📷 Image: Times of India (via NewsData)
Technology
OpenAI claims 10,000 AI agents solve 90‑year‑old Navier‑Stokes problem in 88 hours
✍️ Times of India
🗓 09 Sep 2026, 11:38 AM
👁 28
OpenAI says an unreleased model, using about 10,000 AI agents, resolved the Navier‑Stokes existence and smoothness problem within 88 hours, though a NYU professor disputes the claim.
OpenAI announced that an internal, unreleased AI model succeeded in solving the Navier‑Stokes existence and smoothness problem, one of the seven Millennium Prize Problems set by the Clay Mathematics Institute. The company said the breakthrough came after deploying roughly 10,000 autonomous AI agents that explored candidate proofs over an 88‑hour run.
If the result holds up, it would qualify for the $1 million prize attached to the problem, but the claim has drawn immediate skepticism. Tristan Buckmaster, a professor of applied mathematics at New York University, noted that OpenAI’s approach mirrors a line of reasoning he had been developing in unpublished drafts that later informed OpenAI’s Codex system, suggesting the AI may have effectively "cheated" by using his ideas without attribution.
OpenAI responded that it does not use private user data to train its models, although it cannot entirely rule out that anonymized training material contributed to the outcome. The firm has not released the detailed proof or the underlying code, and independent verification by the mathematics community is still pending.
The episode highlights the growing debate over the role of large‑scale AI in mathematical research and raises ethical questions about the use of proprietary or unpublished academic work in training data.
If the result holds up, it would qualify for the $1 million prize attached to the problem, but the claim has drawn immediate skepticism. Tristan Buckmaster, a professor of applied mathematics at New York University, noted that OpenAI’s approach mirrors a line of reasoning he had been developing in unpublished drafts that later informed OpenAI’s Codex system, suggesting the AI may have effectively "cheated" by using his ideas without attribution.
OpenAI responded that it does not use private user data to train its models, although it cannot entirely rule out that anonymized training material contributed to the outcome. The firm has not released the detailed proof or the underlying code, and independent verification by the mathematics community is still pending.
The episode highlights the growing debate over the role of large‑scale AI in mathematical research and raises ethical questions about the use of proprietary or unpublished academic work in training data.