OpenAI's NS formalization: 10k agents, 88 hours, ~130B output tokens
nrehiew_ · x · 2026-09-09
nrehiew lays out the rough timeline of OpenAI's involvement in formalizing the Navier-Stokes (Millennium Prize) solution:
- Aug 15: Buckmaster and Alpoge solved forced Euler; formalized by Aug 22.
- Aug 28: OpenAI began (post-)training an internal math model that "looks insane."
- Sept 2: After hearing two Millennium Prize problems were solved, OpenAI attempted a batch of problems, not just NS.
- Sept 3: Buckmaster heard the rumor and reached out; around then OpenAI had a solution to unforced Euler.
- Sept 5: O(10,000) agents ran 88 hours, sent 2.7M messages and used 130B output tokens for formalization, done with Astra (not the internal model).
The timeline sketches the actual scale of OpenAI's internal math model plus massive agent orchestration in frontier math formalization.
Related event: 10,000 Agents, 88 Hours: Inside OpenAI's Navier-Stokes Formalization(3 posts)→
More from Models
- OpenAI claims its next-gen model agents solved the 90-year-old Navier-Stokes Millennium Prize Problem — _arohan_ · 2026-09-09
- OpenAI says 10,000 coordinating AI agents solved Navier-Stokes in 88 hours — CatAstro_Piyush · 2026-09-09
- inclusionAI open-sources vision model Ling-3.0-flash-VL with BF16 and FP8 weights — FellMentKE · 2026-09-09
- OpenAI May Already Be Pretraining GPT-6.5 That Beats Astra on Evals — willdepue · 2026-09-09
- After the drama, the real question: is OpenAI's 'internal model' already GPT-6.5 pretraining? — willdepue · 2026-09-09
- Ex-OpenAI researcher disputes Cohere CEO: rewritten ZDR data can't be used for training — willdepue · 2026-09-09