OpenAI's Navier-Stokes solution claim overshadowed by academic objections

Summary

OpenAI announced a formalized solution to the Navier-Stokes equations, a 200-year-old Clay Millennium Prize problem, generated by AI and formalized in Lean. However, this claim is overshadowed by allegations that the company accelerated its efforts after learning of parallel work by mathematicians Buckmaster and Alpöge, while attempting to control credit attribution.

Why it matters

If verified, this would be the first AI-generated solution to a Millennium Prize problem, marking a turning point for both mathematics and AI capabilities. More urgently, the controversy exposes a governance gap. When labs can deploy over 10,000 agents at a cost of millions of dollars, traditional academic priority norms become fragile. Additionally, the line between 'inspiration' and 'adaptation' via anonymized training data is dangerously blurred.

Highlights

• OpenAI deployed up to 10,000 AI agents for over 50 hours to generate a proof formalized in Lean, with the process costing 'millions of dollars.'
• OpenAI's own blog admits it does not 'rule out' that anonymized usage data derived from competitors' Codex sessions improved its models.
• Buckmaster claims OpenAI suggested publishing the solution without naming Alpöge. Bubeck publicly denied this allegation.

Critical take

OpenAI's layered defense—claiming they 'did not see the work' while stating they do not 'rule out' benefits from anonymized data—does not amount to an apology. This stance effectively leaves the central question (did prompts or partial proofs leak into training data?) unanswered. Attempting to resolve the dispute through social media posts rather than peer review or independent audit also undermines the rigor that the Navier-Stokes solution should represent.