
OpenAI hits $1.25T, Anthropic's AI builds AI, and Nvidia's China mess
OpenAI keeps printing money, Anthropic is using AI to accelerate its own AI, and the US apparently approved Nvidia chip sales to an empty office in China.
Subscribe free →
OpenAI Is Now Worth More Than Most Countries' GDPs
OpenAI has reportedly raised another $11 billion, pushing its valuation to $1.25 trillion — up from $852 billion just eight months ago. The round hasn't closed yet but is expected to in coming weeks, and Bloomberg says OpenAI is still pursuing even more capital on top of this.
Why it matters: A company that didn't exist 10 years ago is now valued higher than ExxonMobil, JPMorgan, and Walmart — the AI money printer shows zero signs of slowing.

Anthropic Is Using AI to Build Better AI — Twice as Fast
A new report from AI safety org METR, based on an 18-month evaluation, found that Anthropic is using its own models to automate AI development. The result: its models are improving roughly twice as fast as they were a year ago, with no signs of deceleration.
Why it matters: This is the recursive self-improvement loop researchers have long warned about — and it's not theoretical anymore, it's happening inside a leading AI lab right now.

The US Approved Nvidia AI Chip Sales to a Literal Empty Office in China
The Commerce Department's vetting process for Chinese firms approved to handle Nvidia's restricted H200 chips is falling apart. A New York Times investigation found that "New Matrix," one of the approved firms, operates out of an empty gray shell of an office with wires dangling from the ceiling — no employees, no equipment, nothing.
Why it matters: If the export control regime meant to keep cutting-edge AI chips out of China can be gamed this easily, it's largely theater.

Google Swoops In on Crusoe's Troubled Wyoming Data Center
Crusoe Energy, the AI cloud startup that pivoted from flare-gas mining to data centers, has sold its troubled Wyoming data center project to Google. The project ran into serious development headwinds, underscoring how brutal it is to actually build massive AI infrastructure at speed and scale.
Why it matters: Even well-funded AI infrastructure startups are struggling to execute on gigawatt-scale data center ambitions — Google's willingness to absorb troubled projects shows how desperate hyperscalers are for capacity.

NYT Cooking's AI Assistant: Kitchen Savior or Disaster Waiting to Happen
The New York Times Cooking app now has a built-in AI assistant, and The Verge put it through its paces. It turned out to be genuinely useful for weeknight dinner improvisation — but gave dangerously wrong guidance on a baking recipe, the kind of confident hallucination that could ruin a birthday cake or worse.
Why it matters: AI cooking assistants are going mainstream, and the gap between "impressively helpful" and "confidently wrong" is exactly why stakes matter — savory dishes are forgiving, baking chemistry is not.

Hackers Hit Chainalysis With AI-Powered Attack, Stole All Customer Credentials
Blockchain analytics firm Chainalysis disclosed a "multifront cyberattack" that included AI agents and an on-path data interception. The attackers made off with credentials and session tokens for every single one of its customers — a catastrophic breach for a company whose entire business is tracking illicit crypto flows.
Why it matters: AI-augmented attacks are graduating from theoretical threat to real-world execution at scale, and when the victim is the firm that helps governments trace crypto crime, the downstream damage could be enormous.
Quick Hits
- →The midterms just flipped the political math for Big Tech — companies that spent two years cozying up to the White House now need to make friends with Democrats fast. TechCrunch
- →Elon Musk confirmed Tesla is "winding down" Cybertruck production next year to make room for the Cybercab — the truck is dead, announced via late-night tweet. The Verge
- →Apple says 100,000 apps are now optimized for the Duo form factor, with thousands more arriving daily. TechCrunch
- →Sequoia is shutting down its Scouts program, signaling that the peer-referral VC model has hit a wall. TechCrunch
AI TLDR