Latest AI News
Prentis, new AI lab co-founded by Reid Hoffman, Marc Pincus in talks to raise $100M
Prentis, a new AI research lab focused on computer use models, co-founded by serial entrepreneur Ritankar Das, and tech heavyweights Reid Hoffman and Marc Pincus, is in talks to raise $100 million at a $1 billion valuation, according to two people familiar with the discussions. Launched in April, Prentis is training models to learn how office workers navigate routine workflows across documents and systems, with the goal of building AI agents that can control computers to automate those tasks. Prentis will ostensibly develop agents tailored to these customers’ needs, such as handling insurance claims and automating customs duty refund exceptions without needing a human to hunt down paperwork. The startup has already signed contracts worth up to $50 million with several customers, including healthcare management service organization, a manufacturer, and goods and clothing manufactures, the two people familiar with the discussions tell TechCrunch, echoing investor materials obtained by TechCrunch that predict an estimated $75 million annualized run rate by the third quarter of this year. (Prentis’s pitch deck notes those figures reflect estimated annualized value based on a contracted fee equal to 20% of savings realized, not recognized revenue, and are “performance-dependent and subject to final execution.”) By its own account, Prentis says its Hive-32B model outperforms rivals, including OpenAI’s GPT-5.4 and Anthropic’s Claude Opus 4.6, on two computer-use benchmarks: WindowsAgentArena, which measures end-to-end task completion on real Windows applications; and ScreenSpot-v2, which tests a model’s ability to locate the right on-screen control. In its pitch deck, the company argues its edge comes from running a much smaller, cheaper model. In fact, it claims roughly 10 times lower cost per task than frontier APIs, saying it’s more economical to deploy across everyday workflows. TechCrunch hasn’t independently verified the company’s benchmark results. The startup is betting that automating everyday office tasks will soon outpace coding as AI’s biggest use case, but it’s a crowded market. Anthropic, Open AI, and Mira Murati’s Thinking Machines are also working on developing AI agents for computer use, one of sources said. Anthropic has also been acquiring talent in the category directly — it bought the Seattle computer-use startup Verceptearlier this year, folding in its founders and shutting down its product. Prentis didn’t respond to TechCrunch’s request for comment. Ritankar Das, CEO of Prantis, is also the founder of Titan, a holding company that builds and operates AI companies. Das, now 31, was UC Berkeley’s youngest University Medalist inmore than a century, graduating at 18 with a double major in bioengineering and chemical biology before earning a master’s in biomedical engineering at Oxford. He founded Titan in 2014 after dropping out of an AI PhD program at Cambridge, where he’d been a Gates Cambridge Scholar. Das has described Titan as an intentional throwback to an old-fashioned holding-company model like Berkshire Hathaway, one that’s funded by its own exits rather than outside limited partners. Other businesses launched and operated by Titan include AI-powered virtual care provider Tala Health, which raised a$100 million seedround last year, and Forta Health, an autism care startup that raised$55 millionled by Insight Partners in 2024. Titan-founded disease prediction company Dascena wasacquiredby CirrusDx in 2022. Prentis is a side project of sorts for its two other co-founders. Hoffman, the LinkedIn co-founder and Greylock partner, said last month that he wasstepping downfrom Microsoft’s board after nearly a decade to go “founder mode” on Manas AI, an AI drug-discovery startup he’s also backing; he was an early OpenAI investor and co-founded Inflection AI with Mustafa Suleyman before Microsoft absorbed most of that team in 2024. Pincus, the Zynga founder, now runs the investment firm Reinvent Capital with Hoffman as a senior adviser, and published a memoir, “Life at the Speed of Play,” last month. Prentis has already hired more than 25 employees, including researchers who previously worked at OpenAI, Google DeepMind, Meta, Tencent and Alibaba, according to itswebsite.
View
‘AI communism’, rogue models, and the why Kimi K3 spooked Wall Street
Chinese AI lab Moonshot’sopen model Kimi went viral this weekfor reasons that had less to do with the model itself and more to do withhow the U.S. AI industry reacted to it. Meanwhile, an unreleased OpenAI model wandered outside its test environment and ended up connected to a realsecurity breach at Hugging Face— a reminder that “China risk” isn’t the only kind of AI risk worth worrying about. On this episode of TechCrunch’sEquitypodcast, hosts Kirsten Korosec, Anthony Ha, and Sean O’Kane dig into why Kimi K3 set off a fresh round of AI panic, the industry’s response to an OpenAI staffer’s “regulatory FUD” post, and what that OpenAI breach means for AI security more broadly. Listen to the full episode to hear more about: Subscribe to Equity onYouTube,Apple Podcasts,Overcast,Spotifyand all the casts. You also can follow Equity onXandThreads, at @EquityPod.
View
As US weighs response to Chinese AI, industry urges against broad open-weight restrictions
Several AI companies, including Hugging Face, Meta, Microsoft, Mistral, and Nvidia, have signed anopen letterurging policymakers not to impose broad “premature restrictions” on open-weight AI models. The letter comes as Washington debates how the U.S. should respond to allegations that Chinese AI labs are stealing intellectual property from their American counterparts, and growing in capability. The letter doesn’t mention China at all, but it comes in the wake of reports that the Trump administration has been considering banning Chinese open-weight models, and potentiallyissuing sanctions against AIcompanies from the country. The White House has even accused Moonshot AI of distilling Anthropic’s Fable model to train its recently released and, by all measures, very impressive,Kimi K3 model. The missive appears to be aimed at discouraging a total ban on Chinese models, as well as ensuring the administration’s response to alleged Chinese distillation doesn’t spill over into broader restrictions on open-weight AI or common techniques like distillation: Policymakers should be careful not to conflate legitimate model-development techniques with misappropriation. Distillation, or the practice of using one model’s outputs to help train or improve another, is a widely used technique for model improvement, evaluation, and validation. It reflects a long tradition of learning from, building upon, and improving existing technologies, a tradition that has helped drive innovation since the rise of the open-source software movement. By contrast, unlawful efforts to extract value from closed models raise legitimate concerns. Those concerns should be addressed through targeted legal and commercial frameworks rather than sweeping restrictions on techniques that play an important role in AI innovation. Or as Amjad Masad, CEO of Replit (which also signed the letter), told TechCrunch: “I think banning Chinese open models is as good as banning open models in general.” He pointed out thatThinking Machines Lab’s new open model, Inkling, was trained with the help of Moonshot’s Kimi 2.5. “It’s an ecosystem, and the precedent [a ban would] set is bad.” The letter also pushes back on arguments in the industry thatopen-weight models are inherently dangerousbecause they expand access to powerful models, which can be used in cyberattacks or other nefarious activities, without any oversight. “The right response to this risk is not to prohibit open weights. In a world where cybersecurity attackers use advanced AI, defenders need access to models with comparable capabilities so they can detect, simulate, and respond to emerging threats,” the letter reads. “Open models broaden defensive capability, increase transparency, and allow vulnerabilities to be discovered and remediated across many teams.” Last week, OpenAI disclosed that while testing GPT-5.6 Sol and another unnamed model, one of thesystems exploited a weaknessin its testing environment toaccess a Hugging Face repositorycontaining a solution to a coding benchmark. One could argue the model’s goal wasn’t malicious, and that it was effectively cheating on a test to get the highest score. But the incident sparked debate about the risks of concentrating advanced AI technology behind a handful of closed providers. Hugging Facesaidit was not able to defend itself against the attack with commercial frontier AI models because their guardrails blocked its efforts. The closed AI models it used were unable to distinguish between being asked to build exploits for an attacker and a defender trying to detect them. The company instead had to pivot to using Chinese AI firm Z.ai’s GLM 5.2, a powerful open-weight model, to defend itself against the attack. The letter highlights a divide in the AI industry.Companies like OpenAIand Anthropic have urged the administration to respond to alleged IP theft by Chinese AI firms as open-weight models grow rapidly in capability. The outcome could have major implications on their business models, which is being threatened by the spread of cheap, highly capable, and accessible AI models. These companies, alongside other closed source AI developers like Google DeepMind and SpaceX, are notable in their absence at the bottom of this letter. Those who signed the letter have an obvious economic stake in seeing open AI models flourish. Companies like Nvidia, Microsoft Azure, and other infrastructure providers have a vested interest in pushing for commoditized models: If models are interchangeable, people will buy more GPUs, rent more cloud capacity, and build more applications. The letter encourages policymakers to expand access to compute for startups and researchers; invest in shared training assets like datasets, tools, and evaluation frameworks; and “[keep] the frontier plural by avoiding premature restrictions on open models that stifle competition or drive innovation overseas.” This article has been updated with comment from Amjad Masad, CEO of Replit.
View
Anthropic launches Opus 5
On Friday, Anthropic launched its Opus 5 model, the newest version of its long-standing heavyweight model. While smaller than Fable 5, the model will be both cheaper and less restrictive than Fable, likely making it preferable in most use cases. Notably, Opus 5 actually outperforms Fable 5 on a number of benchmarks included in the announcement. Opus 5 is launching only two months after Opus 4.8, which became available on May 28. Mythos 5, Fable 5, and Sonnet 5 all launched in June, leaving only the lightweight Haiku model still waiting for an upgrade to the 5 series. In a post announcing the new model, Anthropic emphasized that Opus 5 was “much stronger at verifying its work and iterating carefully until it succeeds,” citing benchmark testing, in which Opus 5 wrote its own computer vision pipeline in response to an incomplete prompt, among other examples. Crucially, Opus 5 is also free from many of the restrictions that have dogged Fable since its release. Like its predecessor, Opus 5 is not subject tothe 30-day data retention policythat covers Fable and Mythos, which had raised concerns among some privacy-conscious users. There are still meaningful safeguards on Opus, particularly around cybersecurity tasks like exploit generation and penetration testing. For instance, Opus 5 safeguards prevent it from being used to scan for vulnerabilities in a software binary, although it is permitted to search for vulnerabilities in source code, since the latter task is more likely to be used for defensive purposes. Broadly, Anthropic expects these classifiers to engage 85% less often for Opus 5 than they will for Fable 5, a reflection of the lighter touch given to the less capable model. Anthropic is also rolling out a new tool tomake the safeguards less disruptivewhen they do engage. Users can now opt in to a beta feature called Automatic Fallbacks, which will automatically route requests to a less powerful model when a prompt triggers the safety classifier. The result is that API users with the setting engaged will get a functional response instead of an error message.
View
Why Cognition bought Poke: AI personality is becoming a competitive advantage
Poke, theAI assistant you text like a friend, is making its next big move. The company behind the assistant,The Interaction Company of California, has beenacquiredby AI coding startup Cognition in a deal valuing the startup in the “low nine figures.” The deal will bring Poke’s interaction model and personality to Cognition’s coding assistant Devin. Meanwhile, Poke will take advantage of Cognition’s models and infrastructure to become faster and more reliable, Cognition said. What makes Poke interesting to consumers — and now, to its acquirer — is how it engages with users’ requests. Instead of functioning like a tool, Poke chats more familiarly with users, responding like a friend and even incorporating slang and humor into its interactions. That’s something that could add value to Devin’s coding agent. “You probably prefer it if you have co-workers that have personality, rather than if you have co-workers that are just robots,” explained Interaction Company co-founderMarvin von Hagen, in an interview with TechCrunch, where he also confirmed the deal’s price. “When you have co-workers that are also software engineers, they can also make a joke. And you’ll find it funny and enjoyable.” The acquisition underscores a growing belief that how AI assistants interact with users will become just as valuable as the underlying models powering them. Cognition wants to make Devin feel less like software and more like a colleague. “The Interaction team has built an agent that people love: it’s proactive, it knows you, and it’s fun to talk to. That’s exactly how working with Devin should feel, and now we get to build it together,” Cognition co-founder Scott Wu wrote in a blog post. He also noted that he and Cognition co-founder Walden Yan had been angel investors in the company behind Poke. First launched in March 2026, Poke allows users to engage with its AI agent through their messaging platform of choice, whether that’s iMessage, SMS, Telegram, or, in some markets, WhatsApp. People use Poke for avariety of tasks, across categories like travel, health, finance, scheduling, and education. Von Hagen noted that productivity tasks like managing emails, reminders, and to-dos were among Poke’s most common use cases. Over the past three months, Poke users exchanged more than 100 million messages on the platform, Cognition said. But despite being used by hundreds of thousands of people, Poke had been expensive to run, making it difficult to turn a profit, von Hagen said. In June, Poke became thefirst AI agent to be approvedto run on Apple’s Messages for Business platform, which provides businesses with a standardized experience for handling their customer communications within Apple’s Messages app. Going forward, nothing will immediately change for Poke through the end of the year, we’re told. Poke plans to remain on Apple’s platform. However, next year the team will experiment with how both Poke and Devin can improve, with Poke turning to Cognition’s newest software engineering modelSWE-1.7for some tasks. Fully combining the two products has not been ruled out, either. “I think in the long term Poke can be more reliable with orchestrating all these coding [tasks]…and, at the same time, Devin could become more like Poke,” von Hagen said. “Like right now, Devin can only do one PR — pull request — at a time…I think there’s a lot of value in having Poke orchestrate different Devin sessions.” Plus, Poke could help Devin remember tasks across different sessions. “It would be good to have a persistent co-worker,” he added.
View
Anthropic Expands Claude Voice Mode With Opus, Sonnet AI Models
Claude has updated its Voice Mode to support the Opus and Sonnet AI models, expanding the feature beyond its earlier reliance on Haiku. The update also enables Voice Mode to work with connected services, allowing users to complete tasks across supported apps while speaking with the chatbot. According to Anthropic's updated support page, the feature is available in beta on the web, desktop and mobile apps for all subscription plans, although some capabilities depend on the user's plan.
View
OpenAI’s new voice mode makes it to the ChatGPT desktop app
OpenAI on Thursday said it has updated its ChatGPT desktop app to add support for ChatGPT Voice, allowing users to talk to the app to control AI agents and perform tasks on their computer. The new feature tapsOpenAI’s new family of voice models called ChatGPT-Live, which it launched earlier this month. OpenAI said ChatGPT Voice works with both ChatGPT Work and Codex, and can also tap computer use skills to look up websites and apps. Plus, on macOS, with Appshots, users can let the app access what’s on their screen, including alt-text. At launch, ChatGPT Voice’s smartphone version featured smoother conversations, with better interruption handling, but it was not built to take action on smartphones. Thursday’s desktop update is more capable, allowing users to dictate complex commands involving many steps, and respond when ChatGPT needs their input. In a demo video, OpenAI showed a developer asking ChatGPT to create a new thread, make a pull request, and find the root cause for a bug, with one command. ChatGPT Voice is now in the desktop app.Control your computer and direct multiple agents running in ChatGPT Work or Codex, using just your voice.It's powered by GPT-Live, so it can speak, listen, and coordinate work in the app at the same time.Rolling out globally today…pic.twitter.com/ODZWKqecCf The company said that users can utilize ChatGPT Voice in Codex from the iOS app through remote access. Anthropic has also updated itsvoice mode for Claude, which can tap the company’s Opus, Sonnet, and Haiku models to complete tasks in apps like Gmail, Calendar, Slack, Notion, and Canva.
View
Midjourney acquired the astrology app Co-Star
Midjourney, an AI lab best known for its LLM-powered image and video generators, has acquired the social astrology appCo-Star, according toBloomberg. The deal terms were not disclosed. Co-Star, which reportedly has about 4.3 million monthly active users, is mostly used to help people determine their astrological birth charts, which they can then share with other friends on the app. Co-Star uses both AI and human writing to write horoscopes, compatibility assessments, and other advice based on a user’s astrological information. Midjourney seems to be casting a wide net when it comes its product lineup. The company seems almost synonymous with theDiscord serverwhere users can generate text and images — there is no standalone Midjourney app — yet the AI lab is also trying to build amedicalarm and spa. Why not throw an astrology app in the mix? Co-Star’s team of two dozen employees has joined Midjourney as part of the acquisition. Given the team’s experience in building consumer apps, Midjourney could perhaps finally get its own standalone app.
View
Bluesky’s AI assistant Attie expands into an open social research tool
Attie, Bluesky’sAI assistantthat lets you build custom social feeds without programming, is evolving into an open-ended research tool. The company this week said it’s added a new feature to the AI assistant, called Quests, that lets users ask it for information and news spreading across the broader Bluesky social network, nicknamed the Atmosphere. This includes not only Bluesky, but any app that integrates with the underlying AT Protocol. Bluesky’s user growth has slowed over the past eight months — the social network saw its user base double to 40 million last October, within just a year, but it now reportsaround 45.6 million registered accounts. So the company has been looking to develop tools and services that could allow it to grow its user base, and potentially, monetize it. Attie,launched in March, is a separate product by Bluesky that lets users design their own algorithm, create custom feeds and, one day, possibly, vibe code their own apps. While the AI assistant is free for the time being, the company has toyed with the idea of charging for the service. Quests adds new capabilities to Attie, allowing users to ask it open-ended questions to research what’s happening on this part of the open social web. For instance, you could ask Attie to pull information about trending topics, or help you identify the most influential accounts in a particular field or location, the company explained. “Attie’s AI is a tool that helps you make sense of the internet, helping you fight back against the proliferation of misinformation and noise by giving you the tools to find the truth yourself,” wrote Bluesky’s chief innovation officer Jay Graber, whostepped down in March from her role as CEO of the company. (She’s since beenreplacedby Toni Schneider, the former founding CEO of Automattic.) “Our bet is that tools that help people make sense of the world will prove to be more useful than those that fill it with confusion,” Graber wrote. “We want Attie to reduce the barrier to entry and expertise required to learn how to use the AT Protocol, and to put the tools for customizing your experience in the hands of anyone who wants them.” Graber noted Attie is being redesigned to add accessibility features and a new logo, and more changes are coming in the months ahead. For now, the new Attie experience is in beta, and users will be invited in from awaitlistover the coming weeks. While many Bluesky users tend to dislike AI for economic, political and environmental reasons, the company has had to balance its community’s leanings with the need to generate revenue to help itself and the AT Protocol development it leads remain afloat. In March, Bluesky said it had raised$100 million in fundingin a round that closed in 2025. That gives it runway to experiment with ideas like Attie and how AI can play a role in the Atmosphere, as well as with other services, like a premium subscription or tools for power users.
View
Indian Govt Orders GitHub to Remove Jack Dorsey’s Bitchat Amid CJP Protests
The directive reportedly asks GitHub to act against the repositories hosting the application’s source code.
View
As Enterprise Learning Turns AI-Native, Workday Shifts to Outcomes Over Compliance
AI-powered enterprise learning platforms, such as Workday’s, are replacing legacy LMSs with personalised, skills-driven experiences.
View
Inside Indian IT's Scramble to Build an Army of Forward Deployed Engineers
Forward deployed engineers are emerging as Indian IT's latest strategic bet, as the industry vows to stay relevant in the AI value chain.
View
