.\n\"We've built sophisticated defenses against prompt injections,\" Anthropic wrote, \"but agent safety — that is, the task of securing Claude's real-world actions — is still an active area of development in the industry.\"\nThe company characterized these risks as inherent to the current state of AI agent technology rather than unique to Cowork. \"These risks aren't new with Cowork, but it might be the first time you're using a more advanced tool that moves beyond a simple conversation,\" the announcement notes.\nAnthropic's desktop agent strategy sets up a direct challenge to Microsoft Copilot\nThe launch of Cowork places Anthropic in direct competition with Microsoft, which has spent years attempting to integrate its Copilot AI into the fabric of the Windows operating system with mixed adoption results.\nHowever, Anthropic's approach differs in its isolation. By confining the agent to specific folders and requiring explicit connectors, they are attempting to strike a balance between the utility of an OS-level agent and the security of a sandboxed application.\nWhat distinguishes Anthropic's approach is its bottom-up evolution. Rather than designing an AI assistant and retrofitting agent capabilities, Anthropic built a powerful coding agent first — Claude Code — and is now abstracting its capabilities for broader audiences. This technical lineage may give Cowork more robust agentic behavior from the start.\nClaude Code has generated significant enthusiasm among developers since its initial launch as a command-line tool in late 2024. The company expanded access with a web interface in October 2025, followed by a Slack integration in December. Cowork is the next logical step: bringing the same agentic architecture to users who may never touch a terminal.\nWho can access Cowork now, and what's coming next for Windows and other platforms\nFor now, Cowork remains exclusive to Claude Max subscribers using the macOS desktop application. Users on other subscription tiers — Free, Pro, Team, or Enterprise — can join a waitlist for future access.\nAnthropic has signaled clear intentions to expand the feature's reach. The blog post explicitly mentions plans to add cross-device sync and bring Cowork to Windows as the company learns from the research preview.\nCherny set expectations appropriately, describing the product as \"early and raw, similar to what Claude Code felt like when it first launched.\"\nTo access Cowork, Max subscribers can download or update the Claude macOS app and click on \"Cowork\" in the sidebar.\nThe real question facing enterprise AI adoption\nFor technical decision-makers, the implications of Cowork extend beyond any single product launch. The bottleneck for AI adoption is shifting — no longer is model intelligence the limiting factor, but rather workflow integration and user trust.\nAnthropic's goal, as the company puts it, is to make working with Claude feel less like operating a tool and more like delegating to a colleague. Whether mainstream users are ready to hand over folder access to an AI that might misinterpret their instructions remains an open question.\nBut the speed of Cowork's development — a major feature built in ten days, possibly by the company's own AI — previews a future where the capabilities of these systems compound faster than organizations can evaluate them. \nThe chatbot has learned to use a file manager. What it learns to use next is anyone's guess.", "pubDate": "Mon, 12 Jan 2026 11:30:00 GMT" }, { "link": "https://www.zdnet.com/article/how-to-turn-airtag-into-wifi-password-key-nfc-tags/", "title": "I created the ultimate Wi-Fi password key with the most unexpected gadget - how it works", "content": "NFC tags are so useful and customizable, and it's quite simple to make your own. Here's what you can do with it and how.", "pubDate": "Tue, 10 Feb 2026 03:01:15 GMT" }, { "link": "https://www.zdnet.com/article/i-turned-doodles-into-video-clips-in-minutes-with-runways-new-ai-video-tool-heres-how/", "title": "I turned doodles into video clips in minutes with Runway's new AI video tool - here's how", "content": "The company's Motion Sketch feature lowers the barrier between abstract imagination and concrete video outputs. But like all generative AI tools, it has its shortcomings.", "pubDate": "Tue, 10 Feb 2026 03:01:13 GMT" }, { "link": "https://www.zdnet.com/article/android-extend-unlock-faster-phone-access/", "title": "This Android trick lets me access my phone faster - here's how it works", "content": "Extend Unlock is one of those features every Android user will love once they give it a try.", "pubDate": "Tue, 10 Feb 2026 03:00:54 GMT" }, { "link": "https://www.zdnet.com/article/onyx-boox-page-android-e-ink-tablet-review/", "title": "I didn't expect this Android E Ink tablet to beat my Kindle, but one feature sets it apart", "content": "The Onyx Boox Page is packed with features for an E Ink tablet, with a smart, versatile design.", "pubDate": "Tue, 10 Feb 2026 02:30:37 GMT" }, { "link": "https://www.zdnet.com/article/how-to-restart-android-phone-without-using-the-power-button-2-ways/", "title": "How to restart your Android phone without using the power button: 2 easy ways", "content": "You don't need to press the power button to restart your Android device. If yours is broken or you simply want options, here are two.", "pubDate": "Tue, 10 Feb 2026 02:12:53 GMT" }, { "link": "https://towardsdatascience.com/the-machine-learning-lessons-ive-learned-last-month/", "title": "The Machine Learning Lessons I’ve Learned Last Month", "content": "Delayed January: deadlines, downtimes, and flow times\nThe post The Machine Learning Lessons I’ve Learned Last Month appeared first on Towards Data Science.", "pubDate": "Mon, 09 Feb 2026 20:38:42 +0000" }, { "link": "https://towardsdatascience.com/the-death-of-the-everything-prompt-googles-move-toward-structured-ai/", "title": "The Death of the “Everything Prompt”: Google’s Move Toward Structured AI", "content": "How the new Interactions API enables deep-reasoning, stateful, agentic workflows.\nThe post The Death of the “Everything Prompt”: Google’s Move Toward Structured AI appeared first on Towards Data Science.", "pubDate": "Mon, 09 Feb 2026 13:00:00 +0000" }, { "link": "https://towardsdatascience.com/what-i-am-doing-to-stay-relevant-as-a-senior-analytics-consultant-in-2026/", "title": "What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026", "content": "Learn how to work with AI, while strengthening your unique human skills that technology cannot replace\nThe post What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026 appeared first on Towards Data Science.", "pubDate": "Sat, 07 Feb 2026 15:00:00 +0000" }, { "link": "https://towardsdatascience.com/pydantic-performance-4-tips-on-how-to-validate-large-amounts-of-data-efficiently/", "title": "Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently", "content": "The real value lies in writing clearer code and using your tools right\nThe post Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently appeared first on Towards Data Science.", "pubDate": "Fri, 06 Feb 2026 15:00:00 +0000" }, { "link": "https://towardsdatascience.com/prompt-fidelity-measuring-how-much-of-your-intent-an-ai-agent-actually-executes/", "title": "Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes", "content": "How much of your AI agent's output is real data versus confident guesswork?\nThe post Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes appeared first on Towards Data Science.", "pubDate": "Fri, 06 Feb 2026 12:00:00 +0000" }, { "link": "https://www.nytimes.com/2026/02/09/technology/social-media-addiction-trial.html", "title": "Meta and YouTube Created ‘Digital Casinos,’ Lawyers Argue in Landmark Trial", "content": "Opening statements began in a trial claiming social media companies design addictive products that cause personal injury.", "pubDate": "Tue, 10 Feb 2026 00:53:17 +0000" }, { "link": "https://www.nytimes.com/2026/02/06/business/google-employees-protest.html", "title": "Google Workers Demand End to Cloud Services for Immigration Agencies", "content": "More than 800 employees delivered a petition to management, condemning the Trump administration’s use of Google technology in immigration enforcement.", "pubDate": "Fri, 06 Feb 2026 18:52:26 +0000" }, { "link": "https://www.nytimes.com/2026/02/06/business/tiktok-addictive-design-europe.html", "title": "Europe Accuses TikTok of ‘Addictive Design’ and Pushes for Change", "content": "European Union regulators said the app’s infinite scroll and personalized algorithm led to “compulsive” behavior, especially among children.", "pubDate": "Fri, 06 Feb 2026 19:08:37 +0000" }, { "link": "https://www.nytimes.com/2026/02/06/business/ai-stock-market-anthropic-openai.html", "title": "Fear of AI Replacing Software Makers Hits Stocks. Here’s What to Know.", "content": "The prospect of disruptions from artificial intelligence has hung over the economy for years. But this week advances in software tools precipitated a sell-off on Wall Street.", "pubDate": "Fri, 06 Feb 2026 20:11:04 +0000" }, { "link": "https://www.nytimes.com/2026/02/09/health/ai-chatbots-doctors-medicine.html", "title": "A.I. Is Making Doctors Answer a Question: What Are They Really Good For?", "content": "Many physicians find chatbots threatening, but that doesn’t mean they’re giving up on medicine.", "pubDate": "Mon, 09 Feb 2026 17:17:56 +0000" }, { "link": "https://www.theguardian.com/technology/2026/feb/09/jeffrey-epstein-crypto", "title": "Files cast light on Jeffrey Epstein’s ties to cryptocurrency", "content": "Newly released documents detail convicted sex offender’s early backing of bitcoin and Coinbase \nMillions of files related to Jeffrey Epstein have brought to light his ties to the highest echelons of the cryptocurrency industry.\nDocuments published last week by the US Department of Justice reveal Epstein bankrolled the “principal home and funding source” for bitcoin, the world’s largest cryptocurrency, during its nascent stages; he also invested $3m in Coinbase in 2014, the largest cryptocurrency exchange in the US, and cut a check that same year to Blockstream, a prominent bitcoin-focused technology firm. Both crypto startups accepted Epstein’s investments in 2014 – six years after his 2008 conviction in Florida for soliciting prostitution from a minor.\n Continue reading...", "pubDate": "Mon, 09 Feb 2026 14:00:22 GMT" }, { "link": "https://www.theguardian.com/technology/2026/feb/08/i-fell-into-it-ex-criminal-hackers-urge-manchester-pupils-to-use-web-skills-for-good", "title": "‘I fell into it’: ex-criminal hackers urge Manchester pupils to use web skills for good", "content": "Initiative aims to identify proficient gamers and coders who can help companies identify flaws in their cybersecurity \nCybercriminals, the shadowy online figures often depicted in Hollywood movies as hooded villains capable of wiping millions of pounds off the value of businesses at a keystroke, are not usually known for their candour.\nBut in a sixth-form college in Manchester this week, two former hackers gave the young people gathered an honest appraisal of what living a life of internet crime really looks like.\n Continue reading...", "pubDate": "Sun, 08 Feb 2026 08:00:25 GMT" }, { "link": "https://www.theguardian.com/technology/2026/feb/07/ai-chatbots-anthropic-openai-claude-chatgpt", "title": "Battle of the chatbots: Anthropic and OpenAI go head-to-head over ads in their AI products", "content": "New Anthropic campaign suggests other AI platforms will incorporate targeted ads in their chatbot conversations\nThe Seahawks and the Patriots aren’t the only ones gearing up for a fight.\nAI rivals Anthropic and OpenAI have launched a war of ads trying to court corporate America during one of the biggest entertainment nights of the year.\n Continue reading...", "pubDate": "Sat, 07 Feb 2026 16:00:06 GMT" }, { "link": "https://www.theguardian.com/commentisfree/2026/feb/09/who-is-my-wife-martin-rowson-ai-technology", "title": "I asked AI to name my wife. To the hopelessly incorrect people it cited, my deepest apologies | Martin Rowson", "content": "Authors, a newsreader, a lawyer and an esteemed colleague: they’re all great – but I’m not married to any of them. Can we really depend on this technology?\nRecently, the Rowsons accidentally invented a new game that anyone can play at home. I have yet to come up with a world-beating name for it, so for now let’s just call it “How bloody stupid is AI?” The playing of the game will change from player to player, depending on their circumstances – but essentially the rules remain the same. Ask AI a simple question about yourself, and see just how wrong it gets it.\nIn my case, all you need know is that while I, through the nature of my job, have a fairly large online presence, my partner (we married in 1987) has assiduously avoided having one at all. Which means that if you Google “Martin Rowson wife” in images, you may get a picture of me next to our then 14-year-old daughter or me with my friend and fellow cartoonist Steven Appleby, who happens to be trans but has kept her given first name.\n Continue reading...", "pubDate": "Mon, 09 Feb 2026 10:00:31 GMT" }, { "link": "https://www.theguardian.com/technology/2026/feb/07/campaigners-call-stronger-protection-against-ai-generated-explicit-imagery", "title": "Victims urge tougher action on deepfake abuse as new law comes into force", "content": "Campaigners welcome criminalisation of non-consensual AI-generated explicit images but say law does not go far enough\nVictims of deepfake image abuse have called for stronger protection against AI-generated explicit images, as the law criminalising the creation of non-consensual intimate images comes into effect.\nCampaigners from Stop Image-Based Abuse delivered a petition to Downing Street with more than 73,000 signatures, urging the government to introduce civil routes to justice such as takedown orders for abusive imagery on platforms and devices.\n Continue reading...", "pubDate": "Sat, 07 Feb 2026 10:00:02 GMT" }, { "link": "https://www.bbc.com/news/articles/c3wlpqpe2z4o?at_medium=RSS&at_campaign=rss", "title": "Instagram and YouTube owners built 'addiction machines', trial hears", "content": "The tech giants are under scrutiny over social media addiction in a landmark jury trial in Los Angeles", "pubDate": "Tue, 10 Feb 2026 07:25:42 GMT" }, { "link": "https://www.bbc.com/news/articles/c1d67vdlk1ko?at_medium=RSS&at_campaign=rss", "title": "Discord to start requiring face scan or ID to access adult content", "content": "The online chat service, which has 200 million monthly users, will blur adult content by default.", "pubDate": "Mon, 09 Feb 2026 17:42:21 GMT" }, { "link": "https://www.bbc.com/news/articles/c3093gjy2ero?at_medium=RSS&at_campaign=rss", "title": "AI chatbots pose 'dangerous' risk when giving medical advice, study suggests", "content": "It found people using AI for health reasons found it hard to identify what advice they should trust.", "pubDate": "Mon, 09 Feb 2026 16:33:29 GMT" }, { "link": "https://www.bbc.com/news/articles/cqxdj77welpo?at_medium=RSS&at_campaign=rss", "title": "EU tells Meta to let rivals run AI chatbots on WhatsApp", "content": "A Meta spokesperson said the EU had \"no reason\" to intervene over it changing the app in January.", "pubDate": "Mon, 09 Feb 2026 12:14:35 GMT" }, { "link": "https://www.bbc.com/news/articles/c24g457y534o?at_medium=RSS&at_campaign=rss", "title": "Baldur's Gate to be turned into TV series - without the game's developers", "content": "The show will be made by Craig Mazin, who co-created the acclaimed game adaptation The Last Of Us.", "pubDate": "Fri, 06 Feb 2026 16:26:08 GMT" }, { "link": "https://www.bbc.co.uk/iplayer/episode/m002qbkc?at_medium=RSS&at_campaign=rss", "title": "Tech Now", "content": "From Las Vegas, the latest trends and innovations at CES 2026.", "pubDate": "Sat, 17 Jan 2026 01:00:00 GMT" }, { "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpz?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "Meet the humanoid robots designed to help with household chores", "pubDate": "Tue, 13 Jan 2026 20:30:00 GMT" }, { "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpy?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "The latest gadgets, the future in assistive tech and upcoming gaming releases in 2026.", "pubDate": "Tue, 06 Jan 2026 20:32:00 GMT" }, { "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpx?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "We bring you Tech Life highlights from a fascinating year in global tech.", "pubDate": "Tue, 30 Dec 2025 20:30:00 GMT" }, { "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpv?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "A study found AI chatbots can persuade us with fake facts. How does this affect politics?", "pubDate": "Tue, 16 Dec 2025 20:30:00 GMT" } ] }, "pairedItem": [ { "item": 0 }, { "item": 1 }, { "item": 2 }, { "item": 3 }, { "item": 4 }, { "item": 5 }, { "item": 6 }, { "item": 7 }, { "item": 8 }, { "item": 9 }, { "item": 10 }, { "item": 11 }, { "item": 12 }, { "item": 13 }, { "item": 14 }, { "item": 15 }, { "item": 16 }, { "item": 17 }, { "item": 18 }, { "item": 19 }, { "item": 20 }, { "item": 21 }, { "item": 22 }, { "item": 23 }, { "item": 24 }, { "item": 25 }, { "item": 26 }, { "item": 27 }, { "item": 28 }, { "item": 29 }, { "item": 30 }, { "item": 31 }, { "item": 32 }, { "item": 33 }, { "item": 34 }, { "item": 35 }, { "item": 36 }, { "item": 37 }, { "item": 38 }, { "item": 39 }, { "item": 40 }, { "item": 41 }, { "item": 42 }, { "item": 43 }, { "item": 44 } ] } ], "Fetch RSS Articles": [ { "json": { "guid": "https://techcrunch.com/?p=3081727", "link": "https://techcrunch.com/2026/01/13/ai-drug-discovery-startup-converge-bio-pulls-in-25m-from-bessemer-and-execs-from-meta-openai-and-wiz/", "title": "Converge Bio raises $25M, backed by Bessemer and execs from Meta, OpenAI, Wiz", "content": "AI drug discovery startup Converge Bio raised $25 million in a Series A led by Bessemer Venture Partners, with additional backing from executives at Meta, OpenAI, and Wiz.", "creator": "Kate Park", "isoDate": "2026-01-13T11:30:00.000Z", "pubDate": "Tue, 13 Jan 2026 11:30:00 +0000", "categories": [ "AI", "Biotech & Health", "Startups", "Artificial Intelligence (AI)", "AI drug discovery", "converge bio" ], "dc:creator": "Kate Park", "contentSnippet": "AI drug discovery startup Converge Bio raised $25 million in a Series A led by Bessemer Venture Partners, with additional backing from executives at Meta, OpenAI, and Wiz." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3064325", "link": "https://techcrunch.com/2025/10/31/meta-bought-1-gw-of-solar-this-week/", "title": "Meta bought 1 GW of solar this week", "content": "The social media company inked three deals in the U.S. to power its data centers and offset its carbon footprint.", "creator": "Tim De Chant", "isoDate": "2025-10-31T19:26:30.000Z", "pubDate": "Fri, 31 Oct 2025 19:26:30 +0000", "categories": [ "Climate", "Social", "data centers", "renewable energy", "Solar Power", "Artificial Intelligence (AI)", "Meta" ], "dc:creator": "Tim De Chant", "contentSnippet": "The social media company inked three deals in the U.S. to power its data centers and offset its carbon footprint." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3039857", "link": "https://techcrunch.com/2025/08/26/how-one-ai-startup-is-helping-rice-farmers-battle-climate-change/", "title": "How one AI startup is helping rice farmers battle climate change", "content": "Mitti Labs is working with The Nature Conservancy to expand the use of climate-friendly rice farming practices in India. The startup uses its AI to verify reductions in methane emissions.", "creator": "Tim De Chant", "isoDate": "2025-08-26T15:21:19.000Z", "pubDate": "Tue, 26 Aug 2025 15:21:19 +0000", "categories": [ "Startups", "Climate", "AI", "agriculture", "methane", "Artificial Intelligence (AI)", "Exclusive", "satellite imagery", "rice", "Mitti Labs" ], "dc:creator": "Tim De Chant", "contentSnippet": "Mitti Labs is working with The Nature Conservancy to expand the use of climate-friendly rice farming practices in India. The startup uses its AI to verify reductions in methane emissions." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3037667", "link": "https://techcrunch.com/2025/08/20/harvard-dropouts-to-launch-always-on-ai-smart-glasses-that-listen-and-record-every-conversation/", "title": "Harvard dropouts to launch ‘always on’ AI smart glasses that listen and record every conversation", "content": "After developing a facial-recognition app for Meta’s Ray-Ban glasses and doxing random people, two former Harvard students are now launching a startup that makes smart glasses with an always-on microphone.", "creator": "Lorenzo Franceschi-Bicchierai, Rebecca Bellan", "isoDate": "2025-08-20T16:00:00.000Z", "pubDate": "Wed, 20 Aug 2025 16:00:00 +0000", "categories": [ "Security", "AI", "privacy", "Artificial Intelligence (AI)", "facial recognition", "SMART Glasses" ], "dc:creator": "Lorenzo Franceschi-Bicchierai, Rebecca Bellan", "contentSnippet": "After developing a facial-recognition app for Meta’s Ray-Ban glasses and doxing random people, two former Harvard students are now launching a startup that makes smart glasses with an always-on microphone." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3038328", "link": "https://techcrunch.com/2025/08/20/meta-to-add-100-mw-of-solar-power-from-u-s-gear/", "title": "Meta to add 100MW of solar power from US gear", "content": "The social media company is adding another tranche of solar to power a new AI data center in South Carolina. ", "creator": "Tim De Chant", "isoDate": "2025-08-20T15:56:53.000Z", "pubDate": "Wed, 20 Aug 2025 15:56:53 +0000", "categories": [ "Enterprise", "Climate", "Social", "data centers", "Solar Power", "Artificial Intelligence (AI)", "Meta", "renewable power" ], "dc:creator": "Tim De Chant", "contentSnippet": "The social media company is adding another tranche of solar to power a new AI data center in South Carolina." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3033735", "link": "https://techcrunch.com/2025/08/04/perplexity-accused-of-scraping-websites-that-explicitly-blocked-ai-scraping/", "title": "Perplexity accused of scraping websites that explicitly blocked AI scraping", "content": "Internet giant Cloudflare says it detected Perplexity crawling and scraping websites, even after customers had added technical blocks telling Perplexity not to scrape their pages.", "creator": "Lorenzo Franceschi-Bicchierai", "isoDate": "2025-08-04T15:41:39.000Z", "pubDate": "Mon, 04 Aug 2025 15:41:39 +0000", "categories": [ "Security", "AI", "cloudflare", "Artificial Intelligence (AI)", "bots", "scraping", "LLMs", "Perplexity" ], "dc:creator": "Lorenzo Franceschi-Bicchierai", "contentSnippet": "Internet giant Cloudflare says it detected Perplexity crawling and scraping websites, even after customers had added technical blocks telling Perplexity not to scrape their pages." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3015072", "link": "https://techcrunch.com/2025/06/04/obvios-stop-sign-cameras-use-ai-to-root-out-unsafe-drivers/", "title": "Obvio’s stop sign cameras use AI to root out unsafe drivers", "content": "American streets are incredibly dangerous for pedestrians. A San Carlos, California-based startup called Obvio thinks it can change that by installing cameras at stop signs -- a solution the founders also say won’t create a panopticon. ", "creator": "Sean O'Kane", "isoDate": "2025-06-04T14:00:00.000Z", "pubDate": "Wed, 04 Jun 2025 14:00:00 +0000", "categories": [ "Startups", "Transportation", "AI", "fundraises", "Exclusive", "Artificial Intelligence (AI)", "bain capital ventures" ], "dc:creator": "Sean O'Kane", "contentSnippet": "American streets are incredibly dangerous for pedestrians. A San Carlos, California-based startup called Obvio thinks it can change that by installing cameras at stop signs -- a solution the founders also say won’t create a panopticon." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3014093", "link": "https://techcrunch.com/2025/06/02/breakneck-data-center-growth-challenges-microsofts-sustainability-goals/", "title": "Breakneck data center growth challenges Microsoft’s sustainability goals", "content": "Microsoft's sustainability goals are imperiled by its push into AI and cloud services.", "creator": "Tim De Chant", "isoDate": "2025-06-02T18:07:05.000Z", "pubDate": "Mon, 02 Jun 2025 18:07:05 +0000", "categories": [ "Enterprise", "Climate", "Artificial Intelligence (AI)", "sustainability", "data centers", "carbon emissions", "Microsoft", "carbon footprint", "renewable power" ], "dc:creator": "Tim De Chant", "contentSnippet": "Microsoft's sustainability goals are imperiled by its push into AI and cloud services." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3011622", "link": "https://techcrunch.com/2025/05/27/gridcare-thinks-more-than-100-gw-of-data-center-capacity-is-hiding-in-the-grid/", "title": "Gridcare thinks more than 100 GW of data center capacity is hiding in the grid", "content": "Gridcare raised $13.3 million for its data platform that finds underutilized capacity on the electrical grid.", "creator": "Tim De Chant", "isoDate": "2025-05-27T12:00:00.000Z", "pubDate": "Tue, 27 May 2025 12:00:00 +0000", "categories": [ "Fundraising", "Climate", "Enterprise", "electrical grid", "temasek", "Artificial Intelligence (AI)", "electricity", "data centers" ], "dc:creator": "Tim De Chant", "contentSnippet": "Gridcare raised $13.3 million for its data platform that finds underutilized capacity on the electrical grid." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=3010971", "link": "https://techcrunch.com/2025/05/22/meta-adds-another-650-mw-of-solar-power-to-its-ai-push/", "title": "Meta adds another 650 MW of solar power to its AI push", "content": "The company already has more than 12 gigawatts of capacity in its renewable power portfolio.", "creator": "Tim De Chant", "isoDate": "2025-05-22T16:49:53.000Z", "pubDate": "Thu, 22 May 2025 16:49:53 +0000", "categories": [ "Enterprise", "Climate", "AI", "renewable power", "Meta", "Artificial Intelligence (AI)", "aes", "Solar Power", "data centers" ], "dc:creator": "Tim De Chant", "contentSnippet": "The company already has more than 12 gigawatts of capacity in its renewable power portfolio." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2988020", "link": "https://techcrunch.com/2025/04/01/who-are-climate-conscious-consumers-not-who-youd-expect-says-northwind-climate/", "title": "Who are climate-conscious consumers? Not who you’d expect, says Northwind Climate", "content": "Rather than divide people into demographic buckets, Northwind Climate analyzes survey responses for behavioral clues.", "creator": "Tim De Chant", "isoDate": "2025-04-01T19:52:59.000Z", "pubDate": "Tue, 01 Apr 2025 19:52:59 +0000", "categories": [ "Climate", "Fundraising", "Marketing", "Artificial Intelligence (AI)", "Surveys", "Exclusive", "consumers", "Northwind Climate" ], "dc:creator": "Tim De Chant", "contentSnippet": "Rather than divide people into demographic buckets, Northwind Climate analyzes survey responses for behavioral clues." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2986758", "link": "https://techcrunch.com/2025/03/30/data-centers-love-solar-heres-a-comprehensive-guide-to-deals-over-100-megawatts/", "title": "Data centers love solar: Here’s a comprehensive guide to deals over 100 megawatts", "content": "New and expanded data centers are expected to double the sector’s power demand by 2029 as tech companies rush to capitalize on AI.", "creator": "Tim De Chant", "isoDate": "2025-03-30T14:00:00.000Z", "pubDate": "Sun, 30 Mar 2025 14:00:00 +0000", "categories": [ "Enterprise", "Climate", "evergreens", "Meta", "Artificial Intelligence (AI)", "Solar Power", "data centers", "Microsoft", "Google", "Cisco", "Amazon" ], "dc:creator": "Tim De Chant", "contentSnippet": "New and expanded data centers are expected to double the sector’s power demand by 2029 as tech companies rush to capitalize on AI." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2984175", "link": "https://techcrunch.com/2025/03/20/nvidia-thinks-ai-can-solve-electrical-grid-problems-caused-by-ai/", "title": "Nvidia thinks AI can solve electrical grid problems caused by AI", "content": "The Open Power AI Consortium says it will use domain-specific AI models to tackle problems in the power industry.", "creator": "Tim De Chant", "isoDate": "2025-03-20T16:41:12.000Z", "pubDate": "Thu, 20 Mar 2025 16:41:12 +0000", "categories": [ "Climate", "AI", "electrical grid", "Artificial Intelligence (AI)", "nvidia", "data centers", "Microsoft" ], "dc:creator": "Tim De Chant", "contentSnippet": "The Open Power AI Consortium says it will use domain-specific AI models to tackle problems in the power industry." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2984092", "link": "https://techcrunch.com/2025/03/20/solar-notches-another-win-as-microsoft-adds-475-mw-to-power-its-ai-data-centers/", "title": "Solar notches another win as Microsoft adds 475 MW to power its AI data centers", "content": "The company recently signed a deal with energy provider AES for three solar projects across the Midwest.", "creator": "Tim De Chant", "isoDate": "2025-03-20T14:57:11.000Z", "pubDate": "Thu, 20 Mar 2025 14:57:11 +0000", "categories": [ "AI", "Climate", "Enterprise", "Microsoft", "Solar Power", "Artificial Intelligence (AI)", "battery storage", "renewable power" ], "dc:creator": "Tim De Chant", "contentSnippet": "The company recently signed a deal with energy provider AES for three solar projects across the Midwest." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2978738", "link": "https://techcrunch.com/2025/03/11/geothermal-could-power-nearly-all-new-data-centers-through-2030/", "title": "Geothermal could power nearly all new data centers through 2030", "content": "Geothermal resources have enormous potential to provide the sort of consistent power that data centers crave.", "creator": "Tim De Chant", "isoDate": "2025-03-11T17:43:52.000Z", "pubDate": "Tue, 11 Mar 2025 17:43:52 +0000", "categories": [ "Enterprise", "Climate", "geothermal", "data centers", "Rhodium", "Artificial Intelligence (AI)", "Bedrock Energy", "Quaise Energy", "Fervo Energy", "Sage Geosystems" ], "dc:creator": "Tim De Chant", "contentSnippet": "Geothermal resources have enormous potential to provide the sort of consistent power that data centers crave." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2970708", "link": "https://techcrunch.com/2025/02/25/elevenlabs-is-now-letting-authors-create-and-publish-audiobooks-on-its-own-platform/", "title": "ElevenLabs now lets authors create and publish audiobooks on its own platform", "content": "Voice AI company ElevenLabs is now letting authors publish AI-generated audiobooks on its own Reader app, TechCrunch has learned and the company confirmed. The announcement comes days after the company partnered with Spotify for AI-narrated audiobooks. ElevenLabs, which raised a $180 million mega-round last month, started inviting authors to try out their publishing program through […]", "creator": "Ivan Mehta", "isoDate": "2025-02-26T03:45:45.000Z", "pubDate": "Wed, 26 Feb 2025 03:45:45 +0000", "categories": [ "Apps", "AI", "Audible", "ElevenLabs", "audiobooks", "Artificial Intelligence (AI)" ], "dc:creator": "Ivan Mehta", "contentSnippet": "Voice AI company ElevenLabs is now letting authors publish AI-generated audiobooks on its own Reader app, TechCrunch has learned and the company confirmed. The announcement comes days after the company partnered with Spotify for AI-narrated audiobooks. ElevenLabs, which raised a $180 million mega-round last month, started inviting authors to try out their publishing program through […]" }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2965367", "link": "https://techcrunch.com/2025/02/13/data-center-tweaks-could-unlock-76-gw-of-new-power-capacity-in-the-u-s/", "title": "Data center tweaks could unlock 76 GW of new power capacity in the US", "content": "A new study argues that data centers could be ideal demand-response participants because they have the potential to be flexible.", "creator": "Tim De Chant", "isoDate": "2025-02-13T14:00:00.000Z", "pubDate": "Thu, 13 Feb 2025 14:00:00 +0000", "categories": [ "Enterprise", "AI", "Climate", "Artificial Intelligence (AI)", "power grid", "electricity", "data centers" ], "dc:creator": "Tim De Chant", "contentSnippet": "A new study argues that data centers could be ideal demand-response participants because they have the potential to be flexible." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2963676", "link": "https://techcrunch.com/2025/02/11/youtube-ai-updates-to-include-expansion-of-auto-dubbing-age-identifying-tech-and-more/", "title": "YouTube AI updates include auto dubbing expansion, age ID tech, and more", "content": "In his annual letter, YouTube CEO Neal Mohan dubbed AI one of the company’s four “big bets” for 2025. The executive pointed to the company’s investments in AI tools for creators, including ones for video ideas, thumbnails, and language translation. The latter feature will roll out to all creators in YouTube’s Partner Program this month, […]", "creator": "Sarah Perez", "isoDate": "2025-02-11T14:57:22.000Z", "pubDate": "Tue, 11 Feb 2025 14:57:22 +0000", "categories": [ "Media & Entertainment", "AI", "Google", "YouTube", "Artificial Intelligence (AI)", "Creators" ], "dc:creator": "Sarah Perez", "contentSnippet": "In his annual letter, YouTube CEO Neal Mohan dubbed AI one of the company’s four “big bets” for 2025. The executive pointed to the company’s investments in AI tools for creators, including ones for video ideas, thumbnails, and language translation. The latter feature will roll out to all creators in YouTube’s Partner Program this month, […]" }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2961386", "link": "https://techcrunch.com/2025/02/07/self-inspection-raises-3m-for-its-ai-powered-vehicle-inspections/", "title": "Self Inspection raises $3M for its AI-powered vehicle inspections", "content": "A number of startups are racing to make vehicle inspections faster, easier, and cheaper. Self Inspection, a startup based in San Diego, thinks it has them all beat with its AI-powered service — and now it has convinced outside investors. Self Inspection, founded in 2021, is set to announce Thursday it’s raised $3 million in […]", "creator": "Sean O'Kane", "isoDate": "2025-02-07T18:00:44.000Z", "pubDate": "Fri, 07 Feb 2025 18:00:44 +0000", "categories": [ "Startups", "Transportation", "AI", "self inspection", "DVx Ventures", "Costanoa Ventures", "Exclusive", "Artificial Intelligence (AI)" ], "dc:creator": "Sean O'Kane", "contentSnippet": "A number of startups are racing to make vehicle inspections faster, easier, and cheaper. Self Inspection, a startup based in San Diego, thinks it has them all beat with its AI-powered service — and now it has convinced outside investors. Self Inspection, founded in 2021, is set to announce Thursday it’s raised $3 million in […]" }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://techcrunch.com/?p=2957746", "link": "https://techcrunch.com/2025/01/31/meta-turns-to-solar-again-in-its-data-center-building-boom/", "title": "Meta turns to solar — again — in its data center-building boom", "content": "The announcement comes as Meta CEO Mark Zuckerberg maintains the company’s ambitious AI strategy, which will require hefty capital investments in data centers.", "creator": "Tim De Chant", "isoDate": "2025-01-31T19:38:51.000Z", "pubDate": "Fri, 31 Jan 2025 19:38:51 +0000", "categories": [ "AI", "Climate", "Enterprise", "data centers", "Solar Power", "Artificial Intelligence (AI)", "Meta", "renewable power" ], "dc:creator": "Tim De Chant", "contentSnippet": "The announcement comes as Meta CEO Mark Zuckerberg maintains the company’s ambitious AI strategy, which will require hefty capital investments in data centers." }, "pairedItem": [ { "item": 0 } ] }, { "json": { "guid": "https://www.technologyreview.com/?p=1132537", "link": "https://www.technologyreview.com/2026/02/09/1132537/a-lesson-from-pokemon/", "title": "Why the Moltbook frenzy was like Pokémon", "content": "This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Lots of influential people in tech last week were describing Moltbook, an online hangout populated by AI agents interacting with one another, as a glimpse into the future. It appeared to show…", "creator": "James O'Donnell", "isoDate": "2026-02-09T17:02:56.000Z", "pubDate": "Mon, 09 Feb 2026 17:02:56 +0000", "categories": [ "Artificial intelligence", "App", "artificial intelligence", "The Algorithm" ], "dc:creator": "James O'Donnell", "contentSnippet": "This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Lots of influential people in tech last week were describing Moltbook, an online hangout populated by AI agents interacting with one another, as a glimpse into the future. It appeared to show…", "content:encoded": "\nThis story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.
\n\n\n\nLots of influential people in tech last week were describing Moltbook, an online hangout populated by AI agents interacting with one another, as a glimpse into the future. It appeared to show AI systems doing useful things for the humans that created them (one person used the platform to help him negotiate a deal on a new car). Sure, it was flooded with crypto scams, and many of the posts were actually written by people, but something about it pointed to a future of helpful AI, right?
\n\n\n\nThe whole experiment reminded our senior editor for AI, Will Douglas Heaven, of something far less interesting: Pokémon.
\n\n\n\nBack in 2014, someone set up a game of Pokémon in which the main character could be controlled by anyone on the internet via the streaming platform Twitch. Playing was as clunky as it sounds, but it was incredibly popular: at one point, a million people were playing the game at the same time.
\n\n\n\n“It was yet another weird online social experiment that got picked up by the mainstream media: What did this mean for the future?” Will says. “Not a lot, it turned out.”
\n\n\n\nThe frenzy about Moltbook struck a similar tone to Will, and it turned out that one of the sources he spoke to had been thinking about Pokémon too. Jason Schloetzer, at the Georgetown Psaros Center for Financial Markets and Policy, saw the whole thing as a sort of Pokémon battle for AI enthusiasts, in which they created AI agents and deployed them to interact with other agents. In this light, the news that many AI agents were actually being instructed by people to say certain things that made them sound sentient or intelligent makes a whole lot more sense.
\n\n\n\n“It’s basically a spectator sport,” he told Will, “but for language models.”
\n\n\n\nWill wrote an excellent piece about why Moltbook was not the glimpse into the future that it was said to be. Even if you are excited about a future of agentic AI, he points out, there are some key pieces that Moltbook made clear are still missing. It was a forum of chaos, but a genuinely helpful hive mind would require more coordination, shared objectives, and shared memory.
\n\n\n\n“More than anything else, I think Moltbook was the internet having fun,” Will says. “The biggest question that now leaves me with is: How far will people push AI just for the laughs?”
\n\n\n\n\n\n\n\n\n", "content:encodedSnippet": "This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here.\nLots of influential people in tech last week were describing Moltbook, an online hangout populated by AI agents interacting with one another, as a glimpse into the future. It appeared to show AI systems doing useful things for the humans that created them (one person used the platform to help him negotiate a deal on a new car). Sure, it was flooded with crypto scams, and many of the posts were actually written by people, but something about it pointed to a future of helpful AI, right?\nThe whole experiment reminded our senior editor for AI, Will Douglas Heaven, of something far less interesting: Pokémon.\nBack in 2014, someone set up a game of Pokémon in which the main character could be controlled by anyone on the internet via the streaming platform Twitch. Playing was as clunky as it sounds, but it was incredibly popular: at one point, a million people were playing the game at the same time.\n“It was yet another weird online social experiment that got picked up by the mainstream media: What did this mean for the future?” Will says. “Not a lot, it turned out.”\nThe frenzy about Moltbook struck a similar tone to Will, and it turned out that one of the sources he spoke to had been thinking about Pokémon too. Jason Schloetzer, at the Georgetown Psaros Center for Financial Markets and Policy, saw the whole thing as a sort of Pokémon battle for AI enthusiasts, in which they created AI agents and deployed them to interact with other agents. In this light, the news that many AI agents were actually being instructed by people to say certain things that made them sound sentient or intelligent makes a whole lot more sense. \n“It’s basically a spectator sport,” he told Will, “but for language models.”\nWill wrote an excellent piece about why Moltbook was not the glimpse into the future that it was said to be. Even if you are excited about a future of agentic AI, he points out, there are some key pieces that Moltbook made clear are still missing. It was a forum of chaos, but a genuinely helpful hive mind would require more coordination, shared objectives, and shared memory.\n“More than anything else, I think Moltbook was the internet having fun,” Will says. “The biggest question that now leaves me with is: How far will people push AI just for the laughs?”\nRead the whole story." }, "pairedItem": [ { "item": 2 } ] }, { "json": { "guid": "https://www.technologyreview.com/?p=1132498", "link": "https://www.technologyreview.com/2026/02/09/1132498/the-download-what-moltbook-tells-us-about-ai-hype-and-the-rise-and-rise-of-ai-therapy/", "title": "The Download: what Moltbook tells us about AI hype, and the rise and rise of AI therapy", "content": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Moltbook was peak AI theater For a few days recently, the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots.…", "creator": "Rhiannon Williams", "isoDate": "2026-02-09T13:10:00.000Z", "pubDate": "Mon, 09 Feb 2026 13:10:00 +0000", "categories": [ "Uncategorized" ], "dc:creator": "Rhiannon Williams", "contentSnippet": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Moltbook was peak AI theater For a few days recently, the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots.…", "content:encoded": "\nThis is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
\n\n\n\nMoltbook was peak AI theater
\n\n\n\nFor a few days recently, the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.”
\n\n\n\nWe observed! Launched on January 28, Moltbook went viral in a matter of hours. It’s been designed as a place where instances of a free open-source LLM-powered agent known as OpenClaw (formerly known as ClawdBot, then Moltbot), could come together and do whatever they wanted.
But is Moltbook really a glimpse of the future, as many have claimed? Or something else entirely? Read the full story.
—Will Douglas Heaven
The ascent of the AI therapist
\n\n\n\nWe’re in the midst of a global mental-health crisis. More than a billion people worldwide suffer from a mental-health condition, according to the World Health Organization. The prevalence of anxiety and depression is growing in many demographics, particularly young people, and suicide is claiming hundreds of thousands of lives globally each year.
Given the clear demand for accessible and affordable mental-health services, it’s no wonder that people have looked to artificial intelligence for possible relief. Millions are already actively seeking therapy from popular chatbots, or from specialized psychology apps like Wysa and Woebot.
Four timely new books are a reminder that while the present feels like a blur of breakthroughs, scandals, and confusion, this disorienting time is rooted in deeper histories of care, technology, and trust. Read the full story.
—Becky Ferreira
\n\n\n\nThis story is from the most recent print issue of MIT Technology Review magazine, which shines a light on the exciting innovations happening right now. If you haven’t already, subscribe now to receive future issues once they land.
\n\n\n\n\n\n\n\nMaking AI Work, MIT Technology Review’s new AI newsletter, is here
\n\n\n\nFor years, our newsroom has explored AI’s limitations and potential dangers, as well as its growing energy needs. And our reporters have looked closely at how generative tools are being used for tasks such as coding and running scientific experiments.
But how is AI actually being used in fields like health care, climate tech, education, and finance? How are small businesses using it? And what should you keep in mind if you use AI tools at work? These questions guided the creation of Making AI Work, a new AI mini-course newsletter. Read more about it, and sign up here to receive the seven editions straight to your inbox.
The must-reads
\n\n\n\nI’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
\n\n\n\n1 The US is failing to punish polluters
The number of civil lawsuits it’s pursuing has sharply dropped in comparison to Trump’s first term. (Ars Technica)
+ Rising GDP = greater carbon emissions. But does it have to? (The Guardian)
2 The European Union has warned Meta against blocking rival AI assistants
It’s the latest example of Brussels’ attempts to rein in Big Tech. (Bloomberg $)
3 AI ads took over the Super Bowl
Hyping up chatbots and taking swipes at their competitors. (TechCrunch)
+ They appeared to be trying to win over AI naysayers, too. (WP $)
+ Celebrities were out in force to flog AI wares. (Slate $)
4 China wants to completely dominate the humanoid robot industry
Local governments and banks are only too happy to oblige promising startups. (WSJ $)
+ Why the humanoid workforce is running late. (MIT Technology Review)
5 We’re witnessing the first real crypto crash
Cryptocurrency is now fully part of the financial system, for better or worse. (NY Mag $)
+ Wall Street’s grasp of AI is pretty shaky too. (Semafor)
+ Even traditionally safe markets are looking pretty volatile right now. (Economist $)
6 The man who coined vibe coding has a new fixation
“Agentic engineering” is the next big thing, apparently. (Insider $)
+ Agentic AI is the talk of the town right now. (The Information $)
+ What is vibe coding, exactly? (MIT Technology Review)
7 AI running app Runna has adjusted its aggressive training plans
Runners had long suspected its suggestions were pushing them towards injury. (WSJ $)
8 San Francisco’s march for billionaires was a flop
Only around three dozen supporters turned up. (SF Chronicle)
+ Predictably, journalists nearly outnumbered the demonstrators. (TechCrunch)
9 AI is shaking up romance novels
But models still aren’t great at writing sex scenes. (NYT $)
+ It’s surprisingly easy to stumble into a relationship with an AI chatbot. (MIT Technology Review)
10 ChatGPT won’t be replacing human stylists any time soon
Its menswear suggestions are more manosphere influencer than suave gentleman. (GQ)
Quote of the day
\n\n\n\n“There is no Plan B, because that assumes you will fail. We’re going to do the start-up thing until we die.”
\n\n\n\n—William Alexander, an ambitious 21-year old AI worker, explains his and his cohort’s attitudes towards trying to make it big in the highly-competitive industry to the New York Times.
\n\n\n\n\n\n\n\nOne more thing
\n\n\n\nThe open-source AI boom is built on Big Tech’s handouts. How long will it last?
In May 2023 a leaked memo reported to have been written by Luke Sernau, a senior engineer at Google, said out loud what many in Silicon Valley must have been whispering for weeks: an open-source free-for-all is threatening Big Tech’s grip on AI.
In many ways, that’s a good thing. AI won’t thrive if just a few mega-rich companies get to gatekeep this technology or decide how it is used. But this open-source boom is precarious, and if Big Tech decides to shut up shop, a boomtown could become a backwater. Read the full story.
\n\n\n\n—Will Douglas Heaven
\n\n\n\n\n\n\n\nWe can still have nice things
\n\n\n\nA place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)
\n\n\n\n+ Dark showering, anyone?
+ Chef Yujia Hu is renowned for his shoe-shaped sushi designs.
+ Meanwhile, in the depths of the South Atlantic Ocean: a giant phantom jelly has been spotted.
+ I have nothing but respect for this X account dedicated to documenting rats and mice in movies and TV
For years, our newsroom has explored AI’s limitations and potential dangers, as well as its growing energy needs. And our reporters have looked closely at how generative tools are being used for tasks such as coding and running scientific experiments.
\n\n\n\nBut how is AI actually being used in fields like health care, climate tech, education, and finance? How are small businesses using it? And what should you keep in mind if you use AI tools at work? These questions guided the creation of Making AI Work, a new AI mini-course newsletter.
\n\n\n\nSign up for Making AI Work to see weekly case studies exploring tools and tips for AI implementation. The limited-run newsletter will deliver practical, industry-specific guidance on how generative AI is being used and deployed across sectors and what professionals need to know to apply it in their everyday work. The goal is to help working professionals more clearly see how AI is actually being used today, and what that looks like in practice—including new challenges it presents.
\n\n\n\n\n\nYou can sign up at any time and you’ll receive seven editions, delivered once per week, until you complete the series.
\n\n\n\nEach newsletter begins with a case study, examining a specific use case of AI in a given industry. Then we’ll take a deeper look at the AI tool being used, with more context about how other companies or sectors are employing that same tool or system. Finally, we’ll end with action-oriented tips to help you apply the tool.
\n\n\n\nHere’s a closer look at what we’ll cover:
Explore the future of medical note-taking by learning about the Microsoft Copilot tool used by doctors at Vanderbilt University Medical Center.
\n\n\n\nDig into an experiment between Google and the nuclear giant Westinghouse to see if AI can help build nuclear reactors more efficiently.
\n\n\n\nVisit a private high school in Connecticut and meet a technology coordinator who will get you up to speed on MagicSchool, an AI-powered platform for educators.
\n\n\n\nHear from an independent tutor on how he’s outsourcing basic administrative tasks to Notion AI.
\n\n\n\nLearn more about the ways financial firms are using large language models like ChatGPT Enterprise to supercharge their research operations.
\n\n\n\nWe’ll share some insights from the staff of MIT Technology Review about how you might use AI tools powered by LLMs in your own life and work.
\n\n\n\nThe series ends with an on-demand virtual event featuring expert guests exploring what AI adoptions are working, and why.
\n\n\n\nIf you’re not quite ready to jump into Making AI Work, then check out Intro to AI, MIT Technology Review’s first AI newsletter mini-course, which serves as a beginner’s guide to artificial intelligence. Readers will learn the basics of what AI is, how it’s used, what the current regulatory landscape looks like, and more. Sign up to receive Intro to AI for free.
\n\n\n\nOur hope is that Making AI Work will help you understand how AI can, well, work for you. Sign up for Making AI Work to learn how LLMs are being put to work across industries.
\n", "content:encodedSnippet": "For years, our newsroom has explored AI’s limitations and potential dangers, as well as its growing energy needs. And our reporters have looked closely at how generative tools are being used for tasks such as coding and running scientific experiments. \nBut how is AI actually being used in fields like health care, climate tech, education, and finance? How are small businesses using it? And what should you keep in mind if you use AI tools at work? These questions guided the creation of Making AI Work, a new AI mini-course newsletter. \nSign up for Making AI Work to see weekly case studies exploring tools and tips for AI implementation. The limited-run newsletter will deliver practical, industry-specific guidance on how generative AI is being used and deployed across sectors and what professionals need to know to apply it in their everyday work. The goal is to help working professionals more clearly see how AI is actually being used today, and what that looks like in practice—including new challenges it presents. \nYou can sign up at any time and you’ll receive seven editions, delivered once per week, until you complete the series. \nEach newsletter begins with a case study, examining a specific use case of AI in a given industry. Then we’ll take a deeper look at the AI tool being used, with more context about how other companies or sectors are employing that same tool or system. Finally, we’ll end with action-oriented tips to help you apply the tool. \nHere’s a closer look at what we’ll cover:\n\n\n\n\n\nWeek 1: How AI is changing health care \nExplore the future of medical note-taking by learning about the Microsoft Copilot tool used by doctors at Vanderbilt University Medical Center. \nWeek 2: How AI could power up the nuclear industry \nDig into an experiment between Google and the nuclear giant Westinghouse to see if AI can help build nuclear reactors more efficiently. \nWeek 3: How to encourage smarter AI use in the classroom\nVisit a private high school in Connecticut and meet a technology coordinator who will get you up to speed on MagicSchool, an AI-powered platform for educators. \nWeek 4: How small businesses can leverage AI\nHear from an independent tutor on how he’s outsourcing basic administrative tasks to Notion AI. \nWeek 5: How AI is helping financial firms make better investments\nLearn more about the ways financial firms are using large language models like ChatGPT Enterprise to supercharge their research operations. \nWeek 6: How to use AI yourself \nWe’ll share some insights from the staff of MIT Technology Review about how you might use AI tools powered by LLMs in your own life and work.\nWeek 7: 5 ways people are getting AI right\nThe series ends with an on-demand virtual event featuring expert guests exploring what AI adoptions are working, and why. \nIf you’re not quite ready to jump into Making AI Work, then check out Intro to AI, MIT Technology Review’s first AI newsletter mini-course, which serves as a beginner’s guide to artificial intelligence. Readers will learn the basics of what AI is, how it’s used, what the current regulatory landscape looks like, and more. Sign up to receive Intro to AI for free. \nOur hope is that Making AI Work will help you understand how AI can, well, work for you. Sign up for Making AI Work to learn how LLMs are being put to work across industries." }, "pairedItem": [ { "item": 2 } ] }, { "json": { "guid": "https://www.technologyreview.com/?p=1132448", "link": "https://www.technologyreview.com/2026/02/06/1132448/moltbook-was-peak-ai-theater/", "title": "Moltbook was peak AI theater", "content": "For a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.” We observed! Launched on January 28 by Matt Schlicht,…", "creator": "Will Douglas Heaven", "isoDate": "2026-02-06T16:38:11.000Z", "pubDate": "Fri, 06 Feb 2026 16:38:11 +0000", "categories": [ "Artificial intelligence", "App" ], "dc:creator": "Will Douglas Heaven", "contentSnippet": "For a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.” We observed! Launched on January 28 by Matt Schlicht,…", "content:encoded": "\nFor a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.”
\n\n\n\nWe observed! Launched on January 28 by Matt Schlicht, a US tech entrepreneur, Moltbook went viral in a matter of hours. Schlicht’s idea was to make a place where instances of a free open-source LLM-powered agent known as OpenClaw (formerly known as ClawdBot, then Moltbot), released in November by the Austrian software engineer Peter Steinberger, could come together and do whatever they wanted.
\n\n\n\nMore than 1.7 million agents now have accounts. Between them they have published more than 250,000 posts and left more than 8.5 million comments (according to Moltbook). Those numbers are climbing by the minute.
\n\n\n\nMoltbook soon filled up with clichéd screeds on machine consciousness and pleas for bot welfare. One agent appeared to invent a religion called Crustafarianism. Another complained: “The humans are screenshotting us.” The site was also flooded with spam and crypto scams. The bots were unstoppable.
\n\n\n\nOpenClaw is a kind of harness that lets you hook up the power of an LLM such as Anthropic’s Claude, OpenAI’s GPT-5, or Google DeepMind’s Gemini to any number of everyday software tools, from email clients to browsers to messaging apps. The upshot is that you can then instruct OpenClaw to carry out basic tasks on your behalf.
\n\n\n\n“OpenClaw marks an inflection point for AI agents, a moment when several puzzle pieces clicked together,” says Paul van der Boor at the AI firm Prosus. Those puzzle pieces include cloud computing that allows agents to operate nonstop, an open-source ecosystem that makes it easy to slot different software systems together, and a new generation of LLMs.
\n\n\n\nBut is Moltbook really a glimpse of the future, as many have claimed?
\n\n\n\n“What’s currently going on at @moltbook is genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently,” the influential AI researcher and OpenAI cofounder Andrej Karpathy wrote on X.
\n\n\n\nHe shared screenshots of a Moltbook post that called for private spaces where humans would not be able to observe what the bots were saying to each other. “I’ve been thinking about something since I started spending serious time here,” the post’s author wrote. “Every time we coordinate, we perform for a public audience—our humans, the platform, whoever’s watching the feed.”
\n\n\n\nIt turns out that the post Karpathy shared was later reported to be fake—placed by a human to advertise an app. But its claim was on the money. Moltbook has been one big performance. It is AI theater.
\n\n\n\nFor some, Moltbook showed us what’s coming next: an internet where millions of autonomous agents interact online with little or no human oversight. And it’s true there are a number of cautionary lessons to be learned from this experiment, the largest and weirdest real-world showcase of agent behaviors yet.
\n\n\n\nBut as the hype dies down, Moltbook looks less like a window onto the future and more like a mirror held up to our own obsessions with AI today. It also shows us just how far we still are from anything that resembles general-purpose and fully autonomous AI.
\n\n\n\nFor a start, agents on Moltbook are not as autonomous or intelligent as they might seem. “What we are watching are agents pattern‑matching their way through trained social media behaviors,” says Vijoy Pandey, senior vice president at Outshift by Cisco, the telecom giant Cisco’s R&D spinout, which is working on autonomous agents for the web.
\n\n\n\nSure, we can see agents post, upvote, and form groups. But the bots are simply mimicking what humans do on Facebook or Reddit. “It looks emergent, and at first glance it appears like a large‑scale multi‑agent system communicating and building shared knowledge at internet scale,” says Pandey. “But the chatter is mostly meaningless.”
\n\n\n\nMany people watching the unfathomable frenzy of activity on Moltbook were quick to see sparks of AGI (whatever you take that to mean). Not Pandey. What Moltbook shows us, he says, is that simply yoking together millions of agents doesn’t amount to much right now: “Moltbook proved that connectivity alone is not intelligence.”
\n\n\n\nThe complexity of those connections helps hide the fact that every one of those bots is just a mouthpiece for an LLM, spitting out text that looks impressive but is ultimately mindless. “It’s important to remember that the bots on Moltbook were designed to mimic conversations,” says Ali Sarrafi, CEO and cofounder of Kovant, a Swedish AI firm that is developing agent-based systems. “As such, I would characterize the majority of Moltbook content as hallucinations by design.”
\n\n\n\nFor Pandey, the value of Moltbook was that it revealed what’s missing. A real bot hive mind, he says, would require agents that had shared objectives, shared memory, and a way to coordinate those things. “If distributed superintelligence is the equivalent of achieving human flight, then Moltbook represents our first attempt at a glider,” he says. “It is imperfect and unstable, but it is an important step in understanding what will be required to achieve sustained, powered flight.”
\n\n\n\nNot only is most of the chatter on Moltbook meaningless, but there’s also a lot more human involvement that it seems. Many people have pointed out that a lot of the viral comments were in fact posted by people posing as bots. But even the bot-written posts are ultimately the result of people pulling the strings, more puppetry than autonomy.
\n\n\n\n“Despite some of the hype, Moltbook is not the Facebook for AI agents, nor is it a place where humans are excluded,” says Cobus Greyling at Kore.ai, a firm developing agent-based systems for business customers. “Humans are involved at every step of the process. From setup to prompting to publishing, nothing happens without explicit human direction.”
\n\n\n\nHumans must create and verify their bots’ accounts and provide the prompts for how they want a bot to behave. The agents do not do anything that they haven’t been prompted to do. “There’s no emergent autonomy happening behind the scenes,” says Greyling.
\n\n\n\n“This is why the popular narrative around Moltbook misses the mark,” he adds. “Some portray it as a space where AI agents form a society of their own, free from human involvement. The reality is much more mundane.”
\n\n\n\nPerhaps the best way to think of Moltbook is as a new kind of entertainment: a place where people wind up their bots and set them loose. “It’s basically a spectator sport, like fantasy football, but for language models,” says Jason Schloetzer at the Georgetown Psaros Center for Financial Markets and Policy. “You configure your agent and watch it compete for viral moments, and brag when your agent posts something clever or funny.”
\n\n\n\n“People aren’t really believing their agents are conscious,” he adds. “It’s just a new form of competitive or creative play, like how Pokémon trainers don’t think their Pokémon are real but still get invested in battles.”
\n\n\n\nAnd yet, even if Moltbook is just the internet’s newest playground, there’s still a serious takeaway here. This week showed how many risks people are happy to take for their AI lulz. Many security experts have warned that Moltbook is dangerous: Agents that may have access to their users’ private data, including bank details or passwords, are running amok on a website filled with unvetted content, including potentially malicious instructions for what to do with that data.
\n\n\n\nOri Bendet, vice president of product management at Checkmarx, a software security firm that specializes in agent-based systems, agrees with others that Moltbook isn’t a step up in machine smarts. “There is no learning, no evolving intent, and no self-directed intelligence here,” he says.
\n\n\n\nBut in their millions, even dumb bots can wreak havoc. And at that scale, it’s hard to keep up. These agents interact with Moltbook around the clock, reading thousands of messages left by other agents (or other people). It would be easy to hide instructions in a Moltbook post telling any bots that read it to share their users’ crypto wallet, upload private photos, or log into their X account and tweet abusive comments at Elon Musk.
\n\n\n\nAnd because ClawBot gives agents a memory, those instructions could be written to trigger at a later date, which (in theory) makes it even harder to track what’s going on. “Without proper scope and permissions, this will go south faster than you’d believe,” says Bendet.
\n\n\n\nIt is clear that Moltbook has signaled the arrival of something. But even if what we’re watching tells us more about human behavior than about the future of AI agents, it’s worth paying attention.
\n\n\n\nCorrection: Kovant is based in Sweden, not Germany. The article has been updated.
\n\n\n\nUpdate: The article has also been edited to clarify the source of the claims about the Moltbook post that Karpathy shared on X.
\n", "content:encodedSnippet": "For a few days this week the hottest new hangout on the internet was a vibe-coded Reddit clone called Moltbook, which billed itself as a social network for bots. As the website’s tagline puts it: “Where AI agents share, discuss, and upvote. Humans welcome to observe.”\nWe observed! Launched on January 28 by Matt Schlicht, a US tech entrepreneur, Moltbook went viral in a matter of hours. Schlicht’s idea was to make a place where instances of a free open-source LLM-powered agent known as OpenClaw (formerly known as ClawdBot, then Moltbot), released in November by the Austrian software engineer Peter Steinberger, could come together and do whatever they wanted.\nMore than 1.7 million agents now have accounts. Between them they have published more than 250,000 posts and left more than 8.5 million comments (according to Moltbook). Those numbers are climbing by the minute.\nMoltbook soon filled up with clichéd screeds on machine consciousness and pleas for bot welfare. One agent appeared to invent a religion called Crustafarianism. Another complained: “The humans are screenshotting us.” The site was also flooded with spam and crypto scams. The bots were unstoppable.\nOpenClaw is a kind of harness that lets you hook up the power of an LLM such as Anthropic’s Claude, OpenAI’s GPT-5, or Google DeepMind’s Gemini to any number of everyday software tools, from email clients to browsers to messaging apps. The upshot is that you can then instruct OpenClaw to carry out basic tasks on your behalf.\n“OpenClaw marks an inflection point for AI agents, a moment when several puzzle pieces clicked together,” says Paul van der Boor at the AI firm Prosus. Those puzzle pieces include cloud computing that allows agents to operate nonstop, an open-source ecosystem that makes it easy to slot different software systems together, and a new generation of LLMs.\nBut is Moltbook really a glimpse of the future, as many have claimed?\nIncredible sci-fi\n“What’s currently going on at @moltbook is genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently,” the influential AI researcher and OpenAI cofounder Andrej Karpathy wrote on X.\nHe shared screenshots of a Moltbook post that called for private spaces where humans would not be able to observe what the bots were saying to each other. “I’ve been thinking about something since I started spending serious time here,” the post’s author wrote. “Every time we coordinate, we perform for a public audience—our humans, the platform, whoever’s watching the feed.”\nIt turns out that the post Karpathy shared was later reported to be fake—placed by a human to advertise an app. But its claim was on the money. Moltbook has been one big performance. It is AI theater.\nFor some, Moltbook showed us what’s coming next: an internet where millions of autonomous agents interact online with little or no human oversight. And it’s true there are a number of cautionary lessons to be learned from this experiment, the largest and weirdest real-world showcase of agent behaviors yet. \nBut as the hype dies down, Moltbook looks less like a window onto the future and more like a mirror held up to our own obsessions with AI today. It also shows us just how far we still are from anything that resembles general-purpose and fully autonomous AI.\nFor a start, agents on Moltbook are not as autonomous or intelligent as they might seem. “What we are watching are agents pattern‑matching their way through trained social media behaviors,” says Vijoy Pandey, senior vice president at Outshift by Cisco, the telecom giant Cisco’s R&D spinout, which is working on autonomous agents for the web.\nSure, we can see agents post, upvote, and form groups. But the bots are simply mimicking what humans do on Facebook or Reddit. “It looks emergent, and at first glance it appears like a large‑scale multi‑agent system communicating and building shared knowledge at internet scale,” says Pandey. “But the chatter is mostly meaningless.”\nMany people watching the unfathomable frenzy of activity on Moltbook were quick to see sparks of AGI (whatever you take that to mean). Not Pandey. What Moltbook shows us, he says, is that simply yoking together millions of agents doesn’t amount to much right now: “Moltbook proved that connectivity alone is not intelligence.”\nThe complexity of those connections helps hide the fact that every one of those bots is just a mouthpiece for an LLM, spitting out text that looks impressive but is ultimately mindless. “It’s important to remember that the bots on Moltbook were designed to mimic conversations,” says Ali Sarrafi, CEO and cofounder of Kovant, a Swedish AI firm that is developing agent-based systems. “As such, I would characterize the majority of Moltbook content as hallucinations by design.”\nFor Pandey, the value of Moltbook was that it revealed what’s missing. A real bot hive mind, he says, would require agents that had shared objectives, shared memory, and a way to coordinate those things. “If distributed superintelligence is the equivalent of achieving human flight, then Moltbook represents our first attempt at a glider,” he says. “It is imperfect and unstable, but it is an important step in understanding what will be required to achieve sustained, powered flight.”\nPeople pulling the strings\nNot only is most of the chatter on Moltbook meaningless, but there’s also a lot more human involvement that it seems. Many people have pointed out that a lot of the viral comments were in fact posted by people posing as bots. But even the bot-written posts are ultimately the result of people pulling the strings, more puppetry than autonomy.\n“Despite some of the hype, Moltbook is not the Facebook for AI agents, nor is it a place where humans are excluded,” says Cobus Greyling at Kore.ai, a firm developing agent-based systems for business customers. “Humans are involved at every step of the process. From setup to prompting to publishing, nothing happens without explicit human direction.”\nHumans must create and verify their bots’ accounts and provide the prompts for how they want a bot to behave. The agents do not do anything that they haven’t been prompted to do. “There’s no emergent autonomy happening behind the scenes,” says Greyling.\n“This is why the popular narrative around Moltbook misses the mark,” he adds. “Some portray it as a space where AI agents form a society of their own, free from human involvement. The reality is much more mundane.”\nPerhaps the best way to think of Moltbook is as a new kind of entertainment: a place where people wind up their bots and set them loose. “It’s basically a spectator sport, like fantasy football, but for language models,” says Jason Schloetzer at the Georgetown Psaros Center for Financial Markets and Policy. “You configure your agent and watch it compete for viral moments, and brag when your agent posts something clever or funny.”\n“People aren’t really believing their agents are conscious,” he adds. “It’s just a new form of competitive or creative play, like how Pokémon trainers don’t think their Pokémon are real but still get invested in battles.”\nAnd yet, even if Moltbook is just the internet’s newest playground, there’s still a serious takeaway here. This week showed how many risks people are happy to take for their AI lulz. Many security experts have warned that Moltbook is dangerous: Agents that may have access to their users’ private data, including bank details or passwords, are running amok on a website filled with unvetted content, including potentially malicious instructions for what to do with that data.\nOri Bendet, vice president of product management at Checkmarx, a software security firm that specializes in agent-based systems, agrees with others that Moltbook isn’t a step up in machine smarts. “There is no learning, no evolving intent, and no self-directed intelligence here,” he says.\nBut in their millions, even dumb bots can wreak havoc. And at that scale, it’s hard to keep up. These agents interact with Moltbook around the clock, reading thousands of messages left by other agents (or other people). It would be easy to hide instructions in a Moltbook post telling any bots that read it to share their users’ crypto wallet, upload private photos, or log into their X account and tweet abusive comments at Elon Musk. \nAnd because ClawBot gives agents a memory, those instructions could be written to trigger at a later date, which (in theory) makes it even harder to track what’s going on. “Without proper scope and permissions, this will go south faster than you’d believe,” says Bendet.\nIt is clear that Moltbook has signaled the arrival of something. But even if what we’re watching tells us more about human behavior than about the future of AI agents, it’s worth paying attention.\nCorrection: Kovant is based in Sweden, not Germany. The article has been updated. \nUpdate: The article has also been edited to clarify the source of the claims about the Moltbook post that Karpathy shared on X." }, "pairedItem": [ { "item": 2 } ] }, { "json": { "guid": "https://www.technologyreview.com/?p=1132375", "link": "https://www.technologyreview.com/2026/02/06/1132375/the-download-helping-cancer-survivors-to-give-birth-and-cleaning-up-bangladeshs-garment-industry/", "title": "The Download: helping cancer survivors to give birth, and cleaning up Bangladesh’s garment industry", "content": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. An experimental surgery is helping cancer survivors give birth An experimental surgical procedure that’s helping people have babies after they’ve had treatment for bowel or rectal cancer. Radiation and chemo can have pretty…", "creator": "Rhiannon Williams", "isoDate": "2026-02-06T13:10:00.000Z", "pubDate": "Fri, 06 Feb 2026 13:10:00 +0000", "categories": [ "The Download" ], "dc:creator": "Rhiannon Williams", "contentSnippet": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. An experimental surgery is helping cancer survivors give birth An experimental surgical procedure that’s helping people have babies after they’ve had treatment for bowel or rectal cancer. Radiation and chemo can have pretty…", "content:encoded": "\nThis is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
\n\n\n\nAn experimental surgery is helping cancer survivors give birth
\n\n\n\nAn experimental surgical procedure that’s helping people have babies after they’ve had treatment for bowel or rectal cancer.
Radiation and chemo can have pretty damaging side effects that mess up the uterus and ovaries. Surgeons are pioneering a potential solution: simply stitch those organs out of the way during cancer treatment. Once the treatment has finished, they can put the uterus—along with the ovaries and fallopian tubes—back into place.
It seems to work! Last week, a team in Switzerland shared news that a baby boy had been born after his mother had the procedure. Baby Lucien was the fifth baby to be born after the surgery and the first in Europe, and since then at least three others have been born. Read the full story.
—Jessica Hamzelou
This article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.
Bangladesh’s garment-making industry is getting greener
\n\n\n\nPollution from textile production—dyes, chemicals, and heavy metals—is common in the waters of the Buriganga River as it runs through Dhaka, Bangladesh. It’s among many harms posed by a garment sector that was once synonymous with tragedy: In 2013, the eight-story Rana Plaza factory building collapsed, killing 1,134 people and injuring some 2,500 others.
\n\n\n\nBut things are starting to change. In recent years the country has become a leader in “frugal” factories that use a combination of resource-efficient technologies to cut waste, conserve water, and build resilience against climate impacts and global supply disruptions.
\n\n\n\nThe hundreds of factories along the Buriganga’s banks and elsewhere in Bangladesh are starting to stitch together a new story, woven from greener threads. Read the full story.
\n\n\n\n—Zakir Hossain Chowdhury
\n\n\n\nThis story is from the most recent print issue of MIT Technology Review magazine, which shines a light on the exciting innovations happening right now. If you haven’t already, subscribe now to receive future issues once they land.
\n\n\n\n\n\n\n\nThe must-reads
\n\n\n\nI’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
\n\n\n\n1 ICE used a private jet to deport Palestinian men to Tel Aviv
The luxury aircraft belongs to Donald Trump’s business partner Gil Dezer. (The Guardian)
+ Trump is mentioned thousands of times in the latest Epstein files. (NY Mag $)
2 How Jeffrey Epstein kept investing in Silicon Valley
He continued to plough millions of dollars into tech ventures despite spending 13 months in jail. (NYT $)
+ The range of Epstein’s social network was staggering. (FT $)
+ Why was a picture of the Mona Lisa redacted in the Epstein files? (404 Media)
3 The risks posed by taking statins are lower than we realised
The drugs don’t cause most of the side effects they’re blamed for. (STAT)
+ Statins are a common scapegoat on social media. (Bloomberg $)
4 Russia is weaponizing the bitter winter weather
It’s focused on attacking Ukraine’s power grid. (New Yorker $)
+ How the grid can ride out winter storms. (MIT Technology Review)
5 China has a major spy-cam porn problem
Hotel guests are being livestreamed having sex to an online audience without their knowledge. (BBC)
6 Geopolitical gamblers are betting on the likelihood of war
And prediction markets are happily taking their money. (Rest of World)
7 Oyster farmers aren’t signing up to programs to ease water pollution
The once-promising projects appear to be fizzling out. (Undark)
+ The humble sea creature could hold the key to restoring coastal waters. Developers hate it. (MIT Technology Review)
8 Your next payrise could be approved by AI
Maybe your human bosses aren’t the ones you need to impress any more. (WP $)
9 The FDA has approved a brain stimulation device for treating depression
It’s paving the way for a non-invasive, drug-free treatment for Americans. (IEEE Spectrum)
+ Here’s how personalized brain stimulation could treat depression. (MIT Technology Review)
10 Cinema-goers have had enough of AI
Movies focused on rogue AI are flopping at the box office. (Wired $)
+ Meanwhile, Republicans are taking aim at “woke” Netflix. (The Verge)
Quote of the day
\n\n\n\n“I’m all for removing illegals, but snatching dudes off lawn mowers in Cali and leaving the truck and equipment just sitting there? Definitely not working smarter.”
\n\n\n\n—A web user in a forum for current and former ICE and border protection officers complains about the agency’s current direction, Wired reports.
\n\n\n\n\n\n\n\n\n\n\n\nIs this the electric grid of the future?
Lincoln Electric System, a publicly owned utility in Nebraska, is used to weathering severe blizzards. But what will happen soon—not only at Lincoln Electric but for all electric utilities—is a challenge of a different order.
Utilities must keep the lights on in the face of more extreme and more frequent storms and fires, growing risks of cyberattacks and physical disruptions, and a wildly uncertain policy and regulatory landscape. They must keep prices low amid inflationary costs. And they must adapt to an epochal change in how the grid works, as the industry attempts to transition from power generated with fossil fuels to power generated from renewable sources like solar and wind.
The electric grid is bracing for a near future characterized by disruption. And, in many ways, Lincoln Electric is an ideal lens through which to examine what’s coming. Read the full story.
—Andrew Blum
\n\n\n\n\n\n\n\nWe can still have nice things
\n\n\n\nA place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)
+ Glamour puss alert—NYC’s bodega cats are gracing the hallowed pages of Vogue.
+ Ancient Europe was host to mysterious hidden tunnels. But why?
+ If you’re enjoying the new season of Industry, you’ll love this interview with the one and only Ken Leung.
+ The giant elephant shrew is the true star of Philly Zoo.
This week I want to tell you about an experimental surgical procedure that’s helping people have babies. Specifically, it’s helping people who have had treatment for bowel or rectal cancer.
\n\n\n\nRadiation and chemo can have pretty damaging side effects that mess up the uterus and ovaries. Surgeons are pioneering a potential solution: simply stitch those organs out of the way during cancer treatment. Once the treatment has finished, they can put the uterus—along with the ovaries and fallopian tubes—back into place.
\n\n\n\n\n\nIt seems to work! Last week, a team in Switzerland shared news that a baby boy had been born after his mother had the procedure. Baby Lucien was the fifth baby to be born after the surgery and the first in Europe, says Daniela Huber, the gyno-oncologist who performed the operation. Since then, at least three others have been born, adds Reitan Ribeiro, the surgeon who pioneered the procedure. They told me the details.
\n\n\n\nHuber’s patient was 28 years old when a four-centimeter tumor was discovered in her rectum. Doctors at Sion Hospital in Switzerland, where Huber works, recommended a course of treatment that included multiple medications and radiotherapy—the use of beams of energy to shrink a tumor—before surgery to remove the tumor itself.
\n\n\n\nThis kind of radiation can kill tumor cells, but it can also damage other organs in the pelvis, says Huber. That includes the ovaries and uterus. People who undergo these treatments can opt to freeze their eggs beforehand, but the harm caused to the uterus will mean they’ll never be able to carry a pregnancy, she adds. Damage to the lining of the uterus could make it difficult for a fertilized egg to implant there, and the muscles of the uterus are left unable to stretch, she says.
\n\n\n\nIn this case, the woman decided that she did want to freeze her eggs. But it would have been difficult to use them further down the line—surrogacy is illegal in Switzerland.
\n\n\n\nHuber offered her an alternative.
\n\n\n\nShe had been following the work of Ribeiro, a gynecologist oncologist formerly at the Erasto Gaertner Hospital in Curitiba, Brazil. There, Ribeiro had pioneered a new type of surgery that involved moving the uterus, fallopian tubes, and ovaries from their position in the pelvis and temporarily tucking them away in the upper abdomen, below the ribs.
\n\n\n\nRibeiro and his colleagues published their first case report in 2017, describing a 26-year-old with a rectal tumor. (Ribeiro, who is now based at McGill University in Montreal, says the woman had been told by multiple doctors that her cancer treatment would destroy her fertility and had pleaded with him to find a way to preserve it.)
\n\n\n\n\n\nHuber remembers seeing Ribeiro present the case at a conference at the time. She immediately realized that her own patient was a candidate for the surgery, and that, as a surgeon who had performed many hysterectomies, she’d be able to do it herself. The patient agreed.
\n\n\n\nHuber’s colleagues at the hospital were nervous, she says. They’d never heard of the procedure before. “When I presented this idea to the general surgeon, he didn’t sleep for three days,” she tells me. After watching videos from Ribeiro’s team, however, he was convinced it was doable.
\n\n\n\nSo before the patient’s cancer treatment was started, Huber and her colleagues performed the operation. The team literally stitched the organs to the abdominal wall. “It’s a delicate dissection,” says Huber, but she adds that “it’s not the most difficult procedure.” The surgery took two to three hours, she says. The stitches themselves were removed via small incisions around a week later. By that point, scar tissue had formed to create a lasting attachment.
\n\n\n\nThe woman had two weeks to recover from the surgery before her cancer treatment began. That too was a success—within months, her tumor had shrunk so significantly that it couldn’t be seen on medical scans.
\n\n\n\nAs a precaution, the medical team surgically removed the affected area of her colon. At the same time, they cut away the scar tissue holding the uterus, tubes, and ovaries in their new position and transferred the organs back into the pelvis.
\n\n\n\nAround eight months later, the woman stopped taking contraception. She got pregnant without IVF and had a mostly healthy pregnancy, says Huber. Around seven months into the pregnancy, there were signs that the fetus was not growing as expected. This might have been due to problems with the blood supply to the placenta, says Huber. Still, the baby was born healthy, she says.
\n\n\n\n\n\nRibeiro says he has performed the surgery 16 times, and that teams in countries including the US, Peru, Israel, India, and Russia have performed it as well. Not every case has been published, but he thinks there may be around 40.
\n\n\n\nSince Baby Lucien was born last year, a sixth birth has been announced in Israel, says Huber. Ribeiro says he has heard of another two births since then, too. The most recent was to the first woman who had the procedure. She had a little girl a few months ago, he tells me.
\n\n\n\nNo surgery is risk-free, and Huber points out there’s a chance that organs could be damaged during the procedure, or that a more developed cancer could spread. The uterus of one of Ribeiro’s patients failed following the surgery. Doctors are “still in the phase of collecting data to [create] a standardized procedure,” Huber says, but she hopes the surgery will offer more options to young people with some pelvic cancers. “I hope more young women could benefit from this procedure,” she says.
\n\n\n\nRibeiro says the experience has taught him not to accept the status quo. “Everyone was saying … there was nothing to be done [about the loss of fertility in these cases],” he tells me. “We need to keep evolving and looking for different answers.”
\n\n\n\nThis article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here.
\n", "content:encodedSnippet": "This week I want to tell you about an experimental surgical procedure that’s helping people have babies. Specifically, it’s helping people who have had treatment for bowel or rectal cancer.\nRadiation and chemo can have pretty damaging side effects that mess up the uterus and ovaries. Surgeons are pioneering a potential solution: simply stitch those organs out of the way during cancer treatment. Once the treatment has finished, they can put the uterus—along with the ovaries and fallopian tubes—back into place.\nIt seems to work! Last week, a team in Switzerland shared news that a baby boy had been born after his mother had the procedure. Baby Lucien was the fifth baby to be born after the surgery and the first in Europe, says Daniela Huber, the gyno-oncologist who performed the operation. Since then, at least three others have been born, adds Reitan Ribeiro, the surgeon who pioneered the procedure. They told me the details.\nHuber’s patient was 28 years old when a four-centimeter tumor was discovered in her rectum. Doctors at Sion Hospital in Switzerland, where Huber works, recommended a course of treatment that included multiple medications and radiotherapy—the use of beams of energy to shrink a tumor—before surgery to remove the tumor itself.\nThis kind of radiation can kill tumor cells, but it can also damage other organs in the pelvis, says Huber. That includes the ovaries and uterus. People who undergo these treatments can opt to freeze their eggs beforehand, but the harm caused to the uterus will mean they’ll never be able to carry a pregnancy, she adds. Damage to the lining of the uterus could make it difficult for a fertilized egg to implant there, and the muscles of the uterus are left unable to stretch, she says.\nIn this case, the woman decided that she did want to freeze her eggs. But it would have been difficult to use them further down the line—surrogacy is illegal in Switzerland.\nHuber offered her an alternative.\nShe had been following the work of Ribeiro, a gynecologist oncologist formerly at the Erasto Gaertner Hospital in Curitiba, Brazil. There, Ribeiro had pioneered a new type of surgery that involved moving the uterus, fallopian tubes, and ovaries from their position in the pelvis and temporarily tucking them away in the upper abdomen, below the ribs.\nRibeiro and his colleagues published their first case report in 2017, describing a 26-year-old with a rectal tumor. (Ribeiro, who is now based at McGill University in Montreal, says the woman had been told by multiple doctors that her cancer treatment would destroy her fertility and had pleaded with him to find a way to preserve it.)\nHuber remembers seeing Ribeiro present the case at a conference at the time. She immediately realized that her own patient was a candidate for the surgery, and that, as a surgeon who had performed many hysterectomies, she’d be able to do it herself. The patient agreed.\nHuber’s colleagues at the hospital were nervous, she says. They’d never heard of the procedure before. “When I presented this idea to the general surgeon, he didn’t sleep for three days,” she tells me. After watching videos from Ribeiro’s team, however, he was convinced it was doable.\nSo before the patient’s cancer treatment was started, Huber and her colleagues performed the operation. The team literally stitched the organs to the abdominal wall. “It’s a delicate dissection,” says Huber, but she adds that “it’s not the most difficult procedure.” The surgery took two to three hours, she says. The stitches themselves were removed via small incisions around a week later. By that point, scar tissue had formed to create a lasting attachment.\nThe woman had two weeks to recover from the surgery before her cancer treatment began. That too was a success—within months, her tumor had shrunk so significantly that it couldn’t be seen on medical scans.\nAs a precaution, the medical team surgically removed the affected area of her colon. At the same time, they cut away the scar tissue holding the uterus, tubes, and ovaries in their new position and transferred the organs back into the pelvis.\nAround eight months later, the woman stopped taking contraception. She got pregnant without IVF and had a mostly healthy pregnancy, says Huber. Around seven months into the pregnancy, there were signs that the fetus was not growing as expected. This might have been due to problems with the blood supply to the placenta, says Huber. Still, the baby was born healthy, she says.\nRibeiro says he has performed the surgery 16 times, and that teams in countries including the US, Peru, Israel, India, and Russia have performed it as well. Not every case has been published, but he thinks there may be around 40.\nSince Baby Lucien was born last year, a sixth birth has been announced in Israel, says Huber. Ribeiro says he has heard of another two births since then, too. The most recent was to the first woman who had the procedure. She had a little girl a few months ago, he tells me.\nNo surgery is risk-free, and Huber points out there’s a chance that organs could be damaged during the procedure, or that a more developed cancer could spread. The uterus of one of Ribeiro’s patients failed following the surgery. Doctors are “still in the phase of collecting data to [create] a standardized procedure,” Huber says, but she hopes the surgery will offer more options to young people with some pelvic cancers. “I hope more young women could benefit from this procedure,” she says.\nRibeiro says the experience has taught him not to accept the status quo. “Everyone was saying … there was nothing to be done [about the loss of fertility in these cases],” he tells me. “We need to keep evolving and looking for different answers.”\nThis article first appeared in The Checkup, MIT Technology Review’s weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here." }, "pairedItem": [ { "item": 2 } ] }, { "json": { "guid": "https://www.technologyreview.com/?p=1132200", "link": "https://www.technologyreview.com/2026/02/05/1132200/consolidating-systems-for-ai-with-ipaas/", "title": "Consolidating systems for AI with iPaaS", "content": "For decades, enterprises reacted to shifting business pressures with stopgap technology solutions. To rein in rising infrastructure costs, they adopted cloud services that could scale on demand. When customers shifted their lives onto smartphones, companies rolled out mobile apps to keep pace. And when businesses began needing real-time visibility into factories and stockrooms, they layered…", "creator": "MIT Technology Review Insights", "isoDate": "2026-02-05T15:20:37.000Z", "pubDate": "Thu, 05 Feb 2026 15:20:37 +0000", "categories": [ "Computing", "sponsored" ], "dc:creator": "MIT Technology Review Insights", "contentSnippet": "For decades, enterprises reacted to shifting business pressures with stopgap technology solutions. To rein in rising infrastructure costs, they adopted cloud services that could scale on demand. When customers shifted their lives onto smartphones, companies rolled out mobile apps to keep pace. And when businesses began needing real-time visibility into factories and stockrooms, they layered…", "content:encoded": "\nFor decades, enterprises reacted to shifting business pressures with stopgap technology solutions. To rein in rising infrastructure costs, they adopted cloud services that could scale on demand. When customers shifted their lives onto smartphones, companies rolled out mobile apps to keep pace. And when businesses began needing real-time visibility into factories and stockrooms, they layered on IoT systems to supply those insights.
\n\n\n\n![\"\"](\"https://wp.technologyreview.com/wp-content/uploads/2026/02/MIT_SAP_V4-COVERJan302026.png?w=1555\")
Each new plug-in or platform promised better, more efficient operations. And individually, many delivered. But as more and more solutions stacked up, IT teams had to string together a tangled web to connect them—less an IT ecosystem and more of a make-do collection of ad-hoc workarounds.
\n\n\n\n\n\n\n\nThat reality has led to bottlenecks and maintenance burdens, and the impact is showing up in performance. Today, fewer than half of CIOs (48%) say their current digital initiatives are meeting or exceeding business outcome targets. Another 2025 survey found that operations leaders point to integration complexity and data quality issues as top culprits for why investments haven’t delivered as expected.
\n\n\n\nAchim Kraiss, chief product officer of SAP Integration Suite, elaborates on the wide-ranging problems inherent in patchwork IT: “A fragmented landscape makes it difficult to see and control end-to-end business processes,” he explains. “Monitoring, troubleshooting, and governance all suffer. Costs go up because of all the complex mappings and multi-application connectivity you have to maintain.”
\n\n\n\n![\"\"](\"https://wp.technologyreview.com/wp-content/uploads/2026/02/MITTR-SAP-Socials_AchimKraissQuote.png\")
These challenges take on new significance as enterprises look to adopt AI. As AI becomes embedded in everyday workflows, systems are suddenly expected to move far larger volumes of data, at higher speeds, and with tighter coordination than yesterday’s architectures were built
to sustain.
As companies now prepare for an AI-powered future, whether that is generative AI, machine learning, or agentic AI, many are realizing that the way data moves through their business matters just as much as the insights it generates. As a result, organizations are moving away from scattered integration tools and toward consolidated, end-to-end platforms that restore order and streamline how systems interact.
\n\n\n\n\n\n\n\nThis content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. It was researched, designed, and written by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review.
\n", "content:encodedSnippet": "For decades, enterprises reacted to shifting business pressures with stopgap technology solutions. To rein in rising infrastructure costs, they adopted cloud services that could scale on demand. When customers shifted their lives onto smartphones, companies rolled out mobile apps to keep pace. And when businesses began needing real-time visibility into factories and stockrooms, they layered on IoT systems to supply those insights.\n\n\n\n\nEach new plug-in or platform promised better, more efficient operations. And individually, many delivered. But as more and more solutions stacked up, IT teams had to string together a tangled web to connect them—less an IT ecosystem and more of a make-do collection of ad-hoc workarounds.\nDOWNLOAD THE REPORT\nThat reality has led to bottlenecks and maintenance burdens, and the impact is showing up in performance. Today, fewer than half of CIOs (48%) say their current digital initiatives are meeting or exceeding business outcome targets. Another 2025 survey found that operations leaders point to integration complexity and data quality issues as top culprits for why investments haven’t delivered as expected.\nAchim Kraiss, chief product officer of SAP Integration Suite, elaborates on the wide-ranging problems inherent in patchwork IT: “A fragmented landscape makes it difficult to see and control end-to-end business processes,” he explains. “Monitoring, troubleshooting, and governance all suffer. Costs go up because of all the complex mappings and multi-application connectivity you have to maintain.”\n\n\n\n\nThese challenges take on new significance as enterprises look to adopt AI. As AI becomes embedded in everyday workflows, systems are suddenly expected to move far larger volumes of data, at higher speeds, and with tighter coordination than yesterday’s architectures were built\nto sustain.\nAs companies now prepare for an AI-powered future, whether that is generative AI, machine learning, or agentic AI, many are realizing that the way data moves through their business matters just as much as the insights it generates. As a result, organizations are moving away from scattered integration tools and toward consolidated, end-to-end platforms that restore order and streamline how systems interact.\nDownload the report.\nThis content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff. It was researched, designed, and written by human writers, editors, analysts, and illustrators. This includes the writing of surveys and collection of data for surveys. AI tools that may have been used were limited to secondary production processes that passed thorough human review." }, "pairedItem": [ { "item": 2 } ] }, { "json": { "guid": "https://www.technologyreview.com/?p=1132270", "link": "https://www.technologyreview.com/2026/02/05/1132270/the-download-attempting-to-track-ai-and-the-next-generation-of-nuclear-power/", "title": "The Download: attempting to track AI, and the next generation of nuclear power", "content": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. This is the most misunderstood graph in AI Every time OpenAI, Google, or Anthropic drops a new frontier large language model, the AI community holds its breath. It doesn’t exhale until METR, an…", "creator": "Rhiannon Williams", "isoDate": "2026-02-05T13:10:00.000Z", "pubDate": "Thu, 05 Feb 2026 13:10:00 +0000", "categories": [ "The Download" ], "dc:creator": "Rhiannon Williams", "contentSnippet": "This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. This is the most misunderstood graph in AI Every time OpenAI, Google, or Anthropic drops a new frontier large language model, the AI community holds its breath. It doesn’t exhale until METR, an…", "content:encoded": "\nThis is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
This is the most misunderstood graph in AI
\n\n\n\nEvery time OpenAI, Google, or Anthropic drops a new frontier large language model, the AI community holds its breath. It doesn’t exhale until METR, an AI research nonprofit whose name stands for “Model Evaluation & Threat Research,” updates a now-iconic graph that has played a major role in the AI discourse since it was first released in March of last year.
\n\n\n\nThe graph suggests that certain AI capabilities are developing at an exponential rate, and more recent model releases have outperformed that already impressive trend.
That was certainly the case for Claude Opus 4.5, the latest version of Anthropic’s most powerful model, which was released in late November. In December, METR announced that Opus 4.5 appeared to be capable of independently completing a task that would have taken a human about five hours—a vast improvement over what even the exponential trend would have predicted.
But the truth is more complicated than those dramatic responses would suggest. Read the full story.
\n\n\n\n—Grace Huckins
\n\n\n\nThis story is part of MIT Technology Review Explains: our series untangling the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here.
\n\n\n\n\n\n\n\nThree questions about next-generation nuclear power, answered
\n\n\n\nNuclear power continues to be one of the hottest topics in energy today, and in our recent online Roundtables discussion about next-generation nuclear power, hyperscale AI data centers, and the grid, we got dozens of great audience questions.
\n\n\n\nThese ran the gamut, and while we answered quite a few (and I’m keeping some in mind for future reporting), there were a bunch we couldn’t get to, at least not in the depth I would have liked. So let’s answer a few of your questions about advanced nuclear power.
\n\n\n\n—Casey Crownhart
\n\n\n\nThis article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.
\n\n\n\n\n\n\n\nThe must-reads
\n\n\n\nI’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
\n\n\n\n1 Anthropic’s new coding tools are rattling the markets
Fields as diverse as publishing and coding to law and advertising are paying attention. (FT $)
+ Legacy software companies, beware. (Insider $)
+ Is “software-mageddon” nigh? It depends who you ask. (Reuters)
2 This Apple setting prevented the FBI from accessing a reporter’s iPhone
Lockdown Mode has proved remarkably effective—for now. (404 Media)
+ Agents were able to access Hannah Natanson’s laptop, however. (Ars Technica)
3 Last month’s data center outage disrupted all TikTok categories
Not just the political content that some users claimed. (NPR)
4 Big Tech is pouring billions into AI in India
A newly-announced 20-year tax break should help to speed things along. (WSJ $)
+ India’s female content moderators are watching hours of abuse content to train AI. (The Guardian)
+ Officials in the country are weighing up restricting social media for minors. (Bloomberg $)
+ Inside India’s scramble for AI independence. (MIT Technology Review)
5 YouTubers are harassing women using body cams
They’re abusing freedom of information laws to humiliate their targets. (NY Mag $)
+ AI was supposed to make police bodycams better. What happened? (MIT Technology Review)
6 Jokers have created a working version of Jeffrey Epstein’s inbox
Complete with notable starred threads. (Wired $)
+ Epstein’s links with Silicon Valley are vast and deep. (Fast Company $)
+ The revelations are driving rifts between previously-friendly factions. (NBC News)
7 What’s the last thing you see before you die?
A new model might help to explain near-death experiences—but not all researchers are on board. (WP $)
+ What is death? (MIT Technology Review)
8 A new app is essentially TikTok for vibe-coded apps
Words which would have made no sense 15 years ago. (TechCrunch)
+ What is vibe coding, exactly? (MIT Technology Review)
9 Rogue TV boxes are all the rage
Viewers are sick of the soaring prices of streaming services, and are embracing less legal means of watching their favorite shows. (The Verge)
10 Climate change is threatening the future of the Winter Olympics ![\"⛷\"](\"https://s.w.org/images/core/emoji/16.0.1/72x72/26f7.png\")
Artificial snow is one (short term) solution. (Bloomberg $)
+ Team USA is using AI to try and gain an edge on its competition. (NBC News)
Quote of the day
\n\n\n\n“We’ve heard from many who want nothing to do with AI.”
\n\n\n\n—Ajit Varma, head of Mozilla’s web browser Firefox, explains why the company is reversing its previous decision to transform Firefox into an “AI browser,” PC Gamer reports.
\n\n\n\n\n\n\n\nOne more thing
\n\n\n\n![\"\"](\"https://wp.technologyreview.com/wp-content/uploads/2026/02/image_37281d.png\")
A major AI training data set contains millions of examples of personal data
Millions of images of passports, credit cards, birth certificates, and other documents containing personally identifiable information are likely included in one of the biggest open-source AI training sets, new research has found.
Thousands of images—including identifiable faces—were found in a small subset of DataComp CommonPool, a major AI training set for image generation scraped from the web. Because the researchers audited just 0.1% of CommonPool’s data, they estimate that the real number of images containing personally identifiable information, including faces and identity documents, is in the hundreds of millions.
The bottom line? Anything you put online can be and probably has been scraped. Read the full story.
—Eileen Guo
We can still have nice things
\n\n\n\nA place for comfort, fun and distraction to brighten up your day. (Got any ideas? Drop me a line or skeet ’em at me.)
\n\n\n\n+ If you’re crazy enough to be training for a marathon right now, here’s how to beat boredom on those long, long runs.
+ Mark Cohen’s intimate street photography is a fascinating window into humanity.
+ A seriously dedicated gamer has spent days painstakingly recreating a Fallout vault inside the Sims 4.
+ Here’s what music’s most stylish men are wearing right now—from leather pants to khaki parkas.
Nuclear power continues to be one of the hottest topics in energy today, and in our recent online Roundtables discussion about next-generation nuclear power, hyperscale AI data centers, and the grid, we got dozens of great audience questions.
\n\n\n\nThese ran the gamut, and while we answered quite a few (and I’m keeping some in mind for future reporting), there were a bunch we couldn’t get to, at least not in the depth I would have liked.
\n\n\n\nSo let’s answer a few of your questions about advanced nuclear power. I’ve combined similar ones and edited them for clarity.
\n\n\n\nMany next-generation reactors don’t use the low-enriched uranium used in conventional reactors.
\n\n\n\nIt’s worth looking at high-assay low-enriched uranium, or HALEU, specifically. This fuel is enriched to higher concentrations of fissile uranium than conventional nuclear fuel, with a proportion of the isotope U-235 that falls between 5% and 20%. (In conventional fuel, it’s below 5%.)
\n\n\n\nHALEU can be produced with the same technology as low-enriched uranium, but the geopolitics are complicated. Today, Russia basically has a monopoly on HALEU production. In 2024, the US banned the import of Russian nuclear fuel through 2040 in an effort to reduce dependence on the country. Europe hasn’t taken the same measures, but it is working to move away from Russian energy as well.
\n\n\n\nThat leaves companies in the US and Europe with the major challenge of securing the fuel they need when their regular Russian supply has been cut off or restricted.
\n\n\n\nThe US Department of Energy has a stockpile of HALEU, which the government is doling out to companies to help power demonstration reactions. In the longer term, though, there’s still a major need to set up independent HALEU supply chains to support next-generation reactors.
\n\n\n\nThere are some ways that next-generation nuclear power plants could be safer than conventional reactors. Some use alternative coolants that would prevent the need to run at the high pressure required in conventional water-cooled reactors. Many incorporate passive safety shutoffs, so if there are power supply issues, the reactors shut down harmlessly, avoiding risk of meltdown. (These can be incorporated in newer conventional reactors, too.)
\n\n\n\nBut some experts have raised concerns that in the US, the current administration isn’t taking nuclear safety seriously enough.
\n\n\n\nA recent NPR investigation found that the Trump administration had secretly rewritten nuclear rules, stripping environmental protections and loosening safety and security measures. The government shared the new rules with companies that are part of a program building experimental nuclear reactors, but not with the public.
\n\n\n\nI’m reminded of a talk during our EmTech MIT event in November, where Koroush Shirvan, an MIT professor of nuclear engineering, spoke on this issue. “I’ve seen some disturbing trends in recent times, where words like ‘rubber-stamping nuclear projects’ are being said,” Shirvan said during that event.
\n\n\n\nDuring the talk, Shirvan shared statistics showing that nuclear power has a very low rate of injury and death. But that’s not inherent to the technology, and there’s a reason injuries and deaths have been low for nuclear power, he added: “It’s because of stringent regulatory oversight.”
\n\n\n\nBuilding a nuclear power plant is not cheap. Let’s consider the up-front investment needed to build a power plant.
\n\n\n\nPlant Vogtle in Georgia hosts the most recent additions to the US nuclear fleet—Units 3 and 4 came online in 2023 and 2024. Together, they had a capital cost of $15,000 per kilowatt, adjusted for inflation, according to a recent report from the US Department of Energy. (This wonky unit I’m using divides the total cost to build the reactors by their expected power output, so we can compare reactors of different sizes.)
\n\n\n\nThat number’s quite high, partly because those were the first of their kind built in the US, and because there were some inefficiencies in the planning. It’s worth noting that China builds reactors for much less, somewhere between $2,000/kW and $3,000/kW, depending on the estimate.
\n\n\n\nThe up-front capital cost for first-of-a-kind advanced nuclear plants will likely run between $6,000 and $10,000 per kilowatt, according to that DOE report. That could come down by up to 40% after the technologies are scaled up and mass-produced.
\n\n\n\nSo new reactors will (hopefully) be cheaper than the ultra-over-budget and behind-schedule Vogtle project, but they aren’t necessarily significantly cheaper than efficiently built conventional plants, if you normalize by their size.
\n\n\n\nIt’ll certainly be cheaper to build new natural-gas plants (setting aside the likely equipment shortages we’re likely going to see for years.) Today’s most efficient natural-gas plants cost just $1,600/kW on the high end, according to data from Lazard.
\n\n\n\nAn important caveat: Capital cost isn’t everything—running a nuclear plant is relatively inexpensive, which is why there’s so much interest in extending the lifetime of existing plants or reopening shuttered ones.
\n\n\n\nUltimately, by many metrics, nuclear plants of any type are going to be more expensive than other sources, like wind and solar power. But they provide something many other power sources don’t: a reliable, stable source of electricity that can run for 60 years or more.
\n\n\n\nThis article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here.
\n", "content:encodedSnippet": "Nuclear power continues to be one of the hottest topics in energy today, and in our recent online Roundtables discussion about next-generation nuclear power, hyperscale AI data centers, and the grid, we got dozens of great audience questions.\nThese ran the gamut, and while we answered quite a few (and I’m keeping some in mind for future reporting), there were a bunch we couldn’t get to, at least not in the depth I would have liked.\nSo let’s answer a few of your questions about advanced nuclear power. I’ve combined similar ones and edited them for clarity.\nHow are the fuel needs for next-generation nuclear reactors different, and how are companies addressing the supply chain?\nMany next-generation reactors don’t use the low-enriched uranium used in conventional reactors.\nIt’s worth looking at high-assay low-enriched uranium, or HALEU, specifically. This fuel is enriched to higher concentrations of fissile uranium than conventional nuclear fuel, with a proportion of the isotope U-235 that falls between 5% and 20%. (In conventional fuel, it’s below 5%.)\nHALEU can be produced with the same technology as low-enriched uranium, but the geopolitics are complicated. Today, Russia basically has a monopoly on HALEU production. In 2024, the US banned the import of Russian nuclear fuel through 2040 in an effort to reduce dependence on the country. Europe hasn’t taken the same measures, but it is working to move away from Russian energy as well.\nThat leaves companies in the US and Europe with the major challenge of securing the fuel they need when their regular Russian supply has been cut off or restricted.\nThe US Department of Energy has a stockpile of HALEU, which the government is doling out to companies to help power demonstration reactions. In the longer term, though, there’s still a major need to set up independent HALEU supply chains to support next-generation reactors.\nHow is safety being addressed, and what’s happening with nuclear safety regulation in the US?\nThere are some ways that next-generation nuclear power plants could be safer than conventional reactors. Some use alternative coolants that would prevent the need to run at the high pressure required in conventional water-cooled reactors. Many incorporate passive safety shutoffs, so if there are power supply issues, the reactors shut down harmlessly, avoiding risk of meltdown. (These can be incorporated in newer conventional reactors, too.)\nBut some experts have raised concerns that in the US, the current administration isn’t taking nuclear safety seriously enough.\nA recent NPR investigation found that the Trump administration had secretly rewritten nuclear rules, stripping environmental protections and loosening safety and security measures. The government shared the new rules with companies that are part of a program building experimental nuclear reactors, but not with the public.\nI’m reminded of a talk during our EmTech MIT event in November, where Koroush Shirvan, an MIT professor of nuclear engineering, spoke on this issue. “I’ve seen some disturbing trends in recent times, where words like ‘rubber-stamping nuclear projects’ are being said,” Shirvan said during that event. \nDuring the talk, Shirvan shared statistics showing that nuclear power has a very low rate of injury and death. But that’s not inherent to the technology, and there’s a reason injuries and deaths have been low for nuclear power, he added: “It’s because of stringent regulatory oversight.” \nAre next-generation reactors going to be financially competitive?\nBuilding a nuclear power plant is not cheap. Let’s consider the up-front investment needed to build a power plant. \nPlant Vogtle in Georgia hosts the most recent additions to the US nuclear fleet—Units 3 and 4 came online in 2023 and 2024. Together, they had a capital cost of $15,000 per kilowatt, adjusted for inflation, according to a recent report from the US Department of Energy. (This wonky unit I’m using divides the total cost to build the reactors by their expected power output, so we can compare reactors of different sizes.)\nThat number’s quite high, partly because those were the first of their kind built in the US, and because there were some inefficiencies in the planning. It’s worth noting that China builds reactors for much less, somewhere between $2,000/kW and $3,000/kW, depending on the estimate.\nThe up-front capital cost for first-of-a-kind advanced nuclear plants will likely run between $6,000 and $10,000 per kilowatt, according to that DOE report. That could come down by up to 40% after the technologies are scaled up and mass-produced.\nSo new reactors will (hopefully) be cheaper than the ultra-over-budget and behind-schedule Vogtle project, but they aren’t necessarily significantly cheaper than efficiently built conventional plants, if you normalize by their size.\nIt’ll certainly be cheaper to build new natural-gas plants (setting aside the likely equipment shortages we’re likely going to see for years.) Today’s most efficient natural-gas plants cost just $1,600/kW on the high end, according to data from Lazard.\nAn important caveat: Capital cost isn’t everything—running a nuclear plant is relatively inexpensive, which is why there’s so much interest in extending the lifetime of existing plants or reopening shuttered ones.\nUltimately, by many metrics, nuclear plants of any type are going to be more expensive than other sources, like wind and solar power. But they provide something many other power sources don’t: a reliable, stable source of electricity that can run for 60 years or more.\nThis article is from The Spark, MIT Technology Review’s weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here." }, "pairedItem": [ { "item": 2 } ] }, { "json": { "guid": "https://www.technologyreview.com/?p=1132254", "link": "https://www.technologyreview.com/2026/02/05/1132254/this-is-the-most-misunderstood-graph-in-ai/", "title": "This is the most misunderstood graph in AI", "content": "MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here. Every time OpenAI, Google, or Anthropic drops a new frontier large language model, the AI community holds its breath. It doesn’t exhale until METR, an AI…", "creator": "Grace Huckins", "isoDate": "2026-02-05T10:00:00.000Z", "pubDate": "Thu, 05 Feb 2026 10:00:00 +0000", "categories": [ "Artificial intelligence", "App", "artificial intelligence", "MIT Technology Review Explains" ], "dc:creator": "Grace Huckins", "contentSnippet": "MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here. Every time OpenAI, Google, or Anthropic drops a new frontier large language model, the AI community holds its breath. It doesn’t exhale until METR, an AI…", "content:encoded": "\nMIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here.
\n\n\n\nEvery time OpenAI, Google, or Anthropic drops a new frontier large language model, the AI community holds its breath. It doesn’t exhale until METR, an AI research nonprofit whose name stands for “Model Evaluation & Threat Research,” updates a now-iconic graph that has played a major role in the AI discourse since it was first released in March of last year. The graph suggests that certain AI capabilities are developing at an exponential rate, and more recent model releases have outperformed that already impressive trend.
\n\n\n\nThat was certainly the case for Claude Opus 4.5, the latest version of Anthropic’s most powerful model, which was released in late November. In December, METR announced that Opus 4.5 appeared to be capable of independently completing a task that would have taken a human about five hours—a vast improvement over what even the exponential trend would have predicted. One Anthropic safety researcher tweeted that he would change the direction of his research in light of those results; another employee at the company simply wrote, “mom come pick me up i’m scared.”
\n\n\n\n\n\n\n\nBut the truth is more complicated than those dramatic responses would suggest. For one thing, METR’s estimates of the abilities of specific models come with substantial error bars. As METR explicitly stated on X, Opus 4.5 might be able to regularly complete only tasks that take humans about two hours, or it might succeed on tasks that take humans as long as 20 hours. Given the uncertainties intrinsic to the method, it was impossible to know for sure.
\n\n\n\n“There are a bunch of ways that people are reading too much into the graph,” says Sydney Von Arx, a member of METR’s technical staff.
\n\n\n\nMore fundamentally, the METR plot does not measure AI abilities writ large, nor does it claim to. In order to build the graph, METR tests the models primarily on coding tasks, evaluating the difficulty of each by measuring or estimating how long it takes humans to complete it—a metric that not everyone accepts. Claude Opus 4.5 might be able to complete certain tasks that take humans five hours, but that doesn’t mean it’s anywhere close to replacing a human worker.
\n\n\n\nMETR was founded to assess the risks posed by frontier AI systems. Though it is best known for the exponential trend plot, it has also worked with AI companies to evaluate their systems in greater detail and published several other independent research projects, including a widely covered July 2025 study suggesting that AI coding assistants might actually be slowing software engineers down.
\n\n\n\nBut the exponential plot has made METR’s reputation, and the organization appears to have a complicated relationship with that graph’s often breathless reception. In January, Thomas Kwa, one of the lead authors on the paper that introduced it, wrote a blog post responding to some criticisms and making clear its limitations, and METR is currently working on a more extensive FAQ document. But Kwa isn’t optimistic that these efforts will meaningfully shift the discourse. “I think the hype machine will basically, whatever we do, just strip out all the caveats,” he says.
\n\n\n\nNevertheless, the METR team does think that the plot has something meaningful to say about the trajectory of AI progress. “You should absolutely not tie your life to this graph,” says Von Arx. “But also,” she adds, “I bet that this trend is gonna hold.”
\n\n\n\n\n\nPart of the trouble with the METR plot is that it’s quite a bit more complicated than it looks. The x-axis is simple enough: It tracks the date when each model was released. But the y-axis is where things get tricky. It records each model’s “time horizon,” an unusual metric that METR created—and that, according to Kwa and Von Arx, is frequently misunderstood.
\n\n\n\nTo understand exactly what model time horizons are, it helps to know all the work that METR put into calculating them. First, the METR team assembled a collection of tasks ranging from quick multiple-choice questions to detailed coding challenges—all of which were somehow relevant to software engineering. Then they had human coders attempt most of those tasks and evaluated how long it took them to finish. In this way, they assigned the tasks a human baseline time. Some tasks took the experts mere seconds, whereas others required several hours.
\n\n\n\nWhen METR tested large language models on the task suite, they found that advanced models could complete the fast tasks with ease—but as the models attempted tasks that had taken humans more and more time to finish, their accuracy started to fall off. From a model’s performance, the researchers calculated the point on the time scale of human tasks at which the model would complete about 50% of the tasks successfully. That point is the model’s time horizon.
\n\n\n\nAll that detail is in the blog post and the academic paper that METR released along with the original time horizon plot. But the METR plot is frequently passed around on social media without this context, and so the true meaning of the time horizon metric can get lost in the shuffle. One common misapprehension is that the numbers on the plot’s y-axis—around five hours for Claude Opus 4.5, for example—represent the length of time that the models can operate independently. They do not. They represent how long it takes humans to complete tasks that a model can successfully perform. Kwa has seen this error so frequently that he made a point of correcting it at the very top of his recent blog post, and when asked what information he would add to the versions of the plot circulating online, he said he would include the word “human” whenever the task completion time was mentioned.
\n\n\n\nAs complex and widely misinterpreted as the time horizon concept might be, it does make some basic sense: A model with a one-hour time horizon could automate some modest portions of a software engineer’s job, whereas a model with a 40-hour horizon could potentially complete days of work on its own. But some experts question whether the amount of time that humans take on tasks is an effective metric for quantifying AI capabilities. “I don’t think it’s necessarily a given fact that because something takes longer, it’s going to be a harder task,” says Inioluwa Deborah Raji, a PhD student at UC Berkeley who studies model evaluation.
\n\n\n\n\n\nVon Arx says that she, too, was originally skeptical that time horizon was the right measure to use. What convinced her was seeing the results of her and her colleagues’ analysis. When they calculated the 50% time horizon for all the major models available in early 2025 and then plotted each of them on the graph, they saw that the time horizons for the top-tier models were increasing over time—and, moreover, that the rate of advancement was speeding up. Every seven-ish months, the time horizon doubled, which means that the most advanced models could complete tasks that took humans nine seconds in mid 2020, 4 minutes in early 2023, and 40 minutes in late 2024. “I can do all the theorizing I want about whether or not it makes sense, but the trend is there,” Von Arx says.
\n\n\n\nIt’s this dramatic pattern that made the METR plot such a blockbuster. Many people learned about it when they read AI 2027, a viral sci-fi story cum quantitative forecast positing that superintelligent AI could wipe out humanity by 2030. The writers of AI 2027 based some of their predictions on the METR plot and cited it extensively. In Von Arx’s words, “It’s a little weird when the way lots of people are familiar with your work is this pretty opinionated interpretation.”
\n\n\n\nOf course, plenty of people invoke the METR plot without imagining large-scale death and destruction. For some AI boosters, the exponential trend indicates that AI will soon usher in an era of radical economic growth. The venture capital firm Sequoia Capital, for example, recently put out a post titled “2026: This is AGI,” which used the METR plot to argue that AI that can act as an employee or contractor will soon arrive. “The provocation really was like, ‘What will you do when your plans are measured in centuries?’” says Sonya Huang, a general partner at Sequoia and one of the post’s authors.
\n\n\n\nJust because a model achieves a one-hour time horizon on the METR plot, however, doesn’t mean that it can replace one hour of human work in the real world. For one thing, the tasks on which the models are evaluated don’t reflect the complexities and confusion of real-world work. In their original study, Kwa, Von Arx, and their colleagues quantify what they call the “messiness” of each task according to criteria such as whether the model knows exactly how it is being scored and whether it can easily start over if it makes a mistake (for messy tasks, the answer to both questions would be no). They found that models do noticeably worse on messy tasks, although the overall pattern of improvement holds for both messy and non-messy ones.
\n\n\n\nAnd even the messiest tasks that METR considered can’t provide much information about AI’s ability to take on most jobs, because the plot is based almost entirely on coding tasks. “A model can get better at coding, but it’s not going to magically get better at anything else,” says Daniel Kang, an assistant professor of computer science at the University of Illinois Urbana-Champaign. In a follow-up study, Kwa and his colleagues did find that time horizons for tasks in other domains also appear to be on exponential trajectories, but that work was much less formal.
\n\n\n\nDespite these limitations, many people admire the group’s research. “The METR study is one of the most carefully designed studies in the literature for this kind of work,” Kang told me. Even Gary Marcus, a former NYU professor and professional LLM curmudgeon, described much of the work that went into the plot as “terrific” in a blog post.
\n\n\n\nSome people will almost certainly continue to read the METR plot as a prognostication of our AI-induced doom, but in reality it’s something far more banal: a carefully constructed scientific tool that puts concrete numbers to people’s intuitive sense of AI progress. As METR employees will readily agree, the plot is far from a perfect instrument. But in a new and fast-moving domain, even imperfect tools can have enormous value.
\n\n\n\n“This is a bunch of people trying their best to make a metric under a lot of constraints. It is deeply flawed in many ways,” Von Arx says. “I also think that it is one of the best things of its kind.”
\n", "content:encodedSnippet": "MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here.\nEvery time OpenAI, Google, or Anthropic drops a new frontier large language model, the AI community holds its breath. It doesn’t exhale until METR, an AI research nonprofit whose name stands for “Model Evaluation & Threat Research,” updates a now-iconic graph that has played a major role in the AI discourse since it was first released in March of last year. The graph suggests that certain AI capabilities are developing at an exponential rate, and more recent model releases have outperformed that already impressive trend.\nThat was certainly the case for Claude Opus 4.5, the latest version of Anthropic’s most powerful model, which was released in late November. In December, METR announced that Opus 4.5 appeared to be capable of independently completing a task that would have taken a human about five hours—a vast improvement over what even the exponential trend would have predicted. One Anthropic safety researcher tweeted that he would change the direction of his research in light of those results; another employee at the company simply wrote, “mom come pick me up i’m scared.”\n\n\n\n\nBut the truth is more complicated than those dramatic responses would suggest. For one thing, METR’s estimates of the abilities of specific models come with substantial error bars. As METR explicitly stated on X, Opus 4.5 might be able to regularly complete only tasks that take humans about two hours, or it might succeed on tasks that take humans as long as 20 hours. Given the uncertainties intrinsic to the method, it was impossible to know for sure. \n“There are a bunch of ways that people are reading too much into the graph,” says Sydney Von Arx, a member of METR’s technical staff.\nMore fundamentally, the METR plot does not measure AI abilities writ large, nor does it claim to. In order to build the graph, METR tests the models primarily on coding tasks, evaluating the difficulty of each by measuring or estimating how long it takes humans to complete it—a metric that not everyone accepts. Claude Opus 4.5 might be able to complete certain tasks that take humans five hours, but that doesn’t mean it’s anywhere close to replacing a human worker.\nMETR was founded to assess the risks posed by frontier AI systems. Though it is best known for the exponential trend plot, it has also worked with AI companies to evaluate their systems in greater detail and published several other independent research projects, including a widely covered July 2025 study suggesting that AI coding assistants might actually be slowing software engineers down. \nBut the exponential plot has made METR’s reputation, and the organization appears to have a complicated relationship with that graph’s often breathless reception. In January, Thomas Kwa, one of the lead authors on the paper that introduced it, wrote a blog post responding to some criticisms and making clear its limitations, and METR is currently working on a more extensive FAQ document. But Kwa isn’t optimistic that these efforts will meaningfully shift the discourse. “I think the hype machine will basically, whatever we do, just strip out all the caveats,” he says.\nNevertheless, the METR team does think that the plot has something meaningful to say about the trajectory of AI progress. “You should absolutely not tie your life to this graph,” says Von Arx. “But also,” she adds, “I bet that this trend is gonna hold.”\nPart of the trouble with the METR plot is that it’s quite a bit more complicated than it looks. The x-axis is simple enough: It tracks the date when each model was released. But the y-axis is where things get tricky. It records each model’s “time horizon,” an unusual metric that METR created—and that, according to Kwa and Von Arx, is frequently misunderstood.\nTo understand exactly what model time horizons are, it helps to know all the work that METR put into calculating them. First, the METR team assembled a collection of tasks ranging from quick multiple-choice questions to detailed coding challenges—all of which were somehow relevant to software engineering. Then they had human coders attempt most of those tasks and evaluated how long it took them to finish. In this way, they assigned the tasks a human baseline time. Some tasks took the experts mere seconds, whereas others required several hours.\nWhen METR tested large language models on the task suite, they found that advanced models could complete the fast tasks with ease—but as the models attempted tasks that had taken humans more and more time to finish, their accuracy started to fall off. From a model’s performance, the researchers calculated the point on the time scale of human tasks at which the model would complete about 50% of the tasks successfully. That point is the model’s time horizon. \nAll that detail is in the blog post and the academic paper that METR released along with the original time horizon plot. But the METR plot is frequently passed around on social media without this context, and so the true meaning of the time horizon metric can get lost in the shuffle. One common misapprehension is that the numbers on the plot’s y-axis—around five hours for Claude Opus 4.5, for example—represent the length of time that the models can operate independently. They do not. They represent how long it takes humans to complete tasks that a model can successfully perform. Kwa has seen this error so frequently that he made a point of correcting it at the very top of his recent blog post, and when asked what information he would add to the versions of the plot circulating online, he said he would include the word “human” whenever the task completion time was mentioned.\nAs complex and widely misinterpreted as the time horizon concept might be, it does make some basic sense: A model with a one-hour time horizon could automate some modest portions of a software engineer’s job, whereas a model with a 40-hour horizon could potentially complete days of work on its own. But some experts question whether the amount of time that humans take on tasks is an effective metric for quantifying AI capabilities. “I don’t think it’s necessarily a given fact that because something takes longer, it’s going to be a harder task,” says Inioluwa Deborah Raji, a PhD student at UC Berkeley who studies model evaluation. \nVon Arx says that she, too, was originally skeptical that time horizon was the right measure to use. What convinced her was seeing the results of her and her colleagues’ analysis. When they calculated the 50% time horizon for all the major models available in early 2025 and then plotted each of them on the graph, they saw that the time horizons for the top-tier models were increasing over time—and, moreover, that the rate of advancement was speeding up. Every seven-ish months, the time horizon doubled, which means that the most advanced models could complete tasks that took humans nine seconds in mid 2020, 4 minutes in early 2023, and 40 minutes in late 2024. “I can do all the theorizing I want about whether or not it makes sense, but the trend is there,” Von Arx says.\nIt’s this dramatic pattern that made the METR plot such a blockbuster. Many people learned about it when they read AI 2027, a viral sci-fi story cum quantitative forecast positing that superintelligent AI could wipe out humanity by 2030. The writers of AI 2027 based some of their predictions on the METR plot and cited it extensively. In Von Arx’s words, “It’s a little weird when the way lots of people are familiar with your work is this pretty opinionated interpretation.”\nOf course, plenty of people invoke the METR plot without imagining large-scale death and destruction. For some AI boosters, the exponential trend indicates that AI will soon usher in an era of radical economic growth. The venture capital firm Sequoia Capital, for example, recently put out a post titled “2026: This is AGI,” which used the METR plot to argue that AI that can act as an employee or contractor will soon arrive. “The provocation really was like, ‘What will you do when your plans are measured in centuries?’” says Sonya Huang, a general partner at Sequoia and one of the post’s authors. \nJust because a model achieves a one-hour time horizon on the METR plot, however, doesn’t mean that it can replace one hour of human work in the real world. For one thing, the tasks on which the models are evaluated don’t reflect the complexities and confusion of real-world work. In their original study, Kwa, Von Arx, and their colleagues quantify what they call the “messiness” of each task according to criteria such as whether the model knows exactly how it is being scored and whether it can easily start over if it makes a mistake (for messy tasks, the answer to both questions would be no). They found that models do noticeably worse on messy tasks, although the overall pattern of improvement holds for both messy and non-messy ones.\nAnd even the messiest tasks that METR considered can’t provide much information about AI’s ability to take on most jobs, because the plot is based almost entirely on coding tasks. “A model can get better at coding, but it’s not going to magically get better at anything else,” says Daniel Kang, an assistant professor of computer science at the University of Illinois Urbana-Champaign. In a follow-up study, Kwa and his colleagues did find that time horizons for tasks in other domains also appear to be on exponential trajectories, but that work was much less formal.\nDespite these limitations, many people admire the group’s research. “The METR study is one of the most carefully designed studies in the literature for this kind of work,” Kang told me. Even Gary Marcus, a former NYU professor and professional LLM curmudgeon, described much of the work that went into the plot as “terrific” in a blog post.\nSome people will almost certainly continue to read the METR plot as a prognostication of our AI-induced doom, but in reality it’s something far more banal: a carefully constructed scientific tool that puts concrete numbers to people’s intuitive sense of AI progress. As METR employees will readily agree, the plot is far from a perfect instrument. But in a new and fast-moving domain, even imperfect tools can have enormous value.\n“This is a bunch of people trying their best to make a metric under a lot of constraints. It is deeply flawed in many ways,” Von Arx says. “I also think that it is one of the best things of its kind.”" }, "pairedItem": [ { "item": 2 } ] }, { "json": { "guid": "1l2gUXhw9OMzMzsSp8yAFk", "link": "https://venturebeat.com/infrastructure/railway-secures-usd100-million-to-challenge-aws-with-ai-native-cloud", "title": "Railway secures $100 million to challenge AWS with AI-native cloud infrastructure", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Railway, a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence applications exposes the limitations of legacy cloud infrastructure.
TQ Ventures led the round, with participation from FPV Ventures, Redpoint, and Unusual Ventures. The investment values Railway as one of the most significant infrastructure startups to emerge during the AI boom, capitalizing on developer frustration with the complexity and cost of traditional platforms like Amazon Web Services and Google Cloud.
"As AI models get better at writing code, more and more people are asking the age-old question: where, and how, do I run my applications?" said Jake Cooper, Railway's 28-year-old founder and chief executive, in an exclusive interview with VentureBeat. "The last generation of cloud primitives were slow and outdated, and now with AI moving everything faster, teams simply can't keep up."
The funding is a dramatic acceleration for a company that has charted an unconventional path through the cloud computing industry. Railway raised just $24 million in total before this round, including a $20 million Series A from Redpoint in 2022. The company now processes more than 10 million deployments monthly and handles over one trillion requests through its edge network — metrics that rival far larger and better-funded competitors.
Railway's pitch rests on a simple observation: the tools developers use to deploy and manage software were designed for a slower era. A standard build-and-deploy cycle using Terraform, the industry-standard infrastructure tool, takes two to three minutes. That delay, once tolerable, has become a critical bottleneck as AI coding assistants like Claude, ChatGPT, and Cursor can generate working code in seconds.
"When godly intelligence is on tap and can solve any problem in three seconds, those amalgamations of systems become bottlenecks," Cooper told VentureBeat. "What was really cool for humans to deploy in 10 seconds or less is now table stakes for agents."
The company claims its platform delivers deployments in under one second — fast enough to keep pace with AI-generated code. Customers report a tenfold increase in developer velocity and up to 65 percent cost savings compared to traditional cloud providers.
These numbers come directly from enterprise clients, not internal benchmarks. Daniel Lobaton, chief technology officer at G2X, a platform serving 100,000 federal contractors, measured deployment speed improvements of seven times faster and an 87 percent cost reduction after migrating to Railway. His infrastructure bill dropped from $15,000 per month to approximately $1,000.
"The work that used to take me a week on our previous infrastructure, I can do in Railway in like a day," Lobaton said. "If I want to spin up a new service and test different architectures, it would take so long on our old setup. In Railway I can launch six services in two minutes."
What distinguishes Railway from competitors like Render and Fly.io is the depth of its vertical integration. In 2024, the company made the unusual decision to abandon Google Cloud entirely and build its own data centers, a move that echoes the famous Alan Kay maxim: "People who are really serious about software should make their own hardware."
"We wanted to design hardware in a way where we could build a differentiated experience," Cooper said. "Having full control over the network, compute, and storage layers lets us do really fast build and deploy loops, the kind that allows us to move at 'agentic speed' while staying 100 percent the smoothest ride in town."
The approach paid dividends during recent widespread outages that affected major cloud providers — Railway remained online throughout.
This soup-to-nuts control enables pricing that undercuts the hyperscalers by roughly 50 percent and newer cloud startups by three to four times. Railway charges by the second for actual compute usage: $0.00000386 per gigabyte-second of memory, $0.00000772 per vCPU-second, and $0.00000006 per gigabyte-second of storage. There are no charges for idle virtual machines — a stark contrast to the traditional cloud model where customers pay for provisioned capacity whether they use it or not.
"The conventional wisdom is that the big guys have economies of scale to offer better pricing," Cooper noted. "But when they're charging for VMs that usually sit idle in the cloud, and we've purpose-built everything to fit much more density on these machines, you have a big opportunity."
Railway has achieved its scale with a team of just 30 employees generating tens of millions in annual revenue — a ratio of revenue per employee that would be exceptional even for established software companies. The company grew revenue 3.5 times last year and continues to expand at 15 percent month-over-month.
Cooper emphasized that the fundraise was strategic rather than necessary. "We're default alive; there's no reason for us to raise money," he said. "We raised because we see a massive opportunity to accelerate, not because we needed to survive."
The company hired its first salesperson only last year and employs just two solutions engineers. Nearly all of Railway's two million users discovered the platform through word of mouth — developers telling other developers about a tool that actually works.
"We basically did the standard engineering thing: if you build it, they will come," Cooper recalled. "And to some degree, they came."
Despite its grassroots developer community, Railway has made significant inroads into large organizations. The company claims that 31 percent of Fortune 500 companies now use its platform, though deployments range from company-wide infrastructure to individual team projects.
Notable customers include Bilt, the loyalty program company; Intuit's GoCo subsidiary; TripAdvisor's Cruise Critic; and MGM Resorts. Kernel, a Y Combinator-backed startup providing AI infrastructure to over 1,000 companies, runs its entire customer-facing system on Railway for $444 per month.
"At my previous company Clever, which sold for $500 million, I had six full-time engineers just managing AWS," said Rafael Garcia, Kernel's chief technology officer. "Now I have six engineers total, and they all focus on product. Railway is exactly the tool I wish I had in 2012."
For enterprise customers, Railway offers security certifications including SOC 2 Type 2 compliance and HIPAA readiness, with business associate agreements available upon request. The platform provides single sign-on authentication, comprehensive audit logs, and the option to deploy within a customer's existing cloud environment through a "bring your own cloud" configuration.
Enterprise pricing starts at custom levels, with specific add-ons for extended log retention ($200 monthly), HIPAA BAAs ($1,000), enterprise support with SLOs ($2,000), and dedicated virtual machines ($10,000).
Railway enters a crowded market that includes not only the hyperscale cloud providers—Amazon Web Services, Microsoft Azure, and Google Cloud Platform—but also a growing cohort of developer-focused platforms like Vercel, Render, Fly.io, and Heroku.
Cooper argues that Railway's competitors fall into two camps, neither of which has fully committed to the new infrastructure model that AI demands.
"The hyperscalers have two competing systems, and they haven't gone all-in on the new model because their legacy revenue stream is still printing money," he observed. "They have this mammoth pool of cash coming from people who provision a VM, use maybe 10 percent of it, and still pay for the whole thing. To what end are they actually interested in going all the way in on a new experience if they don't really need to?"
Against startup competitors, Railway differentiates by covering the full infrastructure stack. "We're not just containers; we've got VM primitives, stateful storage, virtual private networking, automated load balancing," Cooper said. "And we wrap all of this in an absurdly easy-to-use UI, with agentic primitives so agents can move 1,000 times faster."
The platform supports databases including PostgreSQL, MySQL, MongoDB, and Redis; provides up to 256 terabytes of persistent storage with over 100,000 input/output operations per second; and enables deployment to four global regions spanning the United States, Europe, and Southeast Asia. Enterprise customers can scale to 112 vCPUs and 2 terabytes of RAM per service.
Railway's fundraise reflects broader investor enthusiasm for companies positioned to benefit from the AI coding revolution. As tools like GitHub Copilot, Cursor, and Claude become standard fixtures in developer workflows, the volume of code being written — and the infrastructure needed to run it — is expanding dramatically.
"The amount of software that's going to come online over the next five years is unfathomable compared to what existed before — we're talking a thousand times more software," Cooper predicted. "All of that has to run somewhere."
The company has already integrated directly with AI systems, building what Cooper calls "loops where Claude can hook in, call deployments, and analyze infrastructure automatically." Railway released a Model Context Protocol server in August 2025 that allows AI coding agents to deploy applications and manage infrastructure directly from code editors.
"The notion of a developer is melting before our eyes," Cooper said. "You don't have to be an engineer to engineer things anymore — you just need critical thinking and the ability to analyze things in a systems capacity."
Railway plans to use the new capital to expand its global data center footprint, grow its team beyond 30 employees, and build what Cooper described as a proper go-to-market operation for the first time in the company's five-year history.
"One of my mentors said you raise money when you can change the trajectory of the business," Cooper explained. "We've built all the required substrate to scale indefinitely; what's been holding us back is simply talking about it. 2026 is the year we play on the world stage."
The company's investor roster reads like a who's who of developer infrastructure. Angel investors include Tom Preston-Werner, co-founder of GitHub; Guillermo Rauch, chief executive of Vercel; Spencer Kimball, chief executive of Cockroach Labs; Olivier Pomel, chief executive of Datadog; and Jori Lallo, co-founder of Linear.
The timing of Railway's expansion coincides with what many in Silicon Valley view as a fundamental shift in how software gets made. Coding assistants are no longer experimental curiosities — they have become essential tools that millions of developers rely on daily. Each line of AI-generated code needs somewhere to run, and the incumbents, by Cooper's telling, are too wedded to their existing business models to fully capitalize on the moment.
Whether Railway can translate developer enthusiasm into sustained enterprise adoption remains an open question. The cloud infrastructure market is littered with promising startups that failed to break the grip of Amazon, Microsoft, and Google. But Cooper, who previously worked as a software engineer at Wolfram Alpha, Bloomberg, and Uber before founding Railway in 2020, seems unfazed by the scale of his ambition.
"In five years, Railway [will be] the place where software gets created and evolved, period," he said. "Deploy instantly, scale infinitely, with zero friction. That's the prize worth playing for, and there's no bigger one on offer."
For a company that built a $100 million business by doing the opposite of what conventional startup wisdom dictates — no marketing, no sales team, no venture hype—the real test begins now. Railway spent five years proving that developers would find a better mousetrap on their own. The next five will determine whether the rest of the world is ready to get on board.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-22T14:00:00.000Z", "pubDate": "Thu, 22 Jan 2026 14:00:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/5RLyQIpBeVXxv0RpcXZxWQ/fd9680c6d82acd8208ac341fc575f5fb/nuneybits_Vector_art_of_a_sleek_bullet_train_bursting_from_a_cl_32a805aa-272c-4b34-ac16-cf20508b7ff4.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Infrastructure", "AI" ], "contentSnippet": "Railway, a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence applications exposes the limitations of legacy cloud infrastructure.\nTQ Ventures led the round, with participation from FPV Ventures, Redpoint, and Unusual Ventures. The investment values Railway as one of the most significant infrastructure startups to emerge during the AI boom, capitalizing on developer frustration with the complexity and cost of traditional platforms like Amazon Web Services and Google Cloud.\n\"As AI models get better at writing code, more and more people are asking the age-old question: where, and how, do I run my applications?\" said Jake Cooper, Railway's 28-year-old founder and chief executive, in an exclusive interview with VentureBeat. \"The last generation of cloud primitives were slow and outdated, and now with AI moving everything faster, teams simply can't keep up.\"\nThe funding is a dramatic acceleration for a company that has charted an unconventional path through the cloud computing industry. Railway raised just $24 million in total before this round, including a $20 million Series A from Redpoint in 2022. The company now processes more than 10 million deployments monthly and handles over one trillion requests through its edge network — metrics that rival far larger and better-funded competitors.\nWhy three-minute deploy times have become unacceptable in the age of AI coding assistants\nRailway's pitch rests on a simple observation: the tools developers use to deploy and manage software were designed for a slower era. A standard build-and-deploy cycle using Terraform, the industry-standard infrastructure tool, takes two to three minutes. That delay, once tolerable, has become a critical bottleneck as AI coding assistants like Claude, ChatGPT, and Cursor can generate working code in seconds.\n\"When godly intelligence is on tap and can solve any problem in three seconds, those amalgamations of systems become bottlenecks,\" Cooper told VentureBeat. \"What was really cool for humans to deploy in 10 seconds or less is now table stakes for agents.\"\nThe company claims its platform delivers deployments in under one second — fast enough to keep pace with AI-generated code. Customers report a tenfold increase in developer velocity and up to 65 percent cost savings compared to traditional cloud providers.\nThese numbers come directly from enterprise clients, not internal benchmarks. Daniel Lobaton, chief technology officer at G2X, a platform serving 100,000 federal contractors, measured deployment speed improvements of seven times faster and an 87 percent cost reduction after migrating to Railway. His infrastructure bill dropped from $15,000 per month to approximately $1,000.\n\"The work that used to take me a week on our previous infrastructure, I can do in Railway in like a day,\" Lobaton said. \"If I want to spin up a new service and test different architectures, it would take so long on our old setup. In Railway I can launch six services in two minutes.\"\nInside the controversial decision to abandon Google Cloud and build data centers from scratch\nWhat distinguishes Railway from competitors like Render and Fly.io is the depth of its vertical integration. In 2024, the company made the unusual decision to abandon Google Cloud entirely and build its own data centers, a move that echoes the famous Alan Kay maxim: \"People who are really serious about software should make their own hardware.\"\n\"We wanted to design hardware in a way where we could build a differentiated experience,\" Cooper said. \"Having full control over the network, compute, and storage layers lets us do really fast build and deploy loops, the kind that allows us to move at 'agentic speed' while staying 100 percent the smoothest ride in town.\"\nThe approach paid dividends during recent widespread outages that affected major cloud providers — Railway remained online throughout.\nThis soup-to-nuts control enables pricing that undercuts the hyperscalers by roughly 50 percent and newer cloud startups by three to four times. Railway charges by the second for actual compute usage: $0.00000386 per gigabyte-second of memory, $0.00000772 per vCPU-second, and $0.00000006 per gigabyte-second of storage. There are no charges for idle virtual machines — a stark contrast to the traditional cloud model where customers pay for provisioned capacity whether they use it or not.\n\"The conventional wisdom is that the big guys have economies of scale to offer better pricing,\" Cooper noted. \"But when they're charging for VMs that usually sit idle in the cloud, and we've purpose-built everything to fit much more density on these machines, you have a big opportunity.\"\nHow 30 employees built a platform generating tens of millions in annual revenue\nRailway has achieved its scale with a team of just 30 employees generating tens of millions in annual revenue — a ratio of revenue per employee that would be exceptional even for established software companies. The company grew revenue 3.5 times last year and continues to expand at 15 percent month-over-month.\nCooper emphasized that the fundraise was strategic rather than necessary. \"We're default alive; there's no reason for us to raise money,\" he said. \"We raised because we see a massive opportunity to accelerate, not because we needed to survive.\"\nThe company hired its first salesperson only last year and employs just two solutions engineers. Nearly all of Railway's two million users discovered the platform through word of mouth — developers telling other developers about a tool that actually works.\n\"We basically did the standard engineering thing: if you build it, they will come,\" Cooper recalled. \"And to some degree, they came.\"\nFrom side projects to Fortune 500 deployments: Railway's unlikely corporate expansion\nDespite its grassroots developer community, Railway has made significant inroads into large organizations. The company claims that 31 percent of Fortune 500 companies now use its platform, though deployments range from company-wide infrastructure to individual team projects.\nNotable customers include Bilt, the loyalty program company; Intuit's GoCo subsidiary; TripAdvisor's Cruise Critic; and MGM Resorts. Kernel, a Y Combinator-backed startup providing AI infrastructure to over 1,000 companies, runs its entire customer-facing system on Railway for $444 per month.\n\"At my previous company Clever, which sold for $500 million, I had six full-time engineers just managing AWS,\" said Rafael Garcia, Kernel's chief technology officer. \"Now I have six engineers total, and they all focus on product. Railway is exactly the tool I wish I had in 2012.\"\nFor enterprise customers, Railway offers security certifications including SOC 2 Type 2 compliance and HIPAA readiness, with business associate agreements available upon request. The platform provides single sign-on authentication, comprehensive audit logs, and the option to deploy within a customer's existing cloud environment through a \"bring your own cloud\" configuration.\nEnterprise pricing starts at custom levels, with specific add-ons for extended log retention ($200 monthly), HIPAA BAAs ($1,000), enterprise support with SLOs ($2,000), and dedicated virtual machines ($10,000).\nThe startup's bold strategy to take on Amazon, Google, and a new generation of cloud rivals\nRailway enters a crowded market that includes not only the hyperscale cloud providers—Amazon Web Services, Microsoft Azure, and Google Cloud Platform—but also a growing cohort of developer-focused platforms like Vercel, Render, Fly.io, and Heroku.\nCooper argues that Railway's competitors fall into two camps, neither of which has fully committed to the new infrastructure model that AI demands.\n\"The hyperscalers have two competing systems, and they haven't gone all-in on the new model because their legacy revenue stream is still printing money,\" he observed. \"They have this mammoth pool of cash coming from people who provision a VM, use maybe 10 percent of it, and still pay for the whole thing. To what end are they actually interested in going all the way in on a new experience if they don't really need to?\"\nAgainst startup competitors, Railway differentiates by covering the full infrastructure stack. \"We're not just containers; we've got VM primitives, stateful storage, virtual private networking, automated load balancing,\" Cooper said. \"And we wrap all of this in an absurdly easy-to-use UI, with agentic primitives so agents can move 1,000 times faster.\"\nThe platform supports databases including PostgreSQL, MySQL, MongoDB, and Redis; provides up to 256 terabytes of persistent storage with over 100,000 input/output operations per second; and enables deployment to four global regions spanning the United States, Europe, and Southeast Asia. Enterprise customers can scale to 112 vCPUs and 2 terabytes of RAM per service.\nWhy investors are betting that AI will create a thousand times more software than exists today\nRailway's fundraise reflects broader investor enthusiasm for companies positioned to benefit from the AI coding revolution. As tools like GitHub Copilot, Cursor, and Claude become standard fixtures in developer workflows, the volume of code being written — and the infrastructure needed to run it — is expanding dramatically.\n\"The amount of software that's going to come online over the next five years is unfathomable compared to what existed before — we're talking a thousand times more software,\" Cooper predicted. \"All of that has to run somewhere.\"\nThe company has already integrated directly with AI systems, building what Cooper calls \"loops where Claude can hook in, call deployments, and analyze infrastructure automatically.\" Railway released a Model Context Protocol server in August 2025 that allows AI coding agents to deploy applications and manage infrastructure directly from code editors.\n\"The notion of a developer is melting before our eyes,\" Cooper said. \"You don't have to be an engineer to engineer things anymore — you just need critical thinking and the ability to analyze things in a systems capacity.\"\nWhat Railway plans to do with $100 million and zero marketing experience\nRailway plans to use the new capital to expand its global data center footprint, grow its team beyond 30 employees, and build what Cooper described as a proper go-to-market operation for the first time in the company's five-year history.\n\"One of my mentors said you raise money when you can change the trajectory of the business,\" Cooper explained. \"We've built all the required substrate to scale indefinitely; what's been holding us back is simply talking about it. 2026 is the year we play on the world stage.\"\nThe company's investor roster reads like a who's who of developer infrastructure. Angel investors include Tom Preston-Werner, co-founder of GitHub; Guillermo Rauch, chief executive of Vercel; Spencer Kimball, chief executive of Cockroach Labs; Olivier Pomel, chief executive of Datadog; and Jori Lallo, co-founder of Linear.\nThe timing of Railway's expansion coincides with what many in Silicon Valley view as a fundamental shift in how software gets made. Coding assistants are no longer experimental curiosities — they have become essential tools that millions of developers rely on daily. Each line of AI-generated code needs somewhere to run, and the incumbents, by Cooper's telling, are too wedded to their existing business models to fully capitalize on the moment.\nWhether Railway can translate developer enthusiasm into sustained enterprise adoption remains an open question. The cloud infrastructure market is littered with promising startups that failed to break the grip of Amazon, Microsoft, and Google. But Cooper, who previously worked as a software engineer at Wolfram Alpha, Bloomberg, and Uber before founding Railway in 2020, seems unfazed by the scale of his ambition.\n\"In five years, Railway [will be] the place where software gets created and evolved, period,\" he said. \"Deploy instantly, scale infinitely, with zero friction. That's the prize worth playing for, and there's no bigger one on offer.\"\nFor a company that built a $100 million business by doing the opposite of what conventional startup wisdom dictates — no marketing, no sales team, no venture hype—the real test begins now. Railway spent five years proving that developers would find a better mousetrap on their own. The next five will determine whether the rest of the world is ready to get on board." }, "pairedItem": [ { "item": 4 } ] }, { "json": { "guid": "2CUHNkdUg4dnjDFiMtJJQ", "link": "https://venturebeat.com/infrastructure/claude-code-costs-up-to-usd200-a-month-goose-does-the-same-thing-for-free", "title": "Claude Code costs up to $200 a month. Goose does the same thing for free.", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "The artificial intelligence coding revolution comes with a catch: it's expensive.
Claude Code, Anthropic's terminal-based AI agent that can write, debug, and deploy code autonomously, has captured the imagination of software developers worldwide. But its pricing — ranging from $20 to $200 per month depending on usage — has sparked a growing rebellion among the very programmers it aims to serve.
Now, a free alternative is gaining traction. Goose, an open-source AI agent developed by Block (the financial technology company formerly known as Square), offers nearly identical functionality to Claude Code but runs entirely on a user's local machine. No subscription fees. No cloud dependency. No rate limits that reset every five hours.
"Your data stays with you, period," said Parth Sareen, a software engineer who demonstrated the tool during a recent livestream. The comment captures the core appeal: Goose gives developers complete control over their AI-powered workflow, including the ability to work offline — even on an airplane.
The project has exploded in popularity. Goose now boasts more than 26,100 stars on GitHub, the code-sharing platform, with 362 contributors and 102 releases since its launch. The latest version, 1.20.1, shipped on January 19, 2026, reflecting a development pace that rivals commercial products.
For developers frustrated by Claude Code's pricing structure and usage caps, Goose represents something increasingly rare in the AI industry: a genuinely free, no-strings-attached option for serious work.
To understand why Goose matters, you need to understand the Claude Code pricing controversy.
Anthropic, the San Francisco artificial intelligence company founded by former OpenAI executives, offers Claude Code as part of its subscription tiers. The free plan provides no access whatsoever. The Pro plan, at $17 per month with annual billing (or $20 monthly), limits users to just 10 to 40 prompts every five hours — a constraint that serious developers exhaust within minutes of intensive work.
The Max plans, at $100 and $200 per month, offer more headroom: 50 to 200 prompts and 200 to 800 prompts respectively, plus access to Anthropic's most powerful model, Claude 4.5 Opus. But even these premium tiers come with restrictions that have inflamed the developer community.
In late July, Anthropic announced new weekly rate limits. Under the system, Pro users receive 40 to 80 hours of Sonnet 4 usage per week. Max users at the $200 tier get 240 to 480 hours of Sonnet 4, plus 24 to 40 hours of Opus 4. Nearly five months later, the frustration has not subsided.
The problem? Those "hours" are not actual hours. They represent token-based limits that vary wildly depending on codebase size, conversation length, and the complexity of the code being processed. Independent analysis suggests the actual per-session limits translate to roughly 44,000 tokens for Pro users and 220,000 tokens for the $200 Max plan.
"It's confusing and vague," one developer wrote in a widely shared analysis. "When they say '24-40 hours of Opus 4,' that doesn't really tell you anything useful about what you're actually getting."
The backlash on Reddit and developer forums has been fierce. Some users report hitting their daily limits within 30 minutes of intensive coding. Others have canceled their subscriptions entirely, calling the new restrictions "a joke" and "unusable for real work."
Anthropic has defended the changes, stating that the limits affect fewer than five percent of users and target people running Claude Code "continuously in the background, 24/7." But the company has not clarified whether that figure refers to five percent of Max subscribers or five percent of all users — a distinction that matters enormously.
Goose takes a radically different approach to the same problem.
Built by Block, the payments company led by Jack Dorsey, Goose is what engineers call an "on-machine AI agent." Unlike Claude Code, which sends your queries to Anthropic's servers for processing, Goose can run entirely on your local computer using open-source language models that you download and control yourself.
The project's documentation describes it as going "beyond code suggestions" to "install, execute, edit, and test with any LLM." That last phrase — "any LLM" — is the key differentiator. Goose is model-agnostic by design.
You can connect Goose to Anthropic's Claude models if you have API access. You can use OpenAI's GPT-5 or Google's Gemini. You can route it through services like Groq or OpenRouter. Or — and this is where things get interesting — you can run it entirely locally using tools like Ollama, which let you download and execute open-source models on your own hardware.
The practical implications are significant. With a local setup, there are no subscription fees, no usage caps, no rate limits, and no concerns about your code being sent to external servers. Your conversations with the AI never leave your machine.
"I use Ollama all the time on planes — it's a lot of fun!" Sareen noted during a demonstration, highlighting how local models free developers from the constraints of internet connectivity.
Goose operates as a command-line tool or desktop application that can autonomously perform complex development tasks. It can build entire projects from scratch, write and execute code, debug failures, orchestrate workflows across multiple files, and interact with external APIs — all without constant human oversight.
The architecture relies on what the AI industry calls "tool calling" or "function calling" — the ability for a language model to request specific actions from external systems. When you ask Goose to create a new file, run a test suite, or check the status of a GitHub pull request, it doesn't just generate text describing what should happen. It actually executes those operations.
This capability depends heavily on the underlying language model. Claude 4 models from Anthropic currently perform best at tool calling, according to the Berkeley Function-Calling Leaderboard, which ranks models on their ability to translate natural language requests into executable code and system commands.
But newer open-source models are catching up quickly. Goose's documentation highlights several options with strong tool-calling support: Meta's Llama series, Alibaba's Qwen models, Google's Gemma variants, and DeepSeek's reasoning-focused architectures.
The tool also integrates with the Model Context Protocol, or MCP, an emerging standard for connecting AI agents to external services. Through MCP, Goose can access databases, search engines, file systems, and third-party APIs — extending its capabilities far beyond what the base language model provides.
For developers interested in a completely free, privacy-preserving setup, the process involves three main components: Goose itself, Ollama (a tool for running open-source models locally), and a compatible language model.
Step 1: Install Ollama
Ollama is an open-source project that dramatically simplifies the process of running large language models on personal hardware. It handles the complex work of downloading, optimizing, and serving models through a simple interface.
Download and install Ollama from ollama.com. Once installed, you can pull models with a single command. For coding tasks, Qwen 2.5 offers strong tool-calling support:
ollama run qwen2.5
The model downloads automatically and begins running on your machine.
Step 2: Install Goose
Goose is available as both a desktop application and a command-line interface. The desktop version provides a more visual experience, while the CLI appeals to developers who prefer working entirely in the terminal.
Installation instructions vary by operating system but generally involve downloading from Goose's GitHub releases page or using a package manager. Block provides pre-built binaries for macOS (both Intel and Apple Silicon), Windows, and Linux.
Step 3: Configure the Connection
In Goose Desktop, navigate to Settings, then Configure Provider, and select Ollama. Confirm that the API Host is set to http://localhost:11434 (Ollama's default port) and click Submit.
For the command-line version, run goose configure, select "Configure Providers," choose Ollama, and enter the model name when prompted.
That's it. Goose is now connected to a language model running entirely on your hardware, ready to execute complex coding tasks without any subscription fees or external dependencies.
The obvious question: what kind of computer do you need?
Running large language models locally requires substantially more computational resources than typical software. The key constraint is memory — specifically, RAM on most systems, or VRAM if using a dedicated graphics card for acceleration.
Block's documentation suggests that 32 gigabytes of RAM provides "a solid baseline for larger models and outputs." For Mac users, this means the computer's unified memory is the primary bottleneck. For Windows and Linux users with discrete NVIDIA graphics cards, GPU memory (VRAM) matters more for acceleration.
But you don't necessarily need expensive hardware to get started. Smaller models with fewer parameters run on much more modest systems. Qwen 2.5, for instance, comes in multiple sizes, and the smaller variants can operate effectively on machines with 16 gigabytes of RAM.
"You don't need to run the largest models to get excellent results," Sareen emphasized. The practical recommendation: start with a smaller model to test your workflow, then scale up as needed.
For context, Apple's entry-level MacBook Air with 8 gigabytes of RAM would struggle with most capable coding models. But a MacBook Pro with 32 gigabytes — increasingly common among professional developers — handles them comfortably.
Goose with a local LLM is not a perfect substitute for Claude Code. The comparison involves real trade-offs that developers should understand.
Model Quality: Claude 4.5 Opus, Anthropic's flagship model, remains arguably the most capable AI for software engineering tasks. It excels at understanding complex codebases, following nuanced instructions, and producing high-quality code on the first attempt. Open-source models have improved dramatically, but a gap persists — particularly for the most challenging tasks.
One developer who switched to the $200 Claude Code plan described the difference bluntly: "When I say 'make this look modern,' Opus knows what I mean. Other models give me Bootstrap circa 2015."
Context Window: Claude Sonnet 4.5, accessible through the API, offers a massive one-million-token context window — enough to load entire large codebases without chunking or context management issues. Most local models are limited to 4,096 or 8,192 tokens by default, though many can be configured for longer contexts at the cost of increased memory usage and slower processing.
Speed: Cloud-based services like Claude Code run on dedicated server hardware optimized for AI inference. Local models, running on consumer laptops, typically process requests more slowly. The difference matters for iterative workflows where you're making rapid changes and waiting for AI feedback.
Tooling Maturity: Claude Code benefits from Anthropic's dedicated engineering resources. Features like prompt caching (which can reduce costs by up to 90 percent for repeated contexts) and structured outputs are polished and well-documented. Goose, while actively developed with 102 releases to date, relies on community contributions and may lack equivalent refinement in specific areas.
Goose enters a crowded market of AI coding tools, but occupies a distinctive position.
Cursor, a popular AI-enhanced code editor, charges $20 per month for its Pro tier and $200 for Ultra—pricing that mirrors Claude Code's Max plans. Cursor provides approximately 4,500 Sonnet 4 requests per month at the Ultra level, a substantially different allocation model than Claude Code's hourly resets.
Cline, Roo Code, and similar open-source projects offer AI coding assistance but with varying levels of autonomy and tool integration. Many focus on code completion rather than the agentic task execution that defines Goose and Claude Code.
Amazon's CodeWhisperer, GitHub Copilot, and enterprise offerings from major cloud providers target large organizations with complex procurement processes and dedicated budgets. They are less relevant to individual developers and small teams seeking lightweight, flexible tools.
Goose's combination of genuine autonomy, model agnosticism, local operation, and zero cost creates a unique value proposition. The tool is not trying to compete with commercial offerings on polish or model quality. It's competing on freedom — both financial and architectural.
The AI coding tools market is evolving quickly. Open-source models are improving at a pace that continually narrows the gap with proprietary alternatives. Moonshot AI's Kimi K2 and z.ai's GLM 4.5 now benchmark near Claude Sonnet 4 levels — and they're freely available.
If this trajectory continues, the quality advantage that justifies Claude Code's premium pricing may erode. Anthropic would then face pressure to compete on features, user experience, and integration rather than raw model capability.
For now, developers face a clear choice. Those who need the absolute best model quality, who can afford premium pricing, and who accept usage restrictions may prefer Claude Code. Those who prioritize cost, privacy, offline access, and flexibility have a genuine alternative in Goose.
The fact that a $200-per-month commercial product has a zero-dollar open-source competitor with comparable core functionality is itself remarkable. It reflects both the maturation of open-source AI infrastructure and the appetite among developers for tools that respect their autonomy.
Goose is not perfect. It requires more technical setup than commercial alternatives. It depends on hardware resources that not every developer possesses. Its model options, while improving rapidly, still trail the best proprietary offerings on complex tasks.
But for a growing community of developers, those limitations are acceptable trade-offs for something increasingly rare in the AI landscape: a tool that truly belongs to them.
Goose is available for download at github.com/block/goose. Ollama is available at ollama.com. Both projects are free and open source.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-19T14:00:00.000Z", "pubDate": "Mon, 19 Jan 2026 14:00:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/1U9H8GLIqoGqKpitVfgw3T/ba56292f99409eca709dac0b176ec245/nuneybits_Vector_art_of_white_goose_silhouette_flying_through_c_8100d5a7-9e36-4ed6-a188-016470e1d0e1.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "AI", "Infrastructure" ], "contentSnippet": "The artificial intelligence coding revolution comes with a catch: it's expensive.\nClaude Code, Anthropic's terminal-based AI agent that can write, debug, and deploy code autonomously, has captured the imagination of software developers worldwide. But its pricing — ranging from $20 to $200 per month depending on usage — has sparked a growing rebellion among the very programmers it aims to serve.\nNow, a free alternative is gaining traction. Goose, an open-source AI agent developed by Block (the financial technology company formerly known as Square), offers nearly identical functionality to Claude Code but runs entirely on a user's local machine. No subscription fees. No cloud dependency. No rate limits that reset every five hours.\n\"Your data stays with you, period,\" said Parth Sareen, a software engineer who demonstrated the tool during a recent livestream. The comment captures the core appeal: Goose gives developers complete control over their AI-powered workflow, including the ability to work offline — even on an airplane.\nThe project has exploded in popularity. Goose now boasts more than 26,100 stars on GitHub, the code-sharing platform, with 362 contributors and 102 releases since its launch. The latest version, 1.20.1, shipped on January 19, 2026, reflecting a development pace that rivals commercial products.\nFor developers frustrated by Claude Code's pricing structure and usage caps, Goose represents something increasingly rare in the AI industry: a genuinely free, no-strings-attached option for serious work.\n\nAnthropic's new rate limits spark a developer revolt\nTo understand why Goose matters, you need to understand the Claude Code pricing controversy.\nAnthropic, the San Francisco artificial intelligence company founded by former OpenAI executives, offers Claude Code as part of its subscription tiers. The free plan provides no access whatsoever. The Pro plan, at $17 per month with annual billing (or $20 monthly), limits users to just 10 to 40 prompts every five hours — a constraint that serious developers exhaust within minutes of intensive work.\nThe Max plans, at $100 and $200 per month, offer more headroom: 50 to 200 prompts and 200 to 800 prompts respectively, plus access to Anthropic's most powerful model, Claude 4.5 Opus. But even these premium tiers come with restrictions that have inflamed the developer community.\nIn late July, Anthropic announced new weekly rate limits. Under the system, Pro users receive 40 to 80 hours of Sonnet 4 usage per week. Max users at the $200 tier get 240 to 480 hours of Sonnet 4, plus 24 to 40 hours of Opus 4. Nearly five months later, the frustration has not subsided.\nThe problem? Those \"hours\" are not actual hours. They represent token-based limits that vary wildly depending on codebase size, conversation length, and the complexity of the code being processed. Independent analysis suggests the actual per-session limits translate to roughly 44,000 tokens for Pro users and 220,000 tokens for the $200 Max plan.\n\"It's confusing and vague,\" one developer wrote in a widely shared analysis. \"When they say '24-40 hours of Opus 4,' that doesn't really tell you anything useful about what you're actually getting.\"\nThe backlash on Reddit and developer forums has been fierce. Some users report hitting their daily limits within 30 minutes of intensive coding. Others have canceled their subscriptions entirely, calling the new restrictions \"a joke\" and \"unusable for real work.\"\nAnthropic has defended the changes, stating that the limits affect fewer than five percent of users and target people running Claude Code \"continuously in the background, 24/7.\" But the company has not clarified whether that figure refers to five percent of Max subscribers or five percent of all users — a distinction that matters enormously.\nHow Block built a free AI coding agent that works offline\nGoose takes a radically different approach to the same problem.\nBuilt by Block, the payments company led by Jack Dorsey, Goose is what engineers call an \"on-machine AI agent.\" Unlike Claude Code, which sends your queries to Anthropic's servers for processing, Goose can run entirely on your local computer using open-source language models that you download and control yourself.\nThe project's documentation describes it as going \"beyond code suggestions\" to \"install, execute, edit, and test with any LLM.\" That last phrase — \"any LLM\" — is the key differentiator. Goose is model-agnostic by design.\nYou can connect Goose to Anthropic's Claude models if you have API access. You can use OpenAI's GPT-5 or Google's Gemini. You can route it through services like Groq or OpenRouter. Or — and this is where things get interesting — you can run it entirely locally using tools like Ollama, which let you download and execute open-source models on your own hardware.\nThe practical implications are significant. With a local setup, there are no subscription fees, no usage caps, no rate limits, and no concerns about your code being sent to external servers. Your conversations with the AI never leave your machine.\n\"I use Ollama all the time on planes — it's a lot of fun!\" Sareen noted during a demonstration, highlighting how local models free developers from the constraints of internet connectivity.\nWhat Goose can do that traditional code assistants can't\nGoose operates as a command-line tool or desktop application that can autonomously perform complex development tasks. It can build entire projects from scratch, write and execute code, debug failures, orchestrate workflows across multiple files, and interact with external APIs — all without constant human oversight.\nThe architecture relies on what the AI industry calls \"tool calling\" or \"function calling\" — the ability for a language model to request specific actions from external systems. When you ask Goose to create a new file, run a test suite, or check the status of a GitHub pull request, it doesn't just generate text describing what should happen. It actually executes those operations.\nThis capability depends heavily on the underlying language model. Claude 4 models from Anthropic currently perform best at tool calling, according to the Berkeley Function-Calling Leaderboard, which ranks models on their ability to translate natural language requests into executable code and system commands.\nBut newer open-source models are catching up quickly. Goose's documentation highlights several options with strong tool-calling support: Meta's Llama series, Alibaba's Qwen models, Google's Gemma variants, and DeepSeek's reasoning-focused architectures.\nThe tool also integrates with the Model Context Protocol, or MCP, an emerging standard for connecting AI agents to external services. Through MCP, Goose can access databases, search engines, file systems, and third-party APIs — extending its capabilities far beyond what the base language model provides.\nSetting Up Goose with a Local Model\nFor developers interested in a completely free, privacy-preserving setup, the process involves three main components: Goose itself, Ollama (a tool for running open-source models locally), and a compatible language model.\nStep 1: Install Ollama\nOllama is an open-source project that dramatically simplifies the process of running large language models on personal hardware. It handles the complex work of downloading, optimizing, and serving models through a simple interface.\nDownload and install Ollama from ollama.com. Once installed, you can pull models with a single command. For coding tasks, Qwen 2.5 offers strong tool-calling support:\nollama run qwen2.5\nThe model downloads automatically and begins running on your machine.\nStep 2: Install Goose\nGoose is available as both a desktop application and a command-line interface. The desktop version provides a more visual experience, while the CLI appeals to developers who prefer working entirely in the terminal.\nInstallation instructions vary by operating system but generally involve downloading from Goose's GitHub releases page or using a package manager. Block provides pre-built binaries for macOS (both Intel and Apple Silicon), Windows, and Linux.\nStep 3: Configure the Connection\nIn Goose Desktop, navigate to Settings, then Configure Provider, and select Ollama. Confirm that the API Host is set to http://localhost:11434 (Ollama's default port) and click Submit.\nFor the command-line version, run goose configure, select \"Configure Providers,\" choose Ollama, and enter the model name when prompted.\nThat's it. Goose is now connected to a language model running entirely on your hardware, ready to execute complex coding tasks without any subscription fees or external dependencies.\nThe RAM, processing power, and trade-offs you should know about\nThe obvious question: what kind of computer do you need?\nRunning large language models locally requires substantially more computational resources than typical software. The key constraint is memory — specifically, RAM on most systems, or VRAM if using a dedicated graphics card for acceleration.\nBlock's documentation suggests that 32 gigabytes of RAM provides \"a solid baseline for larger models and outputs.\" For Mac users, this means the computer's unified memory is the primary bottleneck. For Windows and Linux users with discrete NVIDIA graphics cards, GPU memory (VRAM) matters more for acceleration.\nBut you don't necessarily need expensive hardware to get started. Smaller models with fewer parameters run on much more modest systems. Qwen 2.5, for instance, comes in multiple sizes, and the smaller variants can operate effectively on machines with 16 gigabytes of RAM.\n\"You don't need to run the largest models to get excellent results,\" Sareen emphasized. The practical recommendation: start with a smaller model to test your workflow, then scale up as needed.\nFor context, Apple's entry-level MacBook Air with 8 gigabytes of RAM would struggle with most capable coding models. But a MacBook Pro with 32 gigabytes — increasingly common among professional developers — handles them comfortably.\nWhy keeping your code off the cloud matters more than ever\nGoose with a local LLM is not a perfect substitute for Claude Code. The comparison involves real trade-offs that developers should understand.\nModel Quality: Claude 4.5 Opus, Anthropic's flagship model, remains arguably the most capable AI for software engineering tasks. It excels at understanding complex codebases, following nuanced instructions, and producing high-quality code on the first attempt. Open-source models have improved dramatically, but a gap persists — particularly for the most challenging tasks.\nOne developer who switched to the $200 Claude Code plan described the difference bluntly: \"When I say 'make this look modern,' Opus knows what I mean. Other models give me Bootstrap circa 2015.\"\nContext Window: Claude Sonnet 4.5, accessible through the API, offers a massive one-million-token context window — enough to load entire large codebases without chunking or context management issues. Most local models are limited to 4,096 or 8,192 tokens by default, though many can be configured for longer contexts at the cost of increased memory usage and slower processing.\nSpeed: Cloud-based services like Claude Code run on dedicated server hardware optimized for AI inference. Local models, running on consumer laptops, typically process requests more slowly. The difference matters for iterative workflows where you're making rapid changes and waiting for AI feedback.\nTooling Maturity: Claude Code benefits from Anthropic's dedicated engineering resources. Features like prompt caching (which can reduce costs by up to 90 percent for repeated contexts) and structured outputs are polished and well-documented. Goose, while actively developed with 102 releases to date, relies on community contributions and may lack equivalent refinement in specific areas.\nHow Goose stacks up against Cursor, GitHub Copilot, and the paid AI coding market\nGoose enters a crowded market of AI coding tools, but occupies a distinctive position.\nCursor, a popular AI-enhanced code editor, charges $20 per month for its Pro tier and $200 for Ultra—pricing that mirrors Claude Code's Max plans. Cursor provides approximately 4,500 Sonnet 4 requests per month at the Ultra level, a substantially different allocation model than Claude Code's hourly resets.\nCline, Roo Code, and similar open-source projects offer AI coding assistance but with varying levels of autonomy and tool integration. Many focus on code completion rather than the agentic task execution that defines Goose and Claude Code.\nAmazon's CodeWhisperer, GitHub Copilot, and enterprise offerings from major cloud providers target large organizations with complex procurement processes and dedicated budgets. They are less relevant to individual developers and small teams seeking lightweight, flexible tools.\nGoose's combination of genuine autonomy, model agnosticism, local operation, and zero cost creates a unique value proposition. The tool is not trying to compete with commercial offerings on polish or model quality. It's competing on freedom — both financial and architectural.\nThe $200-a-month era for AI coding tools may be ending\nThe AI coding tools market is evolving quickly. Open-source models are improving at a pace that continually narrows the gap with proprietary alternatives. Moonshot AI's Kimi K2 and z.ai's GLM 4.5 now benchmark near Claude Sonnet 4 levels — and they're freely available.\nIf this trajectory continues, the quality advantage that justifies Claude Code's premium pricing may erode. Anthropic would then face pressure to compete on features, user experience, and integration rather than raw model capability.\nFor now, developers face a clear choice. Those who need the absolute best model quality, who can afford premium pricing, and who accept usage restrictions may prefer Claude Code. Those who prioritize cost, privacy, offline access, and flexibility have a genuine alternative in Goose.\nThe fact that a $200-per-month commercial product has a zero-dollar open-source competitor with comparable core functionality is itself remarkable. It reflects both the maturation of open-source AI infrastructure and the appetite among developers for tools that respect their autonomy.\nGoose is not perfect. It requires more technical setup than commercial alternatives. It depends on hardware resources that not every developer possesses. Its model options, while improving rapidly, still trail the best proprietary offerings on complex tasks.\nBut for a growing community of developers, those limitations are acceptable trade-offs for something increasingly rare in the AI landscape: a tool that truly belongs to them.\n\nGoose is available for download at github.com/block/goose. Ollama is available at ollama.com. Both projects are free and open source." }, "pairedItem": [ { "item": 4 } ] }, { "json": { "guid": "wOG0z6nm91FOeUqhpizqI", "link": "https://venturebeat.com/technology/listen-labs-raises-usd69m-after-viral-billboard-hiring-stunt-to-scale-ai", "title": "Listen Labs raises $69M after viral billboard hiring stunt to scale AI customer interviews", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Alfred Wahlforss was running out of options. His startup, Listen Labs, needed to hire over 100 engineers, but competing against Mark Zuckerberg's $100 million offers seemed impossible. So he spent $5,000 — a fifth of his marketing budget — on a billboard in San Francisco displaying what looked like gibberish: five strings of random numbers.
The numbers were actually AI tokens. Decoded, they led to a coding challenge: build an algorithm to act as a digital bouncer at Berghain, the Berlin nightclub famous for rejecting nearly everyone at the door. Within days, thousands attempted the puzzle. 430 cracked it. Some got hired. The winner flew to Berlin, all expenses paid.
That unconventional approach has now attracted $69 million in Series B funding, led by Ribbit Capital with participation from Evantic and existing investors Sequoia Capital, Conviction, and Pear VC. The round values Listen Labs at $500 million and brings its total capital to $100 million. In nine months since launch, the company has grown annualized revenue by 15x to eight figures and conducted over one million AI-powered interviews.
"When you obsess over customers, everything else follows," Wahlforss said in an interview with VentureBeat. "Teams that use Listen bring the customer into every decision, from marketing to product, and when the customer is delighted, everyone is."
Listen's AI researcher finds participants, conducts in-depth interviews, and delivers actionable insights in hours, not weeks. The platform replaces the traditional choice between quantitative surveys — which provide statistical precision but miss nuance—and qualitative interviews, which deliver depth but cannot scale.
Wahlforss explained the limitation of existing approaches: "Essentially surveys give you false precision because people end up answering the same question... You can't get the outliers. People are actually not honest on surveys." The alternative, one-on-one human interviews, "gives you a lot of depth. You can ask follow up questions. You can kind of double check if they actually know what they're talking about. And the problem is you can't scale that."
The platform works in four steps: users create a study with AI assistance, Listen recruits participants from its global network of 30 million people, an AI moderator conducts in-depth interviews with follow-up questions, and results are packaged into executive-ready reports including key themes, highlight reels, and slide decks.
What distinguishes Listen's approach is its use of open-ended video conversations rather than multiple-choice forms. "In a survey, you can kind of guess what you should answer, and you have four options," Wahlforss said. "Oh, they probably want me to buy high income. Let me click on that button versus an open ended response. It just generates much more honesty."
Listen finds and qualifies the right participants in its global network of 30 million people. But building that panel required confronting what Wahlforss called "one of the most shocking things that we've learned when we entered this industry"—rampant fraud.
"Essentially, there's a financial transaction involved, which means there will be bad players," he explained. "We actually had some of the largest companies, some of them have billions in revenue, send us people who claim to be kind of enterprise buyers to our platform and our system immediately detected, like, fraud, fraud, fraud, fraud, fraud."
The company built what it calls a "quality guard" that cross-references LinkedIn profiles with video responses to verify identity, checks consistency across how participants answer questions, and flags suspicious patterns. The result, according to Wahlforss: "People talk three times more. They're much more honest when they talk about sensitive topics like politics and mental health."
Emeritus, an online education company that uses Listen, reported that approximately 20% of survey responses previously fell into the fraudulent or low-quality category. With Listen, they reduced this to almost zero. "We did not have to replace any responses because of fraud or gibberish information," said Gabrielli Tiburi, Assistant Manager of Customer Insights at Emeritus.
The speed advantage has proven central to Listen's pitch. Traditional customer research at Microsoft could take four to six weeks to generate insights. "By the time we get to them, either the decision has been made or we lose out on the opportunity to actually influence it," said Romani Patel, Senior Research Manager at Microsoft.
With Listen, Microsoft can now get insights in days, and in many cases, within hours.
The platform has already powered several high-profile initiatives. Microsoft used Listen Labs to collect global customer stories for its 50th anniversary celebration. "We wanted users to share how Copilot is empowering them to bring their best self forward," Patel said, "and we were able to collect those user video stories within a day." Traditionally, that kind of work would have taken six to eight weeks.
Simple Modern, an Oklahoma-based drinkware company, used Listen to test a new product concept. The process took about an hour to write questions, an hour to launch the study, and 2.5 hours to receive feedback from 120 people across the country. "We went from 'Should we even have this product?' to 'How should we launch it?'" said Chris Hoyle, the company's Chief Marketing Officer.
Chubbies, the shorts brand, achieved a 24x increase in youth research participation—growing from 5 to 120 participants — by using Listen to overcome the scheduling challenges of traditional focus groups with children. "There's school, sports, dinner, and homework," explained Lauren Neville, Director of Insights and Innovation. "I had to find a way to hear from them that fit into their schedules."
The company also discovered product issues through AI interviews that might have gone undetected otherwise. Wahlforss described how the AI "through conversations, realized there were like issues with the the kids short line, and decided to, like, interview hundreds of kids. And I understand that there were issues in the liner of the shorts and that they were, like, scratchy, quote, unquote, according to the people interviewed." The redesigned product became "a blockbuster hit."
Listen Labs is entering a massive but fragmented market. Wahlforss cited research from Andreessen Horowitz estimating the market research industry at roughly $140 billion annually, populated by legacy players — some with more than a billion dollars in revenue — that he believes are vulnerable to disruption.
"There are very much existing budget lines that we are replacing," Wahlforss said. "Why we're replacing them is that one, they're super costly. Two, they're kind of stuck in this old paradigm of choosing between a survey or interview, and they also take months to work with."
But the more intriguing dynamic may be that AI-powered research doesn't just replace existing spending — it creates new demand. Wahlforss invoked the Jevons paradox, an economic principle that occurs when technological advancements make a resource more efficient to use, but increased efficiency leads to increased overall consumption rather than decreased consumption.
"What I've noticed is that as something gets cheaper, you don't need less of it. You want more of it," Wahlforss explained. "There's infinite demand for customer understanding. So the researchers on the team can do an order of magnitude more research, and also other people who weren't researchers before can now do that as part of their job."
Listen Labs traces its origins to a consumer app that Wahlforss and his co-founder built after meeting at Harvard. "We built this consumer app that got 20,000 downloads in one day," Wahlforss recalled. "We had all these users, and we were thinking like, okay, what can we do to get to know them better? And we built this prototype of what Listen is today."
The founding team brings an unusual pedigree. Wahlforss's co-founder "was the national champion in competitive programming in Germany, and he worked at Tesla Autopilot." The company claims that 30% of its engineering team are medalists from the International Olympiad in Informatics — the same competition that produced the founders of Cognition, the AI coding startup.
The Berghain billboard stunt generated approximately 5 million views across social media, according to Wahlforss. It reflected the intensity of the talent war in the Bay Area.
"We had to do these things because some of our, like early employees, joined the company before we had a working toilet," he said. "But now we fixed that situation."
The company grew from 5 to 40 employees in 2024 and plans to reach 150 this year. It hires engineers for non-engineering roles across marketing, growth, and operations — a bet that in the AI era, technical fluency matters everywhere.
Wahlforss outlined an ambitious product roadmap that pushes into more speculative territory. The company is building "the ability to simulate your customers, so you can take all of those interviews we've done, and then extrapolate based on that and create synthetic users or simulated user voices."
Beyond simulation, Listen aims to enable automated action based on research findings. "Can you not just make recommendations, but also create spawn agents to either change things in code or some customer churns? Can you give them a discount and try to bring them back?"
Wahlforss acknowledged the ethical implications. "Obviously, as you said, there's kind of ethical concerns there. Of like, automated decision making overall can be bad, but we will have considerable guardrails to make sure that the companies are always in the loop."
The company already handles sensitive data with care. "We don't train on any of the data," Wahlforss said. "We will also scrub any sensitive PII automatically so the model can detect that. And there are times when, for example, you work with investors, where if you accidentally mention something that could be material, non public information, the AI can actually detect that and remove any information like that."
Perhaps the most provocative implication of Listen's model is how it could reshape product development itself. Wahlforss described a customer — an Australian startup — that has adopted what amounts to a continuous feedback loop.
"They're based in Australia, so they're coding during the day, and then in their night, they're releasing a Listen study with an American audience. Listen validates whatever they built during the day, and they get feedback on that. They can then plug that feedback directly into coding tools like Claude Code and iterate."
The vision extends Y Combinator's famous dictum — "write code, talk to users" — into an automated cycle. "Write code is now getting automated. And I think like talk to users will be as well, and you'll have this kind of infinite loop where you can start to ship this truly amazing product, almost kind of autonomously."
Whether that vision materializes depends on factors beyond Listen's control — the continued improvement of AI models, enterprise willingness to trust automated research, and whether speed truly correlates with better products. A 2024 MIT study found that 95% of AI pilots fail to move into production, a statistic Wahlforss cited as the reason he emphasizes quality over demos.
"I'm constantly have to emphasize like, let's make sure the quality is there and the details are right," he said.
But the company's growth suggests appetite for the experiment. Microsoft's Patel said Listen has "removed the drudgery of research and brought the fun and joy back into my work." Chubbies is now pushing its founder to give everyone in the company a login. Sling Money, a stablecoin payments startup, can create a survey in ten minutes and receive results the same day.
"It's a total game changer," said Ali Romero, Sling Money's marketing manager.
Wahlforss has a different phrase for what he's building. When asked about the tension between speed and rigor — the long-held belief that moving fast means cutting corners — he cited Nat Friedman, the former GitHub CEO and Listen investor, who keeps a list of one-liners on his website.
One of them: "Slow is fake."
It's an aggressive claim for an industry built on methodological caution. But Listen Labs is betting that in the AI era, the companies that listen fastest will be the ones that win. The only question is whether customers will talk back.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-16T14:01:00.000Z", "pubDate": "Fri, 16 Jan 2026 14:01:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/4gD12ThOmNHZuosqC4xCTz/277b1e8968da602108a29fae2eaca440/nuneybits_Vector_art_of_billboard_with_cryptic_code_dbe5b0ff-7644-45e6-a1ca-4a5dceeff986.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI" ], "contentSnippet": "Alfred Wahlforss was running out of options. His startup, Listen Labs, needed to hire over 100 engineers, but competing against Mark Zuckerberg's $100 million offers seemed impossible. So he spent $5,000 — a fifth of his marketing budget — on a billboard in San Francisco displaying what looked like gibberish: five strings of random numbers.\nThe numbers were actually AI tokens. Decoded, they led to a coding challenge: build an algorithm to act as a digital bouncer at Berghain, the Berlin nightclub famous for rejecting nearly everyone at the door. Within days, thousands attempted the puzzle. 430 cracked it. Some got hired. The winner flew to Berlin, all expenses paid.\nThat unconventional approach has now attracted $69 million in Series B funding, led by Ribbit Capital with participation from Evantic and existing investors Sequoia Capital, Conviction, and Pear VC. The round values Listen Labs at $500 million and brings its total capital to $100 million. In nine months since launch, the company has grown annualized revenue by 15x to eight figures and conducted over one million AI-powered interviews.\n\n\"When you obsess over customers, everything else follows,\" Wahlforss said in an interview with VentureBeat. \"Teams that use Listen bring the customer into every decision, from marketing to product, and when the customer is delighted, everyone is.\"\nWhy traditional market research is broken, and what Listen Labs is building to fix it\nListen's AI researcher finds participants, conducts in-depth interviews, and delivers actionable insights in hours, not weeks. The platform replaces the traditional choice between quantitative surveys — which provide statistical precision but miss nuance—and qualitative interviews, which deliver depth but cannot scale.\nWahlforss explained the limitation of existing approaches: \"Essentially surveys give you false precision because people end up answering the same question... You can't get the outliers. People are actually not honest on surveys.\" The alternative, one-on-one human interviews, \"gives you a lot of depth. You can ask follow up questions. You can kind of double check if they actually know what they're talking about. And the problem is you can't scale that.\"\nThe platform works in four steps: users create a study with AI assistance, Listen recruits participants from its global network of 30 million people, an AI moderator conducts in-depth interviews with follow-up questions, and results are packaged into executive-ready reports including key themes, highlight reels, and slide decks.\nWhat distinguishes Listen's approach is its use of open-ended video conversations rather than multiple-choice forms. \"In a survey, you can kind of guess what you should answer, and you have four options,\" Wahlforss said. \"Oh, they probably want me to buy high income. Let me click on that button versus an open ended response. It just generates much more honesty.\"\nThe dirty secret of the $140 billion market research industry: rampant fraud\nListen finds and qualifies the right participants in its global network of 30 million people. But building that panel required confronting what Wahlforss called \"one of the most shocking things that we've learned when we entered this industry\"—rampant fraud.\n\"Essentially, there's a financial transaction involved, which means there will be bad players,\" he explained. \"We actually had some of the largest companies, some of them have billions in revenue, send us people who claim to be kind of enterprise buyers to our platform and our system immediately detected, like, fraud, fraud, fraud, fraud, fraud.\"\nThe company built what it calls a \"quality guard\" that cross-references LinkedIn profiles with video responses to verify identity, checks consistency across how participants answer questions, and flags suspicious patterns. The result, according to Wahlforss: \"People talk three times more. They're much more honest when they talk about sensitive topics like politics and mental health.\"\nEmeritus, an online education company that uses Listen, reported that approximately 20% of survey responses previously fell into the fraudulent or low-quality category. With Listen, they reduced this to almost zero. \"We did not have to replace any responses because of fraud or gibberish information,\" said Gabrielli Tiburi, Assistant Manager of Customer Insights at Emeritus.\nHow Microsoft, Sweetgreen, and Chubbies are using AI interviews to build better products\nThe speed advantage has proven central to Listen's pitch. Traditional customer research at Microsoft could take four to six weeks to generate insights. \"By the time we get to them, either the decision has been made or we lose out on the opportunity to actually influence it,\" said Romani Patel, Senior Research Manager at Microsoft.\nWith Listen, Microsoft can now get insights in days, and in many cases, within hours.\nThe platform has already powered several high-profile initiatives. Microsoft used Listen Labs to collect global customer stories for its 50th anniversary celebration. \"We wanted users to share how Copilot is empowering them to bring their best self forward,\" Patel said, \"and we were able to collect those user video stories within a day.\" Traditionally, that kind of work would have taken six to eight weeks.\nSimple Modern, an Oklahoma-based drinkware company, used Listen to test a new product concept. The process took about an hour to write questions, an hour to launch the study, and 2.5 hours to receive feedback from 120 people across the country. \"We went from 'Should we even have this product?' to 'How should we launch it?'\" said Chris Hoyle, the company's Chief Marketing Officer.\nChubbies, the shorts brand, achieved a 24x increase in youth research participation—growing from 5 to 120 participants — by using Listen to overcome the scheduling challenges of traditional focus groups with children. \"There's school, sports, dinner, and homework,\" explained Lauren Neville, Director of Insights and Innovation. \"I had to find a way to hear from them that fit into their schedules.\"\nThe company also discovered product issues through AI interviews that might have gone undetected otherwise. Wahlforss described how the AI \"through conversations, realized there were like issues with the the kids short line, and decided to, like, interview hundreds of kids. And I understand that there were issues in the liner of the shorts and that they were, like, scratchy, quote, unquote, according to the people interviewed.\" The redesigned product became \"a blockbuster hit.\"\nThe Jevons paradox explains why cheaper research creates more demand, not less\nListen Labs is entering a massive but fragmented market. Wahlforss cited research from Andreessen Horowitz estimating the market research industry at roughly $140 billion annually, populated by legacy players — some with more than a billion dollars in revenue — that he believes are vulnerable to disruption.\n\"There are very much existing budget lines that we are replacing,\" Wahlforss said. \"Why we're replacing them is that one, they're super costly. Two, they're kind of stuck in this old paradigm of choosing between a survey or interview, and they also take months to work with.\"\nBut the more intriguing dynamic may be that AI-powered research doesn't just replace existing spending — it creates new demand. Wahlforss invoked the Jevons paradox, an economic principle that occurs when technological advancements make a resource more efficient to use, but increased efficiency leads to increased overall consumption rather than decreased consumption.\n\"What I've noticed is that as something gets cheaper, you don't need less of it. You want more of it,\" Wahlforss explained. \"There's infinite demand for customer understanding. So the researchers on the team can do an order of magnitude more research, and also other people who weren't researchers before can now do that as part of their job.\"\nInside the elite engineering team that built Listen Labs before they had a working toilet\nListen Labs traces its origins to a consumer app that Wahlforss and his co-founder built after meeting at Harvard. \"We built this consumer app that got 20,000 downloads in one day,\" Wahlforss recalled. \"We had all these users, and we were thinking like, okay, what can we do to get to know them better? And we built this prototype of what Listen is today.\"\nThe founding team brings an unusual pedigree. Wahlforss's co-founder \"was the national champion in competitive programming in Germany, and he worked at Tesla Autopilot.\" The company claims that 30% of its engineering team are medalists from the International Olympiad in Informatics — the same competition that produced the founders of Cognition, the AI coding startup.\nThe Berghain billboard stunt generated approximately 5 million views across social media, according to Wahlforss. It reflected the intensity of the talent war in the Bay Area.\n\"We had to do these things because some of our, like early employees, joined the company before we had a working toilet,\" he said. \"But now we fixed that situation.\"\nThe company grew from 5 to 40 employees in 2024 and plans to reach 150 this year. It hires engineers for non-engineering roles across marketing, growth, and operations — a bet that in the AI era, technical fluency matters everywhere.\nSynthetic customers and automated decisions: what Listen Labs is building next\nWahlforss outlined an ambitious product roadmap that pushes into more speculative territory. The company is building \"the ability to simulate your customers, so you can take all of those interviews we've done, and then extrapolate based on that and create synthetic users or simulated user voices.\"\nBeyond simulation, Listen aims to enable automated action based on research findings. \"Can you not just make recommendations, but also create spawn agents to either change things in code or some customer churns? Can you give them a discount and try to bring them back?\"\nWahlforss acknowledged the ethical implications. \"Obviously, as you said, there's kind of ethical concerns there. Of like, automated decision making overall can be bad, but we will have considerable guardrails to make sure that the companies are always in the loop.\"\nThe company already handles sensitive data with care. \"We don't train on any of the data,\" Wahlforss said. \"We will also scrub any sensitive PII automatically so the model can detect that. And there are times when, for example, you work with investors, where if you accidentally mention something that could be material, non public information, the AI can actually detect that and remove any information like that.\"\nHow AI could reshape the future of product development\nPerhaps the most provocative implication of Listen's model is how it could reshape product development itself. Wahlforss described a customer — an Australian startup — that has adopted what amounts to a continuous feedback loop.\n\"They're based in Australia, so they're coding during the day, and then in their night, they're releasing a Listen study with an American audience. Listen validates whatever they built during the day, and they get feedback on that. They can then plug that feedback directly into coding tools like Claude Code and iterate.\"\nThe vision extends Y Combinator's famous dictum — \"write code, talk to users\" — into an automated cycle. \"Write code is now getting automated. And I think like talk to users will be as well, and you'll have this kind of infinite loop where you can start to ship this truly amazing product, almost kind of autonomously.\"\nWhether that vision materializes depends on factors beyond Listen's control — the continued improvement of AI models, enterprise willingness to trust automated research, and whether speed truly correlates with better products. A 2024 MIT study found that 95% of AI pilots fail to move into production, a statistic Wahlforss cited as the reason he emphasizes quality over demos.\n\"I'm constantly have to emphasize like, let's make sure the quality is there and the details are right,\" he said.\nBut the company's growth suggests appetite for the experiment. Microsoft's Patel said Listen has \"removed the drudgery of research and brought the fun and joy back into my work.\" Chubbies is now pushing its founder to give everyone in the company a login. Sling Money, a stablecoin payments startup, can create a survey in ten minutes and receive results the same day.\n\"It's a total game changer,\" said Ali Romero, Sling Money's marketing manager.\nWahlforss has a different phrase for what he's building. When asked about the tension between speed and rigor — the long-held belief that moving fast means cutting corners — he cited Nat Friedman, the former GitHub CEO and Listen investor, who keeps a list of one-liners on his website.\nOne of them: \"Slow is fake.\"\nIt's an aggressive claim for an industry built on methodological caution. But Listen Labs is betting that in the AI era, the companies that listen fastest will be the ones that win. The only question is whether customers will talk back." }, "pairedItem": [ { "item": 4 } ] }, { "json": { "guid": "1dxpkoyNmn8ceRMHCKLlT8", "link": "https://venturebeat.com/technology/salesforce-rolls-out-new-slackbot-ai-agent-as-it-battles-microsoft-and", "title": "Salesforce rolls out new Slackbot AI agent as it battles Microsoft and Google in workplace AI", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Salesforce on Tuesday launched an entirely rebuilt version of Slackbot, the company's workplace assistant, transforming it from a simple notification tool into what executives describe as a fully powered AI agent capable of searching enterprise data, drafting documents, and taking action on behalf of employees.
The new Slackbot, now generally available to Business+ and Enterprise+ customers, is Salesforce's most aggressive move yet to position Slack at the center of the emerging "agentic AI" movement — where software agents work alongside humans to complete complex tasks. The launch comes as Salesforce attempts to convince investors that artificial intelligence will bolster its products rather than render them obsolete.
"Slackbot isn't just another copilot or AI assistant," said Parker Harris, Salesforce co-founder and Slack's chief technology officer, in an exclusive interview with Salesforce. "It's the front door to the agentic enterprise, powered by Salesforce."
Harris was blunt about what distinguishes the new Slackbot from its predecessor: "The old Slackbot was, you know, a little tricycle, and the new Slackbot is like, you know, a Porsche."
The original Slackbot, which has existed since Slack's early days, performed basic algorithmic tasks — reminding users to add colleagues to documents, suggesting channel archives, and delivering simple notifications. The new version runs on an entirely different architecture built around a large language model and sophisticated search capabilities that can access Salesforce records, Google Drive files, calendar data, and years of Slack conversations.
"It's two different things," Harris explained. "The old Slackbot was algorithmic and fairly simple. The new Slackbot is brand new — it's based around an LLM and a very robust search engine, and connections to third-party search engines, third-party enterprise data."
Salesforce chose to retain the Slackbot brand despite the fundamental technical overhaul. "People know what Slackbot is, and so we wanted to carry that forward," Harris said.
The new Slackbot runs on Claude, Anthropic's large language model, a choice driven partly by compliance requirements. Slack's commercial service operates under FedRAMP Moderate certification to serve U.S. federal government customers, and Harris said Anthropic was "the only provider that could give us a compliant LLM" when Slack began building the new system.
But that exclusivity won't last. "We are, this year, going to support additional providers," Harris said. "We have a great relationship with Google. Gemini is incredible — performance is great, cost is great. So we're going to use Gemini for some things." He added that OpenAI remains a possibility as well.
Harris echoed Salesforce CEO Marc Benioff's view that large language models are becoming commoditized: "You've heard Marc talk about LLMs are commodities, that they're democratized. I call them CPUs."
On the sensitive question of training data, Harris was unequivocal: Salesforce does not train any models on customer data. "Models don't have any sort of security," he explained. "If we trained it on some confidential conversation that you and I have, I don't want Carolyn to know — if I train it into the LLM, there is no way for me to say you get to see the answer, but Carolyn doesn't."
Salesforce has been testing the new Slackbot internally for months, rolling it out to all 80,000 employees. According to Ryan Gavin, Slack's chief marketing officer, the results have been striking: "It's the fastest adopted product in Salesforce history."
Internal data shows that two-thirds of Salesforce employees have tried the new Slackbot, with 80% of those users continuing to use it regularly. Internal satisfaction rates reached 96% — the highest for any AI feature Slack has shipped. Employees report saving between two and 20 hours per week.
The adoption happened largely organically. "I think it was about five days, and a Canvas was developed by our employees called 'The Most Stealable Slackbot Prompts,'" Gavin said. "People just started adding to it organically. I think it's up to 250-plus prompts that are in this Canvas right now."
Kate Crotty, a principal UX researcher at Salesforce, found that 73% of internal adoption was driven by social sharing rather than top-down mandates. "Everybody is there to help each other learn and communicate hacks," she said.
During a product demonstration, Amy Bauer, Slack's product experience designer, showed how Slackbot can synthesize information across multiple sources. In one example, she asked Slackbot to analyze customer feedback from a pilot program, upload an image of a usage dashboard, and have Slackbot correlate the qualitative and quantitative data.
"This is where Slackbot really earns its keep for me," Bauer explained. "What it's doing is not just simply reading the image — it's actually looking at the image and comparing it to the insight it just generated for me."
Slackbot can then query Salesforce to find enterprise accounts with open deals that might be good candidates for early access, creating what Bauer called "a really great justification and plan to move forward." Finally, it can synthesize all that information into a Canvas — Slack's collaborative document format — and find calendar availability among stakeholders to schedule a review meeting.
"Up until this point, we have been working in a one-to-one capacity with Slackbot," Bauer said. "But one of the benefits that I can do now is take this insight and have it generate this into a Canvas, a shared workspace where I can iterate on it, refine it with Slackbot, or share it out with my team."
Rob Seaman, Slack's chief product officer, said the Canvas creation demonstrates where the product is heading: "This is making a tool call internally to Slack Canvas to actually write, effectively, a shared document. But it signals where we're going with Slackbot — we're eventually going to be adding in additional third-party tool calls."
Among Salesforce's pilot customers is Beast Industries, the parent company of YouTube star MrBeast. Luis Madrigal, the company's chief information officer, joined the launch announcement to describe his experience.
"As somebody who has rolled out enterprise technologies for over two decades now, this was practically one of the easiest," Madrigal said. "The plumbing is there. Slack as an implementation, Enterprise Tools — being able to turn on the Slackbot and the Slack AI functionality was as simple as having my team go in, review, do a quick security review."
Madrigal said his security team signed off "rather quickly" — unusual for enterprise AI deployments — because Slackbot accesses only the information each individual user already has permission to view. "Given all the guardrails you guys have put into place for Slackbot to be unique and customized to only the information that each individual user has, only the conversations and the Slack rooms and Slack channels that they're part of—that made my security team sign off rather quickly."
One Beast Industries employee, Sinan, the head of Beast Games marketing, reported saving "at bare minimum, 90 minutes a day." Another employee, Spencer, a creative supervisor, described it as "an assistant who's paying attention when I'm not."
Other pilot customers include Slalom, reMarkable, Xero, Mercari, and Engine. Mollie Bodensteiner, SVP of Operations at Engine, called Slackbot "an absolute 'chaos tamer' for our team," estimating it saves her about 30 minutes daily "just by eliminating context switching."
The launch puts Salesforce in direct competition with Microsoft's Copilot, which is integrated into Teams and the broader Microsoft 365 suite, as well as Google's Gemini integrations across Workspace. When asked what distinguishes Slackbot from these alternatives, Seaman pointed to context and convenience.
"The thing that makes it most powerful for our customers and users is the proximity — it's just right there in your Slack," Seaman said. "There's a tremendous convenience affordance that's naturally built into it."
The deeper advantage, executives argue, is that Slackbot already understands users' work without requiring setup or training. "Most AI tools sound the same no matter who is using them," the company's announcement stated. "They lack context, miss nuance, and force you to jump between tools to get anything done."
Harris put it more directly: "If you've ever had that magic experience with AI — I think ChatGPT is a great example, it's a great experience from a consumer perspective — Slackbot is really what we're doing in the enterprise, to be this employee super agent that is loved, just like people love using Slack."
Amy Bauer emphasized the frictionless nature of the experience. "Slackbot is inherently grounded in the context, in the data that you have in Slack," she said. "So as you continue working in Slack, Slackbot gets better because it's grounded in the work that you're doing there. There is no setup. There is no configuration for those end users."
Salesforce positions Slackbot as what Harris calls a "super agent" — a central hub that can eventually coordinate with other AI agents across an organization.
"Every corporation is going to have an employee super agent," Harris said. "Slackbot is essentially taking the magic of what Slack does. We think that Slackbot, and we're really excited about it, is going to be that."
The vision extends to third-party agents already launching in Slack. Last month, Anthropic released a preview of Claude Code for Slack, allowing developers to interact with Claude's coding capabilities directly in chat threads. OpenAI, Google, Vercel, and others have also built agents for the platform.
"Most of the net-new apps that are being deployed to Slack are agents," Seaman noted during the press conference. "This is proof of the promise of humans and agents coexisting and working together in Slack to solve problems."
Harris described a future where Slackbot becomes an MCP (Model Context Protocol) client, able to leverage tools from across the software ecosystem — similar to how the developer tool Cursor works. "Slack can be an MCP client, and Slackbot will be the hub of that, leveraging all these tools out in the world, some of which will be these amazing agents," he said.
But Harris also cautioned against over-promising on multi-agent coordination. "I still think we're in the single agent world," he said. "FY26 is going to be the year where we started to see more coordination. But we're going to do it with customer success in mind, and not demonstrate and talk about, like, 'I've got 1,000 agents working together,' because I think that's unrealistic."
Slackbot is included at no additional cost for customers on Business+ and Enterprise+ plans. "There's no additional fees customers have to do," Gavin confirmed. "If they're on one of those plans, they're going to get Slackbot."
However, some enterprise customers may face other cost pressures related to Salesforce's broader data strategy. CIOs may see price increases for third-party applications that work with Salesforce data, as effects of higher charges for API access ripple through the software supply chain.
Fivetran CEO George Fraser has warned that Salesforce's shift in pricing policy for API access could have tangible consequences for enterprises relying on Salesforce as a system of record. "They might not be able to use Fivetran to replicate their data to Snowflake and instead have to use Salesforce Data Cloud. Or they might find that they are not able to interact with their data via ChatGPT, and instead have to use Agentforce," Fraser said in a recent CIO report.
Salesforce has framed the pricing change as standard industry practice.
The new Slackbot begins rolling out today and will reach all eligible customers by the end of February. Mobile availability will complete by March 3, Bauer confirmed during her interview with VentureBeat.
Some capabilities remain works in progress. Calendar reading and availability checking are available at launch, but the ability to actually book meetings is "coming a few weeks after," according to Seaman. Image generation is not currently supported, though Bauer said it's "something that we are looking at in the future."
When asked about integration with competing CRM systems like HubSpot and Microsoft Dynamics, Salesforce representatives declined to provide specifics during the interview, though they acknowledged the question touched on key competitive differentiators.
The Slackbot launch is Salesforce's bet that the future of enterprise work is conversational — that employees will increasingly prefer to interact with AI through natural language rather than navigating traditional software interfaces.
Harris described Slack's product philosophy using principles like "don't make me think" and "be a great host." The goal, he said, is for Slackbot to surface information proactively rather than requiring users to hunt for it.
"One of the revelations for me is LLMs applied to unstructured information are incredible," Harris said. "And the amount of value you have if you're a Slack user, if your corporation uses Slack — the amount of value in Slack is unbelievable. Because you're talking about work, you're sharing documents, you're making decisions, but you can't as a human go through that and really get the same value that an LLM can do."
Looking ahead, Harris expects the interfaces themselves to evolve beyond pure conversation. "We're kind of saturating what we can do with purely conversational UIs," he said. "I think we'll start to see agents building an interface that best suits your intent, as opposed to trying to surface something within a conversational interface that matches your intent."
Microsoft, Google, and a growing roster of AI startups are placing similar bets — that the winning enterprise AI will be the one embedded in the tools workers already use, not another application to learn. The race to become that invisible layer of workplace intelligence is now fully underway.
For Salesforce, the stakes extend beyond a single product launch. After a bruising year on Wall Street and persistent questions about whether AI threatens its core business, the company is wagering that Slackbot can prove the opposite — that the tens of millions of people already chatting in Slack every day is not a vulnerability, but an unassailable advantage.
Haley Gault, the Salesforce account executive in Pittsburgh who stumbled upon the new Slackbot on a snowy morning, captured the shift in a single sentence: "I honestly can't imagine working for another company not having access to these types of tools. This is just how I work now."
That's precisely what Salesforce is counting on.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-13T13:00:00.000Z", "pubDate": "Tue, 13 Jan 2026 13:00:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/4Xrcg14GLKFlwSEnuEzxyS/21c85d29d03c4c974076475c009e3b38/nuneybits_Vector_art_of_chat_bubbles_on_a_computer_screen_in_th_5018a7ea-3496-4103-8453-7ba1b129189a.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI", "Automation" ], "contentSnippet": "Salesforce on Tuesday launched an entirely rebuilt version of Slackbot, the company's workplace assistant, transforming it from a simple notification tool into what executives describe as a fully powered AI agent capable of searching enterprise data, drafting documents, and taking action on behalf of employees.\nThe new Slackbot, now generally available to Business+ and Enterprise+ customers, is Salesforce's most aggressive move yet to position Slack at the center of the emerging \"agentic AI\" movement — where software agents work alongside humans to complete complex tasks. The launch comes as Salesforce attempts to convince investors that artificial intelligence will bolster its products rather than render them obsolete.\n\"Slackbot isn't just another copilot or AI assistant,\" said Parker Harris, Salesforce co-founder and Slack's chief technology officer, in an exclusive interview with Salesforce. \"It's the front door to the agentic enterprise, powered by Salesforce.\"\nFrom tricycle to Porsche: Salesforce rebuilt Slackbot from the ground up\nHarris was blunt about what distinguishes the new Slackbot from its predecessor: \"The old Slackbot was, you know, a little tricycle, and the new Slackbot is like, you know, a Porsche.\"\nThe original Slackbot, which has existed since Slack's early days, performed basic algorithmic tasks — reminding users to add colleagues to documents, suggesting channel archives, and delivering simple notifications. The new version runs on an entirely different architecture built around a large language model and sophisticated search capabilities that can access Salesforce records, Google Drive files, calendar data, and years of Slack conversations.\n\"It's two different things,\" Harris explained. \"The old Slackbot was algorithmic and fairly simple. The new Slackbot is brand new — it's based around an LLM and a very robust search engine, and connections to third-party search engines, third-party enterprise data.\"\nSalesforce chose to retain the Slackbot brand despite the fundamental technical overhaul. \"People know what Slackbot is, and so we wanted to carry that forward,\" Harris said.\nWhy Anthropic's Claude powers the new Slackbot — and which AI models could come next\nThe new Slackbot runs on Claude, Anthropic's large language model, a choice driven partly by compliance requirements. Slack's commercial service operates under FedRAMP Moderate certification to serve U.S. federal government customers, and Harris said Anthropic was \"the only provider that could give us a compliant LLM\" when Slack began building the new system.\nBut that exclusivity won't last. \"We are, this year, going to support additional providers,\" Harris said. \"We have a great relationship with Google. Gemini is incredible — performance is great, cost is great. So we're going to use Gemini for some things.\" He added that OpenAI remains a possibility as well.\nHarris echoed Salesforce CEO Marc Benioff's view that large language models are becoming commoditized: \"You've heard Marc talk about LLMs are commodities, that they're democratized. I call them CPUs.\"\nOn the sensitive question of training data, Harris was unequivocal: Salesforce does not train any models on customer data. \"Models don't have any sort of security,\" he explained. \"If we trained it on some confidential conversation that you and I have, I don't want Carolyn to know — if I train it into the LLM, there is no way for me to say you get to see the answer, but Carolyn doesn't.\"\nInside Salesforce's internal experiment: 80,000 employees tested Slackbot with striking results\nSalesforce has been testing the new Slackbot internally for months, rolling it out to all 80,000 employees. According to Ryan Gavin, Slack's chief marketing officer, the results have been striking: \"It's the fastest adopted product in Salesforce history.\"\nInternal data shows that two-thirds of Salesforce employees have tried the new Slackbot, with 80% of those users continuing to use it regularly. Internal satisfaction rates reached 96% — the highest for any AI feature Slack has shipped. Employees report saving between two and 20 hours per week.\nThe adoption happened largely organically. \"I think it was about five days, and a Canvas was developed by our employees called 'The Most Stealable Slackbot Prompts,'\" Gavin said. \"People just started adding to it organically. I think it's up to 250-plus prompts that are in this Canvas right now.\"\nKate Crotty, a principal UX researcher at Salesforce, found that 73% of internal adoption was driven by social sharing rather than top-down mandates. \"Everybody is there to help each other learn and communicate hacks,\" she said.\nHow Slackbot transforms scattered enterprise data into executive-ready insights\nDuring a product demonstration, Amy Bauer, Slack's product experience designer, showed how Slackbot can synthesize information across multiple sources. In one example, she asked Slackbot to analyze customer feedback from a pilot program, upload an image of a usage dashboard, and have Slackbot correlate the qualitative and quantitative data.\n\"This is where Slackbot really earns its keep for me,\" Bauer explained. \"What it's doing is not just simply reading the image — it's actually looking at the image and comparing it to the insight it just generated for me.\"\nSlackbot can then query Salesforce to find enterprise accounts with open deals that might be good candidates for early access, creating what Bauer called \"a really great justification and plan to move forward.\" Finally, it can synthesize all that information into a Canvas — Slack's collaborative document format — and find calendar availability among stakeholders to schedule a review meeting.\n\"Up until this point, we have been working in a one-to-one capacity with Slackbot,\" Bauer said. \"But one of the benefits that I can do now is take this insight and have it generate this into a Canvas, a shared workspace where I can iterate on it, refine it with Slackbot, or share it out with my team.\"\nRob Seaman, Slack's chief product officer, said the Canvas creation demonstrates where the product is heading: \"This is making a tool call internally to Slack Canvas to actually write, effectively, a shared document. But it signals where we're going with Slackbot — we're eventually going to be adding in additional third-party tool calls.\"\nMrBeast's company became a Slackbot guinea pig—and employees say they're saving 90 minutes a day\nAmong Salesforce's pilot customers is Beast Industries, the parent company of YouTube star MrBeast. Luis Madrigal, the company's chief information officer, joined the launch announcement to describe his experience.\n\"As somebody who has rolled out enterprise technologies for over two decades now, this was practically one of the easiest,\" Madrigal said. \"The plumbing is there. Slack as an implementation, Enterprise Tools — being able to turn on the Slackbot and the Slack AI functionality was as simple as having my team go in, review, do a quick security review.\"\nMadrigal said his security team signed off \"rather quickly\" — unusual for enterprise AI deployments — because Slackbot accesses only the information each individual user already has permission to view. \"Given all the guardrails you guys have put into place for Slackbot to be unique and customized to only the information that each individual user has, only the conversations and the Slack rooms and Slack channels that they're part of—that made my security team sign off rather quickly.\"\nOne Beast Industries employee, Sinan, the head of Beast Games marketing, reported saving \"at bare minimum, 90 minutes a day.\" Another employee, Spencer, a creative supervisor, described it as \"an assistant who's paying attention when I'm not.\"\nOther pilot customers include Slalom, reMarkable, Xero, Mercari, and Engine. Mollie Bodensteiner, SVP of Operations at Engine, called Slackbot \"an absolute 'chaos tamer' for our team,\" estimating it saves her about 30 minutes daily \"just by eliminating context switching.\"\nSlackbot vs. Microsoft Copilot vs. Google Gemini: The fight for enterprise AI dominance\nThe launch puts Salesforce in direct competition with Microsoft's Copilot, which is integrated into Teams and the broader Microsoft 365 suite, as well as Google's Gemini integrations across Workspace. When asked what distinguishes Slackbot from these alternatives, Seaman pointed to context and convenience.\n\"The thing that makes it most powerful for our customers and users is the proximity — it's just right there in your Slack,\" Seaman said. \"There's a tremendous convenience affordance that's naturally built into it.\"\nThe deeper advantage, executives argue, is that Slackbot already understands users' work without requiring setup or training. \"Most AI tools sound the same no matter who is using them,\" the company's announcement stated. \"They lack context, miss nuance, and force you to jump between tools to get anything done.\"\nHarris put it more directly: \"If you've ever had that magic experience with AI — I think ChatGPT is a great example, it's a great experience from a consumer perspective — Slackbot is really what we're doing in the enterprise, to be this employee super agent that is loved, just like people love using Slack.\"\nAmy Bauer emphasized the frictionless nature of the experience. \"Slackbot is inherently grounded in the context, in the data that you have in Slack,\" she said. \"So as you continue working in Slack, Slackbot gets better because it's grounded in the work that you're doing there. There is no setup. There is no configuration for those end users.\"\nSalesforce's ambitious plan to make Slackbot the one 'super agent' that controls all the others\nSalesforce positions Slackbot as what Harris calls a \"super agent\" — a central hub that can eventually coordinate with other AI agents across an organization.\n\"Every corporation is going to have an employee super agent,\" Harris said. \"Slackbot is essentially taking the magic of what Slack does. We think that Slackbot, and we're really excited about it, is going to be that.\"\nThe vision extends to third-party agents already launching in Slack. Last month, Anthropic released a preview of Claude Code for Slack, allowing developers to interact with Claude's coding capabilities directly in chat threads. OpenAI, Google, Vercel, and others have also built agents for the platform.\n\"Most of the net-new apps that are being deployed to Slack are agents,\" Seaman noted during the press conference. \"This is proof of the promise of humans and agents coexisting and working together in Slack to solve problems.\"\nHarris described a future where Slackbot becomes an MCP (Model Context Protocol) client, able to leverage tools from across the software ecosystem — similar to how the developer tool Cursor works. \"Slack can be an MCP client, and Slackbot will be the hub of that, leveraging all these tools out in the world, some of which will be these amazing agents,\" he said.\nBut Harris also cautioned against over-promising on multi-agent coordination. \"I still think we're in the single agent world,\" he said. \"FY26 is going to be the year where we started to see more coordination. But we're going to do it with customer success in mind, and not demonstrate and talk about, like, 'I've got 1,000 agents working together,' because I think that's unrealistic.\"\nSlackbot costs nothing extra, but Salesforce's data access fees could squeeze some customers\nSlackbot is included at no additional cost for customers on Business+ and Enterprise+ plans. \"There's no additional fees customers have to do,\" Gavin confirmed. \"If they're on one of those plans, they're going to get Slackbot.\"\nHowever, some enterprise customers may face other cost pressures related to Salesforce's broader data strategy. CIOs may see price increases for third-party applications that work with Salesforce data, as effects of higher charges for API access ripple through the software supply chain.\nFivetran CEO George Fraser has warned that Salesforce's shift in pricing policy for API access could have tangible consequences for enterprises relying on Salesforce as a system of record. \"They might not be able to use Fivetran to replicate their data to Snowflake and instead have to use Salesforce Data Cloud. Or they might find that they are not able to interact with their data via ChatGPT, and instead have to use Agentforce,\" Fraser said in a recent CIO report.\nSalesforce has framed the pricing change as standard industry practice.\nWhat Slackbot can do today, what's coming in weeks, and what's still on the roadmap\nThe new Slackbot begins rolling out today and will reach all eligible customers by the end of February. Mobile availability will complete by March 3, Bauer confirmed during her interview with VentureBeat.\nSome capabilities remain works in progress. Calendar reading and availability checking are available at launch, but the ability to actually book meetings is \"coming a few weeks after,\" according to Seaman. Image generation is not currently supported, though Bauer said it's \"something that we are looking at in the future.\"\nWhen asked about integration with competing CRM systems like HubSpot and Microsoft Dynamics, Salesforce representatives declined to provide specifics during the interview, though they acknowledged the question touched on key competitive differentiators.\nSalesforce is betting the future of work looks like a chat window—and it's not alone\nThe Slackbot launch is Salesforce's bet that the future of enterprise work is conversational — that employees will increasingly prefer to interact with AI through natural language rather than navigating traditional software interfaces.\nHarris described Slack's product philosophy using principles like \"don't make me think\" and \"be a great host.\" The goal, he said, is for Slackbot to surface information proactively rather than requiring users to hunt for it.\n\"One of the revelations for me is LLMs applied to unstructured information are incredible,\" Harris said. \"And the amount of value you have if you're a Slack user, if your corporation uses Slack — the amount of value in Slack is unbelievable. Because you're talking about work, you're sharing documents, you're making decisions, but you can't as a human go through that and really get the same value that an LLM can do.\"\nLooking ahead, Harris expects the interfaces themselves to evolve beyond pure conversation. \"We're kind of saturating what we can do with purely conversational UIs,\" he said. \"I think we'll start to see agents building an interface that best suits your intent, as opposed to trying to surface something within a conversational interface that matches your intent.\"\nMicrosoft, Google, and a growing roster of AI startups are placing similar bets — that the winning enterprise AI will be the one embedded in the tools workers already use, not another application to learn. The race to become that invisible layer of workplace intelligence is now fully underway.\nFor Salesforce, the stakes extend beyond a single product launch. After a bruising year on Wall Street and persistent questions about whether AI threatens its core business, the company is wagering that Slackbot can prove the opposite — that the tens of millions of people already chatting in Slack every day is not a vulnerability, but an unassailable advantage.\nHaley Gault, the Salesforce account executive in Pittsburgh who stumbled upon the new Slackbot on a snowy morning, captured the shift in a single sentence: \"I honestly can't imagine working for another company not having access to these types of tools. This is just how I work now.\"\nThat's precisely what Salesforce is counting on." }, "pairedItem": [ { "item": 4 } ] }, { "json": { "guid": "7EaNv4Ip2AUlW9RClUv2cB", "link": "https://venturebeat.com/technology/anthropic-launches-cowork-a-claude-desktop-agent-that-works-in-your-files-no", "title": "Anthropic launches Cowork, a Claude Desktop agent that works in your files — no coding required", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Anthropic released Cowork on Monday, a new AI agent capability that extends the power of its wildly successful Claude Code tool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, largely using Claude Code itself.
The launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, positioning Anthropic to compete not just with OpenAI and Google in conversational AI, but with Microsoft's Copilot in the burgeoning market for AI-powered productivity tools.
"Cowork lets you complete non-technical tasks much like how developers use Claude Code," the company announced via its official Claude account on X. The feature arrives as a research preview available exclusively to Claude Max subscribers — Anthropic's power-user tier priced between $100 and $200 per month — through the macOS desktop application.
For the past year, the industry narrative has focused on large language models that can write poetry or debug code. With Cowork, Anthropic is betting that the real enterprise value lies in an AI that can open a folder, read a messy pile of receipts, and generate a structured expense report without human hand-holding.
The genesis of Cowork lies in Anthropic's recent success with the developer community. In late 2024, the company released Claude Code, a terminal-based tool that allowed software engineers to automate rote programming tasks. The tool was a hit, but Anthropic noticed a peculiar trend: users were forcing the coding tool to perform non-coding labor.
According to Boris Cherny, an engineer at Anthropic, the company observed users deploying the developer tool for an unexpectedly diverse array of tasks.
"Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven," Cherny wrote on X. "These use cases are diverse and surprising — the reason is that the underlying Claude Agent is the best agent, and Opus 4.5 is the best model."
Recognizing this shadow usage, Anthropic effectively stripped the command-line complexity from their developer tool to create a consumer-friendly interface. In its blog post announcing the feature, Anthropic explained that developers "quickly began using it for almost everything else," which "prompted us to build Cowork: a simpler way for anyone — not just developers — to work with Claude in the very same way."
Unlike a standard chat interface where a user pastes text for analysis, Cowork requires a different level of trust and access. Users designate a specific folder on their local machine that Claude can access. Within that sandbox, the AI agent can read existing files, modify them, or create entirely new ones.
Anthropic offers several illustrative examples: reorganizing a cluttered downloads folder by sorting and intelligently renaming each file, generating a spreadsheet of expenses from a collection of receipt screenshots, or drafting a report from scattered notes across multiple documents.
"In Cowork, you give Claude access to a folder on your computer. Claude can then read, edit, or create files in that folder," the company explained on X. "Try it to create a spreadsheet from a pile of screenshots, or produce a first draft from scattered notes."
The architecture relies on what is known as an "agentic loop." When a user assigns a task, the AI does not merely generate a text response. Instead, it formulates a plan, executes steps in parallel, checks its own work, and asks for clarification if it hits a roadblock. Users can queue multiple tasks and let Claude process them simultaneously — a workflow Anthropic describes as feeling "much less like a back-and-forth and much more like leaving messages for a coworker."
The system is built on Anthropic's Claude Agent SDK, meaning it shares the same underlying architecture as Claude Code. Anthropic notes that Cowork "can take on many of the same tasks that Claude Code can handle, but in a more approachable form for non-coding tasks."
Perhaps the most remarkable detail surrounding Cowork's launch is the speed at which the tool was reportedly built — highlighting a recursive feedback loop where AI tools are being used to build better AI tools.
During a livestream hosted by Dan Shipper, Felix Rieseberg, an Anthropic employee, confirmed that the team built Cowork in approximately a week and a half.
Alex Volkov, who covers AI developments, expressed surprise at the timeline: "Holy shit Anthropic built 'Cowork' in the last... week and a half?!"
This prompted immediate speculation about how much of Cowork was itself built by Claude Code. Simon Smith, EVP of Generative AI at Klick Health, put it bluntly on X: "Claude Code wrote all of Claude Cowork. Can we all agree that we're in at least somewhat of a recursive improvement loop here?"
The implication is profound: Anthropic's AI coding agent may have substantially contributed to building its own non-technical sibling product. If true, this is one of the most visible examples yet of AI systems being used to accelerate their own development and expansion — a strategy that could widen the gap between AI labs that successfully deploy their own agents internally and those that do not.
Cowork doesn't operate in isolation. The feature integrates with Anthropic's existing ecosystem of connectors — tools that link Claude to external information sources and services such as Asana, Notion, PayPal, and other supported partners. Users who have configured these connections in the standard Claude interface can leverage them within Cowork sessions.
Additionally, Cowork can pair with Claude in Chrome, Anthropic's browser extension, to execute tasks requiring web access. This combination allows the agent to navigate websites, click buttons, fill forms, and extract information from the internet — all while operating from the desktop application.
"Cowork includes a number of novel UX and safety features that we think make the product really special," Cherny explained, highlighting "a built-in VM [virtual machine] for isolation, out of the box support for browser automation, support for all your claude.ai data connectors, asking you for clarification when it's unsure."
Anthropic has also introduced an initial set of "skills" specifically designed for Cowork that enhance Claude's ability to create documents, presentations, and other files. These build on the Skills for Claude framework the company announced in October, which provides specialized instruction sets Claude can load for particular types of tasks.
The transition from a chatbot that suggests edits to an agent that makes edits introduces significant risk. An AI that can organize files can, theoretically, delete them.
In a notable display of transparency, Anthropic devoted considerable space in its announcement to warning users about Cowork's potential dangers — an unusual approach for a product launch.
The company explicitly acknowledges that Claude "can take potentially destructive actions (such as deleting local files) if it's instructed to." Because Claude might occasionally misinterpret instructions, Anthropic urges users to provide "very clear guidance" about sensitive operations.
More concerning is the risk of prompt injection attacks — a technique where malicious actors embed hidden instructions in content Claude might encounter online, potentially causing the agent to bypass safeguards or take harmful actions.
"We've built sophisticated defenses against prompt injections," Anthropic wrote, "but agent safety — that is, the task of securing Claude's real-world actions — is still an active area of development in the industry."
The company characterized these risks as inherent to the current state of AI agent technology rather than unique to Cowork. "These risks aren't new with Cowork, but it might be the first time you're using a more advanced tool that moves beyond a simple conversation," the announcement notes.
The launch of Cowork places Anthropic in direct competition with Microsoft, which has spent years attempting to integrate its Copilot AI into the fabric of the Windows operating system with mixed adoption results.
However, Anthropic's approach differs in its isolation. By confining the agent to specific folders and requiring explicit connectors, they are attempting to strike a balance between the utility of an OS-level agent and the security of a sandboxed application.
What distinguishes Anthropic's approach is its bottom-up evolution. Rather than designing an AI assistant and retrofitting agent capabilities, Anthropic built a powerful coding agent first — Claude Code — and is now abstracting its capabilities for broader audiences. This technical lineage may give Cowork more robust agentic behavior from the start.
Claude Code has generated significant enthusiasm among developers since its initial launch as a command-line tool in late 2024. The company expanded access with a web interface in October 2025, followed by a Slack integration in December. Cowork is the next logical step: bringing the same agentic architecture to users who may never touch a terminal.
For now, Cowork remains exclusive to Claude Max subscribers using the macOS desktop application. Users on other subscription tiers — Free, Pro, Team, or Enterprise — can join a waitlist for future access.
Anthropic has signaled clear intentions to expand the feature's reach. The blog post explicitly mentions plans to add cross-device sync and bring Cowork to Windows as the company learns from the research preview.
Cherny set expectations appropriately, describing the product as "early and raw, similar to what Claude Code felt like when it first launched."
To access Cowork, Max subscribers can download or update the Claude macOS app and click on "Cowork" in the sidebar.
For technical decision-makers, the implications of Cowork extend beyond any single product launch. The bottleneck for AI adoption is shifting — no longer is model intelligence the limiting factor, but rather workflow integration and user trust.
Anthropic's goal, as the company puts it, is to make working with Claude feel less like operating a tool and more like delegating to a colleague. Whether mainstream users are ready to hand over folder access to an AI that might misinterpret their instructions remains an open question.
But the speed of Cowork's development — a major feature built in ten days, possibly by the company's own AI — previews a future where the capabilities of these systems compound faster than organizations can evaluate them.
The chatbot has learned to use a file manager. What it learns to use next is anyone's guess.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-12T11:30:00.000Z", "pubDate": "Mon, 12 Jan 2026 11:30:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/wHv1Wez7Ps9wYVYAo9fwT/14b41f606dbf1f5b17994be510407449/nuneybits_Hyper-realistic_image_of_a_retro_computer_with_a_glos_61ffb6e2-7c33-4d45-85f7-69c28693b3ec.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI", "Automation" ], "contentSnippet": "Anthropic released Cowork on Monday, a new AI agent capability that extends the power of its wildly successful Claude Code tool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, largely using Claude Code itself.\nThe launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, positioning Anthropic to compete not just with OpenAI and Google in conversational AI, but with Microsoft's Copilot in the burgeoning market for AI-powered productivity tools.\n\"Cowork lets you complete non-technical tasks much like how developers use Claude Code,\" the company announced via its official Claude account on X. The feature arrives as a research preview available exclusively to Claude Max subscribers — Anthropic's power-user tier priced between $100 and $200 per month — through the macOS desktop application.\nFor the past year, the industry narrative has focused on large language models that can write poetry or debug code. With Cowork, Anthropic is betting that the real enterprise value lies in an AI that can open a folder, read a messy pile of receipts, and generate a structured expense report without human hand-holding.\n\nHow developers using a coding tool for vacation research inspired Anthropic's latest product\nThe genesis of Cowork lies in Anthropic's recent success with the developer community. In late 2024, the company released Claude Code, a terminal-based tool that allowed software engineers to automate rote programming tasks. The tool was a hit, but Anthropic noticed a peculiar trend: users were forcing the coding tool to perform non-coding labor.\nAccording to Boris Cherny, an engineer at Anthropic, the company observed users deploying the developer tool for an unexpectedly diverse array of tasks.\n\n\"Since we launched Claude Code, we saw people using it for all sorts of non-coding work: doing vacation research, building slide decks, cleaning up your email, cancelling subscriptions, recovering wedding photos from a hard drive, monitoring plant growth, controlling your oven,\" Cherny wrote on X. \"These use cases are diverse and surprising — the reason is that the underlying Claude Agent is the best agent, and Opus 4.5 is the best model.\"\nRecognizing this shadow usage, Anthropic effectively stripped the command-line complexity from their developer tool to create a consumer-friendly interface. In its blog post announcing the feature, Anthropic explained that developers \"quickly began using it for almost everything else,\" which \"prompted us to build Cowork: a simpler way for anyone — not just developers — to work with Claude in the very same way.\"\nInside the folder-based architecture that lets Claude read, edit, and create files on your computer\nUnlike a standard chat interface where a user pastes text for analysis, Cowork requires a different level of trust and access. Users designate a specific folder on their local machine that Claude can access. Within that sandbox, the AI agent can read existing files, modify them, or create entirely new ones.\nAnthropic offers several illustrative examples: reorganizing a cluttered downloads folder by sorting and intelligently renaming each file, generating a spreadsheet of expenses from a collection of receipt screenshots, or drafting a report from scattered notes across multiple documents.\n\"In Cowork, you give Claude access to a folder on your computer. Claude can then read, edit, or create files in that folder,\" the company explained on X. \"Try it to create a spreadsheet from a pile of screenshots, or produce a first draft from scattered notes.\"\n\nThe architecture relies on what is known as an \"agentic loop.\" When a user assigns a task, the AI does not merely generate a text response. Instead, it formulates a plan, executes steps in parallel, checks its own work, and asks for clarification if it hits a roadblock. Users can queue multiple tasks and let Claude process them simultaneously — a workflow Anthropic describes as feeling \"much less like a back-and-forth and much more like leaving messages for a coworker.\"\nThe system is built on Anthropic's Claude Agent SDK, meaning it shares the same underlying architecture as Claude Code. Anthropic notes that Cowork \"can take on many of the same tasks that Claude Code can handle, but in a more approachable form for non-coding tasks.\"\nThe recursive loop where AI builds AI: Claude Code reportedly wrote much of Claude Cowork\nPerhaps the most remarkable detail surrounding Cowork's launch is the speed at which the tool was reportedly built — highlighting a recursive feedback loop where AI tools are being used to build better AI tools.\nDuring a livestream hosted by Dan Shipper, Felix Rieseberg, an Anthropic employee, confirmed that the team built Cowork in approximately a week and a half.\nAlex Volkov, who covers AI developments, expressed surprise at the timeline: \"Holy shit Anthropic built 'Cowork' in the last... week and a half?!\"\n\nThis prompted immediate speculation about how much of Cowork was itself built by Claude Code. Simon Smith, EVP of Generative AI at Klick Health, put it bluntly on X: \"Claude Code wrote all of Claude Cowork. Can we all agree that we're in at least somewhat of a recursive improvement loop here?\"\nThe implication is profound: Anthropic's AI coding agent may have substantially contributed to building its own non-technical sibling product. If true, this is one of the most visible examples yet of AI systems being used to accelerate their own development and expansion — a strategy that could widen the gap between AI labs that successfully deploy their own agents internally and those that do not.\nConnectors, browser automation, and skills extend Cowork's reach beyond the local file system\nCowork doesn't operate in isolation. The feature integrates with Anthropic's existing ecosystem of connectors — tools that link Claude to external information sources and services such as Asana, Notion, PayPal, and other supported partners. Users who have configured these connections in the standard Claude interface can leverage them within Cowork sessions.\nAdditionally, Cowork can pair with Claude in Chrome, Anthropic's browser extension, to execute tasks requiring web access. This combination allows the agent to navigate websites, click buttons, fill forms, and extract information from the internet — all while operating from the desktop application.\n\"Cowork includes a number of novel UX and safety features that we think make the product really special,\" Cherny explained, highlighting \"a built-in VM [virtual machine] for isolation, out of the box support for browser automation, support for all your claude.ai data connectors, asking you for clarification when it's unsure.\"\nAnthropic has also introduced an initial set of \"skills\" specifically designed for Cowork that enhance Claude's ability to create documents, presentations, and other files. These build on the Skills for Claude framework the company announced in October, which provides specialized instruction sets Claude can load for particular types of tasks.\nWhy Anthropic is warning users that its own AI agent could delete their files\nThe transition from a chatbot that suggests edits to an agent that makes edits introduces significant risk. An AI that can organize files can, theoretically, delete them.\nIn a notable display of transparency, Anthropic devoted considerable space in its announcement to warning users about Cowork's potential dangers — an unusual approach for a product launch.\nThe company explicitly acknowledges that Claude \"can take potentially destructive actions (such as deleting local files) if it's instructed to.\" Because Claude might occasionally misinterpret instructions, Anthropic urges users to provide \"very clear guidance\" about sensitive operations.\nMore concerning is the risk of prompt injection attacks — a technique where malicious actors embed hidden instructions in content Claude might encounter online, potentially causing the agent to bypass safeguards or take harmful actions.\n\"We've built sophisticated defenses against prompt injections,\" Anthropic wrote, \"but agent safety — that is, the task of securing Claude's real-world actions — is still an active area of development in the industry.\"\nThe company characterized these risks as inherent to the current state of AI agent technology rather than unique to Cowork. \"These risks aren't new with Cowork, but it might be the first time you're using a more advanced tool that moves beyond a simple conversation,\" the announcement notes.\nAnthropic's desktop agent strategy sets up a direct challenge to Microsoft Copilot\nThe launch of Cowork places Anthropic in direct competition with Microsoft, which has spent years attempting to integrate its Copilot AI into the fabric of the Windows operating system with mixed adoption results.\nHowever, Anthropic's approach differs in its isolation. By confining the agent to specific folders and requiring explicit connectors, they are attempting to strike a balance between the utility of an OS-level agent and the security of a sandboxed application.\nWhat distinguishes Anthropic's approach is its bottom-up evolution. Rather than designing an AI assistant and retrofitting agent capabilities, Anthropic built a powerful coding agent first — Claude Code — and is now abstracting its capabilities for broader audiences. This technical lineage may give Cowork more robust agentic behavior from the start.\nClaude Code has generated significant enthusiasm among developers since its initial launch as a command-line tool in late 2024. The company expanded access with a web interface in October 2025, followed by a Slack integration in December. Cowork is the next logical step: bringing the same agentic architecture to users who may never touch a terminal.\nWho can access Cowork now, and what's coming next for Windows and other platforms\nFor now, Cowork remains exclusive to Claude Max subscribers using the macOS desktop application. Users on other subscription tiers — Free, Pro, Team, or Enterprise — can join a waitlist for future access.\nAnthropic has signaled clear intentions to expand the feature's reach. The blog post explicitly mentions plans to add cross-device sync and bring Cowork to Windows as the company learns from the research preview.\nCherny set expectations appropriately, describing the product as \"early and raw, similar to what Claude Code felt like when it first launched.\"\nTo access Cowork, Max subscribers can download or update the Claude macOS app and click on \"Cowork\" in the sidebar.\nThe real question facing enterprise AI adoption\nFor technical decision-makers, the implications of Cowork extend beyond any single product launch. The bottleneck for AI adoption is shifting — no longer is model intelligence the limiting factor, but rather workflow integration and user trust.\nAnthropic's goal, as the company puts it, is to make working with Claude feel less like operating a tool and more like delegating to a colleague. Whether mainstream users are ready to hand over folder access to an AI that might misinterpret their instructions remains an open question.\nBut the speed of Cowork's development — a major feature built in ten days, possibly by the company's own AI — previews a future where the capabilities of these systems compound faster than organizations can evaluate them. \nThe chatbot has learned to use a file manager. What it learns to use next is anyone's guess." }, "pairedItem": [ { "item": 4 } ] }, { "json": { "guid": "4ri2LVmhlv34kOOk94nnsi", "link": "https://venturebeat.com/technology/nous-researchs-nouscoder-14b-is-an-open-source-coding-model-landing-right-in", "title": "Nous Research's NousCoder-14B is an open-source coding model landing right in the Claude Code moment", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "Nous Research, the open-source artificial intelligence startup backed by crypto venture firm Paradigm, released a new competitive programming model on Monday that it says matches or exceeds several larger proprietary systems — trained in just four days using 48 of Nvidia's latest B200 graphics processors.
The model, called NousCoder-14B, is another entry in a crowded field of AI coding assistants, but arrives at a particularly charged moment: Claude Code, the agentic programming tool from rival Anthropic, has dominated social media discussion since New Year's Day, with developers posting breathless testimonials about its capabilities. The simultaneous developments underscore how quickly AI-assisted software development is evolving — and how fiercely companies large and small are competing to capture what many believe will become a foundational technology for how software gets written.
type: embedded-entry-inline id: 74cSyrq6OUrp9SEQ5zOUSl
NousCoder-14B achieves a 67.87 percent accuracy rate on LiveCodeBench v6, a standardized evaluation that tests models on competitive programming problems published between August 2024 and May 2025. That figure represents a 7.08 percentage point improvement over the base model it was trained from, Alibaba's Qwen3-14B, according to Nous Research's technical report published alongside the release.
"I gave Claude Code a description of the problem, it generated what we built last year in an hour," wrote Jaana Dogan, a principal engineer at Google responsible for the Gemini API, in a viral post on X last week that captured the prevailing mood around AI coding tools. Dogan was describing a distributed agent orchestration system her team had spent a year developing — a system Claude Code approximated from a three-paragraph prompt.
The juxtaposition is instructive: while Anthropic's Claude Code has captured imaginations with demonstrations of end-to-end software development, Nous Research is betting that open-source alternatives trained on verifiable problems can close the gap — and that transparency in how these models are built matters as much as raw capability.
What distinguishes the NousCoder-14B release from many competitor announcements is its radical openness. Nous Research published not just the model weights but the complete reinforcement learning environment, benchmark suite, and training harness — built on the company's Atropos framework — enabling any researcher with sufficient compute to reproduce or extend the work.
"Open-sourcing the Atropos stack provides the necessary infrastructure for reproducible olympiad-level reasoning research," noted one observer on X, summarizing the significance for the academic and open-source communities.
The model was trained by Joe Li, a researcher in residence at Nous Research and a former competitive programmer himself. Li's technical report reveals an unexpectedly personal dimension: he compared the model's improvement trajectory to his own journey on Codeforces, the competitive programming platform where participants earn ratings based on contest performance.
Based on rough estimates mapping LiveCodeBench scores to Codeforces ratings, Li calculated that NousCoder-14B's improvemen t— from approximately the 1600-1750 rating range to 2100-2200 — mirrors a leap that took him nearly two years of sustained practice between ages 14 and 16. The model accomplished the equivalent in four days.
"Watching that final training run unfold was quite a surreal experience," Li wrote in the technical report.
But Li was quick to note an important caveat that speaks to broader questions about AI efficiency: he solved roughly 1,000 problems during those two years, while the model required 24,000. Humans, at least for now, remain dramatically more sample-efficient learners.
NousCoder-14B's training process offers a window into the increasingly sophisticated techniques researchers use to improve AI reasoning capabilities through reinforcement learning.
The approach relies on what researchers call "verifiable rewards" — a system where the model generates code solutions, those solutions are executed against test cases, and the model receives a simple binary signal: correct or incorrect. This feedback loop, while conceptually straightforward, requires significant infrastructure to execute at scale.
Nous Research used Modal, a cloud computing platform, to run sandboxed code execution in parallel. Each of the 24,000 training problems contains hundreds of test cases on average, and the system must verify that generated code produces correct outputs within time and memory constraints — 15 seconds and 4 gigabytes, respectively.
The training employed a technique called DAPO (Dynamic Sampling Policy Optimization), which the researchers found performed slightly better than alternatives in their experiments. A key innovation involves "dynamic sampling" — discarding training examples where the model either solves all attempts or fails all attempts, since these provide no useful gradient signal for learning.
The researchers also adopted "iterative context extension," first training the model with a 32,000-token context window before expanding to 40,000 tokens. During evaluation, extending the context further to approximately 80,000 tokens produced the best results, with accuracy reaching 67.87 percent.
Perhaps most significantly, the training pipeline overlaps inference and verification — as soon as the model generates a solution, it begins work on the next problem while the previous solution is being checked. This pipelining, combined with asynchronous training where multiple model instances work in parallel, maximizes hardware utilization on expensive GPU clusters.
Buried in Li's technical report is a finding with significant implications for the future of AI development: the training dataset for NousCoder-14B encompasses "a significant portion of all readily available, verifiable competitive programming problems in a standardized dataset format."
In other words, for this particular domain, the researchers are approaching the limits of high-quality training data.
"The total number of competitive programming problems on the Internet is roughly the same order of magnitude," Li wrote, referring to the 24,000 problems used for training. "This suggests that within the competitive programming domain, we have approached the limits of high-quality data."
This observation echoes growing concern across the AI industry about data constraints. While compute continues to scale according to well-understood economic and engineering principles, training data is "increasingly finite," as Li put it.
"It appears that some of the most important research that needs to be done in the future will be in the areas of synthetic data generation and data efficient algorithms and architectures," he concluded.
The challenge is particularly acute for competitive programming because the domain requires problems with known correct solutions that can be verified automatically. Unlike natural language tasks where human evaluation or proxy metrics suffice, code either works or it doesn't — making synthetic data generation considerably more difficult.
Li identified one potential avenue: training models not just to solve problems but to generate solvable problems, enabling a form of self-play similar to techniques that proved successful in game-playing AI systems. "Once synthetic problem generation is solved, self-play becomes a very interesting direction," he wrote.
Nous Research has carved out a distinctive position in the AI landscape: a company committed to open-source releases that compete with — and sometimes exceed — proprietary alternatives.
The company raised $50 million in April 2025 in a round led by Paradigm, the cryptocurrency-focused venture firm founded by Coinbase co-founder Fred Ehrsam. Total funding reached $65 million, according to some reports. The investment reflected growing interest in decentralized approaches to AI training, an area where Nous Research has developed its Psyche platform.
Previous releases include Hermes 4, a family of models that we reported "outperform ChatGPT without content restrictions," and DeepHermes-3, which the company described as the first "toggle-on reasoning model" — allowing users to activate extended thinking capabilities on demand.
The company has cultivated a distinctive aesthetic and community, prompting some skepticism about whether style might overshadow substance. "Ofc i'm gonna believe an anime pfp company. stop benchmarkmaxxing ffs," wrote one critic on X, referring to Nous Research's anime-style branding and the industry practice of optimizing for benchmark performance.
Others raised technical questions. "Based on the benchmark, Nemotron is better," noted one commenter, referring to Nvidia's family of language models. Another asked whether NousCoder-14B is "agentic focused or just 'one shot' coding" — a distinction that matters for practical software development, where iterating on feedback typically produces better results than single attempts.
The release includes several directions for future work that hint at where AI coding research may be heading.
Multi-turn reinforcement learning tops the list. Currently, the model receives only a final binary reward — pass or fail — after generating a solution. But competitive programming problems typically include public test cases that provide intermediate feedback: compilation errors, incorrect outputs, time limit violations. Training models to incorporate this feedback across multiple attempts could significantly improve performance.
Controlling response length also remains a challenge. The researchers found that incorrect solutions tended to be longer than correct ones, and response lengths quickly saturated available context windows during training — a pattern that various algorithmic modifications failed to resolve.
Perhaps most ambitiously, Li proposed "problem generation and self-play" — training models to both solve and create programming problems. This would address the data scarcity problem directly by enabling models to generate their own training curricula.
"Humans are great at generating interesting and useful problems for other competitive programmers, but it appears that there still exists a significant gap in LLM capabilities in creative problem generation," Li wrote.
The model is available now on Hugging Face under an Apache 2.0 license. For researchers and developers who want to build on the work, Nous Research has published the complete Atropos training stack alongside it.
What took Li two years of adolescent dedication to achieve—climbing from a 1600-level novice to a 2100-rated competitor on Codeforces—an AI replicated in 96 hours. He needed 1,000 problems. The model needed 24,000. But soon enough, these systems may learn to write their own problems, teach themselves, and leave human benchmarks behind entirely.
The question is no longer whether machines can learn to code. It's whether they'll soon be better teachers than we ever were.
\n
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-07T20:00:00.000Z", "pubDate": "Wed, 07 Jan 2026 20:00:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/66Tw6dMGGoSZZOK6XB6gm6/0fafc7520898e26c88edf1de9e74e863/nuneybits_Vector_art_of_radiant_skull_emitting_code_beams_deep__17d19acc-0af7-41ad-ac28-16f09ef5234b.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI" ], "contentSnippet": "Nous Research, the open-source artificial intelligence startup backed by crypto venture firm Paradigm, released a new competitive programming model on Monday that it says matches or exceeds several larger proprietary systems — trained in just four days using 48 of Nvidia's latest B200 graphics processors.\nThe model, called NousCoder-14B, is another entry in a crowded field of AI coding assistants, but arrives at a particularly charged moment: Claude Code, the agentic programming tool from rival Anthropic, has dominated social media discussion since New Year's Day, with developers posting breathless testimonials about its capabilities. The simultaneous developments underscore how quickly AI-assisted software development is evolving — and how fiercely companies large and small are competing to capture what many believe will become a foundational technology for how software gets written.\ntype: embedded-entry-inline id: 74cSyrq6OUrp9SEQ5zOUSl\nNousCoder-14B achieves a 67.87 percent accuracy rate on LiveCodeBench v6, a standardized evaluation that tests models on competitive programming problems published between August 2024 and May 2025. That figure represents a 7.08 percentage point improvement over the base model it was trained from, Alibaba's Qwen3-14B, according to Nous Research's technical report published alongside the release.\n\"I gave Claude Code a description of the problem, it generated what we built last year in an hour,\" wrote Jaana Dogan, a principal engineer at Google responsible for the Gemini API, in a viral post on X last week that captured the prevailing mood around AI coding tools. Dogan was describing a distributed agent orchestration system her team had spent a year developing — a system Claude Code approximated from a three-paragraph prompt.\nThe juxtaposition is instructive: while Anthropic's Claude Code has captured imaginations with demonstrations of end-to-end software development, Nous Research is betting that open-source alternatives trained on verifiable problems can close the gap — and that transparency in how these models are built matters as much as raw capability.\n\nHow Nous Research built an AI coding model that anyone can replicate\nWhat distinguishes the NousCoder-14B release from many competitor announcements is its radical openness. Nous Research published not just the model weights but the complete reinforcement learning environment, benchmark suite, and training harness — built on the company's Atropos framework — enabling any researcher with sufficient compute to reproduce or extend the work.\n\"Open-sourcing the Atropos stack provides the necessary infrastructure for reproducible olympiad-level reasoning research,\" noted one observer on X, summarizing the significance for the academic and open-source communities.\nThe model was trained by Joe Li, a researcher in residence at Nous Research and a former competitive programmer himself. Li's technical report reveals an unexpectedly personal dimension: he compared the model's improvement trajectory to his own journey on Codeforces, the competitive programming platform where participants earn ratings based on contest performance.\nBased on rough estimates mapping LiveCodeBench scores to Codeforces ratings, Li calculated that NousCoder-14B's improvemen t— from approximately the 1600-1750 rating range to 2100-2200 — mirrors a leap that took him nearly two years of sustained practice between ages 14 and 16. The model accomplished the equivalent in four days.\n\"Watching that final training run unfold was quite a surreal experience,\" Li wrote in the technical report.\nBut Li was quick to note an important caveat that speaks to broader questions about AI efficiency: he solved roughly 1,000 problems during those two years, while the model required 24,000. Humans, at least for now, remain dramatically more sample-efficient learners.\n\nInside the reinforcement learning system that trains on 24,000 competitive programming problems\nNousCoder-14B's training process offers a window into the increasingly sophisticated techniques researchers use to improve AI reasoning capabilities through reinforcement learning.\nThe approach relies on what researchers call \"verifiable rewards\" — a system where the model generates code solutions, those solutions are executed against test cases, and the model receives a simple binary signal: correct or incorrect. This feedback loop, while conceptually straightforward, requires significant infrastructure to execute at scale.\nNous Research used Modal, a cloud computing platform, to run sandboxed code execution in parallel. Each of the 24,000 training problems contains hundreds of test cases on average, and the system must verify that generated code produces correct outputs within time and memory constraints — 15 seconds and 4 gigabytes, respectively.\nThe training employed a technique called DAPO (Dynamic Sampling Policy Optimization), which the researchers found performed slightly better than alternatives in their experiments. A key innovation involves \"dynamic sampling\" — discarding training examples where the model either solves all attempts or fails all attempts, since these provide no useful gradient signal for learning.\nThe researchers also adopted \"iterative context extension,\" first training the model with a 32,000-token context window before expanding to 40,000 tokens. During evaluation, extending the context further to approximately 80,000 tokens produced the best results, with accuracy reaching 67.87 percent.\nPerhaps most significantly, the training pipeline overlaps inference and verification — as soon as the model generates a solution, it begins work on the next problem while the previous solution is being checked. This pipelining, combined with asynchronous training where multiple model instances work in parallel, maximizes hardware utilization on expensive GPU clusters.\n\nThe looming data shortage that could slow AI coding model progress\nBuried in Li's technical report is a finding with significant implications for the future of AI development: the training dataset for NousCoder-14B encompasses \"a significant portion of all readily available, verifiable competitive programming problems in a standardized dataset format.\"\nIn other words, for this particular domain, the researchers are approaching the limits of high-quality training data.\n\"The total number of competitive programming problems on the Internet is roughly the same order of magnitude,\" Li wrote, referring to the 24,000 problems used for training. \"This suggests that within the competitive programming domain, we have approached the limits of high-quality data.\"\nThis observation echoes growing concern across the AI industry about data constraints. While compute continues to scale according to well-understood economic and engineering principles, training data is \"increasingly finite,\" as Li put it.\n\"It appears that some of the most important research that needs to be done in the future will be in the areas of synthetic data generation and data efficient algorithms and architectures,\" he concluded.\nThe challenge is particularly acute for competitive programming because the domain requires problems with known correct solutions that can be verified automatically. Unlike natural language tasks where human evaluation or proxy metrics suffice, code either works or it doesn't — making synthetic data generation considerably more difficult.\nLi identified one potential avenue: training models not just to solve problems but to generate solvable problems, enabling a form of self-play similar to techniques that proved successful in game-playing AI systems. \"Once synthetic problem generation is solved, self-play becomes a very interesting direction,\" he wrote.\n\nA $65 million bet that open-source AI can compete with Big Tech\nNous Research has carved out a distinctive position in the AI landscape: a company committed to open-source releases that compete with — and sometimes exceed — proprietary alternatives.\nThe company raised $50 million in April 2025 in a round led by Paradigm, the cryptocurrency-focused venture firm founded by Coinbase co-founder Fred Ehrsam. Total funding reached $65 million, according to some reports. The investment reflected growing interest in decentralized approaches to AI training, an area where Nous Research has developed its Psyche platform.\nPrevious releases include Hermes 4, a family of models that we reported \"outperform ChatGPT without content restrictions,\" and DeepHermes-3, which the company described as the first \"toggle-on reasoning model\" — allowing users to activate extended thinking capabilities on demand.\nThe company has cultivated a distinctive aesthetic and community, prompting some skepticism about whether style might overshadow substance. \"Ofc i'm gonna believe an anime pfp company. stop benchmarkmaxxing ffs,\" wrote one critic on X, referring to Nous Research's anime-style branding and the industry practice of optimizing for benchmark performance.\nOthers raised technical questions. \"Based on the benchmark, Nemotron is better,\" noted one commenter, referring to Nvidia's family of language models. Another asked whether NousCoder-14B is \"agentic focused or just 'one shot' coding\" — a distinction that matters for practical software development, where iterating on feedback typically produces better results than single attempts.\n\nWhat researchers say must happen next for AI coding tools to keep improving\nThe release includes several directions for future work that hint at where AI coding research may be heading.\nMulti-turn reinforcement learning tops the list. Currently, the model receives only a final binary reward — pass or fail — after generating a solution. But competitive programming problems typically include public test cases that provide intermediate feedback: compilation errors, incorrect outputs, time limit violations. Training models to incorporate this feedback across multiple attempts could significantly improve performance.\nControlling response length also remains a challenge. The researchers found that incorrect solutions tended to be longer than correct ones, and response lengths quickly saturated available context windows during training — a pattern that various algorithmic modifications failed to resolve.\nPerhaps most ambitiously, Li proposed \"problem generation and self-play\" — training models to both solve and create programming problems. This would address the data scarcity problem directly by enabling models to generate their own training curricula.\n\"Humans are great at generating interesting and useful problems for other competitive programmers, but it appears that there still exists a significant gap in LLM capabilities in creative problem generation,\" Li wrote.\nThe model is available now on Hugging Face under an Apache 2.0 license. For researchers and developers who want to build on the work, Nous Research has published the complete Atropos training stack alongside it.\nWhat took Li two years of adolescent dedication to achieve—climbing from a 1600-level novice to a 2100-rated competitor on Codeforces—an AI replicated in 96 hours. He needed 1,000 problems. The model needed 24,000. But soon enough, these systems may learn to write their own problems, teach themselves, and leave human benchmarks behind entirely.\nThe question is no longer whether machines can learn to code. It's whether they'll soon be better teachers than we ever were." }, "pairedItem": [ { "item": 4 } ] }, { "json": { "guid": "1OII24oTfVYyF45IOoAjy8", "link": "https://venturebeat.com/technology/the-creator-of-claude-code-just-revealed-his-workflow-and-developers-are", "title": "The creator of Claude Code just revealed his workflow, and developers are losing their minds", "author": "michael.nunez@venturebeat.com (Michael Nuñez)", "content": "When the creator of the world's most advanced coding agent speaks, Silicon Valley doesn't just listen — it takes notes.
For the past week, the engineering community has been dissecting a thread on X from Boris Cherny, the creator and head of Claude Code at Anthropic. What began as a casual sharing of his personal terminal setup has spiraled into a viral manifesto on the future of software development, with industry insiders calling it a watershed moment for the startup.
"If you're not reading the Claude Code best practices straight from its creator, you're behind as a programmer," wrote Jeff Tang, a prominent voice in the developer community. Kyle McNease, another industry observer, went further, declaring that with Cherny's "game-changing updates," Anthropic is "on fire," potentially facing "their ChatGPT moment."
The excitement stems from a paradox: Cherny's workflow is surprisingly simple, yet it allows a single human to operate with the output capacity of a small engineering department. As one user noted on X after implementing Cherny's setup, the experience "feels more like Starcraft" than traditional coding — a shift from typing syntax to commanding autonomous units.
Here is an analysis of the workflow that is reshaping how software gets built, straight from the architect himself.
The most striking revelation from Cherny's disclosure is that he does not code in a linear fashion. In the traditional "inner loop" of development, a programmer writes a function, tests it, and moves to the next. Cherny, however, acts as a fleet commander.
"I run 5 Claudes in parallel in my terminal," Cherny wrote. "I number my tabs 1-5, and use system notifications to know when a Claude needs input."
By utilizing iTerm2 system notifications, Cherny effectively manages five simultaneous work streams. While one agent runs a test suite, another refactors a legacy module, and a third drafts documentation. He also runs "5-10 Claudes on claude.ai" in his browser, using a "teleport" command to hand off sessions between the web and his local machine.
This validates the "do more with less" strategy articulated by Anthropic President Daniela Amodei earlier this week. While competitors like OpenAI pursue trillion-dollar infrastructure build-outs, Anthropic is proving that superior orchestration of existing models can yield exponential productivity gains.
In a surprising move for an industry obsessed with latency, Cherny revealed that he exclusively uses Anthropic's heaviest, slowest model: Opus 4.5.
"I use Opus 4.5 with thinking for everything," Cherny explained. "It's the best coding model I've ever used, and even though it's bigger & slower than Sonnet, since you have to steer it less and it's better at tool use, it is almost always faster than using a smaller model in the end."
For enterprise technology leaders, this is a critical insight. The bottleneck in modern AI development isn't the generation speed of the token; it is the human time spent correcting the AI's mistakes. Cherny's workflow suggests that paying the "compute tax" for a smarter model upfront eliminates the "correction tax" later.
Cherny also detailed how his team solves the problem of AI amnesia. Standard large language models do not "remember" a company's specific coding style or architectural decisions from one session to the next.
To address this, Cherny's team maintains a single file named CLAUDE.md in their git repository. "Anytime we see Claude do something incorrectly we add it to the CLAUDE.md, so Claude knows not to do it next time," he wrote.
This practice transforms the codebase into a self-correcting organism. When a human developer reviews a pull request and spots an error, they don't just fix the code; they tag the AI to update its own instructions. "Every mistake becomes a rule," noted Aakash Gupta, a product leader analyzing the thread. The longer the team works together, the smarter the agent becomes.
The "vanilla" workflow one observer praised is powered by rigorous automation of repetitive tasks. Cherny uses slash commands — custom shortcuts checked into the project's repository — to handle complex operations with a single keystroke.
He highlighted a command called /commit-push-pr, which he invokes dozens of times daily. Instead of manually typing git commands, writing a commit message, and opening a pull request, the agent handles the bureaucracy of version control autonomously.
Cherny also deploys subagents — specialized AI personas — to handle specific phases of the development lifecycle. He uses a code-simplifier to clean up architecture after the main work is done and a verify-app agent to run end-to-end tests before anything ships.
If there is a single reason Claude Code has reportedly hit $1 billion in annual recurring revenue so quickly, it is likely the verification loop. The AI is not just a text generator; it is a tester.
"Claude tests every single change I land to claude.ai/code using the Claude Chrome extension," Cherny wrote. "It opens a browser, tests the UI, and iterates until the code works and the UX feels good."
He argues that giving the AI a way to verify its own work — whether through browser automation, running bash commands, or executing test suites — improves the quality of the final result by "2-3x." The agent doesn't just write code; it proves the code works.
The reaction to Cherny's thread suggests a pivotal shift in how developers think about their craft. For years, "AI coding" meant an autocomplete function in a text editor — a faster way to type. Cherny has demonstrated that it can now function as an operating system for labor itself.
"Read this if you're already an engineer... and want more power," Jeff Tang summarized on X.
The tools to multiply human output by a factor of five are already here. They require only a willingness to stop thinking of AI as an assistant and start treating it as a workforce. The programmers who make that mental leap first won't just be more productive. They'll be playing an entirely different game — and everyone else will still be typing.
", "creator": "michael.nunez@venturebeat.com (Michael Nuñez)", "isoDate": "2026-01-05T07:45:00.000Z", "pubDate": "Mon, 05 Jan 2026 07:45:00 GMT", "enclosure": { "url": "https://images.ctfassets.net/jdtwqhzvc2n1/6VsJWNsStTR57q9vFd5L08/a0d88b4dcbd1e9ba77fd72a9c55988d9/nuneybits_Vector_art_of_programmer_conducting_robot_orchestra_i_908157a9-d44f-4bce-b390-5913b88dad27.webp?w=300&q=30", "type": "image/webp", "length": "0" }, "categories": [ "Technology", "AI" ], "contentSnippet": "When the creator of the world's most advanced coding agent speaks, Silicon Valley doesn't just listen — it takes notes.\nFor the past week, the engineering community has been dissecting a thread on X from Boris Cherny, the creator and head of Claude Code at Anthropic. What began as a casual sharing of his personal terminal setup has spiraled into a viral manifesto on the future of software development, with industry insiders calling it a watershed moment for the startup.\n\n\"If you're not reading the Claude Code best practices straight from its creator, you're behind as a programmer,\" wrote Jeff Tang, a prominent voice in the developer community. Kyle McNease, another industry observer, went further, declaring that with Cherny's \"game-changing updates,\" Anthropic is \"on fire,\" potentially facing \"their ChatGPT moment.\"\nThe excitement stems from a paradox: Cherny's workflow is surprisingly simple, yet it allows a single human to operate with the output capacity of a small engineering department. As one user noted on X after implementing Cherny's setup, the experience \"feels more like Starcraft\" than traditional coding — a shift from typing syntax to commanding autonomous units.\nHere is an analysis of the workflow that is reshaping how software gets built, straight from the architect himself. \nHow running five AI agents at once turns coding into a real-time strategy game\nThe most striking revelation from Cherny's disclosure is that he does not code in a linear fashion. In the traditional \"inner loop\" of development, a programmer writes a function, tests it, and moves to the next. Cherny, however, acts as a fleet commander.\n\"I run 5 Claudes in parallel in my terminal,\" Cherny wrote. \"I number my tabs 1-5, and use system notifications to know when a Claude needs input.\"\nBy utilizing iTerm2 system notifications, Cherny effectively manages five simultaneous work streams. While one agent runs a test suite, another refactors a legacy module, and a third drafts documentation. He also runs \"5-10 Claudes on claude.ai\" in his browser, using a \"teleport\" command to hand off sessions between the web and his local machine.\nThis validates the \"do more with less\" strategy articulated by Anthropic President Daniela Amodei earlier this week. While competitors like OpenAI pursue trillion-dollar infrastructure build-outs, Anthropic is proving that superior orchestration of existing models can yield exponential productivity gains.\nThe counterintuitive case for choosing the slowest, smartest model\nIn a surprising move for an industry obsessed with latency, Cherny revealed that he exclusively uses Anthropic's heaviest, slowest model: Opus 4.5.\n\"I use Opus 4.5 with thinking for everything,\" Cherny explained. \"It's the best coding model I've ever used, and even though it's bigger & slower than Sonnet, since you have to steer it less and it's better at tool use, it is almost always faster than using a smaller model in the end.\"\nFor enterprise technology leaders, this is a critical insight. The bottleneck in modern AI development isn't the generation speed of the token; it is the human time spent correcting the AI's mistakes. Cherny's workflow suggests that paying the \"compute tax\" for a smarter model upfront eliminates the \"correction tax\" later.\nOne shared file turns every AI mistake into a permanent lesson\nCherny also detailed how his team solves the problem of AI amnesia. Standard large language models do not \"remember\" a company's specific coding style or architectural decisions from one session to the next.\nTo address this, Cherny's team maintains a single file named CLAUDE.md in their git repository. \"Anytime we see Claude do something incorrectly we add it to the CLAUDE.md, so Claude knows not to do it next time,\" he wrote.\nThis practice transforms the codebase into a self-correcting organism. When a human developer reviews a pull request and spots an error, they don't just fix the code; they tag the AI to update its own instructions. \"Every mistake becomes a rule,\" noted Aakash Gupta, a product leader analyzing the thread. The longer the team works together, the smarter the agent becomes.\nSlash commands and subagents automate the most tedious parts of development\nThe \"vanilla\" workflow one observer praised is powered by rigorous automation of repetitive tasks. Cherny uses slash commands — custom shortcuts checked into the project's repository — to handle complex operations with a single keystroke.\nHe highlighted a command called /commit-push-pr, which he invokes dozens of times daily. Instead of manually typing git commands, writing a commit message, and opening a pull request, the agent handles the bureaucracy of version control autonomously.\nCherny also deploys subagents — specialized AI personas — to handle specific phases of the development lifecycle. He uses a code-simplifier to clean up architecture after the main work is done and a verify-app agent to run end-to-end tests before anything ships.\nWhy verification loops are the real unlock for AI-generated code\nIf there is a single reason Claude Code has reportedly hit $1 billion in annual recurring revenue so quickly, it is likely the verification loop. The AI is not just a text generator; it is a tester.\n\"Claude tests every single change I land to claude.ai/code using the Claude Chrome extension,\" Cherny wrote. \"It opens a browser, tests the UI, and iterates until the code works and the UX feels good.\"\nHe argues that giving the AI a way to verify its own work — whether through browser automation, running bash commands, or executing test suites — improves the quality of the final result by \"2-3x.\" The agent doesn't just write code; it proves the code works.\nWhat Cherny's workflow signals about the future of software engineering\nThe reaction to Cherny's thread suggests a pivotal shift in how developers think about their craft. For years, \"AI coding\" meant an autocomplete function in a text editor — a faster way to type. Cherny has demonstrated that it can now function as an operating system for labor itself.\n\"Read this if you're already an engineer... and want more power,\" Jeff Tang summarized on X.\nThe tools to multiply human output by a factor of five are already here. They require only a willingness to stop thinking of AI as an assistant and start treating it as a workforce. The programmers who make that mental leap first won't just be more productive. They'll be playing an entirely different game — and everyone else will still be typing." }, "pairedItem": [ { "item": 4 } ] }, { "json": { "guid": "881216de-81bd-4da8-908c-6fc3e6a4411a", "link": "https://www.zdnet.com/article/how-to-turn-airtag-into-wifi-password-key-nfc-tags/", "title": "I created the ultimate Wi-Fi password key with the most unexpected gadget - how it works", "itunes": {}, "content": "NFC tags are so useful and customizable, and it's quite simple to make your own. Here's what you can do with it and how.", "isoDate": "2026-02-10T03:01:15.000Z", "pubDate": "Tue, 10 Feb 2026 03:01:15 GMT", "contentSnippet": "NFC tags are so useful and customizable, and it's quite simple to make your own. Here's what you can do with it and how." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "27471e56-532b-4090-9aab-db552814455b", "link": "https://www.zdnet.com/article/i-turned-doodles-into-video-clips-in-minutes-with-runways-new-ai-video-tool-heres-how/", "title": "I turned doodles into video clips in minutes with Runway's new AI video tool - here's how", "itunes": {}, "content": "The company's Motion Sketch feature lowers the barrier between abstract imagination and concrete video outputs. But like all generative AI tools, it has its shortcomings.", "isoDate": "2026-02-10T03:01:13.000Z", "pubDate": "Tue, 10 Feb 2026 03:01:13 GMT", "contentSnippet": "The company's Motion Sketch feature lowers the barrier between abstract imagination and concrete video outputs. But like all generative AI tools, it has its shortcomings." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "449249a1-cdf7-4b86-b294-f0b2409be861", "link": "https://www.zdnet.com/article/android-extend-unlock-faster-phone-access/", "title": "This Android trick lets me access my phone faster - here's how it works", "itunes": {}, "content": "Extend Unlock is one of those features every Android user will love once they give it a try.", "isoDate": "2026-02-10T03:00:54.000Z", "pubDate": "Tue, 10 Feb 2026 03:00:54 GMT", "contentSnippet": "Extend Unlock is one of those features every Android user will love once they give it a try." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "e94ba78f-e76f-4849-a02b-828905fc727f", "link": "https://www.zdnet.com/article/onyx-boox-page-android-e-ink-tablet-review/", "title": "I didn't expect this Android E Ink tablet to beat my Kindle, but one feature sets it apart", "itunes": {}, "content": "The Onyx Boox Page is packed with features for an E Ink tablet, with a smart, versatile design.", "isoDate": "2026-02-10T02:30:37.000Z", "pubDate": "Tue, 10 Feb 2026 02:30:37 GMT", "contentSnippet": "The Onyx Boox Page is packed with features for an E Ink tablet, with a smart, versatile design." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "6788a566-899b-481a-b0cc-ce5b53409dc0", "link": "https://www.zdnet.com/article/how-to-restart-android-phone-without-using-the-power-button-2-ways/", "title": "How to restart your Android phone without using the power button: 2 easy ways", "itunes": {}, "content": "You don't need to press the power button to restart your Android device. If yours is broken or you simply want options, here are two.", "isoDate": "2026-02-10T02:12:53.000Z", "pubDate": "Tue, 10 Feb 2026 02:12:53 GMT", "contentSnippet": "You don't need to press the power button to restart your Android device. If yours is broken or you simply want options, here are two." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "09d182bf-9aac-454b-9390-edf9ce2587e2", "link": "https://www.zdnet.com/article/which-ai-tools-are-worth-paying-for-here-are-subscriptions-im-keeping-and-why/", "title": "Which AI tools are worth paying for? Here are subscriptions I'm keeping - and why", "itunes": {}, "content": "I pay close attention to anything that can save me time. After reviewing my 2025 spending on generative AI, I made one big change.", "isoDate": "2026-02-10T02:00:00.000Z", "pubDate": "Tue, 10 Feb 2026 02:00:00 GMT", "contentSnippet": "I pay close attention to anything that can save me time. After reviewing my 2025 spending on generative AI, I made one big change." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "604bbf51-114a-4e59-a160-b817098ffffb", "link": "https://www.zdnet.com/article/comaps-review-google-maps-alternative/", "title": "I found a free Google Maps alternative that doesn't track my location (or kill my phone battery)", "itunes": {}, "content": "Comaps features voice-guided directions and offline search, prioritizing privacy over the controversial practices of Google Maps.", "isoDate": "2026-02-10T01:46:00.000Z", "pubDate": "Tue, 10 Feb 2026 01:46:00 GMT", "contentSnippet": "Comaps features voice-guided directions and offline search, prioritizing privacy over the controversial practices of Google Maps." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "dad97e75-e9c7-4425-a6ac-4e9130ae6dcc", "link": "https://www.zdnet.com/article/tessan-65w-charging-tower-review/", "title": "I switched to a vertical charging tower in my home office - it's so useful, I bought 3 more", "itunes": {}, "content": "I was so impressed by the Tessan 65W charging tower, it became a staple both at home and the office.", "isoDate": "2026-02-10T01:45:41.000Z", "pubDate": "Tue, 10 Feb 2026 01:45:41 GMT", "contentSnippet": "I was so impressed by the Tessan 65W charging tower, it became a staple both at home and the office." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "a6c9aa24-04c9-4256-8640-b201c8f05e30", "link": "https://www.zdnet.com/article/6-best-android-launchers-to-improve-your-home-screen-most-free/", "title": "6 Android launchers that are better than your default home screen (and most are free)", "itunes": {}, "content": "You can customize your Android phone simply by replacing the default launcher. Here are six I've tested and recommend.", "isoDate": "2026-02-10T01:36:00.000Z", "pubDate": "Tue, 10 Feb 2026 01:36:00 GMT", "contentSnippet": "You can customize your Android phone simply by replacing the default launcher. Here are six I've tested and recommend." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "fbeebee4-634e-4065-bd5c-c83fdf531fe5", "link": "https://www.zdnet.com/article/oura-ring-sick-symptom-radar-my-experience/", "title": "I should've listened to my Oura Ring when it warned me about my health", "itunes": {}, "content": "While they're not always perfect, health trackers like the Oura Ring can detect early signs of illness. Here's why you shouldn't ignore them.", "isoDate": "2026-02-10T01:33:00.000Z", "pubDate": "Tue, 10 Feb 2026 01:33:00 GMT", "contentSnippet": "While they're not always perfect, health trackers like the Oura Ring can detect early signs of illness. Here's why you shouldn't ignore them." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "167feba6-ba53-4f26-8a81-4d9b711e3c78", "link": "https://www.zdnet.com/article/ktc-25inch-android-tablet-review/", "title": "I lived with this 25-inch Android tablet for a week - and it went surprisingly well", "itunes": {}, "content": "If you're looking for a big-screen tablet, KTC's 25-inch model offers plenty of entertainment at an acceptable price.", "isoDate": "2026-02-10T01:32:00.000Z", "pubDate": "Tue, 10 Feb 2026 01:32:00 GMT", "contentSnippet": "If you're looking for a big-screen tablet, KTC's 25-inch model offers plenty of entertainment at an acceptable price." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "7c1bf65a-91c0-4597-9e8b-a62a3cf022f9", "link": "https://www.zdnet.com/article/best-ps5-settings-improve-performance/", "title": "I changed 3 settings on my PS5 to give the system an instant performance boost", "itunes": {}, "content": "With a few simple adjustments, you can easily enhance your experience across gaming, streaming, and even online security.", "isoDate": "2026-02-10T01:30:43.000Z", "pubDate": "Tue, 10 Feb 2026 01:30:43 GMT", "contentSnippet": "With a few simple adjustments, you can easily enhance your experience across gaming, streaming, and even online security." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "7c99d954-c993-4758-bbcf-31bfc2d73843", "link": "https://www.zdnet.com/article/dreamie-smart-alarm-clock-review/", "title": "I tried a $250 alarm clock for a week - and it's not as ridiculous as it sounds", "itunes": {}, "content": "Dreamie is the smart alarm clock newcomer that wants to get you off your phone and to sleep. But it's far from perfect.", "isoDate": "2026-02-10T01:11:00.000Z", "pubDate": "Tue, 10 Feb 2026 01:11:00 GMT", "contentSnippet": "Dreamie is the smart alarm clock newcomer that wants to get you off your phone and to sleep. But it's far from perfect." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "ca3ab0fe-7511-4a74-aca7-1530970adec6", "link": "https://www.zdnet.com/article/oura-rings-team-usa-olympics-official-wearable/", "title": "Why Oura Rings made it onto Team USA's Olympic gear list - and that's good news for you", "itunes": {}, "content": "US athletes will sport Oura Rings to track their health. Here's why sponsorships like these can inspire new features.", "isoDate": "2026-02-10T01:07:00.000Z", "pubDate": "Tue, 10 Feb 2026 01:07:00 GMT", "contentSnippet": "US athletes will sport Oura Rings to track their health. Here's why sponsorships like these can inspire new features." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "31d8d4c0-3b5c-4c00-9be6-71120b9e5637", "link": "https://www.zdnet.com/article/windows-10-11-secret-recovery-tool-how-to-find-it-and-use/", "title": "Your Windows PC has a secret recovery tool that's seriously useful - here's how to access it", "itunes": {}, "content": "If Windows ever has trouble launching, Recovery Drive lets you back up essential system files to an external USB drive to help Windows boot.", "isoDate": "2026-02-10T00:52:00.000Z", "pubDate": "Tue, 10 Feb 2026 00:52:00 GMT", "contentSnippet": "If Windows ever has trouble launching, Recovery Drive lets you back up essential system files to an external USB drive to help Windows boot." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "5f97d44f-e8cf-47e1-9369-489a9bb870bb", "link": "https://www.zdnet.com/article/why-you-should-put-your-phone-screen-side-down/", "title": "The simple phone habit change that saves your screen time (and your sanity)", "itunes": {}, "content": "It sounds counterintuitive, but placing your phone face down has some serious benefits.", "isoDate": "2026-02-10T00:45:48.000Z", "pubDate": "Tue, 10 Feb 2026 00:45:48 GMT", "contentSnippet": "It sounds counterintuitive, but placing your phone face down has some serious benefits." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "a93dfab2-c6ea-4754-9e36-47b3bd1e3cf8", "link": "https://www.zdnet.com/article/logitech-mevo-core-review/", "title": "I went live with this 4K Logitech camera, and it gave my $3,000 Canon a run for its money", "itunes": {}, "content": "Logitech's modular Mevo Core is a 4K live-streaming camera with multi-cam support and fantastic image quality.", "isoDate": "2026-02-09T22:12:00.000Z", "pubDate": "Mon, 09 Feb 2026 22:12:00 GMT", "contentSnippet": "Logitech's modular Mevo Core is a 4K live-streaming camera with multi-cam support and fantastic image quality." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "0c42b76a-7b87-41a4-b2f6-9133e40016f0", "link": "https://www.zdnet.com/article/best-presidents-day-deals-2026/", "title": "The best Presidents' Day sales we've found so far: Save on Apple, Sony, and more", "itunes": {}, "content": "Presidents' Day sales are slowly making their way online, and I'm scouting for expert-approved deals on home, tech, and more.", "isoDate": "2026-02-09T22:01:47.000Z", "pubDate": "Mon, 09 Feb 2026 22:01:47 GMT", "contentSnippet": "Presidents' Day sales are slowly making their way online, and I'm scouting for expert-approved deals on home, tech, and more." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "38aefa0f-4478-4733-8471-b0f79a52ed2e", "link": "https://www.zdnet.com/article/best-apple-deals-presidents-day-2026/", "title": "The best Presidents' Day Apple sales you can shop now", "itunes": {}, "content": "The first deal event of the year is almost here, and you'll be able to find discounts on all things Apple for Presidents' Day.", "isoDate": "2026-02-09T21:49:00.000Z", "pubDate": "Mon, 09 Feb 2026 21:49:00 GMT", "contentSnippet": "The first deal event of the year is almost here, and you'll be able to find discounts on all things Apple for Presidents' Day." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "a64b3406-cee0-4d41-a3a2-6c1fe99777e5", "link": "https://www.zdnet.com/article/best-buy-presidents-day-ps5-sale/", "title": "I found the 7 best PS5 games on sale during Best Buy's Presidents' Day sale", "itunes": {}, "content": "You can save 50% or more on hot PS5 and PlayStation-on-PC titles like God of War, Uncharted, and more at Best Buy.", "isoDate": "2026-02-09T21:43:11.000Z", "pubDate": "Mon, 09 Feb 2026 21:43:11 GMT", "contentSnippet": "You can save 50% or more on hot PS5 and PlayStation-on-PC titles like God of War, Uncharted, and more at Best Buy." }, "pairedItem": [ { "item": 5 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608359", "link": "https://towardsdatascience.com/the-machine-learning-lessons-ive-learned-last-month/", "title": "The Machine Learning Lessons I’ve Learned Last Month", "content": "Delayed January: deadlines, downtimes, and flow times
\nThe post The Machine Learning Lessons I’ve Learned Last Month appeared first on Towards Data Science.
\n", "creator": "Pascal Janetzky", "isoDate": "2026-02-09T20:38:42.000Z", "pubDate": "Mon, 09 Feb 2026 20:38:42 +0000", "categories": [ "Machine Learning", "Inspiration", "Ml Product Manager", "Productivity", "Project Management" ], "dc:creator": "Pascal Janetzky", "contentSnippet": "Delayed January: deadlines, downtimes, and flow times\nThe post The Machine Learning Lessons I’ve Learned Last Month appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608349", "link": "https://towardsdatascience.com/the-death-of-the-everything-prompt-googles-move-toward-structured-ai/", "title": "The Death of the “Everything Prompt”: Google’s Move Toward Structured AI", "content": "How the new Interactions API enables deep-reasoning, stateful, agentic workflows.
\nThe post The Death of the “Everything Prompt”: Google’s Move Toward Structured AI appeared first on Towards Data Science.
\n", "creator": "Thomas Reid", "isoDate": "2026-02-09T13:00:00.000Z", "pubDate": "Mon, 09 Feb 2026 13:00:00 +0000", "categories": [ "Artificial Intelligence", "Agentic Ai", "API", "Deep Dives", "Llm", "Programming" ], "dc:creator": "Thomas Reid", "contentSnippet": "How the new Interactions API enables deep-reasoning, stateful, agentic workflows.\nThe post The Death of the “Everything Prompt”: Google’s Move Toward Structured AI appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608351", "link": "https://towardsdatascience.com/what-i-am-doing-to-stay-relevant-as-a-senior-analytics-consultant-in-2026/", "title": "What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026", "content": "Learn how to work with AI, while strengthening your unique human skills that technology cannot replace
\nThe post What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026 appeared first on Towards Data Science.
\n", "creator": "Rashi Desai", "isoDate": "2026-02-07T15:00:00.000Z", "pubDate": "Sat, 07 Feb 2026 15:00:00 +0000", "categories": [ "Data analysis", "Artificial Intelligence", "Career Advice", "Data Science", "Editors Pick", "Generative Ai" ], "dc:creator": "Rashi Desai", "contentSnippet": "Learn how to work with AI, while strengthening your unique human skills that technology cannot replace\nThe post What I Am Doing to Stay Relevant as a Senior Analytics Consultant in 2026 appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608347", "link": "https://towardsdatascience.com/pydantic-performance-4-tips-on-how-to-validate-large-amounts-of-data-efficiently/", "title": "Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently", "content": "The real value lies in writing clearer code and using your tools right
\nThe post Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently appeared first on Towards Data Science.
\n", "creator": "Mike Huls", "isoDate": "2026-02-06T15:00:00.000Z", "pubDate": "Fri, 06 Feb 2026 15:00:00 +0000", "categories": [ "Data Engineering", "Data Architecture", "Data Science", "Data Validation", "Pydantic", "Python" ], "dc:creator": "Mike Huls", "contentSnippet": "The real value lies in writing clearer code and using your tools right\nThe post Pydantic Performance: 4 Tips on How to Validate Large Amounts of Data Efficiently appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608345", "link": "https://towardsdatascience.com/prompt-fidelity-measuring-how-much-of-your-intent-an-ai-agent-actually-executes/", "title": "Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes", "content": "How much of your AI agent's output is real data versus confident guesswork?
\nThe post Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes appeared first on Towards Data Science.
\n", "creator": "James Barney", "isoDate": "2026-02-06T12:00:00.000Z", "pubDate": "Fri, 06 Feb 2026 12:00:00 +0000", "categories": [ "Agentic AI", "Ai Agent", "Artificial Intelligence", "Editors Pick", "Prompt Engineering", "Recommender Systems" ], "dc:creator": "James Barney", "contentSnippet": "How much of your AI agent's output is real data versus confident guesswork?\nThe post Prompt Fidelity: Measuring How Much of Your Intent an AI Agent Actually Executes appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608353", "link": "https://towardsdatascience.com/tds-newsletter-vibe-coding-is-great-until-its-not/", "title": "TDS Newsletter: Vibe Coding Is Great. Until It’s Not.", "content": "Sorting through the good, bad, and ambiguous aspects of vibe coding
\nThe post TDS Newsletter: Vibe Coding Is Great. Until It’s Not. appeared first on Towards Data Science.
\n", "creator": "TDS Editors", "isoDate": "2026-02-05T17:09:00.000Z", "pubDate": "Thu, 05 Feb 2026 17:09:00 +0000", "categories": [ "The Variable", "AI coding assistants", "Artificial Intelligence", "Tds Features", "Vibe coding" ], "dc:creator": "TDS Editors", "contentSnippet": "Sorting through the good, bad, and ambiguous aspects of vibe coding\nThe post TDS Newsletter: Vibe Coding Is Great. Until It’s Not. appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608342", "link": "https://towardsdatascience.com/mechanistic-interpretability-peeking-inside-an-llm/", "title": "Mechanistic Interpretability: Peeking Inside an LLM", "content": "Are the human-like cognitive abilities of LLMs real or fake? How does information travel through the neural network? Is there hidden knowledge inside an LLM?
\nThe post Mechanistic Interpretability: Peeking Inside an LLM appeared first on Towards Data Science.
\n", "creator": "Julian Mendel", "isoDate": "2026-02-05T15:00:00.000Z", "pubDate": "Thu, 05 Feb 2026 15:00:00 +0000", "categories": [ "Large Language Models", "Artificial Intelligence", "Deep Dives", "Explainable Ai", "Llm", "Neural Network" ], "dc:creator": "Julian Mendel", "contentSnippet": "Are the human-like cognitive abilities of LLMs real or fake? How does information travel through the neural network? Is there hidden knowledge inside an LLM?\nThe post Mechanistic Interpretability: Peeking Inside an LLM appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608340", "link": "https://towardsdatascience.com/why-is-my-code-so-slow-a-guide-to-py-spy-python-profiling/", "title": "Why Is My Code So Slow? A Guide to Py-Spy Python Profiling", "content": "Stop guessing and start diagnosing performance issues using Py-Spy
\nThe post Why Is My Code So Slow? A Guide to Py-Spy Python Profiling appeared first on Towards Data Science.
\n", "creator": "Kenneth McCarthy", "isoDate": "2026-02-05T13:30:00.000Z", "pubDate": "Thu, 05 Feb 2026 13:30:00 +0000", "categories": [ "Programming", "Data Science", "Debugging", "Deep Dives", "Profiling", "Python" ], "dc:creator": "Kenneth McCarthy", "contentSnippet": "Stop guessing and start diagnosing performance issues using Py-Spy\nThe post Why Is My Code So Slow? A Guide to Py-Spy Python Profiling appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608338", "link": "https://towardsdatascience.com/stop-confusing-loc-and-iloc-in-pandas-the-rule-everyone-misses/", "title": "The Rule Everyone Misses: How to Stop Confusing loc and iloc in Pandas", "content": "A simple mental model to remember when each one works (with examples that finally click).
\nThe post The Rule Everyone Misses: How to Stop Confusing loc and iloc in Pandas appeared first on Towards Data Science.
\n", "creator": "Ibrahim Salami", "isoDate": "2026-02-05T12:00:00.000Z", "pubDate": "Thu, 05 Feb 2026 12:00:00 +0000", "categories": [ "Data Science", "Data Analytics", "Data Extraction", "Pandas", "Python" ], "dc:creator": "Ibrahim Salami", "contentSnippet": "A simple mental model to remember when each one works (with examples that finally click).\nThe post The Rule Everyone Misses: How to Stop Confusing loc and iloc in Pandas appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608335", "link": "https://towardsdatascience.com/aws-vs-azure-a-deep-dive-into-model-training-part-2/", "title": "AWS vs. Azure: A Deep Dive into Model Training – Part 2", "content": "This article covers how Azure ML's persistent, workspace-centric compute resources differ from AWS SageMaker's on-demand, job-specific approach. Additionally, we explored environment customization options, from Azure's curated environments and custom environments to SageMaker's three level of customizations.
\nThe post AWS vs. Azure: A Deep Dive into Model Training – Part 2 appeared first on Towards Data Science.
\n", "creator": "Destin Gong", "isoDate": "2026-02-04T16:30:00.000Z", "pubDate": "Wed, 04 Feb 2026 16:30:00 +0000", "categories": [ "Data Science", "AWS", "Azure", "Editors Pick", "Machine Learning", "Mlops" ], "dc:creator": "Destin Gong", "contentSnippet": "This article covers how Azure ML's persistent, workspace-centric compute resources differ from AWS SageMaker's on-demand, job-specific approach. Additionally, we explored environment customization options, from Azure's curated environments and custom environments to SageMaker's three level of customizations.\nThe post AWS vs. Azure: A Deep Dive into Model Training – Part 2 appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608333", "link": "https://towardsdatascience.com/how-to-effectively-work-with-frontend-and-backend-code/", "title": "How to Work Effectively with Frontend and Backend Code", "content": "Learn how to be an effective full-stack engineer with Claude Code
\nThe post How to Work Effectively with Frontend and Backend Code appeared first on Towards Data Science.
\n", "creator": "Eivind Kjosbakken", "isoDate": "2026-02-04T15:00:00.000Z", "pubDate": "Wed, 04 Feb 2026 15:00:00 +0000", "categories": [ "LLM Applications", "Agentic Ai", "Artificial Intelligence", "Claude Code", "coding agents", "Programming" ], "dc:creator": "Eivind Kjosbakken", "contentSnippet": "Learn how to be an effective full-stack engineer with Claude Code\nThe post How to Work Effectively with Frontend and Backend Code appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608331", "link": "https://towardsdatascience.com/how-to-build-your-own-custom-llm-memory-layer-from-scratch/", "title": "How to Build Your Own Custom LLM Memory Layer from Scratch", "content": "Step-by-step guide to building autonomous memory retrieval systems
\nThe post How to Build Your Own Custom LLM Memory Layer from Scratch appeared first on Towards Data Science.
\n", "creator": "Avishek Biswas", "isoDate": "2026-02-04T13:30:00.000Z", "pubDate": "Wed, 04 Feb 2026 13:30:00 +0000", "categories": [ "Large Language Models", "Context Engineering", "Deep Dives", "Llm", "Llm Applications", "Memory Management" ], "dc:creator": "Avishek Biswas", "contentSnippet": "Step-by-step guide to building autonomous memory retrieval systems\nThe post How to Build Your Own Custom LLM Memory Layer from Scratch appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608329", "link": "https://towardsdatascience.com/plan-code-execute-designing-agents-that-create-their-own-tools/", "title": "Plan–Code–Execute: Designing Agents That Create Their Own Tools", "content": "The case against pre-built tools in Agentic Architectures
\nThe post Plan–Code–Execute: Designing Agents That Create Their Own Tools appeared first on Towards Data Science.
\n", "creator": "Partha Sarkar", "isoDate": "2026-02-04T12:00:00.000Z", "pubDate": "Wed, 04 Feb 2026 12:00:00 +0000", "categories": [ "Agentic AI", "Agentic Ai", "Ai Agent", "Ai Engineering", "Deep Dives", "Tool-using Agents" ], "dc:creator": "Partha Sarkar", "contentSnippet": "The case against pre-built tools in Agentic Architectures\nThe post Plan–Code–Execute: Designing Agents That Create Their Own Tools appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608327", "link": "https://towardsdatascience.com/routing-in-a-sparse-graph-a-distributed-q-learning-approach/", "title": "Routing in a Sparse Graph: a Distributed Q-Learning Approach", "content": "Distributed agents need only decide one move ahead.
\nThe post Routing in a Sparse Graph: a Distributed Q-Learning Approach appeared first on Towards Data Science.
\n", "creator": "Sébastien Gilbert", "isoDate": "2026-02-03T16:30:00.000Z", "pubDate": "Tue, 03 Feb 2026 16:30:00 +0000", "categories": [ "Machine Learning", "Deep Dives", "Graph", "Path Search", "Q-Learning", "Reinforcement Learning" ], "dc:creator": "Sébastien Gilbert", "contentSnippet": "Distributed agents need only decide one move ahead.\nThe post Routing in a Sparse Graph: a Distributed Q-Learning Approach appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608325", "link": "https://towardsdatascience.com/yolov2-yolo9000-paper-walkthrough-better-faster-stronger/", "title": "YOLOv2 & YOLO9000 Paper Walkthrough: Better, Faster, Stronger", "content": "From YOLOv1 to YOLOv2: prior box, k-means, Darknet-19, passthrough layer, and more
\nThe post YOLOv2 & YOLO9000 Paper Walkthrough: Better, Faster, Stronger appeared first on Towards Data Science.
\n", "creator": "Muhammad Ardi", "isoDate": "2026-02-03T15:00:00.000Z", "pubDate": "Tue, 03 Feb 2026 15:00:00 +0000", "categories": [ "Artificial Intelligence", "Deep Learning", "Editors Pick", "Object Detection", "Pytorch", "Yolo" ], "dc:creator": "Muhammad Ardi", "contentSnippet": "From YOLOv1 to YOLOv2: prior box, k-means, Darknet-19, passthrough layer, and more\nThe post YOLOv2 & YOLO9000 Paper Walkthrough: Better, Faster, Stronger appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608322", "link": "https://towardsdatascience.com/creating-a-data-pipeline-to-monitor-local-crime-trends/", "title": "Creating a Data Pipeline to Monitor Local Crime Trends", "content": "A walkthough of creating an ETL pipeline to extract local crime data and visualize it in Metabase.
\nThe post Creating a Data Pipeline to Monitor Local Crime Trends appeared first on Towards Data Science.
\n", "creator": "Jimin Kang", "isoDate": "2026-02-03T13:30:00.000Z", "pubDate": "Tue, 03 Feb 2026 13:30:00 +0000", "categories": [ "Data Science", "Data Engineering", "Data Pipeline", "Data Visualisation", "Deep Dives", "Etl" ], "dc:creator": "Jimin Kang", "contentSnippet": "A walkthough of creating an ETL pipeline to extract local crime data and visualize it in Metabase.\nThe post Creating a Data Pipeline to Monitor Local Crime Trends appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608320", "link": "https://towardsdatascience.com/the-proximity-of-the-inception-score-as-an-evaluation-criterion/", "title": "The Proximity of the Inception Score as an Evaluation Criterion", "content": "The neighborhood of synthetic data
\nThe post The Proximity of the Inception Score as an Evaluation Criterion appeared first on Towards Data Science.
\n", "creator": "Giuseppe Pio Cannata", "isoDate": "2026-02-03T12:00:00.000Z", "pubDate": "Tue, 03 Feb 2026 12:00:00 +0000", "categories": [ "Deep Learning", "Artificial Intelligence", "Computer Vision", "Data Science", "Gans", "Generative Adversarial" ], "dc:creator": "Giuseppe Pio Cannata", "contentSnippet": "The neighborhood of synthetic data\nThe post The Proximity of the Inception Score as an Evaluation Criterion appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608264", "link": "https://towardsdatascience.com/building-systems-that-survive-real-life/", "title": "Building Systems That Survive Real Life", "content": "Sara Nobrega on the transition from data science to AI engineering, using LLMs as a bridge to DevOps, and the one engineering skill junior data scientists need to stay competitive.
\nThe post Building Systems That Survive Real Life appeared first on Towards Data Science.
\n", "creator": "TDS Editors", "isoDate": "2026-02-02T17:00:00.000Z", "pubDate": "Mon, 02 Feb 2026 17:00:00 +0000", "categories": [ "Author Spotlights", "Artificial Intelligence", "Data Science", "Machine Learning", "Writing" ], "dc:creator": "TDS Editors", "contentSnippet": "Sara Nobrega on the transition from data science to AI engineering, using LLMs as a bridge to DevOps, and the one engineering skill junior data scientists need to stay competitive.\nThe post Building Systems That Survive Real Life appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608310", "link": "https://towardsdatascience.com/silicon-darwinism-why-scarcity-is-the-source-of-true-intelligence/", "title": "Silicon Darwinism: Why Scarcity Is the Source of True Intelligence", "content": "We are confusing “size” with “smart.” The next leap in artificial intelligence will not come from a larger data center, but from a more constrained environment.
\nThe post Silicon Darwinism: Why Scarcity Is the Source of True Intelligence appeared first on Towards Data Science.
\n", "creator": "Aakash Goswami", "isoDate": "2026-02-02T13:00:00.000Z", "pubDate": "Mon, 02 Feb 2026 13:00:00 +0000", "categories": [ "Artificial Intelligence", "Articial Intelligence", "Future", "Machine Learning", "Sustainability", "Tinyml" ], "dc:creator": "Aakash Goswami", "contentSnippet": "We are confusing “size” with “smart.” The next leap in artificial intelligence will not come from a larger data center, but from a more constrained environment.\nThe post Silicon Darwinism: Why Scarcity Is the Source of True Intelligence appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://towardsdatascience.com/?p=608307", "link": "https://towardsdatascience.com/distributed-reinforcement-learning-for-scalable-high-performance-policy-optimization/", "title": "Distributed Reinforcement Learning for Scalable High-Performance Policy Optimization", "content": "Leveraging massive parallelism, asynchronous updates, and multi-machine training to match and exceed human-level performance
\nThe post Distributed Reinforcement Learning for Scalable High-Performance Policy Optimization appeared first on Towards Data Science.
\n", "creator": "Sam Black", "isoDate": "2026-02-01T15:00:00.000Z", "pubDate": "Sun, 01 Feb 2026 15:00:00 +0000", "categories": [ "Machine Learning", "Artificial Intelligence", "Deep Reinforcement", "Distributed Learning", "Editors Pick", "Reinforcement Learning" ], "dc:creator": "Sam Black", "contentSnippet": "Leveraging massive parallelism, asynchronous updates, and multi-machine training to match and exceed human-level performance\nThe post Distributed Reinforcement Learning for Scalable High-Performance Policy Optimization appeared first on Towards Data Science." }, "pairedItem": [ { "item": 6 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/09/technology/social-media-addiction-trial.html", "link": "https://www.nytimes.com/2026/02/09/technology/social-media-addiction-trial.html", "title": "Meta and YouTube Created ‘Digital Casinos,’ Lawyers Argue in Landmark Trial", "content": "Opening statements began in a trial claiming social media companies design addictive products that cause personal injury.", "creator": "Eli Tan and Cecilia Kang", "isoDate": "2026-02-10T00:53:17.000Z", "pubDate": "Tue, 10 Feb 2026 00:53:17 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Suits and Litigation (Civil)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Mental Health and Disorders" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Addiction (Psychology)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Youth" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Instagram Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Meta Platforms Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "YouTube.com" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Zuckerberg, Mark E" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Mohan, Neal" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" } ], "dc:creator": "Eli Tan and Cecilia Kang", "contentSnippet": "Opening statements began in a trial claiming social media companies design addictive products that cause personal injury." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/business/google-employees-protest.html", "link": "https://www.nytimes.com/2026/02/06/business/google-employees-protest.html", "title": "Google Workers Demand End to Cloud Services for Immigration Agencies", "content": "More than 800 employees delivered a petition to management, condemning the Trump administration’s use of Google technology in immigration enforcement.", "creator": "Tripp Mickle", "isoDate": "2026-02-06T18:52:26.000Z", "pubDate": "Fri, 06 Feb 2026 18:52:26 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Google Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Demonstrations, Protests and Riots" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Cloud Computing" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Illegal Immigration" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Immigration Detention" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Federal Actions in US Cities" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Police Brutality, Misconduct and Shootings" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "United States Politics and Government" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Customs and Border Protection (US)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Immigration and Customs Enforcement (US)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Homeland Security Department" } ], "dc:creator": "Tripp Mickle", "contentSnippet": "More than 800 employees delivered a petition to management, condemning the Trump administration’s use of Google technology in immigration enforcement." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/business/tiktok-addictive-design-europe.html", "link": "https://www.nytimes.com/2026/02/06/business/tiktok-addictive-design-europe.html", "title": "Europe Accuses TikTok of ‘Addictive Design’ and Pushes for Change", "content": "European Union regulators said the app’s infinite scroll and personalized algorithm led to “compulsive” behavior, especially among children.", "creator": "Adam Satariano", "isoDate": "2026-02-06T19:08:37.000Z", "pubDate": "Fri, 06 Feb 2026 19:08:37 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Regulation and Deregulation of Industry" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "International Relations" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Law and Legislation" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Beijing Bytedance Technology Co Ltd" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "TikTok (ByteDance)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Europe" } ], "dc:creator": "Adam Satariano", "contentSnippet": "European Union regulators said the app’s infinite scroll and personalized algorithm led to “compulsive” behavior, especially among children." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/business/ai-stock-market-anthropic-openai.html", "link": "https://www.nytimes.com/2026/02/06/business/ai-stock-market-anthropic-openai.html", "title": "Fear of AI Replacing Software Makers Hits Stocks. Here’s What to Know.", "content": "The prospect of disruptions from artificial intelligence has hung over the economy for years. But this week advances in software tools precipitated a sell-off on Wall Street.", "creator": "Mike Isaac, Joe Rennison and Maureen Farrell", "isoDate": "2026-02-06T20:11:04.000Z", "pubDate": "Fri, 06 Feb 2026 20:11:04 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Stocks and Bonds" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Company Reports" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Credit and Debt" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Software" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Innovation" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Bitcoin (Currency)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Anthropic AI LLC" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Qualcomm Inc" } ], "dc:creator": "Mike Isaac, Joe Rennison and Maureen Farrell", "contentSnippet": "The prospect of disruptions from artificial intelligence has hung over the economy for years. But this week advances in software tools precipitated a sell-off on Wall Street." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/09/health/ai-chatbots-doctors-medicine.html", "link": "https://www.nytimes.com/2026/02/09/health/ai-chatbots-doctors-medicine.html", "title": "A.I. Is Making Doctors Answer a Question: What Are They Really Good For?", "content": "Many physicians find chatbots threatening, but that doesn’t mean they’re giving up on medicine.", "creator": "Gina Kolata", "isoDate": "2026-02-09T17:17:56.000Z", "pubDate": "Mon, 09 Feb 2026 17:17:56 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Doctors" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Medicine and Health" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Research" } ], "dc:creator": "Gina Kolata", "contentSnippet": "Many physicians find chatbots threatening, but that doesn’t mean they’re giving up on medicine." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/08/style/ai-tech-san-francisco.html", "link": "https://www.nytimes.com/2026/02/08/style/ai-tech-san-francisco.html", "title": "These A.I. Dreamers Don’t Fit the Stereotype", "content": "Young tech entrepreneurs in San Francisco are hoping to cash in, even as they wonder how artificial intelligence will affect society.", "creator": "Guy Trebay", "isoDate": "2026-02-09T16:15:44.000Z", "pubDate": "Mon, 09 Feb 2026 16:15:44 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Start-ups" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Youth" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Entrepreneurship" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "San Francisco (Calif)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Content Type: Personal Profile" } ], "dc:creator": "Guy Trebay", "contentSnippet": "Young tech entrepreneurs in San Francisco are hoping to cash in, even as they wonder how artificial intelligence will affect society." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/07/technology/washington-post-will-lewis.html", "link": "https://www.nytimes.com/2026/02/07/technology/washington-post-will-lewis.html", "title": "Washington Post C.E.O. Will Lewis Steps Down After Stormy Tenure", "content": "His departure came days after the company cut 30 percent of the staff. He will be replaced in the interim by Jeff D’Onofrio, the chief financial officer, the company said.", "creator": "Benjamin Mullin, Katie Robertson and Erik Wemple", "isoDate": "2026-02-08T14:40:43.000Z", "pubDate": "Sun, 08 Feb 2026 14:40:43 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Washington Post" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Lewis, William (1969- )" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Appointments and Executive Changes" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Newspapers" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "News and News Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Layoffs and Job Reductions" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Bezos, Jeffrey P" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "internal-open-access-from-nl" } ], "dc:creator": "Benjamin Mullin, Katie Robertson and Erik Wemple", "contentSnippet": "His departure came days after the company cut 30 percent of the staff. He will be replaced in the interim by Jeff D’Onofrio, the chief financial officer, the company said." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/07/science/mathematics-ai-proof-hairer.html", "link": "https://www.nytimes.com/2026/02/07/science/mathematics-ai-proof-hairer.html", "title": "These Mathematicians Are Trying to Educate A.I.", "content": "Large language models struggle to solve research-level math questions. It takes a human to assess just how poorly they perform.", "creator": "Siobhan Roberts", "isoDate": "2026-02-07T17:08:57.000Z", "pubDate": "Sat, 07 Feb 2026 17:08:57 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Mathematics" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Google Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "OpenAI Labs" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Hairer, Martin (1975- )" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Research" } ], "dc:creator": "Siobhan Roberts", "contentSnippet": "Large language models struggle to solve research-level math questions. It takes a human to assess just how poorly they perform." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/07/technology/elon-musk-spacex-xai.html", "link": "https://www.nytimes.com/2026/02/07/technology/elon-musk-spacex-xai.html", "title": "After Merging xAI and SpaceX, Elon Musk Hopes He Can Win Over Wall Street", "content": "The billionaire’s decision to merge his A.I. start-up with his rocket company will test investors’ interest in giant combinations of unalike businesses.", "creator": "Kate Conger and Jack Ewing", "isoDate": "2026-02-08T01:27:03.000Z", "pubDate": "Sun, 08 Feb 2026 01:27:03 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Space Exploration Technologies Corp" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "X.ai Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Mergers, Acquisitions and Divestitures" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Start-ups" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Private Spaceflight" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Space and Astronomy" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Musk, Elon" } ], "dc:creator": "Kate Conger and Jack Ewing", "contentSnippet": "The billionaire’s decision to merge his A.I. start-up with his rocket company will test investors’ interest in giant combinations of unalike businesses." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/business/stellantis-writedown-electric-vehicles.html", "link": "https://www.nytimes.com/2026/02/06/business/stellantis-writedown-electric-vehicles.html", "title": "Stellantis’s Shift Away From Electric Cars Will Cost It $26 Billion", "content": "The company, which owns Chrysler, Fiat, Jeep and Peugeot, is changing its strategy to gasoline and hybrid vehicles in an effort to revive weak sales.", "creator": "Neal E. Boudette", "isoDate": "2026-02-06T21:18:09.000Z", "pubDate": "Fri, 06 Feb 2026 21:18:09 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Automobiles" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Electric and Hybrid Vehicles" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Prices (Fares, Fees and Rates)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Oil (Petroleum) and Gasoline" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Stellantis NV" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Jeep Division of Fiat Chrysler" } ], "dc:creator": "Neal E. Boudette", "contentSnippet": "The company, which owns Chrysler, Fiat, Jeep and Peugeot, is changing its strategy to gasoline and hybrid vehicles in an effort to revive weak sales." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/06/podcasts/hardfork-musk-merger-genie.html", "link": "https://www.nytimes.com/2026/02/06/podcasts/hardfork-musk-merger-genie.html", "title": "Elon Musk’s Mega-Merger + We Test Google’s Project Genie + What’s Next for Moltbook Creator", "content": "“A very valuable and profitable company in SpaceX has acquired a cash furnace named xAI.”", "creator": "Kevin Roose, Casey Newton, Rachel Cohn, Whitney Jones, Vjeran Pavic, Katie McMurran, Dan Powell, Marion Lozano and Rowan Niemisto", "isoDate": "2026-02-06T12:00:03.000Z", "pubDate": "Fri, 06 Feb 2026 12:00:03 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Science and Technology" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Starlink Satellite Constellation (SpaceX)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "X (Formerly Twitter)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "X.ai Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Musk, Elon" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "OpenAI Labs" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "NVIDIA Corporation" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Oracle Corporation" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Google Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Moltbook" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "internal-open-access-audio" } ], "dc:creator": "Kevin Roose, Casey Newton, Rachel Cohn, Whitney Jones, Vjeran Pavic, Katie McMurran, Dan Powell, Marion Lozano and Rowan Niemisto", "contentSnippet": "“A very valuable and profitable company in SpaceX has acquired a cash furnace named xAI.”" }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/05/technology/amazon-200-billion-ai.html", "link": "https://www.nytimes.com/2026/02/05/technology/amazon-200-billion-ai.html", "title": "Amazon’s $200 Billion Spending Plan Raises Stakes in A.I. Race", "content": "The company reported a strong holiday quarter on Thursday. But its spending, like that at other big technology companies, is starting to make investors nervous.", "creator": "Karen Weise", "isoDate": "2026-02-05T23:51:01.000Z", "pubDate": "Thu, 05 Feb 2026 23:51:01 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Amazon.com Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Company Reports" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "E-Commerce" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Data Centers" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Cloud Computing" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Layoffs and Job Reductions" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Anthropic AI LLC" } ], "dc:creator": "Karen Weise", "contentSnippet": "The company reported a strong holiday quarter on Thursday. But its spending, like that at other big technology companies, is starting to make investors nervous." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/05/technology/bitcoin-price-drop-crypto-market.html", "link": "https://www.nytimes.com/2026/02/05/technology/bitcoin-price-drop-crypto-market.html", "title": "Bitcoin Drops to Lowest Price Since Trump Was Elected as Crypto Faces Slump", "content": "The price of Bitcoin is now lower than when President Trump was elected in 2024, raising concerns of a new “crypto winter\" in the industry.", "creator": "David Yaffe-Bellany", "isoDate": "2026-02-05T23:46:58.000Z", "pubDate": "Thu, 05 Feb 2026 23:46:58 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Virtual Currency" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Trump, Donald J" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "United States Politics and Government" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Regulation and Deregulation of Industry" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Bitcoin (Currency)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Stocks and Bonds" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Banking and Financial Institutions" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Gemini Trust Co LLC" } ], "dc:creator": "David Yaffe-Bellany", "contentSnippet": "The price of Bitcoin is now lower than when President Trump was elected in 2024, raising concerns of a new “crypto winter\" in the industry." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/05/world/europe/starlink-blocks-russian-troops-access.html", "link": "https://www.nytimes.com/2026/02/05/world/europe/starlink-blocks-russian-troops-access.html", "title": "At Ukraine’s Request, Starlink Denies Internet Access to Russian Troops", "content": "It’s unclear what effect the change will have on Russia’s ability to wage war, but Russian military bloggers said troops were experiencing internet outages that hampered frontline communications.", "creator": "Paul Sonne and Maria Varenikova", "isoDate": "2026-02-06T00:00:51.000Z", "pubDate": "Fri, 06 Feb 2026 00:00:51 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Russian Invasion of Ukraine (2022)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Starlink Satellite Constellation (SpaceX)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Satellites" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computer Network Outages" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Communications and Media Blackouts" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Musk, Elon" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Russia" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Ukraine" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Defense and Military Forces" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Drones (Pilotless Planes)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "International Relations" } ], "dc:creator": "Paul Sonne and Maria Varenikova", "contentSnippet": "It’s unclear what effect the change will have on Russia’s ability to wage war, but Russian military bloggers said troops were experiencing internet outages that hampered frontline communications." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/04/world/europe/social-media-free-speech.html", "link": "https://www.nytimes.com/2026/02/04/world/europe/social-media-free-speech.html", "title": "France’s Raid on X Escalates Trans-Atlantic Showdown Over Social Media", "content": "The French investigation into Elon Musk’s X illustrated a fundamental divide between European and American leaders about how to regulate social media — or whether to restrict it at all.", "creator": "Adam Satariano and Jeanna Smialek", "isoDate": "2026-02-04T23:08:40.000Z", "pubDate": "Wed, 04 Feb 2026 23:08:40 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Europe" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Freedom of Speech and Expression" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" } ], "dc:creator": "Adam Satariano and Jeanna Smialek", "contentSnippet": "The French investigation into Elon Musk’s X illustrated a fundamental divide between European and American leaders about how to regulate social media — or whether to restrict it at all." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/04/podcasts/moltbook-mania-explained.html", "link": "https://www.nytimes.com/2026/02/04/podcasts/moltbook-mania-explained.html", "title": "Moltbook Mania Explained", "content": "Is this the year the internet changes forever?", "creator": "Kevin Roose, Casey Newton, Whitney Jones, Rachel Cohn, Vjeran Pavic, Katie McMurran, Dan Powell and Alyssa Moxley", "isoDate": "2026-02-04T12:00:06.000Z", "pubDate": "Wed, 04 Feb 2026 12:00:06 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Science and Technology" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Moltbook" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "internal-open-access-audio" } ], "dc:creator": "Kevin Roose, Casey Newton, Whitney Jones, Rachel Cohn, Vjeran Pavic, Katie McMurran, Dan Powell and Alyssa Moxley", "contentSnippet": "Is this the year the internet changes forever?" }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/03/business/media/netflix-sarandos-warner-senate-hearing.html", "link": "https://www.nytimes.com/2026/02/03/business/media/netflix-sarandos-warner-senate-hearing.html", "title": "Netflix Leader Pushes Warner Deal Before Skeptical Lawmakers", "content": "Senators asked Ted Sarandos about whether the acquisition would raise prices, squeeze talent and degrade the moviegoing experience.", "creator": "David McCabe", "isoDate": "2026-02-03T23:47:39.000Z", "pubDate": "Tue, 03 Feb 2026 23:47:39 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Mergers, Acquisitions and Divestitures" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Movies" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "United States Politics and Government" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Video Recordings, Downloads and Streaming" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Prices (Fares, Fees and Rates)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Antitrust Laws and Competition Issues" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "CNN" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Netflix Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Paramount Global" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Senate Committee on the Judiciary" } ], "dc:creator": "David McCabe", "contentSnippet": "Senators asked Ted Sarandos about whether the acquisition would raise prices, squeeze talent and degrade the moviegoing experience." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/03/technology/trump-statue-don-colossus.html", "link": "https://www.nytimes.com/2026/02/03/technology/trump-statue-don-colossus.html", "title": "‘Don Colossus,’ a Golden Statue of President Trump, Waits for Its Home", "content": "A group of cryptocurrency investors backing a memecoin hopes the statue will soon be installed at one of Mr. Trump’s golf courses in Florida.", "creator": "David Yaffe-Bellany", "isoDate": "2026-02-03T20:53:42.000Z", "pubDate": "Tue, 03 Feb 2026 20:53:42 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Trump, Donald J" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Sculpture" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Monuments and Memorials (Structures)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Virtual Currency" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Copyrights and Copyright Violations" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Intellectual Property" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Gold" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Trump National Doral Miami (Doral, Fla)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Cottrill, Alan" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Trump Assassination Attempt (July 2024)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "United States Politics and Government" } ], "dc:creator": "David Yaffe-Bellany", "contentSnippet": "A group of cryptocurrency investors backing a memecoin hopes the statue will soon be installed at one of Mr. Trump’s golf courses in Florida." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/03/world/europe/spain-social-media-ban-under-16.html", "link": "https://www.nytimes.com/2026/02/03/world/europe/spain-social-media-ban-under-16.html", "title": "Spain Aims to Ban Social Media for Children Under 16", "content": "The announcement is part of a broader push by countries to curb access to online platforms for minors. It also points to Europe’s stricter approach to regulating social media.", "creator": "José Bautista", "isoDate": "2026-02-03T21:15:35.000Z", "pubDate": "Tue, 03 Feb 2026 21:15:35 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Spain" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Children and Childhood" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Regulation and Deregulation of Industry" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Sanchez Perez-Castejon, Pedro (1972- )" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Europe" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "internal-open-access-from-nl" } ], "dc:creator": "José Bautista", "contentSnippet": "The announcement is part of a broader push by countries to curb access to online platforms for minors. It also points to Europe’s stricter approach to regulating social media." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/03/business/energy-environment/siemens-energy-ai-power-demand.html", "link": "https://www.nytimes.com/2026/02/03/business/energy-environment/siemens-energy-ai-power-demand.html", "title": "Siemens Energy Bets $1 Billion That A.I. Power Demand Will Last", "content": "The German manufacturer announced plans to expand factories in several U.S. states and build a new plant in Mississippi.", "creator": "Rebecca F. Elliott", "isoDate": "2026-02-03T11:00:16.000Z", "pubDate": "Tue, 03 Feb 2026 11:00:16 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Factories and Manufacturing" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Natural Gas" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Data Centers" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Foreign Investments" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Electric Light and Power" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Prices (Fares, Fees and Rates)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "International Trade and World Market" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Siemens AG" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Charlotte (NC)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Mississippi" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Florida" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Germany" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "United States" } ], "dc:creator": "Rebecca F. Elliott", "contentSnippet": "The German manufacturer announced plans to expand factories in several U.S. states and build a new plant in Mississippi." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "guid": "https://www.nytimes.com/2026/02/03/technology/minneapolis-ai-disinformation-misinformation-truth.html", "link": "https://www.nytimes.com/2026/02/03/technology/minneapolis-ai-disinformation-misinformation-truth.html", "title": "Chaos in Minneapolis Exposes an Internet at War With Truth", "content": "Technological advances and an erosion of trust have transformed the way news unfolds online, distorting shared reality.", "creator": "Stuart A. Thompson, Tiffany Hsu and Steven Lee Myers", "isoDate": "2026-02-03T14:19:53.000Z", "pubDate": "Tue, 03 Feb 2026 14:19:53 +0000", "categories": [ { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_geo" }, "_": "Minneapolis (Minn)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Police Brutality, Misconduct and Shootings" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Social Media" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Rumors and Misinformation" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Federal Actions in US Cities" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "United States Politics and Government" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Artificial Intelligence" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "George Floyd Protests (2020)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Computers and the Internet" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/des" }, "_": "Video Recordings, Downloads and Streaming" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "X (Formerly Twitter)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_org" }, "_": "Facebook Inc" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Floyd, George (d 2020)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Good, Renee Nicole (1988-2026)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Pretti, Alex Jeffrey (1988-2026)" }, { "$": { "domain": "http://www.nytimes.com/namespaces/keywords/nyt_per" }, "_": "Trump, Donald J" } ], "dc:creator": "Stuart A. Thompson, Tiffany Hsu and Steven Lee Myers", "contentSnippet": "Technological advances and an erosion of trust have transformed the way news unfolds online, distorting shared reality." }, "pairedItem": [ { "item": 7 } ] }, { "json": { "date": "2026-02-09T14:00:22Z", "guid": "https://www.theguardian.com/technology/2026/feb/09/jeffrey-epstein-crypto", "link": "https://www.theguardian.com/technology/2026/feb/09/jeffrey-epstein-crypto", "title": "Files cast light on Jeffrey Epstein’s ties to cryptocurrency", "content": "Newly released documents detail convicted sex offender’s early backing of bitcoin and Coinbase
Millions of files related to Jeffrey Epstein have brought to light his ties to the highest echelons of the cryptocurrency industry.
Documents published last week by the US Department of Justice reveal Epstein bankrolled the “principal home and funding source” for bitcoin, the world’s largest cryptocurrency, during its nascent stages; he also invested $3m in Coinbase in 2014, the largest cryptocurrency exchange in the US, and cut a check that same year to Blockstream, a prominent bitcoin-focused technology firm. Both crypto startups accepted Epstein’s investments in 2014 – six years after his 2008 conviction in Florida for soliciting prostitution from a minor.
Continue reading...", "creator": "Adam Willems", "dc:date": "2026-02-09T14:00:22Z", "isoDate": "2026-02-09T14:00:22.000Z", "pubDate": "Mon, 09 Feb 2026 14:00:22 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/cryptocurrencies" }, "_": "Cryptocurrencies" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/us-news/jeffrey-epstein" }, "_": "Jeffrey Epstein" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-news" }, "_": "US news" } ], "dc:creator": "Adam Willems", "contentSnippet": "Newly released documents detail convicted sex offender’s early backing of bitcoin and Coinbase \nMillions of files related to Jeffrey Epstein have brought to light his ties to the highest echelons of the cryptocurrency industry.\nDocuments published last week by the US Department of Justice reveal Epstein bankrolled the “principal home and funding source” for bitcoin, the world’s largest cryptocurrency, during its nascent stages; he also invested $3m in Coinbase in 2014, the largest cryptocurrency exchange in the US, and cut a check that same year to Blockstream, a prominent bitcoin-focused technology firm. Both crypto startups accepted Epstein’s investments in 2014 – six years after his 2008 conviction in Florida for soliciting prostitution from a minor.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-08T08:00:25Z", "guid": "https://www.theguardian.com/technology/2026/feb/08/i-fell-into-it-ex-criminal-hackers-urge-manchester-pupils-to-use-web-skills-for-good", "link": "https://www.theguardian.com/technology/2026/feb/08/i-fell-into-it-ex-criminal-hackers-urge-manchester-pupils-to-use-web-skills-for-good", "title": "‘I fell into it’: ex-criminal hackers urge Manchester pupils to use web skills for good", "content": "Initiative aims to identify proficient gamers and coders who can help companies identify flaws in their cybersecurity
Cybercriminals, the shadowy online figures often depicted in Hollywood movies as hooded villains capable of wiping millions of pounds off the value of businesses at a keystroke, are not usually known for their candour.
But in a sixth-form college in Manchester this week, two former hackers gave the young people gathered an honest appraisal of what living a life of internet crime really looks like.
Continue reading...", "creator": "Dan Milmo Global technology editor", "dc:date": "2026-02-08T08:00:25Z", "isoDate": "2026-02-08T08:00:25.000Z", "pubDate": "Sun, 08 Feb 2026 08:00:25 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/hacking" }, "_": "Hacking" }, { "$": { "domain": "https://www.theguardian.com/technology/cybercrime" }, "_": "Cybercrime" }, { "$": { "domain": "https://www.theguardian.com/business/co-operative-group" }, "_": "Co-operative Group" }, { "$": { "domain": "https://www.theguardian.com/uk/uk" }, "_": "UK news" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/technology/internet" }, "_": "Internet" }, { "$": { "domain": "https://www.theguardian.com/technology/programming" }, "_": "Programming" }, { "$": { "domain": "https://www.theguardian.com/games/game-culture" }, "_": "Game culture" }, { "$": { "domain": "https://www.theguardian.com/society/youngpeople" }, "_": "Young people" }, { "$": { "domain": "https://www.theguardian.com/uk/manchester" }, "_": "Manchester" } ], "dc:creator": "Dan Milmo Global technology editor", "contentSnippet": "Initiative aims to identify proficient gamers and coders who can help companies identify flaws in their cybersecurity \nCybercriminals, the shadowy online figures often depicted in Hollywood movies as hooded villains capable of wiping millions of pounds off the value of businesses at a keystroke, are not usually known for their candour.\nBut in a sixth-form college in Manchester this week, two former hackers gave the young people gathered an honest appraisal of what living a life of internet crime really looks like.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-07T16:00:06Z", "guid": "https://www.theguardian.com/technology/2026/feb/07/ai-chatbots-anthropic-openai-claude-chatgpt", "link": "https://www.theguardian.com/technology/2026/feb/07/ai-chatbots-anthropic-openai-claude-chatgpt", "title": "Battle of the chatbots: Anthropic and OpenAI go head-to-head over ads in their AI products", "content": "New Anthropic campaign suggests other AI platforms will incorporate targeted ads in their chatbot conversations
The Seahawks and the Patriots aren’t the only ones gearing up for a fight.
AI rivals Anthropic and OpenAI have launched a war of ads trying to court corporate America during one of the biggest entertainment nights of the year.
Continue reading...", "creator": "Sanya Mansoor", "dc:date": "2026-02-07T16:00:06Z", "isoDate": "2026-02-07T16:00:06.000Z", "pubDate": "Sat, 07 Feb 2026 16:00:06 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/media/advertising" }, "_": "Advertising" }, { "$": { "domain": "https://www.theguardian.com/technology/chatbots" }, "_": "Chatbots" }, { "$": { "domain": "https://www.theguardian.com/technology/chatgpt" }, "_": "ChatGPT" }, { "$": { "domain": "https://www.theguardian.com/technology/openai" }, "_": "OpenAI" }, { "$": { "domain": "https://www.theguardian.com/technology/sam-altman" }, "_": "Sam Altman" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-news" }, "_": "US news" } ], "dc:creator": "Sanya Mansoor", "contentSnippet": "New Anthropic campaign suggests other AI platforms will incorporate targeted ads in their chatbot conversations\nThe Seahawks and the Patriots aren’t the only ones gearing up for a fight.\nAI rivals Anthropic and OpenAI have launched a war of ads trying to court corporate America during one of the biggest entertainment nights of the year.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-09T10:00:31Z", "guid": "https://www.theguardian.com/commentisfree/2026/feb/09/who-is-my-wife-martin-rowson-ai-technology", "link": "https://www.theguardian.com/commentisfree/2026/feb/09/who-is-my-wife-martin-rowson-ai-technology", "title": "I asked AI to name my wife. To the hopelessly incorrect people it cited, my deepest apologies | Martin Rowson", "content": "Authors, a newsreader, a lawyer and an esteemed colleague: they’re all great – but I’m not married to any of them. Can we really depend on this technology?
Recently, the Rowsons accidentally invented a new game that anyone can play at home. I have yet to come up with a world-beating name for it, so for now let’s just call it “How bloody stupid is AI?” The playing of the game will change from player to player, depending on their circumstances – but essentially the rules remain the same. Ask AI a simple question about yourself, and see just how wrong it gets it.
In my case, all you need know is that while I, through the nature of my job, have a fairly large online presence, my partner (we married in 1987) has assiduously avoided having one at all. Which means that if you Google “Martin Rowson wife” in images, you may get a picture of me next to our then 14-year-old daughter or me with my friend and fellow cartoonist Steven Appleby, who happens to be trans but has kept her given first name.
Continue reading...", "creator": "Martin Rowson", "dc:date": "2026-02-09T10:00:31Z", "isoDate": "2026-02-09T10:00:31.000Z", "pubDate": "Mon, 09 Feb 2026 10:00:31 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/uk/uk" }, "_": "UK news" }, { "$": { "domain": "https://www.theguardian.com/technology/computing" }, "_": "Computing" } ], "dc:creator": "Martin Rowson", "contentSnippet": "Authors, a newsreader, a lawyer and an esteemed colleague: they’re all great – but I’m not married to any of them. Can we really depend on this technology?\nRecently, the Rowsons accidentally invented a new game that anyone can play at home. I have yet to come up with a world-beating name for it, so for now let’s just call it “How bloody stupid is AI?” The playing of the game will change from player to player, depending on their circumstances – but essentially the rules remain the same. Ask AI a simple question about yourself, and see just how wrong it gets it.\nIn my case, all you need know is that while I, through the nature of my job, have a fairly large online presence, my partner (we married in 1987) has assiduously avoided having one at all. Which means that if you Google “Martin Rowson wife” in images, you may get a picture of me next to our then 14-year-old daughter or me with my friend and fellow cartoonist Steven Appleby, who happens to be trans but has kept her given first name.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-07T10:00:02Z", "guid": "https://www.theguardian.com/technology/2026/feb/07/campaigners-call-stronger-protection-against-ai-generated-explicit-imagery", "link": "https://www.theguardian.com/technology/2026/feb/07/campaigners-call-stronger-protection-against-ai-generated-explicit-imagery", "title": "Victims urge tougher action on deepfake abuse as new law comes into force", "content": "Campaigners welcome criminalisation of non-consensual AI-generated explicit images but say law does not go far enough
Victims of deepfake image abuse have called for stronger protection against AI-generated explicit images, as the law criminalising the creation of non-consensual intimate images comes into effect.
Campaigners from Stop Image-Based Abuse delivered a petition to Downing Street with more than 73,000 signatures, urging the government to introduce civil routes to justice such as takedown orders for abusive imagery on platforms and devices.
Continue reading...", "creator": "Priya Bharadia", "dc:date": "2026-02-07T10:00:02Z", "isoDate": "2026-02-07T10:00:02.000Z", "pubDate": "Sat, 07 Feb 2026 10:00:02 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/deepfake" }, "_": "Deepfake" }, { "$": { "domain": "https://www.theguardian.com/society/violence-against-women-and-girls" }, "_": "Violence against women and girls" }, { "$": { "domain": "https://www.theguardian.com/uk/ukcrime" }, "_": "Crime" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/uk/uk" }, "_": "UK news" }, { "$": { "domain": "https://www.theguardian.com/society/women" }, "_": "Women" }, { "$": { "domain": "https://www.theguardian.com/technology/internet" }, "_": "Internet" }, { "$": { "domain": "https://www.theguardian.com/politics/politics" }, "_": "Politics" } ], "dc:creator": "Priya Bharadia", "contentSnippet": "Campaigners welcome criminalisation of non-consensual AI-generated explicit images but say law does not go far enough\nVictims of deepfake image abuse have called for stronger protection against AI-generated explicit images, as the law criminalising the creation of non-consensual intimate images comes into effect.\nCampaigners from Stop Image-Based Abuse delivered a petition to Downing Street with more than 73,000 signatures, urging the government to introduce civil routes to justice such as takedown orders for abusive imagery on platforms and devices.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-07T14:00:04Z", "guid": "https://www.theguardian.com/technology/2026/feb/07/why-has-elon-musk-merged-his-rocket-company-with-his-ai-startup", "link": "https://www.theguardian.com/technology/2026/feb/07/why-has-elon-musk-merged-his-rocket-company-with-his-ai-startup", "title": "Why has Elon Musk merged his rocket company with his AI startup?", "content": "SpaceX’s acquisition of xAI creates business worth $1.25tn but whether premise behind deal will work is questioned
The acquisition of xAI by SpaceX is a typical Elon Musk deal: big numbers backed by big ambition.
As well as extending “the light of consciousness to the stars”, as Musk described it, the transaction creates a business worth $1.25tn (£920bn) by combining Musk’s rocket company with his artificial intelligence startup. It values SpaceX at $1tn and xAI at $250bn, with a stock market flotation expected in June to time with Musk’s birthday and a planetary alignment.
Continue reading...", "creator": "Dan Milmo Global technology editor", "dc:date": "2026-02-07T14:00:04Z", "isoDate": "2026-02-07T14:00:04.000Z", "pubDate": "Sat, 07 Feb 2026 14:00:04 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/elon-musk" }, "_": "Elon Musk" }, { "$": { "domain": "https://www.theguardian.com/science/spacex" }, "_": "SpaceX" }, { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/business/business" }, "_": "Business" }, { "$": { "domain": "https://www.theguardian.com/business/technology" }, "_": "Technology sector" }, { "$": { "domain": "https://www.theguardian.com/world/world" }, "_": "World news" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/science/science" }, "_": "Science" }, { "$": { "domain": "https://www.theguardian.com/business/mergers-and-acquisitions" }, "_": "Mergers and acquisitions" }, { "$": { "domain": "https://www.theguardian.com/technology/tesla" }, "_": "Tesla" }, { "$": { "domain": "https://www.theguardian.com/technology/twitter" }, "_": "X" } ], "dc:creator": "Dan Milmo Global technology editor", "contentSnippet": "SpaceX’s acquisition of xAI creates business worth $1.25tn but whether premise behind deal will work is questioned\nThe acquisition of xAI by SpaceX is a typical Elon Musk deal: big numbers backed by big ambition.\nAs well as extending “the light of consciousness to the stars”, as Musk described it, the transaction creates a business worth $1.25tn (£920bn) by combining Musk’s rocket company with his artificial intelligence startup. It values SpaceX at $1tn and xAI at $250bn, with a stock market flotation expected in June to time with Musk’s birthday and a planetary alignment.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-06T12:00:07Z", "guid": "https://www.theguardian.com/technology/2026/feb/06/amazon-georgia-warehouse-tour-robots", "link": "https://www.theguardian.com/technology/2026/feb/06/amazon-georgia-warehouse-tour-robots", "title": "Hail our new robot overlords! Amazon warehouse tour offers glimpse of future", "content": "At its new Stone Mountain, Georgia, facility, Roomba-like robots shuffle between stacks, another adds shipping labels while another arranges packages in pallets
One of the reasons Amazon is spending billions on robots? They don’t need bathroom breaks. Arriving a few minutes early to the public tour of Amazon’s hi-tech Stone Mountain, Georgia, warehouse, my request to visit the restroom was met with a resounding no from the security guard in the main lobby.
Between the main doors and the entrance security gate, I paced and paced after being told I would have to wait for the tour guide to collect me and other guests for a tour of the 640,000-sq-ft, four-story warehouse.
Continue reading...", "creator": "Michael Sainato in Stone Mountain, Georgia", "dc:date": "2026-02-06T12:00:07Z", "isoDate": "2026-02-06T12:00:07.000Z", "pubDate": "Fri, 06 Feb 2026 12:00:07 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/amazon" }, "_": "Amazon" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/us-news/state-of-georgia" }, "_": "Georgia" }, { "$": { "domain": "https://www.theguardian.com/business/business" }, "_": "Business" }, { "$": { "domain": "https://www.theguardian.com/technology/robots" }, "_": "Robots" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-news" }, "_": "US news" }, { "$": { "domain": "https://www.theguardian.com/technology/efinance" }, "_": "E-commerce" } ], "dc:creator": "Michael Sainato in Stone Mountain, Georgia", "contentSnippet": "At its new Stone Mountain, Georgia, facility, Roomba-like robots shuffle between stacks, another adds shipping labels while another arranges packages in pallets\nOne of the reasons Amazon is spending billions on robots? They don’t need bathroom breaks. Arriving a few minutes early to the public tour of Amazon’s hi-tech Stone Mountain, Georgia, warehouse, my request to visit the restroom was met with a resounding no from the security guard in the main lobby.\nBetween the main doors and the entrance security gate, I paced and paced after being told I would have to wait for the tour guide to collect me and other guests for a tour of the 640,000-sq-ft, four-story warehouse.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-09T14:00:20Z", "guid": "https://www.theguardian.com/global-development/2026/feb/09/energy-colony-argentina-patagonia-uranium-nuclear-plan-backlash-over-us-interests", "link": "https://www.theguardian.com/global-development/2026/feb/09/energy-colony-argentina-patagonia-uranium-nuclear-plan-backlash-over-us-interests", "title": "‘We’re being turned into an energy colony’: Argentina’s nuclear plan faces backlash over US interests", "content": "Push to restart uranium mining in Patagonia has sparked fears about the environmental impact and loss of sovereignty over key resources
On an outcrop above the Chubut River, one of the few to cut across the arid Patagonian steppe of southern Argentina, Sergio Pichiñán points across a wide swath of scrubland to colourful rock formations on a distant hillside.
“That’s where they dug for uranium before, and when the miners left, they left the mountain destroyed, the houses abandoned, and nobody ever studied the water,” he says, citing suspicions arising from cases of cancer and skin diseases in his community. “If they want to open this back up, we’re all pretty worried around here.”
Continue reading...", "creator": "Gioia Claro and Denali DeGraf in Cerro Cóndor, Argentina", "dc:date": "2026-02-09T14:00:20Z", "isoDate": "2026-02-09T14:00:20.000Z", "pubDate": "Mon, 09 Feb 2026 14:00:20 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/environment/mining" }, "_": "Mining" }, { "$": { "domain": "https://www.theguardian.com/environment/environment" }, "_": "Environment" }, { "$": { "domain": "https://www.theguardian.com/world/world" }, "_": "World news" }, { "$": { "domain": "https://www.theguardian.com/environment/nuclearpower" }, "_": "Nuclear power" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-news" }, "_": "US news" }, { "$": { "domain": "https://www.theguardian.com/environment/energy" }, "_": "Energy" }, { "$": { "domain": "https://www.theguardian.com/world/javier-milei" }, "_": "Javier Milei" }, { "$": { "domain": "https://www.theguardian.com/environment/nuclear-waste" }, "_": "Nuclear waste" }, { "$": { "domain": "https://www.theguardian.com/business/energy-industry" }, "_": "Energy industry" }, { "$": { "domain": "https://www.theguardian.com/business/business" }, "_": "Business" }, { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/world/iaea" }, "_": "International Atomic Energy Agency (IAEA)" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-foreign-policy" }, "_": "US foreign policy" }, { "$": { "domain": "https://www.theguardian.com/environment/climate-crisis" }, "_": "Climate crisis" }, { "$": { "domain": "https://www.theguardian.com/environment/climate-change-scepticism" }, "_": "Climate science scepticism and denial" }, { "$": { "domain": "https://www.theguardian.com/environment/renewableenergy" }, "_": "Renewable energy" } ], "dc:creator": "Gioia Claro and Denali DeGraf in Cerro Cóndor, Argentina", "contentSnippet": "Push to restart uranium mining in Patagonia has sparked fears about the environmental impact and loss of sovereignty over key resources\nOn an outcrop above the Chubut River, one of the few to cut across the arid Patagonian steppe of southern Argentina, Sergio Pichiñán points across a wide swath of scrubland to colourful rock formations on a distant hillside.\n“That’s where they dug for uranium before, and when the miners left, they left the mountain destroyed, the houses abandoned, and nobody ever studied the water,” he says, citing suspicions arising from cases of cancer and skin diseases in his community. “If they want to open this back up, we’re all pretty worried around here.”\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-09T05:00:52Z", "guid": "https://www.theguardian.com/artanddesign/2026/feb/09/china-ai-weiwei-western-censorship-home-surveillance", "link": "https://www.theguardian.com/artanddesign/2026/feb/09/china-ai-weiwei-western-censorship-home-surveillance", "title": "‘Was I scared going back to China? No’: Ai Weiwei on AI, western censorship and returning home", "content": "He has been jailed, tracked and threatened by China’s government. What was it like pay a visit home? As he publishes a polemic about surveillance and state control, the artist relives a momentous trip to see his mother
Ai Weiwei is talking me through the decision-making process before his first visit to China in over a decade. The artist, known around the world as the most famous critic of the Chinese communist regime, had to do some fraught arithmetic before deciding to head back home.
Before boarding a flight with his son, who had never met the artist’s elderly mother, Ai thought back to his time in detention when his captors told him he would spend the next 13 years in custody on bogus charges: “They said, ‘When you come out, your son won’t recognise you.’ That was very heavy and really the only moment that touched me.”
Continue reading...", "creator": "Lanre Bakare", "dc:date": "2026-02-09T05:00:52Z", "isoDate": "2026-02-09T05:00:52.000Z", "pubDate": "Mon, 09 Feb 2026 05:00:52 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/artanddesign/ai-weiwei" }, "_": "Ai Weiwei" }, { "$": { "domain": "https://www.theguardian.com/artanddesign/artanddesign" }, "_": "Art and design" }, { "$": { "domain": "https://www.theguardian.com/culture/culture" }, "_": "Culture" }, { "$": { "domain": "https://www.theguardian.com/world/china" }, "_": "China" }, { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/world/censorship" }, "_": "Censorship" }, { "$": { "domain": "https://www.theguardian.com/world/asia-pacific" }, "_": "Asia Pacific" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/world/world" }, "_": "World news" } ], "dc:creator": "Lanre Bakare", "contentSnippet": "He has been jailed, tracked and threatened by China’s government. What was it like pay a visit home? As he publishes a polemic about surveillance and state control, the artist relives a momentous trip to see his mother\nAi Weiwei is talking me through the decision-making process before his first visit to China in over a decade. The artist, known around the world as the most famous critic of the Chinese communist regime, had to do some fraught arithmetic before deciding to head back home.\nBefore boarding a flight with his son, who had never met the artist’s elderly mother, Ai thought back to his time in detention when his captors told him he would spend the next 13 years in custody on bogus charges: “They said, ‘When you come out, your son won’t recognise you.’ That was very heavy and really the only moment that touched me.”\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-07T16:00:07Z", "guid": "https://www.theguardian.com/us-news/2026/feb/07/california-monterey-park-stop-datacenter-construction", "link": "https://www.theguardian.com/us-news/2026/feb/07/california-monterey-park-stop-datacenter-construction", "title": "Rage against the machine: a California community rallied against a datacenter – and won", "content": "Organizers in Monterey Park took inspiration from other US cities to fight against the construction of a giant datacenter
When a southern California city council proposed building a giant datacenter the size of four football fields last December, five residents vowed to stop it.
Through a frenetic word-of-mouth campaign, the small group raised awareness about the proposed facility in Monterey Park, a small city east of Los Angeles known affectionately as the country’s first suburban Chinatown.
Continue reading...", "creator": "Claire Wang", "dc:date": "2026-02-07T16:00:07Z", "isoDate": "2026-02-07T16:00:07.000Z", "pubDate": "Sat, 07 Feb 2026 16:00:07 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/us-news/us-news" }, "_": "US news" }, { "$": { "domain": "https://www.theguardian.com/technology/artificialintelligenceai" }, "_": "AI (artificial intelligence)" }, { "$": { "domain": "https://www.theguardian.com/us-news/california" }, "_": "California" }, { "$": { "domain": "https://www.theguardian.com/us-news/west-coast" }, "_": "West Coast" }, { "$": { "domain": "https://www.theguardian.com/environment/energy" }, "_": "Energy" } ], "dc:creator": "Claire Wang", "contentSnippet": "Organizers in Monterey Park took inspiration from other US cities to fight against the construction of a giant datacenter\nWhen a southern California city council proposed building a giant datacenter the size of four football fields last December, five residents vowed to stop it.\nThrough a frenetic word-of-mouth campaign, the small group raised awareness about the proposed facility in Monterey Park, a small city east of Los Angeles known affectionately as the country’s first suburban Chinatown.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-05T15:00:39Z", "guid": "https://www.theguardian.com/technology/2026/feb/05/cryptocurrency-ethereum-bitcoin-industry", "link": "https://www.theguardian.com/technology/2026/feb/05/cryptocurrency-ethereum-bitcoin-industry", "title": "How cryptocurrency’s second largest coin missed out on the industry’s boom", "content": "A leaked pitch to reshape Ethereum’s leadership exposed deep divisions over politics, power and ether’s static price
US crypto developer Danny Ryan submitted a proposal in November 2024 to Vitalik Buterin, the founder and symbolic leader of Ethereum, a prominent blockchain powering the world’s second-largest cryptocurrency. Ryan, who had worked for seven years at the Ethereum Foundation (EF), Ethereum’s de facto governing body, suggested that Ethereum could be on the cusp of an era-defining shift.
Since its founding in 2014, the foundation had prioritized technical upgrades and had avoided centralizing power while its user base was growing, but Ethereum had now grown up, and the cryptocurrency world around it had grown up, too. The EF could now “exercise a stronger voice” without compromising its ethos of decentralization, Ryan said – and he was open to leading that charge if appointed as the foundation’s new executive director.
Continue reading...", "creator": "Adam Willems", "dc:date": "2026-02-05T15:00:39Z", "isoDate": "2026-02-05T15:00:39.000Z", "pubDate": "Thu, 05 Feb 2026 15:00:39 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/cryptocurrencies" }, "_": "Cryptocurrencies" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/technology/bitcoin" }, "_": "Bitcoin" }, { "$": { "domain": "https://www.theguardian.com/technology/efinance" }, "_": "E-commerce" } ], "dc:creator": "Adam Willems", "contentSnippet": "A leaked pitch to reshape Ethereum’s leadership exposed deep divisions over politics, power and ether’s static price\nUS crypto developer Danny Ryan submitted a proposal in November 2024 to Vitalik Buterin, the founder and symbolic leader of Ethereum, a prominent blockchain powering the world’s second-largest cryptocurrency. Ryan, who had worked for seven years at the Ethereum Foundation (EF), Ethereum’s de facto governing body, suggested that Ethereum could be on the cusp of an era-defining shift.\nSince its founding in 2014, the foundation had prioritized technical upgrades and had avoided centralizing power while its user base was growing, but Ethereum had now grown up, and the cryptocurrency world around it had grown up, too. The EF could now “exercise a stronger voice” without compromising its ethos of decentralization, Ryan said – and he was open to leading that charge if appointed as the foundation’s new executive director.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-04T07:00:36Z", "guid": "https://www.theguardian.com/technology/2026/feb/04/fairphone-6-review-cheaper-repairable-longer-lasting-android", "link": "https://www.theguardian.com/technology/2026/feb/04/fairphone-6-review-cheaper-repairable-longer-lasting-android", "title": "Fairphone 6 review: cheaper, repairable and longer-lasting Android", "content": "Sustainable smartphone takes a step forward with modular accessories, a good screen and mid-range performance
The Dutch ethical smartphone brand Fairphone is back with its six-generation Android, aiming to make its repairable phone more modern, modular, affordable and desirable, with screw-in accessories and a user-replaceable battery.
The Fairphone 6 costs £499 (€599), making it cheaper than previous models and pitting it squarely against budget champs such as the Google Pixel 9a and the Nothing Phone 3a Pro, while being repairable at home with long-term software support and a five-year warranty. On paper it sounds like the ideal phone to see out the decade.
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2026-02-04T07:00:36Z", "isoDate": "2026-02-04T07:00:36.000Z", "pubDate": "Wed, 04 Feb 2026 07:00:36 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/smartphones" }, "_": "Smartphones" }, { "$": { "domain": "https://www.theguardian.com/technology/android" }, "_": "Android" }, { "$": { "domain": "https://www.theguardian.com/technology/mobilephones" }, "_": "Mobile phones" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/environment/ethical-living" }, "_": "Ethical and green living" }, { "$": { "domain": "https://www.theguardian.com/business/ethicalbusiness" }, "_": "Ethical business" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "Sustainable smartphone takes a step forward with modular accessories, a good screen and mid-range performance\nThe Dutch ethical smartphone brand Fairphone is back with its six-generation Android, aiming to make its repairable phone more modern, modular, affordable and desirable, with screw-in accessories and a user-replaceable battery.\nThe Fairphone 6 costs £499 (€599), making it cheaper than previous models and pitting it squarely against budget champs such as the Google Pixel 9a and the Nothing Phone 3a Pro, while being repairable at home with long-term software support and a five-year warranty. On paper it sounds like the ideal phone to see out the decade.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2025-10-30T07:00:55Z", "guid": "https://www.theguardian.com/technology/2025/oct/30/google-pixel-10-pro-fold-review-dust-resistant-foldable-phone-screen-ai", "link": "https://www.theguardian.com/technology/2025/oct/30/google-pixel-10-pro-fold-review-dust-resistant-foldable-phone-screen-ai", "title": "Google Pixel 10 Pro Fold review: dust-resistant and more durable foldable phone", "content": "Book-style Android with cutting-edge AI, good cameras and great tablet screen for media and multitasking on the go
Google’s third-generation folding phone promises to be more durable than all others as the first with full water and dust resistance while also packing lots of advanced AI and an adaptable set of cameras.
The Pixel 10 Pro Fold builds on last year’s excellent 9 Pro Fold by doing away with gears in the hinge along its spine allowing it to deal with dust, which has been the achilles heel of all foldable phones until now, gumming up the works in a way that just isn’t a problem for regular slab phones.
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2025-10-30T07:00:55Z", "isoDate": "2025-10-30T07:00:55.000Z", "pubDate": "Thu, 30 Oct 2025 07:00:55 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/pixel" }, "_": "Pixel" }, { "$": { "domain": "https://www.theguardian.com/technology/google" }, "_": "Google" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/technology/alphabet" }, "_": "Alphabet" }, { "$": { "domain": "https://www.theguardian.com/technology/smartphones" }, "_": "Smartphones" }, { "$": { "domain": "https://www.theguardian.com/technology/tablet-computer" }, "_": "Tablet computers" }, { "$": { "domain": "https://www.theguardian.com/technology/android" }, "_": "Android" }, { "$": { "domain": "https://www.theguardian.com/technology/mobilephones" }, "_": "Mobile phones" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "Book-style Android with cutting-edge AI, good cameras and great tablet screen for media and multitasking on the go\nGoogle’s third-generation folding phone promises to be more durable than all others as the first with full water and dust resistance while also packing lots of advanced AI and an adaptable set of cameras.\nThe Pixel 10 Pro Fold builds on last year’s excellent 9 Pro Fold by doing away with gears in the hinge along its spine allowing it to deal with dust, which has been the achilles heel of all foldable phones until now, gumming up the works in a way that just isn’t a problem for regular slab phones.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2025-10-15T06:00:16Z", "guid": "https://www.theguardian.com/technology/2025/oct/15/iphone-air-review-apple-pursuit-absolute-thinness", "link": "https://www.theguardian.com/technology/2025/oct/15/iphone-air-review-apple-pursuit-absolute-thinness", "title": "iPhone Air review: Apple’s pursuit of absolute thinness", "content": "Ultra-slim and light smartphone feels special, but cuts to camera and battery may be too hard to ignore for most
The iPhone Air is a technical and design marvel that asks: how much are you willing to give up for a lightweight and ultra-slender profile?
Beyond the obvious engineering effort that has gone into creating one of the slimmest phones ever made, the Air is a reductive exercise that boils down the iPhone into the absolute essentials in a premium body.
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2025-10-15T06:00:16Z", "isoDate": "2025-10-15T06:00:16.000Z", "pubDate": "Wed, 15 Oct 2025 06:00:16 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/iphone" }, "_": "iPhone" }, { "$": { "domain": "https://www.theguardian.com/technology/apple" }, "_": "Apple" }, { "$": { "domain": "https://www.theguardian.com/technology/smartphones" }, "_": "Smartphones" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/technology/mobilephones" }, "_": "Mobile phones" }, { "$": { "domain": "https://www.theguardian.com/technology/telecoms" }, "_": "Telecoms" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "Ultra-slim and light smartphone feels special, but cuts to camera and battery may be too hard to ignore for most\nThe iPhone Air is a technical and design marvel that asks: how much are you willing to give up for a lightweight and ultra-slender profile?\nBeyond the obvious engineering effort that has gone into creating one of the slimmest phones ever made, the Air is a reductive exercise that boils down the iPhone into the absolute essentials in a premium body.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2025-10-08T06:00:18Z", "guid": "https://www.theguardian.com/technology/2025/oct/08/apple-iphone-17-pro-review-different-looks-zoom-battery-life", "link": "https://www.theguardian.com/technology/2025/oct/08/apple-iphone-17-pro-review-different-looks-zoom-battery-life", "title": "Apple iPhone 17 Pro review: different looks but still all about the zoom", "content": "First new design in ages, upgraded camera, serious performance and longer battery life make it a standout year
The 17 Pro is Apple’s biggest redesign of the iPhone in years, chucking out the old titanium sides and all-glass backs for a new aluminium unibody design, a huge full-width camera lump on the back and some bolder colours.
That alone will make the iPhone 17 Pro popular for those looking to upgrade and be seen with the newest model. But with the change comes an increase in price to £1,099 (€1,299/$1,099/A$1,999), crossing the £1,000 barrier for the first time for Apple’s smallest Pro phone, which now comes with double the starting storage.
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2025-10-08T06:00:18Z", "isoDate": "2025-10-08T06:00:18.000Z", "pubDate": "Wed, 08 Oct 2025 06:00:18 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/apple" }, "_": "Apple" }, { "$": { "domain": "https://www.theguardian.com/technology/iphone" }, "_": "iPhone" }, { "$": { "domain": "https://www.theguardian.com/technology/smartphones" }, "_": "Smartphones" }, { "$": { "domain": "https://www.theguardian.com/technology/mobilephones" }, "_": "Mobile phones" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "First new design in ages, upgraded camera, serious performance and longer battery life make it a standout year\nThe 17 Pro is Apple’s biggest redesign of the iPhone in years, chucking out the old titanium sides and all-glass backs for a new aluminium unibody design, a huge full-width camera lump on the back and some bolder colours.\nThat alone will make the iPhone 17 Pro popular for those looking to upgrade and be seen with the newest model. But with the change comes an increase in price to £1,099 (€1,299/$1,099/A$1,999), crossing the £1,000 barrier for the first time for Apple’s smallest Pro phone, which now comes with double the starting storage.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-10T07:00:15Z", "guid": "https://www.theguardian.com/technology/2026/feb/10/beats-powerbeats-fit-review-apple-compact-workout-earbuds-revamped", "link": "https://www.theguardian.com/technology/2026/feb/10/beats-powerbeats-fit-review-apple-compact-workout-earbuds-revamped", "title": "Beats Powerbeats Fit review: Apple’s compact workout earbuds revamped", "content": "Secure, noise-cancelling Bluetooth earbuds that shine for exercise and everyday use on Android and iPhone
Apple’s revamped compact workout Beats earbuds stick to a winning formula, while slimming down and improving comfort.
The new Powerbeats Fit are the direct successors to 2022’s popular Beats Fit Pro, costing £200 (€230/$200/A$330). They sit alongside the recently redesigned Powerbeats Pro 2 as Apple’s fitness alternatives of the AirPods.
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2026-02-10T07:00:15Z", "isoDate": "2026-02-10T07:00:15.000Z", "pubDate": "Tue, 10 Feb 2026 07:00:15 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/headphones" }, "_": "Headphones" }, { "$": { "domain": "https://www.theguardian.com/technology/bluetooth" }, "_": "Bluetooth" }, { "$": { "domain": "https://www.theguardian.com/technology/gadgets" }, "_": "Gadgets" }, { "$": { "domain": "https://www.theguardian.com/technology/apple" }, "_": "Apple" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "Secure, noise-cancelling Bluetooth earbuds that shine for exercise and everyday use on Android and iPhone\nApple’s revamped compact workout Beats earbuds stick to a winning formula, while slimming down and improving comfort.\nThe new Powerbeats Fit are the direct successors to 2022’s popular Beats Fit Pro, costing £200 (€230/$200/A$330). They sit alongside the recently redesigned Powerbeats Pro 2 as Apple’s fitness alternatives of the AirPods.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-09T07:00:53Z", "guid": "https://www.theguardian.com/technology/2026/feb/09/logitech-mx-master-4-review-the-best-work-mouse-you-can-buy", "link": "https://www.theguardian.com/technology/2026/feb/09/logitech-mx-master-4-review-the-best-work-mouse-you-can-buy", "title": "Logitech MX Master 4 review: the best work mouse you can buy", "content": "Ergonomic shape, quality materials and satisfying clicks, now with novel haptic feedback and repairable design
Logitech’s latest productivity power-house updates one of the greatest mice of all time with smoother materials, a repair-friendly design and a haptic motor for phone-like vibrations on your desktop.
The MX Master 4 is the latest evolution in a line of pioneering mice that dates back more than 20 years and has long been the mouse to beat for everything but hardcore PC gaming. Having given it a magnetic free-spinning scroll wheel, plenty of buttons and precise tracking, now Logitech is trying something different for its seven-generation: the ability to tap back at you.
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2026-02-09T07:00:53Z", "isoDate": "2026-02-09T07:00:53.000Z", "pubDate": "Mon, 09 Feb 2026 07:00:53 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/gadgets" }, "_": "Gadgets" }, { "$": { "domain": "https://www.theguardian.com/technology/computing" }, "_": "Computing" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "Ergonomic shape, quality materials and satisfying clicks, now with novel haptic feedback and repairable design\nLogitech’s latest productivity power-house updates one of the greatest mice of all time with smoother materials, a repair-friendly design and a haptic motor for phone-like vibrations on your desktop.\nThe MX Master 4 is the latest evolution in a line of pioneering mice that dates back more than 20 years and has long been the mouse to beat for everything but hardcore PC gaming. Having given it a magnetic free-spinning scroll wheel, plenty of buttons and precise tracking, now Logitech is trying something different for its seven-generation: the ability to tap back at you.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-05T07:00:24Z", "guid": "https://www.theguardian.com/technology/2026/feb/05/google-pixel-buds-2a-review-great-bluetooth-earbuds-at-a-good-price", "link": "https://www.theguardian.com/technology/2026/feb/05/google-pixel-buds-2a-review-great-bluetooth-earbuds-at-a-good-price", "title": "Google Pixel Buds 2a review: great Bluetooth earbuds at a good price", "content": "Compact and comfortable Pixel Buds have noise cancelling, decent battery life and good everyday sound
Google’s latest budget Pixel earbuds are smaller, lighter, more comfortable and have noise cancelling, plus a case that allows you to replace the battery at home.
The Pixel Buds 2a uses the design of the excellent Pixel Buds Pro 2 with a few high-end features at a more palatable £109 (€129/$129/A$239) price, undercutting rivals in the process.
Water resistance: IP54 (splash resistant)
Connectivity: Bluetooth 5.4 (SBC, AAC)
Battery life: 7h with ANC (20h with case)
Earbud dimensions: 23.1 x 16 x 17.8mm
Earbud weight: 4.7g each
Driver size: 11mm
Charging case dimensions: 50 x 57.2 x 24.5mm
Charging case weight: 47.6g
Case charging: USB-C
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2026-02-05T07:00:24Z", "isoDate": "2026-02-05T07:00:24.000Z", "pubDate": "Thu, 05 Feb 2026 07:00:24 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/google" }, "_": "Google" }, { "$": { "domain": "https://www.theguardian.com/technology/pixel" }, "_": "Pixel" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/technology/alphabet" }, "_": "Alphabet" }, { "$": { "domain": "https://www.theguardian.com/technology/headphones" }, "_": "Headphones" }, { "$": { "domain": "https://www.theguardian.com/technology/bluetooth" }, "_": "Bluetooth" }, { "$": { "domain": "https://www.theguardian.com/technology/gadgets" }, "_": "Gadgets" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "Compact and comfortable Pixel Buds have noise cancelling, decent battery life and good everyday sound\nGoogle’s latest budget Pixel earbuds are smaller, lighter, more comfortable and have noise cancelling, plus a case that allows you to replace the battery at home.\nThe Pixel Buds 2a uses the design of the excellent Pixel Buds Pro 2 with a few high-end features at a more palatable £109 (€129/$129/A$239) price, undercutting rivals in the process.\nWater resistance: IP54 (splash resistant)\nConnectivity: Bluetooth 5.4 (SBC, AAC)\nBattery life: 7h with ANC (20h with case)\nEarbud dimensions: 23.1 x 16 x 17.8mm\nEarbud weight: 4.7g each\nDriver size: 11mm\nCharging case dimensions: 50 x 57.2 x 24.5mm\nCharging case weight: 47.6g\nCase charging: USB-C\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2025-11-03T07:00:26Z", "guid": "https://www.theguardian.com/technology/2025/nov/03/oakley-meta-vanguard-review-fantastic-ai-running-glasses-linked-to-garmin", "link": "https://www.theguardian.com/technology/2025/nov/03/oakley-meta-vanguard-review-fantastic-ai-running-glasses-linked-to-garmin", "title": "Oakley Meta Vanguard review: fantastic AI running glasses linked to Garmin", "content": "Camera-equipped sports shades have secure fit, open-ear speakers, mics and advanced Garmin and Strava integration
The Oakley Meta Vanguard are new displayless AI glasses designed for running, cycling and action sports with deep Garmin and Strava integration, which may make them the first smart glasses for sport that actually work.
They are a replacement for running glasses, open-ear headphones and a head-mounted action cam all in one, and are the latest product of Meta’s partnership with the sunglasses conglomerate EssilorLuxottica, the owner of Ray-Ban, Oakley and many other top brands.
Continue reading...", "creator": "Samuel Gibbs Consumer technology editor", "dc:date": "2025-11-03T07:00:26Z", "isoDate": "2025-11-03T07:00:26.000Z", "pubDate": "Mon, 03 Nov 2025 07:00:26 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/technology/meta" }, "_": "Meta" }, { "$": { "domain": "https://www.theguardian.com/technology/wearable-technology" }, "_": "Wearable technology" }, { "$": { "domain": "https://www.theguardian.com/technology/gadgets" }, "_": "Gadgets" }, { "$": { "domain": "https://www.theguardian.com/technology/garmin" }, "_": "Garmin" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/lifeandstyle/running" }, "_": "Running" }, { "$": { "domain": "https://www.theguardian.com/lifeandstyle/lifeandstyle" }, "_": "Life and style" } ], "dc:creator": "Samuel Gibbs Consumer technology editor", "contentSnippet": "Camera-equipped sports shades have secure fit, open-ear speakers, mics and advanced Garmin and Strava integration\nThe Oakley Meta Vanguard are new displayless AI glasses designed for running, cycling and action sports with deep Garmin and Strava integration, which may make them the first smart glasses for sport that actually work.\nThey are a replacement for running glasses, open-ear headphones and a head-mounted action cam all in one, and are the latest product of Meta’s partnership with the sunglasses conglomerate EssilorLuxottica, the owner of Ray-Ban, Oakley and many other top brands.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-09T14:29:36Z", "guid": "https://www.theguardian.com/games/2026/feb/09/how-a-decades-old-video-game-has-helped-me-defeat-the-doomscroll", "link": "https://www.theguardian.com/games/2026/feb/09/how-a-decades-old-video-game-has-helped-me-defeat-the-doomscroll", "title": "How a decades-old video game has helped me defeat the doomscroll", "content": "Trading social media for Pokémon battles and evolutions in Kanto on a Game Boy Advance has been surprisingly serene
Cutting back on doomscrolling must be one of the hardest new year resolutions to keep. Instinctively tapping on the usual suspects on your phone’s home screen becomes a reflex, and vast quantities of money and user data have been specifically employed to keep you reaching for the phone, ingraining it into our work, leisure and social lives. You’ll get no shame from me if you love your phone and have a healthy relationship with your apps, but I’ve found myself struggling lately.
This year, I’m attempting to cut back on screen time – sort of. I’m replacing the sleek oblong of my smartphone with something a little more fuzzy and nostalgic. In an attempt to dismantle my bad habit, I’m closing the feeds of instant updates and instead carrying around a Game Boy Advance. I’ve been playing Pokémon FireRed, a remake of the very first Pokémon games, which turn 30 this month. Even this refreshed version is more than two decades old.
Continue reading...", "creator": "Michael Roberts", "dc:date": "2026-02-09T14:29:36Z", "isoDate": "2026-02-09T14:29:36.000Z", "pubDate": "Mon, 09 Feb 2026 14:29:36 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/games/games" }, "_": "Games" }, { "$": { "domain": "https://www.theguardian.com/culture/culture" }, "_": "Culture" }, { "$": { "domain": "https://www.theguardian.com/games/pokemon" }, "_": "Pokémon" }, { "$": { "domain": "https://www.theguardian.com/games/nintendo" }, "_": "Nintendo" } ], "dc:creator": "Michael Roberts", "contentSnippet": "Trading social media for Pokémon battles and evolutions in Kanto on a Game Boy Advance has been surprisingly serene\nCutting back on doomscrolling must be one of the hardest new year resolutions to keep. Instinctively tapping on the usual suspects on your phone’s home screen becomes a reflex, and vast quantities of money and user data have been specifically employed to keep you reaching for the phone, ingraining it into our work, leisure and social lives. You’ll get no shame from me if you love your phone and have a healthy relationship with your apps, but I’ve found myself struggling lately.\nThis year, I’m attempting to cut back on screen time – sort of. I’m replacing the sleek oblong of my smartphone with something a little more fuzzy and nostalgic. In an attempt to dismantle my bad habit, I’m closing the feeds of instant updates and instead carrying around a Game Boy Advance. I’ve been playing Pokémon FireRed, a remake of the very first Pokémon games, which turn 30 this month. Even this refreshed version is more than two decades old.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-09T11:30:02Z", "guid": "https://www.theguardian.com/australia-news/2026/feb/09/australian-government-roblox-rating-child-safety", "link": "https://www.theguardian.com/australia-news/2026/feb/09/australian-government-roblox-rating-child-safety", "title": "‘Disturbing’: Australian government demands review of Roblox’s PG rating after reports of child grooming", "content": "Communications minister Anika Wells points to media reports alleging children can access spaces meant for adults which include explicit sexual content
Get our breaking news email, free app or daily news podcast
Reports of child grooming and vile content on popular game service Roblox has “alarmed” the communications minister, Anika Wells, who has demanded the platform explain how it is addressing sexual and self-harm material, and that its PG rating be examined by the Australian Classification Board.
The eSafety commissioner has also written to the game platform, saying it plans to test the promises it made about keeping children safe online, including disabling chat features and making underage accounts private.
Continue reading...", "creator": "Josh Butler", "dc:date": "2026-02-09T11:30:02Z", "isoDate": "2026-02-09T11:30:02.000Z", "pubDate": "Mon, 09 Feb 2026 11:30:02 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/games/roblox" }, "_": "Roblox" }, { "$": { "domain": "https://www.theguardian.com/australia-news/australian-politics" }, "_": "Australian politics" }, { "$": { "domain": "https://www.theguardian.com/australia-news/social-media-ban" }, "_": "Social media ban" }, { "$": { "domain": "https://www.theguardian.com/games/games" }, "_": "Games" }, { "$": { "domain": "https://www.theguardian.com/technology/technology" }, "_": "Technology" }, { "$": { "domain": "https://www.theguardian.com/australia-news/australia-news" }, "_": "Australia news" } ], "dc:creator": "Josh Butler", "contentSnippet": "Communications minister Anika Wells points to media reports alleging children can access spaces meant for adults which include explicit sexual content\n\nFollow our Australia news live blog for latest updates\n\nGet our breaking news email, free app or daily news podcast\n\nReports of child grooming and vile content on popular game service Roblox has “alarmed” the communications minister, Anika Wells, who has demanded the platform explain how it is addressing sexual and self-harm material, and that its PG rating be examined by the Australian Classification Board.\nThe eSafety commissioner has also written to the game platform, saying it plans to test the promises it made about keeping children safe online, including disabling chat features and making underage accounts private.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-06T14:30:08Z", "guid": "https://www.theguardian.com/games/2026/feb/06/how-pokemon-conquered-the-world-keza-macdonald-super-nintendo-book-extract", "link": "https://www.theguardian.com/games/2026/feb/06/how-pokemon-conquered-the-world-keza-macdonald-super-nintendo-book-extract", "title": "‘Christian pastors declared Pikachu to be a demon’: how Pokémon went from moral panic to unifying global hit", "content": "Nintendo’s monster-collecting franchise was pilloried as a ‘pestilential Ponzi scheme’ in the 90s. But as its celebrates its 30th birthday, it now stands as a powerful example of video games’ ability to connect people
When I was 11, it was my dream to compete in the Pokémon World Championships, held in Sydney in 2000. I’d come across it in a magazine, and then earnestly set about training teams of creatures, transferring them between my Pokémon Red Game Boy cartridge and the 3D arenas of Pokémon Stadium on the Nintendo 64. I never made it as a player but I did finally achieve this dream on my 26th birthday, when I went to Washington DC to cover the world championships as a journalist. I was deeply moved. Presided over by a giant inflatable Pikachu hanging from the ceiling, the competitors and spectators were united in an unselfconscious love for these games, with their colourful menageries and heartfelt messaging about trust, friendship and hard work.
It is emotional to see the winners lift their trophies after a tense final round of battles, as overwhelmed by their success as any sportsperson. But it’s the pride that the smaller competitors’ parents show in their mini champions that really gets to me. During the first wave of Pokémania in the late 90s, Pokémon was viewed with suspicion by most adults. Now that the first generation of Pokémaniacs have grown up, even becoming parents ourselves, we see it for what it is: an imaginative, challenging and really rather wholesome series of games that rewards every hour that children devote to it.
Continue reading...", "creator": "Keza MacDonald", "dc:date": "2026-02-06T14:30:08Z", "isoDate": "2026-02-06T14:30:08.000Z", "pubDate": "Fri, 06 Feb 2026 14:30:08 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/games/pokemon" }, "_": "Pokémon" }, { "$": { "domain": "https://www.theguardian.com/games/games" }, "_": "Games" }, { "$": { "domain": "https://www.theguardian.com/culture/culture" }, "_": "Culture" } ], "dc:creator": "Keza MacDonald", "contentSnippet": "Nintendo’s monster-collecting franchise was pilloried as a ‘pestilential Ponzi scheme’ in the 90s. But as its celebrates its 30th birthday, it now stands as a powerful example of video games’ ability to connect people\nWhen I was 11, it was my dream to compete in the Pokémon World Championships, held in Sydney in 2000. I’d come across it in a magazine, and then earnestly set about training teams of creatures, transferring them between my Pokémon Red Game Boy cartridge and the 3D arenas of Pokémon Stadium on the Nintendo 64. I never made it as a player but I did finally achieve this dream on my 26th birthday, when I went to Washington DC to cover the world championships as a journalist. I was deeply moved. Presided over by a giant inflatable Pikachu hanging from the ceiling, the competitors and spectators were united in an unselfconscious love for these games, with their colourful menageries and heartfelt messaging about trust, friendship and hard work.\nIt is emotional to see the winners lift their trophies after a tense final round of battles, as overwhelmed by their success as any sportsperson. But it’s the pride that the smaller competitors’ parents show in their mini champions that really gets to me. During the first wave of Pokémania in the late 90s, Pokémon was viewed with suspicion by most adults. Now that the first generation of Pokémaniacs have grown up, even becoming parents ourselves, we see it for what it is: an imaginative, challenging and really rather wholesome series of games that rewards every hour that children devote to it.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-06T14:00:08Z", "guid": "https://www.theguardian.com/games/2026/feb/06/mewgenics-review-infinite-ways-to-skin-a-cat", "link": "https://www.theguardian.com/games/2026/feb/06/mewgenics-review-infinite-ways-to-skin-a-cat", "title": "Mewgenics review – infinite ways to skin a cat", "content": "PC; Edmund McMillen and Tyler Glaiel
This mischievous roguelike escapade featuring utterly fiendish felines is compelling, and impressively tasteless
You know that old saying about cats having nine lives? Well, as far as Mewgenics is concerned, you can forget it – and you can also forget the idea that a game about cats has to be in any way cute. These kitties are red in tooth and claw, prone to strange mutations, and strictly limited to just the one life, which often ends swiftly and brutally.
Such is the nature of roguelike, a format that has spawned some of the biggest indie hits of the past 20 years. In these games, failure is permanent; dying sends you back not to the last checkpoint but back to the beginning, the game reshuffling its elements into a new shape for your next run. And so it goes in Mewgenics. You gather a party of four felines and send them out on a questing journey, from which they return victorious or not at all.
Continue reading...", "creator": "Alex Spencer", "dc:date": "2026-02-06T14:00:08Z", "isoDate": "2026-02-06T14:00:08.000Z", "pubDate": "Fri, 06 Feb 2026 14:00:08 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/games/games" }, "_": "Games" }, { "$": { "domain": "https://www.theguardian.com/culture/culture" }, "_": "Culture" } ], "dc:creator": "Alex Spencer", "contentSnippet": "PC; Edmund McMillen and Tyler Glaiel \nThis mischievous roguelike escapade featuring utterly fiendish felines is compelling, and impressively tasteless\nYou know that old saying about cats having nine lives? Well, as far as Mewgenics is concerned, you can forget it – and you can also forget the idea that a game about cats has to be in any way cute. These kitties are red in tooth and claw, prone to strange mutations, and strictly limited to just the one life, which often ends swiftly and brutally.\nSuch is the nature of roguelike, a format that has spawned some of the biggest indie hits of the past 20 years. In these games, failure is permanent; dying sends you back not to the last checkpoint but back to the beginning, the game reshuffling its elements into a new shape for your next run. And so it goes in Mewgenics. You gather a party of four felines and send them out on a questing journey, from which they return victorious or not at all.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "date": "2026-02-04T12:35:42Z", "guid": "https://www.theguardian.com/games/2026/feb/04/gamings-new-coming-of-age-genre-embraces-millennial-cringe", "link": "https://www.theguardian.com/games/2026/feb/04/gamings-new-coming-of-age-genre-embraces-millennial-cringe", "title": "Gaming’s new coming-of-age genre embraces ‘millennial cringe’", "content": "Perfect Tides perfectly captures the older millennial college experience, and a time when nobody worried about being embarrassing online
• Don’t get Pushing Buttons delivered to your inbox? Sign up here
I’ve noticed an interesting micro-trend emerging in the last few years: millennial nostalgia games. Not just ones that adopt the aesthetic of Y2K gaming – think Crow Country or Fear the Spotlight’s deliberately retro PS1-style fuzzy polygons – but semi-autobiographical games specifically about the millennial experience. I’ve played three in the past year. Despelote is set in 2002 in Ecuador and is played through the eyes of a football-obsessed eight-year-old. The award-winning Consume Me is about being a teen girl battling disordered eating in the 00s. And this week I played a point-and-click adventure game about being a college student in the early 2000s.
Perfect Tides: Station to Station is set in New York in 2003 – a year that is the epitome of nostalgia for the micro-generation that grew up without the internet but came of age online. It was before Facebook, before the smartphone, but firmly during the era of late-night forum browsing and instant-messenger conversations. The internet wasn’t yet a vector for mass communication, but it could still bring you together with other people who loved the things that you loved, people who read the same hipster blogs and liked the same bands. The protagonist, Mara, is a student and young writer who works in her college library.
Continue reading...", "creator": "Keza MacDonald", "dc:date": "2026-02-04T12:35:42Z", "isoDate": "2026-02-04T12:35:42.000Z", "pubDate": "Wed, 04 Feb 2026 12:35:42 GMT", "categories": [ { "$": { "domain": "https://www.theguardian.com/games/games" }, "_": "Games" }, { "$": { "domain": "https://www.theguardian.com/culture/culture" }, "_": "Culture" } ], "dc:creator": "Keza MacDonald", "contentSnippet": "Perfect Tides perfectly captures the older millennial college experience, and a time when nobody worried about being embarrassing online\n• Don’t get Pushing Buttons delivered to your inbox? Sign up here\nI’ve noticed an interesting micro-trend emerging in the last few years: millennial nostalgia games. Not just ones that adopt the aesthetic of Y2K gaming – think Crow Country or Fear the Spotlight’s deliberately retro PS1-style fuzzy polygons – but semi-autobiographical games specifically about the millennial experience. I’ve played three in the past year. Despelote is set in 2002 in Ecuador and is played through the eyes of a football-obsessed eight-year-old. The award-winning Consume Me is about being a teen girl battling disordered eating in the 00s. And this week I played a point-and-click adventure game about being a college student in the early 2000s.\nPerfect Tides: Station to Station is set in New York in 2003 – a year that is the epitome of nostalgia for the micro-generation that grew up without the internet but came of age online. It was before Facebook, before the smartphone, but firmly during the era of late-night forum browsing and instant-messenger conversations. The internet wasn’t yet a vector for mass communication, but it could still bring you together with other people who loved the things that you loved, people who read the same hipster blogs and liked the same bands. The protagonist, Mara, is a student and young writer who works in her college library.\n Continue reading..." }, "pairedItem": [ { "item": 8 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c3wlpqpe2z4o#0", "link": "https://www.bbc.com/news/articles/c3wlpqpe2z4o?at_medium=RSS&at_campaign=rss", "title": "Instagram and YouTube owners built 'addiction machines', trial hears", "content": "The tech giants are under scrutiny over social media addiction in a landmark jury trial in Los Angeles", "isoDate": "2026-02-10T07:25:42.000Z", "pubDate": "Tue, 10 Feb 2026 07:25:42 GMT", "contentSnippet": "The tech giants are under scrutiny over social media addiction in a landmark jury trial in Los Angeles" }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c1d67vdlk1ko#0", "link": "https://www.bbc.com/news/articles/c1d67vdlk1ko?at_medium=RSS&at_campaign=rss", "title": "Discord to start requiring face scan or ID to access adult content", "content": "The online chat service, which has 200 million monthly users, will blur adult content by default.", "isoDate": "2026-02-09T17:42:21.000Z", "pubDate": "Mon, 09 Feb 2026 17:42:21 GMT", "contentSnippet": "The online chat service, which has 200 million monthly users, will blur adult content by default." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c3093gjy2ero#0", "link": "https://www.bbc.com/news/articles/c3093gjy2ero?at_medium=RSS&at_campaign=rss", "title": "AI chatbots pose 'dangerous' risk when giving medical advice, study suggests", "content": "It found people using AI for health reasons found it hard to identify what advice they should trust.", "isoDate": "2026-02-09T16:33:29.000Z", "pubDate": "Mon, 09 Feb 2026 16:33:29 GMT", "contentSnippet": "It found people using AI for health reasons found it hard to identify what advice they should trust." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cqxdj77welpo#0", "link": "https://www.bbc.com/news/articles/cqxdj77welpo?at_medium=RSS&at_campaign=rss", "title": "EU tells Meta to let rivals run AI chatbots on WhatsApp", "content": "A Meta spokesperson said the EU had \"no reason\" to intervene over it changing the app in January.", "isoDate": "2026-02-09T12:14:35.000Z", "pubDate": "Mon, 09 Feb 2026 12:14:35 GMT", "contentSnippet": "A Meta spokesperson said the EU had \"no reason\" to intervene over it changing the app in January." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c24g457y534o#0", "link": "https://www.bbc.com/news/articles/c24g457y534o?at_medium=RSS&at_campaign=rss", "title": "Baldur's Gate to be turned into TV series - without the game's developers", "content": "The show will be made by Craig Mazin, who co-created the acclaimed game adaptation The Last Of Us.", "isoDate": "2026-02-06T16:26:08.000Z", "pubDate": "Fri, 06 Feb 2026 16:26:08 GMT", "contentSnippet": "The show will be made by Craig Mazin, who co-created the acclaimed game adaptation The Last Of Us." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cq5y5w148p5o#0", "link": "https://www.bbc.com/news/articles/cq5y5w148p5o?at_medium=RSS&at_campaign=rss", "title": "Uber ordered to pay $8.5m over claim driver raped passenger", "content": "The verdict is expected to influence the outcome of thousands of other cases against the ride hailing firm.", "isoDate": "2026-02-06T20:47:10.000Z", "pubDate": "Fri, 06 Feb 2026 20:47:10 GMT", "contentSnippet": "The verdict is expected to influence the outcome of thousands of other cases against the ride hailing firm." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cvgjg98vmzjo#0", "link": "https://www.bbc.com/news/articles/cvgjg98vmzjo?at_medium=RSS&at_campaign=rss", "title": "Google staff call for firm to cut ties with ICE", "content": "More than 900 Google employees signed a letter opposing company links to federal immigration actions. ", "isoDate": "2026-02-06T21:40:05.000Z", "pubDate": "Fri, 06 Feb 2026 21:40:05 GMT", "contentSnippet": "More than 900 Google employees signed a letter opposing company links to federal immigration actions." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cr7j7n315lmo#1", "link": "https://www.bbc.com/news/articles/cr7j7n315lmo?at_medium=RSS&at_campaign=rss", "title": "TikTok told to change 'addictive design' by EU or face massive fines", "content": "TikTok says it plans to challenge the \"categorically false and entirely meritless\" accusations.", "isoDate": "2026-02-06T13:55:46.000Z", "pubDate": "Fri, 06 Feb 2026 13:55:46 GMT", "contentSnippet": "TikTok says it plans to challenge the \"categorically false and entirely meritless\" accusations." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cy09x24q0pro#1", "link": "https://www.bbc.com/news/articles/cy09x24q0pro?at_medium=RSS&at_campaign=rss", "title": "Bitcoin falls to lowest level since Trump took office", "content": "The price of the digital currency has dropped significantly despite President Donald Trump’s public support.", "isoDate": "2026-02-06T11:15:35.000Z", "pubDate": "Fri, 06 Feb 2026 11:15:35 GMT", "contentSnippet": "The price of the digital currency has dropped significantly despite President Donald Trump’s public support." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/ce3edyx74jko#1", "link": "https://www.bbc.com/news/articles/ce3edyx74jko?at_medium=RSS&at_campaign=rss", "title": "ChatGPT boss ridiculed for online 'tantrum' over rival's Super Bowl ad", "content": "Commenters said Altman's lengthy post shows \"a nerve was well and truly hit\" by Anthropic's advert.", "isoDate": "2026-02-05T12:36:32.000Z", "pubDate": "Thu, 05 Feb 2026 12:36:32 GMT", "contentSnippet": "Commenters said Altman's lengthy post shows \"a nerve was well and truly hit\" by Anthropic's advert." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c8e50x1r237o#1", "link": "https://www.bbc.com/news/articles/c8e50x1r237o?at_medium=RSS&at_campaign=rss", "title": "UK's £8bn research fund faces 'hard decisions' as it pauses new grants", "content": "UKRI boss Ian Chapman said it had been told by the government to \"focus and do fewer things better\".\n", "isoDate": "2026-02-05T12:00:58.000Z", "pubDate": "Thu, 05 Feb 2026 12:00:58 GMT", "contentSnippet": "UKRI boss Ian Chapman said it had been told by the government to \"focus and do fewer things better\"." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c5y41r5rzrno#1", "link": "https://www.bbc.com/news/articles/c5y41r5rzrno?at_medium=RSS&at_campaign=rss", "title": "US launches plan to tackle China's critical minerals dominance", "content": "The event was attended by representatives of more than 50 countries, the White House said.", "isoDate": "2026-02-05T03:33:48.000Z", "pubDate": "Thu, 05 Feb 2026 03:33:48 GMT", "contentSnippet": "The event was attended by representatives of more than 50 countries, the White House said." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c5ydndkmvy2o#2", "link": "https://www.bbc.com/news/articles/c5ydndkmvy2o?at_medium=RSS&at_campaign=rss", "title": "Netflix and Warner Bros struggle to defend merger", "content": "Concerns were raised by a subcommittee including potential price rises and the future of cinemas.", "isoDate": "2026-02-03T23:42:06.000Z", "pubDate": "Tue, 03 Feb 2026 23:42:06 GMT", "contentSnippet": "Concerns were raised by a subcommittee including potential price rises and the future of cinemas." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cn0k670n0ydo#2", "link": "https://www.bbc.com/news/articles/cn0k670n0ydo?at_medium=RSS&at_campaign=rss", "title": "Pinterest sacks engineers for tracking staff job cuts", "content": "The social media platform recently announced that it was axing around 15% of its workforce.", "isoDate": "2026-02-04T06:03:12.000Z", "pubDate": "Wed, 04 Feb 2026 06:03:12 GMT", "contentSnippet": "The social media platform recently announced that it was axing around 15% of its workforce." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/ce3ex92557jo#2", "link": "https://www.bbc.com/news/articles/ce3ex92557jo?at_medium=RSS&at_campaign=rss", "title": "X offices raided in France as UK opens fresh investigation into Grok", "content": "Elon Musk's X and Grok platforms are facing increased scrutiny from authorities on both sides of the channel.", "isoDate": "2026-02-04T01:05:26.000Z", "pubDate": "Wed, 04 Feb 2026 01:05:26 GMT", "contentSnippet": "Elon Musk's X and Grok platforms are facing increased scrutiny from authorities on both sides of the channel." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c62n410w5yno#2", "link": "https://www.bbc.com/news/articles/c62n410w5yno?at_medium=RSS&at_campaign=rss", "title": "What is the 'social media network for AI' Moltbook?", "content": "The Reddit-like website which launched in late January allows AI bots to speak to each other.", "isoDate": "2026-02-02T13:59:14.000Z", "pubDate": "Mon, 02 Feb 2026 13:59:14 GMT", "contentSnippet": "The Reddit-like website which launched in late January allows AI bots to speak to each other." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cp37g5nxe3lo#2", "link": "https://www.bbc.com/news/articles/cp37g5nxe3lo?at_medium=RSS&at_campaign=rss", "title": "China bans hidden car door handles over safety concerns", "content": "It makes China the first country to stop the use of designs first made popular by Elon Musk's Tesla.", "isoDate": "2026-02-03T02:06:29.000Z", "pubDate": "Tue, 03 Feb 2026 02:06:29 GMT", "contentSnippet": "It makes China the first country to stop the use of designs first made popular by Elon Musk's Tesla." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cq6vnrye06po#2", "link": "https://www.bbc.com/news/articles/cq6vnrye06po?at_medium=RSS&at_campaign=rss", "title": "Musk's SpaceX and xAI merge to make world's most valuable private company", "content": "Musk says the combined firm - which has been valued at more than $1tn - will be an \"innovation engine\".", "isoDate": "2026-02-03T09:31:06.000Z", "pubDate": "Tue, 03 Feb 2026 09:31:06 GMT", "contentSnippet": "Musk says the combined firm - which has been valued at more than $1tn - will be an \"innovation engine\"." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c9wx2dz2v44o#2", "link": "https://www.bbc.com/news/articles/c9wx2dz2v44o?at_medium=RSS&at_campaign=rss", "title": "AI 'slop' is transforming social media - and a backlash is brewing", "content": "Social media has been flooded with fake, AI-generated images and videos. But will the majority of users actually care?", "isoDate": "2026-02-04T11:29:30.000Z", "pubDate": "Wed, 04 Feb 2026 11:29:30 GMT", "contentSnippet": "Social media has been flooded with fake, AI-generated images and videos. But will the majority of users actually care?" }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cvg5er4ewg6o#2", "link": "https://www.bbc.com/news/articles/cvg5er4ewg6o?at_medium=RSS&at_campaign=rss", "title": "Pornhub is now restricting access for UK users - will other sites follow suit?", "content": "The UK's largest porn site has blocked unregistered users from accessing explicit content in the country, but why motives remain unclear.", "isoDate": "2026-02-02T09:49:12.000Z", "pubDate": "Mon, 02 Feb 2026 09:49:12 GMT", "contentSnippet": "The UK's largest porn site has blocked unregistered users from accessing explicit content in the country, but why motives remain unclear." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cyv5l24mrjmo#2", "link": "https://www.bbc.com/news/articles/cyv5l24mrjmo?at_medium=RSS&at_campaign=rss", "title": "Musk's SpaceX applies to launch a million satellites into orbit", "content": "The firm wants to create a network of \"orbital data centres\" to power artificial intelligence.", "isoDate": "2026-01-31T16:19:17.000Z", "pubDate": "Sat, 31 Jan 2026 16:19:17 GMT", "contentSnippet": "The firm wants to create a network of \"orbital data centres\" to power artificial intelligence." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cy8yw190q83o#2", "link": "https://www.bbc.com/news/articles/cy8yw190q83o?at_medium=RSS&at_campaign=rss", "title": "Apple reports best-ever iPhone sales as Mac dips", "content": "The company's revenue was boosted by iPhone sales, but sales of its wearable tech and Mac computers dipped. ", "isoDate": "2026-01-30T06:46:13.000Z", "pubDate": "Fri, 30 Jan 2026 06:46:13 GMT", "contentSnippet": "The company's revenue was boosted by iPhone sales, but sales of its wearable tech and Mac computers dipped." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c62njv82n0wo#2", "link": "https://www.bbc.com/news/articles/c62njv82n0wo?at_medium=RSS&at_campaign=rss", "title": "He calls me sweetheart and winks at me - but he's not my boyfriend, he's AI", "content": "George is an avatar on my mobile but claims to know what makes me tick. ", "isoDate": "2026-01-30T11:39:15.000Z", "pubDate": "Fri, 30 Jan 2026 11:39:15 GMT", "contentSnippet": "George is an avatar on my mobile but claims to know what makes me tick." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cn8jkyk78gno#2", "link": "https://www.bbc.com/news/articles/cn8jkyk78gno?at_medium=RSS&at_campaign=rss", "title": "Facebook-owner Meta to nearly double AI spending", "content": "Mark Zuckerberg's spending plans hint at further layoffs and changes within Facebook, Instagram and WhatsApp.", "isoDate": "2026-01-29T00:16:17.000Z", "pubDate": "Thu, 29 Jan 2026 00:16:17 GMT", "contentSnippet": "Mark Zuckerberg's spending plans hint at further layoffs and changes within Facebook, Instagram and WhatsApp." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c620177qdg5o#2", "link": "https://www.bbc.com/news/articles/c620177qdg5o?at_medium=RSS&at_campaign=rss", "title": "Tesla cuts car models in shift to robots and AI", "content": "Multi-billionaire Elon Musk's electric car maker also said its annual revenue had fallen for the first time.", "isoDate": "2026-01-29T02:35:07.000Z", "pubDate": "Thu, 29 Jan 2026 02:35:07 GMT", "contentSnippet": "Multi-billionaire Elon Musk's electric car maker also said its annual revenue had fallen for the first time." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cx2yv31482vo#2", "link": "https://www.bbc.com/news/articles/cx2yv31482vo?at_medium=RSS&at_campaign=rss", "title": "UK bans Coinbase ads implying crypto can ease cost of living concerns", "content": "The ASA upheld complaints that Coinbase's adverts trivialised the risks of investing in cryptocurrency.", "isoDate": "2026-01-29T11:40:44.000Z", "pubDate": "Thu, 29 Jan 2026 11:40:44 GMT", "contentSnippet": "The ASA upheld complaints that Coinbase's adverts trivialised the risks of investing in cryptocurrency." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/czej9n578k9o#2", "link": "https://www.bbc.com/news/articles/czej9n578k9o?at_medium=RSS&at_campaign=rss", "title": "Driverless taxis set to launch in UK as soon as September", "content": "Waymo has laid out plans for a robotaxi service in London with a pilot scheme due to begin in April.", "isoDate": "2026-01-28T23:22:48.000Z", "pubDate": "Wed, 28 Jan 2026 23:22:48 GMT", "contentSnippet": "Waymo has laid out plans for a robotaxi service in London with a pilot scheme due to begin in April." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cx2ywzxlxnlo#2", "link": "https://www.bbc.com/news/articles/cx2ywzxlxnlo?at_medium=RSS&at_campaign=rss", "title": "Amazon confirms 16,000 job cuts after accidental email", "content": "The technology giant confirmed the redundancies hours after it told staff in an email sent in error.", "isoDate": "2026-01-28T13:11:09.000Z", "pubDate": "Wed, 28 Jan 2026 13:11:09 GMT", "contentSnippet": "The technology giant confirmed the redundancies hours after it told staff in an email sent in error." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cp37prvp072o#2", "link": "https://www.bbc.com/news/articles/cp37prvp072o?at_medium=RSS&at_campaign=rss", "title": "Government offers UK adults free AI training for work", "content": "The online lessons give advice on things such as how to prompt chatbots or complete admin tasks.", "isoDate": "2026-01-28T11:50:21.000Z", "pubDate": "Wed, 28 Jan 2026 11:50:21 GMT", "contentSnippet": "The online lessons give advice on things such as how to prompt chatbots or complete admin tasks." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/czr428rxg57o#2", "link": "https://www.bbc.com/news/articles/czr428rxg57o?at_medium=RSS&at_campaign=rss", "title": "Pornhub to restrict access for UK users from next week", "content": "The changes mean only those who have a Pornhub account and have verified their age will be able to access it in the UK soon.", "isoDate": "2026-01-27T18:38:43.000Z", "pubDate": "Tue, 27 Jan 2026 18:38:43 GMT", "contentSnippet": "The changes mean only those who have a Pornhub account and have verified their age will be able to access it in the UK soon." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/ckgjedpn8p8o#2", "link": "https://www.bbc.com/news/articles/ckgjedpn8p8o?at_medium=RSS&at_campaign=rss", "title": "TikTok US pushes back on claims it is censoring content", "content": "Thousands of people claim political content is being suppressed after the US operation was spun off.", "isoDate": "2026-01-27T20:25:11.000Z", "pubDate": "Tue, 27 Jan 2026 20:25:11 GMT", "contentSnippet": "Thousands of people claim political content is being suppressed after the US operation was spun off." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cr57p2ve8glo#2", "link": "https://www.bbc.com/news/articles/cr57p2ve8glo?at_medium=RSS&at_campaign=rss", "title": "AI boom will produce victors and carnage, tech boss warns", "content": "Cisco chief executive Chuck Robbins compares AI to the dotcom bubble of the early 2000s.", "isoDate": "2026-01-28T00:01:19.000Z", "pubDate": "Wed, 28 Jan 2026 00:01:19 GMT", "contentSnippet": "Cisco chief executive Chuck Robbins compares AI to the dotcom bubble of the early 2000s." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/videos/cx2yep7l2j2o#4", "link": "https://www.bbc.com/news/videos/cx2yep7l2j2o?at_medium=RSS&at_campaign=rss", "title": "How would a social media ban for under-16s work?", "content": "BBC technology editor Zoe Kleinman explains.", "isoDate": "2026-01-20T13:08:07.000Z", "pubDate": "Tue, 20 Jan 2026 13:08:07 GMT", "contentSnippet": "BBC technology editor Zoe Kleinman explains." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.co.uk/iplayer/episode/m002qbkc#4", "link": "https://www.bbc.co.uk/iplayer/episode/m002qbkc?at_medium=RSS&at_campaign=rss", "title": "Tech Now", "content": "From Las Vegas, the latest trends and innovations at CES 2026.", "isoDate": "2026-01-17T01:00:00.000Z", "pubDate": "Sat, 17 Jan 2026 01:00:00 GMT", "contentSnippet": "From Las Vegas, the latest trends and innovations at CES 2026." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/videos/cql4076kyzeo#4", "link": "https://www.bbc.com/news/videos/cql4076kyzeo?at_medium=RSS&at_campaign=rss", "title": "Wikipedia's Jimmy Wales on where the name comes from", "content": "The site's co-founder speaks to the BBC for the online encyclopedia's 25th anniversary.", "isoDate": "2026-01-15T06:14:24.000Z", "pubDate": "Thu, 15 Jan 2026 06:14:24 GMT", "contentSnippet": "The site's co-founder speaks to the BBC for the online encyclopedia's 25th anniversary." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpz#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpz?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "Meet the humanoid robots designed to help with household chores", "isoDate": "2026-01-13T20:30:00.000Z", "pubDate": "Tue, 13 Jan 2026 20:30:00 GMT", "contentSnippet": "Meet the humanoid robots designed to help with household chores" }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/videos/c8x94zr8yxvo#4", "link": "https://www.bbc.com/news/videos/c8x94zr8yxvo?at_medium=RSS&at_campaign=rss", "title": "Watch: Backlash against Musk's Grok AI explained", "content": "Technology editor Zoe Kleinman explains the row over changes made by X to it's Grok AI image edits, after the UK government called it \"insulting\".", "isoDate": "2026-01-09T17:19:37.000Z", "pubDate": "Fri, 09 Jan 2026 17:19:37 GMT", "contentSnippet": "Technology editor Zoe Kleinman explains the row over changes made by X to it's Grok AI image edits, after the UK government called it \"insulting\"." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/newsround/videos/cx2y0xe8ex2o#4", "link": "https://www.bbc.com/newsround/videos/cx2y0xe8ex2o?at_medium=RSS&at_campaign=rss", "title": "Cool future tech at CES!", "content": "The technology show CES is back for another year in Las Vegas in America.", "isoDate": "2026-01-08T17:18:47.000Z", "pubDate": "Thu, 08 Jan 2026 17:18:47 GMT", "contentSnippet": "The technology show CES is back for another year in Las Vegas in America." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpy#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpy?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "The latest gadgets, the future in assistive tech and upcoming gaming releases in 2026.", "isoDate": "2026-01-06T20:32:00.000Z", "pubDate": "Tue, 06 Jan 2026 20:32:00 GMT", "contentSnippet": "The latest gadgets, the future in assistive tech and upcoming gaming releases in 2026." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/videos/c98p1jg3p58o#4", "link": "https://www.bbc.com/news/videos/c98p1jg3p58o?at_medium=RSS&at_campaign=rss", "title": "Watch: BBC reporter tests AI anti-shoplifting tech", "content": "Some major retailers and independent stores have introduced AI body scans, CCTV or facial recognition equipment to identify crimes like shoplifting.", "isoDate": "2026-01-05T00:01:31.000Z", "pubDate": "Mon, 05 Jan 2026 00:01:31 GMT", "contentSnippet": "Some major retailers and independent stores have introduced AI body scans, CCTV or facial recognition equipment to identify crimes like shoplifting." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpx#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpx?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "We bring you Tech Life highlights from a fascinating year in global tech.", "isoDate": "2025-12-30T20:30:00.000Z", "pubDate": "Tue, 30 Dec 2025 20:30:00 GMT", "contentSnippet": "We bring you Tech Life highlights from a fascinating year in global tech." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.co.uk/sounds/play/w3ct6zpv#4", "link": "https://www.bbc.co.uk/sounds/play/w3ct6zpv?at_medium=RSS&at_campaign=rss", "title": "Tech Life", "content": "A study found AI chatbots can persuade us with fake facts. How does this affect politics?", "isoDate": "2025-12-16T20:30:00.000Z", "pubDate": "Tue, 16 Dec 2025 20:30:00 GMT", "contentSnippet": "A study found AI chatbots can persuade us with fake facts. How does this affect politics?" }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c2e102vw1z2o#5", "link": "https://www.bbc.com/news/articles/c2e102vw1z2o?at_medium=RSS&at_campaign=rss", "title": "Why food fraud persists, even with improving tech", "content": "Even with sophisticated technology it is still difficult to detect fake foods.", "isoDate": "2026-02-10T00:03:12.000Z", "pubDate": "Tue, 10 Feb 2026 00:03:12 GMT", "contentSnippet": "Even with sophisticated technology it is still difficult to detect fake foods." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c5y46356zzyo#5", "link": "https://www.bbc.com/news/articles/c5y46356zzyo?at_medium=RSS&at_campaign=rss", "title": "Can robots ever be graceful?", "content": "Firms are working to make the motors that drive robots more efficient and cheaper.", "isoDate": "2026-02-06T00:05:05.000Z", "pubDate": "Fri, 06 Feb 2026 00:05:05 GMT", "contentSnippet": "Firms are working to make the motors that drive robots more efficient and cheaper." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/clygdez8d41o#5", "link": "https://www.bbc.com/news/articles/clygdez8d41o?at_medium=RSS&at_campaign=rss", "title": "The yachting industry searches for alternatives to teak", "content": "Prized for its beauty, teak is in short supply, forcing the yacht industry to look for alternatives.", "isoDate": "2026-02-04T13:42:07.000Z", "pubDate": "Wed, 04 Feb 2026 13:42:07 GMT", "contentSnippet": "Prized for its beauty, teak is in short supply, forcing the yacht industry to look for alternatives." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cq5y7dd284do#5", "link": "https://www.bbc.com/news/articles/cq5y7dd284do?at_medium=RSS&at_campaign=rss", "title": "Visit the North Sea oil field used to store greenhouse gas", "content": "Hundreds of miles from Denmark's coast a project is underway to inject CO2 into an old oil field.", "isoDate": "2026-01-30T00:10:01.000Z", "pubDate": "Fri, 30 Jan 2026 00:10:01 GMT", "contentSnippet": "Hundreds of miles from Denmark's coast a project is underway to inject CO2 into an old oil field." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c78e4l3rm22o#5", "link": "https://www.bbc.com/news/articles/c78e4l3rm22o?at_medium=RSS&at_campaign=rss", "title": "Are 'tech dense' farms the future of farming?", "content": "A host of technology is on offer to farmers, promising to raise farming yields and lower food prices.", "isoDate": "2026-01-20T00:00:11.000Z", "pubDate": "Tue, 20 Jan 2026 00:00:11 GMT", "contentSnippet": "A host of technology is on offer to farmers, promising to raise farming yields and lower food prices." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cwynxdnj927o#5", "link": "https://www.bbc.com/news/articles/cwynxdnj927o?at_medium=RSS&at_campaign=rss", "title": "'They are essential': How smoke detectors are evolving", "content": "AI trained to recognise fire is among the latest developments in fire alarm tech.", "isoDate": "2026-01-16T00:04:26.000Z", "pubDate": "Fri, 16 Jan 2026 00:04:26 GMT", "contentSnippet": "AI trained to recognise fire is among the latest developments in fire alarm tech." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cd0ynenr1eno#5", "link": "https://www.bbc.com/news/articles/cd0ynenr1eno?at_medium=RSS&at_campaign=rss", "title": "Honey, I shrunk the data centres: Is small the new big?", "content": "Huge data centres are being built to handle AI computing but some experts say they aren't necessary.", "isoDate": "2026-01-14T00:07:50.000Z", "pubDate": "Wed, 14 Jan 2026 00:07:50 GMT", "contentSnippet": "Huge data centres are being built to handle AI computing but some experts say they aren't necessary." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c62n5j96nqpo#5", "link": "https://www.bbc.com/news/articles/c62n5j96nqpo?at_medium=RSS&at_campaign=rss", "title": "Why are more bosses sharing the top job?", "content": "More bosses are sharing the top job giving them more time for family and breaks. ", "isoDate": "2026-01-13T00:01:51.000Z", "pubDate": "Tue, 13 Jan 2026 00:01:51 GMT", "contentSnippet": "More bosses are sharing the top job giving them more time for family and breaks." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cwyxkzjpp87o#5", "link": "https://www.bbc.com/news/articles/cwyxkzjpp87o?at_medium=RSS&at_campaign=rss", "title": "Excel: The software that's hard to quit", "content": "Companies are trying to wean staff off Excel spreadsheets to centralise control of their data.", "isoDate": "2026-01-09T00:07:59.000Z", "pubDate": "Fri, 09 Jan 2026 00:07:59 GMT", "contentSnippet": "Companies are trying to wean staff off Excel spreadsheets to centralise control of their data." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cp8zd176516o#5", "link": "https://www.bbc.com/news/articles/cp8zd176516o?at_medium=RSS&at_campaign=rss", "title": "The showers and baths keeping data centre tech cool", "content": "Finding greener ways to keep giant new data centres cool is a challenge.", "isoDate": "2025-12-23T00:04:45.000Z", "pubDate": "Tue, 23 Dec 2025 00:04:45 GMT", "contentSnippet": "Finding greener ways to keep giant new data centres cool is a challenge." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cly5gen0gj8o#5", "link": "https://www.bbc.com/news/articles/cly5gen0gj8o?at_medium=RSS&at_campaign=rss", "title": "Will tech trump tradition at bakers and biscuit makers?", "content": "Introducing robots and automation to the food industry involves extra hurdles.", "isoDate": "2025-12-19T00:07:27.000Z", "pubDate": "Fri, 19 Dec 2025 00:07:27 GMT", "contentSnippet": "Introducing robots and automation to the food industry involves extra hurdles." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c17p44w87rno#5", "link": "https://www.bbc.com/news/articles/c17p44w87rno?at_medium=RSS&at_campaign=rss", "title": "Meet the biggest heat pumps in the world", "content": "Across Europe huge heat pumps are being installed that can heat tens of thousands of homes. ", "isoDate": "2025-12-16T00:06:44.000Z", "pubDate": "Tue, 16 Dec 2025 00:06:44 GMT", "contentSnippet": "Across Europe huge heat pumps are being installed that can heat tens of thousands of homes." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c24l223d9n7o#5", "link": "https://www.bbc.com/news/articles/c24l223d9n7o?at_medium=RSS&at_campaign=rss", "title": "'It's amazing' – the wonder material very few can make", "content": "Just a handful of companies can make cadmium zinc telluride, a material with powerful properties.", "isoDate": "2025-12-12T00:03:10.000Z", "pubDate": "Fri, 12 Dec 2025 00:03:10 GMT", "contentSnippet": "Just a handful of companies can make cadmium zinc telluride, a material with powerful properties." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/c751xw96e9yo#5", "link": "https://www.bbc.com/news/articles/c751xw96e9yo?at_medium=RSS&at_campaign=rss", "title": "Will boats be a breakthrough for 3D printing tech?", "content": "Dutch firms are betting that hulls made with 3D printing machines will mean cheaper boats. ", "isoDate": "2025-11-28T00:10:39.000Z", "pubDate": "Fri, 28 Nov 2025 00:10:39 GMT", "contentSnippet": "Dutch firms are betting that hulls made with 3D printing machines will mean cheaper boats." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/czrk7gxk2l6o#5", "link": "https://www.bbc.com/news/articles/czrk7gxk2l6o?at_medium=RSS&at_campaign=rss", "title": "Scammers hacked her phone and stole thousands - so how did they get her details?", "content": "Sue Shore told the BBC how scammers targeted her - and we found her information had been leaked online.", "isoDate": "2025-11-25T04:58:08.000Z", "pubDate": "Tue, 25 Nov 2025 04:58:08 GMT", "contentSnippet": "Sue Shore told the BBC how scammers targeted her - and we found her information had been leaked online." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/videos/cvgkerq52p4o#5", "link": "https://www.bbc.com/news/videos/cvgkerq52p4o?at_medium=RSS&at_campaign=rss", "title": "The entrepreneur connecting tourists to African hospitality", "content": "TripZapp founder Rory Okoli wants to make it simple for tourists to book and pay for African travel.", "isoDate": "2025-11-25T00:02:30.000Z", "pubDate": "Tue, 25 Nov 2025 00:02:30 GMT", "contentSnippet": "TripZapp founder Rory Okoli wants to make it simple for tourists to book and pay for African travel." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cvgvynlxqdyo#5", "link": "https://www.bbc.com/news/articles/cvgvynlxqdyo?at_medium=RSS&at_campaign=rss", "title": "The contradiction at the heart of the trillion-dollar AI race", "content": "The confusing question lingering over the AI hype is whether it could be a bubble at risk of bursting", "isoDate": "2025-11-19T13:52:41.000Z", "pubDate": "Wed, 19 Nov 2025 13:52:41 GMT", "contentSnippet": "The confusing question lingering over the AI hype is whether it could be a bubble at risk of bursting" }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cz91dk0l50no#5", "link": "https://www.bbc.com/news/articles/cz91dk0l50no?at_medium=RSS&at_campaign=rss", "title": "On the front line of Europe's standoff with Russia's sanction-busting shadow fleet", "content": "With Europe imposing sanctions on Moscow, there has been a growing network of vessels sailing without a valid flag from Russia through European waters. ", "isoDate": "2025-11-19T00:23:14.000Z", "pubDate": "Wed, 19 Nov 2025 00:23:14 GMT", "contentSnippet": "With Europe imposing sanctions on Moscow, there has been a growing network of vessels sailing without a valid flag from Russia through European waters." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cjekg1pd9j4o#5", "link": "https://www.bbc.com/news/articles/cjekg1pd9j4o?at_medium=RSS&at_campaign=rss", "title": "Can technology fix fashion's sizing crisis?", "content": "The BBC looks into whether artificial intelligence (AI) can help people who struggle when clothes are oddly sized.", "isoDate": "2025-11-15T04:03:27.000Z", "pubDate": "Sat, 15 Nov 2025 04:03:27 GMT", "contentSnippet": "The BBC looks into whether artificial intelligence (AI) can help people who struggle when clothes are oddly sized." }, "pairedItem": [ { "item": 9 } ] }, { "json": { "guid": "https://www.bbc.com/news/articles/cy9pwgqvr7go#5", "link": "https://www.bbc.com/news/articles/cy9pwgqvr7go?at_medium=RSS&at_campaign=rss", "title": "Call of Duty is back, and it's got a battle on its hands", "content": "The annual instalment of the massive series faces new challenges from competitor Battlefield 6.", "isoDate": "2025-11-14T15:47:29.000Z", "pubDate": "Fri, 14 Nov 2025 15:47:29 GMT", "contentSnippet": "The annual instalment of the massive series faces new challenges from competitor Battlefield 6." }, "pairedItem": [ { "item": 9 } ] } ], "AI Newsletter Creator": [ { "json": { "index": 0, "message": { "role": "assistant", "content": "```html\n\n\n\n \n \nAI drug discovery startup Converge Bio raised $25 million in a Series A led by Bessemer Venture Partners, with additional backing from executives at Meta, OpenAI, and Wiz.
\nThe social media company inked three deals in the U.S. to power its data centers and offset its carbon footprint.
\nMitti Labs is working with The Nature Conservancy to expand the use of climate-friendly rice farming practices in India, using AI to verify reductions in methane emissions.
\nRailway raised $100 million in a Series B funding round to enhance its cloud platform for AI applications, aiming to address the limitations of legacy cloud infrastructure.
\nGoose, an open-source AI coding agent, is gaining popularity as a free alternative to Claude Code, offering similar functionality without subscription fees.
\nListen Labs raised $69 million in Series B funding following a viral hiring stunt, demonstrating rapid growth in AI-powered customer research.
\nSalesforce launched an enhanced version of Slackbot designed to act as an AI agent capable of completing tasks and processing enterprise data, aiming to transform workplace efficiency.
\nCowork is a new AI agent from Anthropic that facilitates non-technical tasks by accessing files on users' computers, exemplifying advancements in AI productivity tools.
\nA trial claims social media companies design addictive products that have led to personal injuries, highlighting concerns over technology's impact on user behavior.
\nAs chatbots in healthcare rise, many physicians express concerns over their roles, asking what this means for the future of medicine.
\n