A new Latent Space newsletter reports that OpenAI has claimed a significant mathematical result: a singularity in the Navier-Stokes equations, which describe fluid motion. The work was done using a system called Astra-next, which coordinated roughly 10,000 agents and consumed 130 billion tokens over 88 hours. The cost is estimated at more than $40 million, making it one of the most expensive single AI computations disclosed to date.
The Navier-Stokes existence and smoothness problem is one of the seven Millennium Prize Problems, each carrying a $1 million award. Only one has been solved so far (the Poincaré conjecture). If verified, this would be the second. The newsletter notes that this announcement overshadowed other major AI news of the day, including large funding rounds for Cognition and Mistral, as well as new product releases from Meta and OpenAI itself.
The source does not provide independent verification or peer review details, so the claim remains a report rather than a confirmed result. Still, the scale of the computation—and the potential prize—marks a notable moment in AI-driven mathematics, even if the outcome is not yet settled.