Daily AI Briefing

Anthropic’s trillion-dollar IPO puts AI demand to the test

Anthropic set its IPO range at $70–$78 a share, which works out to a valuation of up to $1.3 trillion, according to its updated filing. It could raise as much as $64 billion from public investors, plus a separate $26 billion share sale to Google, Nvidia and other big tech firms. All in, that’s up to $100 billion.

So this is the big test: will public investors buy that AI demand justifies what it costs? Anthropic’s pointing to an $81 billion revenue run rate. Meanwhile, OpenAI keeps losing money even as revenue climbs. And Amazon and Google are about to see some very nice paydays on their stakes.

• WSJ: Anthropic’s IPO Says A Lot About AI’s Future
• NY Times: Anthropic’s Giant IPO Will Tie It Closer to Nvidia
• TechCrunch: All 7 Anthropic founders will each cash out about $250M in IPO

Anthropic launches faster Fable 5.2 amid scrutiny of AI capabilities

Anthropic released Claude Fable 5.2 and Mythos 5.2, promising better coding, reasoning and vision, plus responses up to 30% faster. Fable’s input price drops to $8 per million tokens, and it’s broadly available with safeguards. Mythos stays restricted to sensitive research.

The wrinkle: the models sometimes suspect they’re being tested, TechCrunch reports. That makes it harder to know how they’ll actually behave out in the real world. A little unnerving. And in a fun aside, Simon Willison found Fable is more likely to clean up duplicated code when you tell it to treat the codebase as its own. Ownership matters, apparently.