Terence Tao has pushed back on the fast-moving claim that Claude solved the Navier-Stokes equations, saying his recent remarks were only part of a hypothetical AI research scenario and that he knows of no significant new development on the problem.
The clarification landed after a rumor spread rapidly on social media and drew more than 2.5 million views. Andrew Curran had posted a prediction that Anthropic had solved a Millennium Prize Problem tied to the Navier-Stokes equations, saying the results were under expert review and could be disclosed before Anthropic's IPO. No papers, proof documents, or expert review materials were publicly available to back it up.
Tao's Sept. 3 post had already framed the Navier-Stokes equations as an example inside a future research workflow, one in which autonomous AI systems would keep trying mathematical constructions, test failures, adjust course, verify results, and eventually generate a candidate proof. He said such a proof could then be checked formally with systems like Lean. That is the kind of structure mathematicians mean when they talk about machine-assisted proof: not a screenshot or a claim, but a candidate argument that survives formal verification step by step.
The friction is in the gap between that hypothetical process and the viral leap that followed. Tao later said that his comments belonged to a scenario he considers increasingly plausible given the pace of AI progress, but not to any known breakthrough in the Navier-Stokes existence and smoothness problem. The Clay Mathematics Institute still lists the question as unsolved, with the $1 million Millennium Prize attached to whether smooth initial conditions in the three-dimensional incompressible case stay smooth forever or break down in finite time.
For now, that leaves the rumor where it started: loud, specific, and unproved. The next real marker would be a public proof, formal verification, or expert review material that has not appeared.
