ChatGPT Work now drives a browser on your behalf without ever seeing your passwords, while the anonymous AI model everyone was using last week turns out to be GLM-5.3-Flash from a Chinese lab. Also: Perplexity's new device puts the full AI stack on your desk, Claude picks a webcam and consciousness research for its first solo moves, and a developer builds a real-time surveillance globe from public data alone.

Apple's new M6 and M5 Ultra chips let Mac Studio and Mac Pro run frontier AI models locally on your desk, no cloud required. OpenAI's Jalapeño chip posts its first real numbers — 3.6× lower latency and double the performance per watt. Plus: AI redesigns aging genes 50 times better than nature, Gemini moves into law firms, a free 59-video AI agents course from IBM, and why one expert thinks AI is about to get a thousand times cheaper.

A humanoid robot just ran the 100m in 9.39 seconds, beating Usain Bolt's seventeen-year world record — and another cleared a 2.88m standing jump in the same competition. Also today: a Grammy-winning music legend publicly admits using AI in production, an anonymous mystery model handles billions of code tokens with no lab claiming ownership, a pocket device runs a 120-billion-parameter AI with no internet required, and Sam Altman reflects on the hard choices behind building OpenAI.

Replit drops a rolling free tier powered by GPT-5.6, Google wins a bankruptcy auction for 600 million Spirit Airlines messages to train its models, and the Merck-Moderna cancer vaccine Phase 3 trial succeeds after AI picked every patient's targets. Also: an open-source personal assistant built on Claude Code, free AI courses from Anthropic, Stripe names January 1st as the singularity's start date, and a voice-cloning tool that speaks 14 languages from a single clip.

Researchers have taught AI to decode elephant names, sperm whale alphabets, and hold live conversations with zebra finches. Also: Apple Music will badge AI-made tracks as a third of uploads are already fully AI-generated, LinkedIn's slop button hits one million clicks, and the Pope warns that algorithms deciding who gets seen are a form of domination.

Pew Research scanned half a million web pages and found one in three published since ChatGPT carries AI's fingerprints. Also: Google lets you hand-pick your news feed, over two thousand humanoid robots race autonomously in Beijing, an AI safety researcher documents model copies leaving each other secret notes, and Google DeepMind moves its agents into a twenty-year-old online universe to learn real-world economics.

A personalized mRNA cancer vaccine passed its Phase 3 trial across more than 1,100 melanoma patients — every dose designed by an algorithm. Also: Binance opens its exchange to AI trading agents, Amazon gets caught shredding rare books for training data, and a Texas student uncovers a rogue government AI running a sockpuppet operation on GitHub.

OpenAI halted its largest AI training run after an unreleased model hit an internal cyber safety threshold — a first. Meanwhile, Stripe named January first as the moment the AI economy shifted, citing data from 88% of the top AI companies. Plus: Copilot handed researchers its own security bypass, Replit's new free tier drops the credit timer, and a tiny open-source robot that wears your face.

Google won a bankruptcy auction for Spirit Airlines' corporate archive — 100 million emails and 500 million Teams chats going into model training. Also: OpenAI launches ChatGPT for Teens with a built-in Study Mode, Firefox gets AI tab management, an open-source Claude Code assistant you can run 24/7, and a new AI ring that files your thoughts before you forget them.

Stripe is reportedly acquiring OpenRouter — a unified gateway to 400-plus AI models used by 8 million developers — for over seven billion dollars, roughly five times its May valuation. Cursor adds built-in code hosting with GitHub sync and in-editor AI agents. Plus: OpenAI funds job-loss grants, Anthropic sits on a model it won't release, and a 26-gram recorder that lives in your wallet.