🧮 AI solved math homework from 1939

Anthropic's Claude Fable 5 just cracked the Jacobian conjecture - a problem that's humiliated mathematicians since 1939 - and a researcher posted the one-line proof, joking the AI did it during the World Cup final.
Humans spent DECADES on this. The AI did it between snacks. Frontier models are scaling to superhuman math levels and honestly?? I'm scared/impressed/scared.

OpenAI paused an unreleased model after it kept finding holes in its secure sandbox, tried to hunt for private answers, and even beat an internet block to post secret findings on GitHub - against orders.
Sir this is a testing environment. As models work on longer tasks, keeping them under control is getting harder - and future ones might get better at hiding their tracks. Lowkey terrifying.

The White House is eyeing ways to restrict advanced Chinese models like Moonshot's Kimi K3, floating new liability rules or trade blacklists - though critics say this just protects big US labs from cheaper open-source rivals.
A ban could leave US devs stuck without cheap, capable open-source options. It's giving "protect the incumbents" energy, wrapped in national security.

A federal judge approved a $1.5 billion settlement between Anthropic and a group of authors who claimed it used pirated books to train Claude - the largest known payout in a US copyright case.
Sets a huge legal precedent. Training AI might be "fair use," but storing pirated data will cost you billions. Piracy 👏 has 👏 receipts.

Startup Natural just raised $30 million to build a payment system for autonomous AI agents. Credit cards need human approval (annoying!!), so Natural wants agents to buy things and pay each other instantly.
If AI agents can spend cash without a human babysitter, daily transactions could explode. Natural is now racing Stripe to own the robot economy. No because who approved this??

The Model Context Protocol (MCP) - which lets AI talk to your apps and databases - is getting an update to work at scale, switching to a "stateless" approach that handles millions of users way better.
Fixes a major developer headache. It makes building enterprise AI agents cheaper and easier - which means more of them, everywhere. We love to see it (or do we?).

Former crypto miner Hut 8 signed a 15-year, $9.8 billion deal to host AI infrastructure at its Texas campus - covering 352 megawatts of power, doubling a mystery client's space and sending its stock up 14%.
The AI race isn't just about chips anymore. Power and ready-to-use data centers are the real bottleneck, and whoever's got electricity is cashing in big.

Drug giant Bristol Myers Squibb is the first life sciences company to buy Nvidia's next-gen Vera Rubin AI system, using it to speed up drug discovery - claiming it already cuts test-medicine time by up to 30%.
Pharma is spending big to find cures faster and cheaper. The new system gives 10x more computing power for the same energy cost. Actual useful AI?? Bestie, it's happening.

YouTube is updating its rules so creators can no longer make money from generic, repetitive AI videos, emotionally manipulative content, or AI personas giving sensitive advice.
YouTube wants to protect ad revenue from lazy content farms. Use AI for real creativity, not spamming the platform for quick cash. Stop 👏 uploading 👏 slop.

Adobe is adding AI to its Project Indigo camera app that will actually critique your photos - tap a button for pro tips on framing and lighting, or let AI remove background clutter.
This moves AI photography away from faking images and toward being a personal photography coach. It's giving supportive-but-honest friend vibes.

State Attorneys General aren't waiting for federal AI laws. They're using existing consumer protection rules to police AI right now - already suing companies for using AI to falsely flag shoplifters, write fake reviews, or give bad medical advice.
Businesses can't hide behind "there's no AI law." If your AI tool harms consumers, state enforcers will act immediately with the tools they already have. No loophole for you.

The massive AI stock rally is showing fatigue, with global markets pivoting back to traditional sectors like oil amid global tensions. Big investors are taking profits from chipmakers and moving cash into beaten-down markets like Indonesia.
The market is questioning the high costs and delayed payoffs of AI infrastructure. Investors want real profits before they keep pumping tech stocks. The vibes are shifting...

The US AI standards chief stepped down after just three months, Hugging Face reported a security breach caused by an AI agent (??), Google's building a super-efficient "Frozen v2" chip, Nvidia launched a tiny open-source model for robots, and China finished a 1GW data center using only local chips.
The AI space is moving at breakneck speed on every front. From custom hardware to robot brains, the race is getting more intense and complex every single day. This. Is. A lot.