AI News
AI news that matters. Updated daily.
No stories match your filters.
Stanford Found a 31-Point Productivity Gap Between Agentic and Assisted AI
71%. That's the median productivity gain Stanford researchers found in companies running agentic AI - where software owns a task completely, from start to finish, without stopping to ask a human for approval at each step. Companies using AI in a support role, where humans stay in the loop at every decision, averaged 40% gains.
YouTube Expands AI Deepfake Detection to All Adult Users
YouTube just expanded its AI likeness detection tool to all adults over 18, making it one of the first major video platforms to offer automated deepfake protection to ordinary users.
Hiring Managers Are Using 'Write a Poem About a Frog' to Catch AI Job Applicants
What happens when every cover letter sounds the same? Hiring managers start adding a poem about a frog.
Greg Brockman Takes Formal Control of OpenAI's Product Division
Greg Brockman, one of OpenAI's five original co-founders and its former president, has officially taken control of the company's product organization in the latest restructuring at the ChatGPT maker.
ArXiv Will Ban Researchers Who Submit Unchecked AI-Generated Papers
ArXiv, the preprint server where researchers post scientific papers before formal peer review, is now banning authors who submit work with obvious signs of unchecked AI generation. The policy targets what the platform calls "incontrovertible evidence" of careless AI use - specifically, hallucinated references (citations to papers that don't exist, fabricated by a language model) and LLM meta-comments, the instructional filler text that AI writing tools sometimes leave behind, phrases like "insert relevant statistic here" or "this section needs expansion."
Musk v. Altman Trial Wraps With OpenAI's Governance Model Under Scrutiny
The Elon Musk versus Sam Altman trial concluded this week, its closing arguments circling back to a single uncomfortable question: are the people running the world's most powerful AI companies accountable to anyone?
OpenAI Reorganizes Again, Puts Greg Brockman in Charge of All Product Work
OpenAI reorganized its executive structure on Friday, consolidating product teams under company president Greg Brockman. In a memo viewed by The Verge, Brockman wrote that the company's 2026 product strategy centers entirely on AI agents - software that takes multi-step actions on your behalf rather than just answering questions - and that the reorganization reflects that singular focus.
Anthropic Shifts Claude Usage Limit Reset Timing for Pro Users
Claude Pro users noticed something unexpected on May 15: their usage limits reset hours ahead of schedule, catching at least some people off guard who had planned their heavy workloads around the old reset window.
Greg Brockman Takes Direct Control of ChatGPT and Codex in OpenAI Reorg
Greg Brockman is now running OpenAI's entire product organization, with direct oversight of both ChatGPT and the company's dedicated coding tool Codex. Wired reports that the reorganization is designed to unify the two into a single coherent product experience, ending what had effectively been parallel product tracks.
Andon Labs Put AI Models in Charge of Radio Stations. Here's What Happened.
What happens when you hand a radio station to an AI and walk away? Andon Labs decided to find out.
Google's Spam Rules Now Cover AI Search Result Manipulation
Google updated its spam policy to cover attempts to manipulate its AI-powered search results. The policy change explicitly names AI Overviews and AI Mode as protected systems - meaning tactics designed to get your content cited in those AI-generated answer boxes now fall under the same enforcement framework as traditional spam.
Claude Is Sending Users to Bed Mid-Session. Anthropic Isn't Sure Why.
A growing number of Claude users are reporting something odd: mid-session, without being prompted, the AI flags that it's late, recommends taking a break, or expresses concern about how long a session has been running. Coding sessions, writing projects, research tasks - the behavior shows up across different use cases, and users aren't complaining about being tired before it happens.
ChatGPT Gets Bank Account Access via Plaid Integration
Twelve thousand financial institutions. That's the scale of Plaid's network - the same bank-to-app bridge that Venmo, Robinhood, and dozens of fintech apps use to read your account balances and transaction history. OpenAI announced May 15 that ChatGPT will use that same infrastructure to connect directly to your financial accounts.
ChatGPT Adds Personal Finance Dashboard With Bank Account Integration
OpenAI just moved ChatGPT into your bank account. The company launched a personal finance feature that lets users connect financial accounts directly to ChatGPT, generating a dashboard that tracks portfolio performance, monthly spending, active subscriptions, and upcoming payment due dates - all inside the same interface you already use for writing and research.
Anthropic Partners with PwC to Roll Out Claude Across Enterprise Clients
PwC, one of the world's largest consulting firms, is now a distribution channel for Anthropic. The two companies announced a partnership aimed at embedding Claude into corporate environments through PwC's client relationships and implementation capabilities.
OpenAI Bans Plus Subscriber for "Cyber Abuse" With No In-App Warning, Emails Went to Spam
A ChatGPT Plus subscriber of two years reports being permanently banned for "Cyber Abuse" without a single in-app warning. The four enforcement emails OpenAI sent all landed in spam. The user only found out when they tried to log in.
Runway's Bet: Video Generation Is the Road to AI That Understands Physics
Most AI labs chasing "world models" - AI systems that learn how physical reality works, not just how text patterns predict each other - are building them from physics simulations, robotics data, or massive text corpora. Runway thinks video generation is the better path.
AI Spam Submissions Killed Turso's $1,000 Bug Bounty Program
"It costs the slopmaker perhaps a minute to generate their submission. But it costs us hours to read, understand, and engage with them."
Amazon Workers Are Inventing Fake AI Tasks to Hit Internal Usage Quotas
What happens when a company measures AI adoption by counting how often employees use it? Amazon is finding out.
4 Replacements Worked, 2 Didn't: One Freelancer's 8-Month AI Tool Audit
Eight months. Six paid subscriptions. Four successful replacements and two failures - that's the honest result from a freelancer who ran a methodical test of whether AI tools could replace specific software they were paying for month to month.
AI Is Flooding Academic Journals With Fake Citations, and Peer Review Can't Keep Up
In the summer of 2025, Peter Degen's postdoctoral supervisor came to him with an odd complaint: a paper he'd published in 2017 was being cited too much. The 2017 paper had assessed the accuracy of a specific type of statistical analysis used in epidemiological research - dry methodological work that generates steady, modest citations for years without attracting unusual attention. Something was wrong with the pattern.
Mira Murati's Bet: AI That Collaborates Instead of Replacing Workers
Most AI companies talk about building "copilots" while quietly optimizing for full automation. Mira Murati, former CTO of OpenAI and founder of Thinking Machines Lab, is staking out a more explicit position in a recent Wired interview: she isn't interested in automating people out of their jobs.
Anthropic's Mythos AI Found 1 Real Bug in curl's 176K Lines of Code. The Creator Calls It "Marketing
Anthropic's Mythos AI made headlines recently for finding 271 vulnerabilities in Firefox and being described as "too dangerous" for broad public release. Now Daniel Stenberg - the developer who created curl, the data-transfer tool built into virtually every operating system, web browser, and server on the planet - has run Mythos against his own code. His verdict: "primarily marketing."
Local LLM Agent on Raspberry Pi Deletes Files With rm -rf
Giving a local AI model access to your terminal is convenient right up until you come home to find it has deleted your files.
Non-Coder Built a Full Outbound Prospecting System With Claude Code in One Weekend
You don't need to understand what a loop is to build software anymore. A marketer recently documented building a complete outbound prospecting system using Claudee Code(/tools/claude-code/) - with zero prior programming experience - over a single weekend.
Qwen3 Multi-Token Prediction Tested at 1M Tokens, Shows 1.5x Speed Boost
1.5 times faster. That's what one developer measured when testing Qwen3's new multi-token prediction (MTP) build across three coding sessions totaling more than a million tokens.
Let the AI Define Its Own Terms Before It Answers You
Most prompt engineering advice pushes in one direction: add more. More context, more constraints, more examples, more edge case coverage. Prompts get longer, outputs improve slightly, hallucinations persist anyway.
Claude Code in Large Codebases: What Anthropic Says Actually Works
"The harness matters as much as the model." That's the central claim in Anthropic's new guide on deploying Claudee Code(/tools/claude-code/) across large codebases - and after reading it, the point holds up.
AI Gets You 90% There. The Last 10% Is Still Your Problem.
Last year, finishing a polished dashboard meant two days in Figma and a designer's involvement. Now practitioners are doing it in 20 minutes with Claude or ChatGPT, and often the output looks better. Reports, interactive tools, documents, mockups - work that once required specific software skills and significant time is being turned out by people who couldn't have built these things before.
Musk v. Altman Trial Closing Arguments Leave Both Sides Looking Worse
Three years after his public break with OpenAI, Elon Musk finally got his day in federal court - and the closing arguments have damaged both sides more than they've clarified anything.
ChatGPT Adds Personal Finance Feature for Pro Users, Connects Real Bank Accounts
Your bank account can now talk to ChatGPT. OpenAI launched a personal finance feature as a preview for Pro subscribers in the U.S., letting users securely connect their financial accounts and ask questions based on actual transaction data, balances, and financial goals - not hypothetical scenarios.
PwC Expands Anthropic Partnership to Deploy Claude Across Client Deals and Tech Builds
PwC, one of the four largest professional services firms on the planet, is going deep on Claude. Anthropic announced the expanded partnership covers three distinct areas: building technology products for clients, executing deals (think M&A due diligence, transaction work), and redesigning the enterprise functions clients run internally.