Full Breakdown
OpenAI’s AI Claim to Solve Navier-Stokes Problem Sparks Academic Controversy
By Drooid · · How we work
Core Event – AI Claim and Allegations
On September 8, 2026 OpenAI announced that a fleet of roughly 10,000 internal AI agents had produced a 166-page proof for the Navier-Stokes existence-and-smoothness problem, one of the seven Clay Mathematics Institute Millennium Prize Problems that carry a $1 million prize. The proof was formally verified in the Lean proof assistant after about 17 hours of additional checking.
Mathematician Tristan Buckmaster (NYU) and Anthropic researcher Levent Alpöge contend that OpenAI learned of their year-long work on the same problem, accelerated its effort, and threatened them not to publicize their research.
Background & Context
The Navier-Stokes equations describe fluid dynamics and have resisted proof of existence and smoothness since the Clay Institute listed them in 2000. In the past year AI systems have contributed to proofs of Erdos problems, verified Fields-Medal-winning results, and advanced dozens of other open questions, prompting debate about AI’s role in mathematics.
Key Figures & Groups
- Tristan Buckmaster – NYU professor, alleges copying.
- Levent Alpöge – Anthropic researcher, collaborator.
- Michael Harris – Columbia professor, called the behavior “predatory.”
- Jacob Tsimerman – University of Toronto professor, warned of a “golden age.”
- Sam Altman – OpenAI CEO.
- Greg Brockman – OpenAI President.
- Peter Scholze – Max Planck Institute co-director, dismissed the claim as PR.
Timeline
- September 8, 2026 – OpenAI announces Navier-Stokes solution; later amends statement after investigation.
- August 1 – OpenAI releases ten results claiming advances on long-standing open problems.
- September 1 – OpenAI learns two Millennium problems have been solved and redirects a larger model to the remaining five.
Data & Statistics
- 10,000 AI agents deployed.
- 166-page proof; 130 billion output tokens consumed.
- $15 million spent on compute.
- Solution emerged after ? 88 hours of agent runs, followed by ? 17 hours of Lean verification.
Official Statements & Responses
- Michael Harris warned of “predatory behavior” and called for a regulatory framework to protect mathematicians.
- Jacob Tsimerman argued that AI could usher in a “golden age” if new mathematics serves society.
Criticism & Opposition
- Buckmaster accused OpenAI of a monopoly that left him “stuck.”
- Scholze emphasized that the purpose of Millennium problems is to generate fresh human insight, which a machine-produced proof does not provide.
Conflicting Reports & Gaps
OpenAI’s claim remains “pending independent verification” according to multiple outlets; no peer-reviewed publication has yet confirmed the proof. Wired reports the investigation’s amendment, while LiveScience highlights the “firestorm” of debate, indicating differing assessments of credibility.
Verbatim Quotes
- “To be honest, whenever anyone talks about 10 years from now, or even five years from now … it just seems unfathomably far away,” — Fields Medalist Jacob Tsimerman
What’s Next
OpenAI has not announced a timeline for external peer review. The controversy continues at venues such as the Heidelberg Laureate Forum, where mathematicians are debating potential regulatory measures and the long-term impact of AI-generated mathematics.