Full Breakdown
OpenAI Claims AI Solves Navier–Stokes Problem Amid Credit Dispute
9/9/2026, 7:47:43 PM
Core Event
On September 8, 2026 OpenAI announced that an internal AI system produced a proof for the Navier–Stokes existence and smoothness problem, one of the seven Millennium Prize Problems. The result was generated by roughly 10,000 concurrent AI agents over 88 hours, followed by a 17-hour formal verification in the Lean proof assistant. OpenAI said it does not intend to claim the $1 million prize.
Background & Context
The Navier–Stokes equations model fluid motion and have been a central open question since the May 24 2000 designation by the Clay Mathematics Institute (CMI). Only the Poincaré Conjecture has been officially solved. Recent AI-assisted breakthroughs on other longstanding conjectures have shown that large language models can contribute to high-level mathematics.
Timeline
- August 28, 2026 – OpenAI began training a new internal model described as “significantly more capable” than GPT-6 Astra.
- September 1, 2026 – OpenAI launched a multi-agent effort on all open Millennium problems.
- September 5, 2026 – The agent swarm produced a Navier–Stokes resolution (88 hours after launch).
- September 6, 2026 – Lean formalization completed; OpenAI reached out to external researchers.
- September 7, 2026 – NYU professor Tristan Buckmaster and Anthropic researcher Levent Alpöge posted related blow-up results for Euler.
- September 8, 2026 – OpenAI issued its public announcement.
Data & Statistics
- ? 10,000 agents operated concurrently.
- ? 2.7 million messages exchanged; ? 130 billion output tokens devoted to Navier–Stokes.
- Compute cost estimated in the millions of dollars.
- Lean verification required an additional 17 hours.
Official Statements & Responses
OpenAI described the proof as showing that an initially smooth fluid at rest, under a smooth external force, can develop a singularity in finite time, satisfying statements C and D of the CMI formulation. The company reiterated it will not claim the prize.
“The process of evaluation is deliberately unhurried, and we shall ensure that it is absolutely rigorous,” — Martin Bridson, president of the Clay Mathematics Institute, to *New Scientist*.
Criticism & Opposition
Terence Tao warned that “It’s a little like going to watch a movie and jumping straight from the first ten minutes to the last ten minutes; technically, all the plot lines are resolved, but most of the value of the experience was lost.”
Buckmaster clarified that he has not seen OpenAI’s proof and made no accusation, stating “I do not know whether our data was used.”
Critics note that the forced-fluid formulation differs from the unforced version traditionally required for the official prize, raising questions about whether the result meets the full Millennium criteria.
On-the-Ground Reports
Tristan Buckmaster, a mathematics professor at New York University, reported that he and Alpöge had used OpenAI’s Codex and Anthropic’s Claude for a year-long collaboration, achieving blow-up results on August 15, 2026 and Lean verification on August 22, 2026. After learning of their progress, OpenAI contacted Buckmaster, offered joint-publication options, and later announced its own proof without having seen the pair’s drafts.
Conflicting Reports & Gaps
- OpenAI claims the proof resolves the problem under the forced formulations (C and D). The Clay Institute’s official statement permits a forced solution, but the mathematical community has not yet accepted it as a complete resolution.
- The institute still lists Navier–Stokes as unsolved, indicating that peer-reviewed publication and a two-year acceptance period are pending.
- OpenAI’s statement that no specific user data was accessed conflicts with Buckmaster’s unanswered question about possible use of Codex session data, leaving the provenance of the model’s knowledge unclear.
