OpenAI claims Navier-Stokes solution found by 10,000 agents

The announcement comes amid a series of major AI product launches, including Meta's Muse agent and OpenAI's Image 2.5. However, the Navier-Stokes claim dominates due to its symbolic weight.

Ethan Knight's disclosure describes a research system architecture focused on scale: approximately 10,000 agents, trained over a year using multi-agent RL, and run with large amounts of unstructured parallel test-time compute. This is less about fluid mechanics and more about a new approach to frontier research.

If the claims hold, it may indicate a shift from training-time scaling to inference-time scaling. However, the lack of artifacts raises questions about reproducibility, and the exact role of humans versus models in the proof remains unclear, highlighting the importance of provenance in such a high-stakes mathematical claim.