OpenAI Claims Navier–Stokes Breakthrough, Sparking Research Dispute

OpenAI claims its artificial intelligence system solved the Navier-Stokes equations, a complex fluid dynamics problem that has eluded mathematicians for nearly 90 years. The AI system, reportedly more capable than OpenAI's GPT-6 Astra, achieved the breakthrough in just 88 hours by using thousands of concurrent agents that generated 130 billion output tokens.

The equation, which describes fluid movement, dates back to the 19th century. It was first mathematically formalized in 1934 by French mathematician Jean Leray, who posed the question of whether smooth fluid motion can develop a finite-time singularity, where speeds grow without bound, with smooth external forces.

The Clay Mathematics Institute offered a $1 million prize for a solution to the problem, but OpenAI's claim has sparked controversy. Independent mathematicians, including NYU's Tristan Buckmaster and Anthropic researcher Levent Alpöge, had been working on related fluid dynamics problems using OpenAI's Codex tool. Buckmaster claims OpenAI offered him sole authorship of the solution if Alpöge's name was removed from the work.

OpenAI denies this, stating they simply acknowledged the use of de-identified data derived from their products to improve their models.

The dispute highlights the significant resource disparity between OpenAI and independent researchers, with the latter estimating the effort would cost around $6 million at retail rates. This has raised concerns about the implications for the future of mathematical research and the value placed on human expertise in an era of advanced AI tools.

OpenAI's approach of using AI for both generating solutions and verifying them could serve as a model for independent checks on high-stakes AI output, emphasizing the need for human oversight and clear policies on the use of proprietary data in AI research.