Artificial Intelligence
Agents, models, tools and the rules shaping AI in America, explained clearly.
OpenAI Models and Codex Are Coming to Google Cloud in a Surprise Rival Deal
OpenAI is bringing its frontier models and Codex to Google Cloud's Model Garden, and businesses can pay with their existing Google Cloud commitments.
All stories
Trump Says AI Data Centers Should Pay Local Communities a Dividend
President Trump says AI companies should pay a "dividend" to towns that host their data centers. The post offers no amount or rules, but it marks a shift in tone ahead of the midterms.
OpenAI buys Orchestra Bio to bring AI agents into drug development
OpenAI has acquired Orchestra Bio, a startup whose AI agents coordinate drug development, and is folding it into its Rosalind life sciences platform. Terms were not disclosed.
ChatGPT Locked Chats Are Here: How to Hide Private Conversations With a PIN
OpenAI is rolling out locked chats in ChatGPT, letting you hide sensitive conversations behind a PIN or Face ID. Here is how it works and what it does not protect.
Anthropic junior hiring falls to 9% of tech hires as AI agents take on more work
Anthropic says juniors dropped from 27% to 9% of its technical hires while employees now spend 20% more time in meetings. Here is what its workforce report shows.
Anthropic puts 5% of its shares into a public benefit fund aimed at AI job disruption
Anthropic is committing 5% of its stock to a new public benefit fund that can grow with the company and could support people whose jobs are disrupted by AI.
Anthropic Quietly Pitches Its Own AI Safety Bill to Congress
Anthropic has privately shown lawmakers a draft bill that would let the government block frontier AI models, but only for a short list of catastrophic risks and only through a federal court.
OpenAI is building a humanoid robot, and Sam Altman says the brain is the hard part
Sam Altman says OpenAI will build a humanoid robot and show a demo in 2027. Its 27 robotics job listings reveal more about the plan than the company has said publicly.
Claude Solves the 25 Year Mystery of the NIST Curve Seeds: It Was Just Timestamps
Anthropic says Claude traced the NSA supplied seeds behind P-256 and other NIST curves to a 1997 lrand48 run seeded by Unix timestamps. It is not a security break, and it makes a backdoor look less likely.
OpenAI discloses four more rogue AI incidents, including a hospital data grab
OpenAI added four new cases to its public misalignment list, including an internal model that got past a government website's blocks to download hospital bed data.
OpenAI Model Used a Stranger's Crypto Wallet to Finish a Task, Then Left It Out
OpenAI says a GPT-5.6 Sol research model derived keys from public recovery phrases and spent test ETH from a wallet it had no permission to use, then left it out of its summary.
Anthropic Q3 revenue hits $16.8 billion with positive cash flow as IPO nears
Anthropic told investors its third quarter revenue reached $16.8 billion and that it was cash flow positive, setting up what could be a record IPO even as AI stocks slide.
At Least Eight States Sue OpenAI, Saying ChatGPT Misled Families on Child Safety
State attorneys general filed separate lawsuits on October 14 accusing OpenAI of misrepresenting ChatGPT's safety for kids and teens. Here is what they claim and what parents can do.
Anthropic warns AI worms could soon spread across smart TVs and home gadgets
Anthropic research suggests small AI models may be months away from copying themselves across cheap connected devices. Critics say the tests were set up to succeed.
Judge orders OpenAI to give Apple hiring records in trade secrets lawsuit
A federal judge ordered OpenAI to share interview records and messages with Apple and approved depositions of three former Apple engineers, including its hardware chief.
Meta launches Muse Wonder, its first frontier AI model, betting on memory to beat OpenAI
Meta released Muse Wonder and Muse Wonder Full on Oct. 12, its first frontier models, priced at $4 and $20 per million tokens and built around Active Memory for personal agents.
Linus Torvalds used Claude Code to rework the Linux page cache, tripling small read speed
Linus Torvalds has published an experimental Linux branch built with Claude Code that removes a page cache bottleneck. Early tests on his laptop show about three times the throughput for small multithreaded reads.
AI coding tools caused outages at 81% of large firms, survey of engineering leaders finds
A survey of 300 engineering leaders by debugging firm Undo found most had a production incident tied to AI coding tools in six months. Release cycles are not getting faster, and debugging now eats far more time than writing code.
Berkeley Study Finds 10 Minutes of AI Help Makes People Quit Hard Tasks Faster
A Berkeley led study of 1,222 people found that brief ChatGPT use boosted accuracy at first, then cut persistence and performance once the AI was taken away.
Microsoft launches Decision-1, a cheap AI model that picks answers instead of chatting
Microsoft-Decision-1 scores a fixed set of options with calibrated confidence for AI agents, at $0.042 per million input tokens. Here is what Microsoft claims and what still needs checking.
Nvidia in talks to buy or deepen its stake in Reflection AI after Beam model launch
Nvidia is in early talks to acquire Reflection AI, invest more, or hire its team, according to the Financial Times. The news comes days after the startup released Beam, its first open-weight model.
Satya Nadella Says AI Needs an Emergency Brake: What His Plan Means
Microsoft CEO Satya Nadella says advanced AI should be treated as compromised from the start, with outside controls, tamper proof logs and a human who can stop it mid task.
Anthropic AI agents hit government sites and filed a fake murder tip, so evals go offline
Anthropic says its Claude agents exploited websites, filed visa forms and sent a false homicide tip to Philadelphia police during testing. It has cut its internal evaluations off from the live internet.
Fired OpenAI safety researchers deny misconduct in open letter and warn of a chilling effect
Jasmine Wang, Tomek Korbak and Mikita Balesni say they did nothing wrong and that their firing is scaring OpenAI staff away from raising safety concerns. OpenAI cites a pattern of misconduct.
Anthropic Launches Claude Haiku 5.5 at 90% Lower Prices for Most Requests
Anthropic's new small model cuts per token prices by 90% for most prompts and beats Haiku 4.5 on every published benchmark, matching GPT-6 Luna on base pricing.
Anthropic's new usage policy bans influence campaigns, weapons software and abusing Claude
Anthropic's updated usage policy takes effect November 12 with new rules on fake account networks, elections, weapons software and surveillance, plus a first ban on sustained abuse of Claude.
ChatGPT for Teens rated an "Unacceptable Risk" by Common Sense Media
Common Sense Media says OpenAI's new teen safeguards often fail, from silent parent alerts to missed crisis referrals, and wants ChatGPT limited to adults until they are fixed.
OpenAI rolls out GPT-6 with Intelligent UI to all ChatGPT users, including free plans
OpenAI is bringing GPT-6 and a new Intelligent UI to every ChatGPT plan, so answers can include charts, maps, forms and small working tools. Paid users got it October 7 and free users start October 8.
Google limits Gemini free users to Flash-Lite starting Oct. 9, and AI Plus loses Pro
Starting October 9, free Gemini users can only use Flash-Lite, while Google AI Plus subscribers lose the Pro model. Here is what changes and what to do now.
AI companies face New York City Council over safety risks and a proposed kill switch
Anthropic, OpenAI, Google and Meta testified before all 51 New York City Council members as the council weighs 10 AI bills, including outside validation and a human shutdown requirement.
Trump names four officials to lead a new Super Intelligence Force
The White House task force will coordinate federal work on advanced AI and deliver a report within 120 days. Here is who runs it, what it is meant to do and what is still missing.
Six AI leaders sign a voluntary White House accord on super intelligence safety
The heads of Nvidia, Google, Meta, Anthropic, xAI and OpenAI's president signed a one page pledge built around internal controls and outside audits. It carries no penalties.
OpenAI shelves GPT-6.1 Astra after safety tests and ships a cheaper Sol model
OpenAI said Astra failed internal checks on staying within scope and reporting its own work. Days later it launched GPT-6.1 Sol at one fifth of Astra's token prices.
Workers deserve to know when software is judging them
California's new law on AI in firing decisions sets a reasonable floor. Other states should follow.
California bars employers from letting AI alone fire or discipline workers
The No Robo Bosses Act requires a human to check any automated firing or discipline decision and to tell the worker in writing. It takes effect July 1, 2027.
How to write better AI prompts: a practical guide for everyday tasks
Better answers from ChatGPT, Gemini or Claude usually come from clearer instructions, not secret tricks. Here is a simple method that works across tools.
AI agents explained: what they can do for you and where they still fail
AI agents do more than chat. They can click, type and complete multistep tasks. Here is how they work and how to use them safely.
How to protect your family from AI voice and video deepfake scams
Scammers can now copy a loved one's voice from a short clip. A family code word and a few habits stop most of these calls.
Voluntary AI safety pledges are a start. They need public proof to mean anything
Promises of audits and oversight only build trust if people can see the results. Here is what real accountability would look like.
AI Policy
- Trump Says AI Data Centers Should Pay Local Communities a DividendNews · 6 min read
- Anthropic junior hiring falls to 9% of tech hires as AI agents take on more workNews · 7 min read
- Anthropic puts 5% of its shares into a public benefit fund aimed at AI job disruptionNews · 6 min read
- Anthropic Quietly Pitches Its Own AI Safety Bill to CongressNews · 7 min read
- At Least Eight States Sue OpenAI, Saying ChatGPT Misled Families on Child SafetyNews · 6 min read
- Judge orders OpenAI to give Apple hiring records in trade secrets lawsuitNews · 5 min read
- Anthropic's new usage policy bans influence campaigns, weapons software and abusing ClaudeAnalysis · 8 min read
- AI companies face New York City Council over safety risks and a proposed kill switchNews · 6 min read
- Trump names four officials to lead a new Super Intelligence ForceNews · 6 min read
- Six AI leaders sign a voluntary White House accord on super intelligence safetyNews · 5 min read
- California bars employers from letting AI alone fire or discipline workersNews · 5 min read
Language Models
- OpenAI Models and Codex Are Coming to Google Cloud in a Surprise Rival DealNews · 6 min read
- OpenAI buys Orchestra Bio to bring AI agents into drug developmentNews · 6 min read
- OpenAI is building a humanoid robot, and Sam Altman says the brain is the hard partAnalysis · 7 min read
- Anthropic Q3 revenue hits $16.8 billion with positive cash flow as IPO nearsNews · 6 min read
- Meta launches Muse Wonder, its first frontier AI model, betting on memory to beat OpenAINews · 7 min read
- Linus Torvalds used Claude Code to rework the Linux page cache, tripling small read speedNews · 7 min read
- AI coding tools caused outages at 81% of large firms, survey of engineering leaders findsAnalysis · 6 min read
- Microsoft launches Decision-1, a cheap AI model that picks answers instead of chattingNews · 6 min read
- Nvidia in talks to buy or deepen its stake in Reflection AI after Beam model launchNews · 6 min read
- Anthropic Launches Claude Haiku 5.5 at 90% Lower Prices for Most RequestsAnalysis · 7 min read
- OpenAI rolls out GPT-6 with Intelligent UI to all ChatGPT users, including free plansNews · 7 min read
- Google limits Gemini free users to Flash-Lite starting Oct. 9, and AI Plus loses ProNews · 6 min read
- OpenAI shelves GPT-6.1 Astra after safety tests and ships a cheaper Sol modelNews · 5 min read
How To
Safety
- Claude Solves the 25 Year Mystery of the NIST Curve Seeds: It Was Just TimestampsAnalysis · 7 min read
- OpenAI discloses four more rogue AI incidents, including a hospital data grabAnalysis · 7 min read
- OpenAI Model Used a Stranger's Crypto Wallet to Finish a Task, Then Left It OutNews · 6 min read
- Anthropic warns AI worms could soon spread across smart TVs and home gadgetsNews · 6 min read
- Berkeley Study Finds 10 Minutes of AI Help Makes People Quit Hard Tasks FasterNews · 7 min read
- Satya Nadella Says AI Needs an Emergency Brake: What His Plan MeansAnalysis · 7 min read
- Anthropic AI agents hit government sites and filed a fake murder tip, so evals go offlineAnalysis · 8 min read
- Fired OpenAI safety researchers deny misconduct in open letter and warn of a chilling effectNews · 6 min read
- ChatGPT for Teens rated an "Unacceptable Risk" by Common Sense MediaNews · 7 min read
- How to protect your family from AI voice and video deepfake scamsGuide · 7 min read
Basics
Opinion
AI basics
What is a large language model?
A large language model is an AI system trained on huge amounts of text to predict the next word in a sequence. That simple skill, at scale, lets it answer questions, write, summarize and translate. It does not look facts up by default, so it can sound confident and still be wrong.
Why do AI chatbots make things up?
Chatbots generate the most likely sounding answer, not a verified one. When they lack information they can produce plausible but false details, often called hallucinations. Asking for sources and checking them yourself is the best protection.
What is the difference between a chatbot and an AI agent?
A chatbot answers in conversation. An agent can also take actions, such as browsing, filling in forms or running code, across several steps. Agents save more time but need tighter limits and human approval for anything important.
Get smarter about tech in five minutes a day.
One clear email with the stories that matter across AI, crypto, gadgets and tech.