OpenAI says a large group of agents produced a 166-page proof addressing the Navier-Stokes existence and smoothness problem, a decades-old question about equations used to describe fluid flow. The result remains under peer review, and experts have not yet confirmed that the argument is correct.
The debate is not only about the final yes-or-no answer. Mathematicians build proofs to expose reusable ideas, connect them to earlier work, and make every step understandable to peers. Critics say the agent-generated document does not clearly explain how it reached the result or reliably situate the argument in the literature, making verification and learning harder even if the proof ultimately holds.
Navier-Stokes is also a case where the pure mathematical question differs from everyday engineering. Solving the abstract problem would not directly produce a better aircraft wing, but the effort around difficult conjectures often creates concepts that later find unexpected uses. OpenAI’s approach may demonstrate that many agents can search an enormous proof space, yet scale is not a substitute for exposition. Until specialists complete review, the claimed solution should be treated as an unverified research result and a test of how mathematical credit, transparency, and understanding will work in AI-assisted discovery.