OpenAI’s purported Navier–Stokes proof faces questions about unpublished work
One post credits 10,000 coordinating agents with producing a proof; another questions OpenAI’s assertion that it did not refer to unpublished work by Buckmaster and colleagues.
TLDR
A post claims OpenAI used 10,000 coordinating agents, 88 hours and 130 billion output tokens to produce a Navier–Stokes proof, followed by 17 more hours to get it accepted by the Lean proof checker. It frames the effort as a parallel research organization rather than a chatbot answering one hard question. Another post questions whether OpenAI’s assertion that it made no reference to unpublished work by Buckmaster and colleagues is credible.

