OpenAI's New Model Solved Navier-Stokes in 88 Hours — Then a Credit Fight Broke Out
Don't Worry About the Vase (Zvi) · rss · 2026-09-13
Zvi dissects the week's biggest AI story: OpenAI's next internal model (stronger than Astra), eight days after starting training, cracked the Navier-Stokes finite-time blowup problem in 88 hours, with Astra then formalizing and verifying it in Lean over 17 more hours. OpenAI framed the move as informing the world about the pace of AI progress.
- Backdrop: Mathematicians Tristan Buckmaster and Levent Alpoge built on a blowup program by Diego Cordoba and Luis Martínez-Zoroa, using LLMs to extend it to smooth forcing and incompressible Euler. They rushed the papers out early — Buckmaster apologized for the write-ups, comparing them to AI slop, and publicly argued Martínez-Zoroa deserves a Fields Medal.
- The trigger: On Sept 1, after hearing a false rumor that two Millennium Problems were solved, OpenAI spent millions of dollars of inference — agents sent 4.9M messages and 300B output tokens overall (130B for Navier-Stokes), roughly $22M at public prices. roon claimed such agent swarms will cost "a buck fifty" within a year.
- The accusation: Buckmaster says OpenAI's Sebastien Bubeck misled him, then proposed either staggered announcements or sole authorship for Buckmaster with Alpoge (who works at Anthropic) removed — with pressure tactics including "Why would you ruin your career?" and "If you don't want me to be nice, then I don't have to be nice." Buckmaster declined and went public, noting he hasn't seen OpenAI's proof and isn't accusing data misuse. Levent highlighted OpenAI's admission it "cannot rule out" that de-identified usage data improved its models.
- Takeaway: OpenAI won't claim the $1M prize; the fight was always about credit, not money.
More from Models
- Ollama's jmorgan: small models now handle most conversational and reasoning use cases — ollama · 2026-09-14
- GLM Flash 5.3 keeps slipping into Chinese mid-conversation, user reports — DevDminGod · 2026-09-14
- AI Claims First Millennium Prize: Navier-Stokes Falls, OpenAI Model Leaps in Math — TheZvi · 2026-09-14
- If embeddings are so powerful, why is retrieval their only mainstream use? — ProposalOrganic1043 · 2026-09-14
- An OOD test worth watching: asking Gemini 4.1 Pro to paint surrealism with pure Python code — teortaxesTex · 2026-09-14
- Leaked GPT-6 Sol Output Impresses, But OpenAI Reportedly Has Stronger Bell Internally — VraserX · 2026-09-14