OpenAI Claims 10,000 Agents Solved Navier-Stokes in 88 Hours
OpenAI claims an in-training internal model (reportedly stronger than the commercial GPT-6 Astra, per @r0ck3t23) used 10,000 cooperating AI agents over 88 hours to close two key bounds of the Navier-Stokes existence and smoothness problem, with frontier-evaluation-grade safety monitoring and isolation throughout. The paper is open on OpenAI's site with Lean formal verification per @AGI Hunt; Quanta Magazine reports the agents found "singularities" in the equation, potentially marking a major shift in mathematical research. Ben Goertzel estimates $22M in compute (@AGI Hunt says millions—both figures circulate). This is the first large-scale AI assault on a top-tier math problem, but currently rests on OpenAI's claims alone. Two Minute Papers has published a video explainer.
Confirmed
- Around Sept 8, OpenAI said the internal model solved part of the Navier-Stokes proof in 88 hours with 10,000 agents, claiming step-change gains on many benchmarks; the paper is open on its site.
- Per @AGI Hunt, the proof carries Lean formal verification; compute cost in the millions of dollars—equivalent to a brilliant person working 100 years nonstop.
- Historical context: Leray proved weak solutions in 1934; smoothness stayed open 90 years; the Clay Mathematics Institute listed it in 2000 as one of seven Millennium Prize Problems worth $1M each. Navier-Stokes governs airflow over wings, blood flow, weather and oceans.
- @bengoertzel assesses OpenAI spent $22M of compute.
Timeline
- Per @haider1: RL training reportedly restarted Aug 28; on Sept 1 the model was tasked with a Millennium problem—only 4 days into training; a better checkpoint a day or two later produced the Navier-Stokes solution before initial training even finished.
Unconfirmed
- The internal model's capabilities, gap versus commercial models, and whether the proof survives peer review and earns the Clay prize all rest on OpenAI's claims pending independent verification; @AlexTensor relayed it as unverified "sci-fi news" predating the paper's release.
- Per @eyishazyer, discussion shifted within hours from math to credit-attribution disputes, with no settled conclusions.
Why it matters
- @kacoef notes the news shares theoretical underpinnings with the ZrTe₅ quantum experiment, suggesting reusable AI math capability across frontier sciences.
- @robleclerc, after a Grok review, finds the result "unsurprising as a math endpoint": Córdoba–Martínez–Zoroa had established the scale-by-scale blow-up plus smooth forcing program, and prior work (e.g., Alpög) paved the way—AI succeeded standing on giants' shoulders rather than creating from nothing.
2026-09-09 ~ 2026-09-11 · 13 related posts
- Episode 1: Rumors claim GPT-6 cracks Navier-Stokes, credibility doubtful(2026-09-05, 2 posts)
- Episode 2: OpenAI Claims 10,000 Agents Solved Navier-Stokes in 88 Hours(2026-09-09, 13 posts)
- Episode 3: Existence Proof Effect: Rumor of Solvability Drove OpenAI's Navier-Stokes Attempt(2026-09-09, 3 posts)
- Episode 4: Mathematicians Accuse OpenAI of Racing to Claim Their Unfinished Proof(2026-09-09, 6 posts)
- Episode 5: Anthropic Faces Ethics Backlash Over Math Proof and Scooped Collaboration(2026-09-09, 2 posts)
- Episode 6: Rumors swirl that OpenAI nears Hodge and BSD conjectures as Millennium Prize problems fall(2026-09-10, 27 posts)
- Episode 7: OpenAI Claims Navier-Stokes Solved, Progress on a Second Millennium Problem(2026-09-10, 21 posts)
- Episode 8: Mathematician accuses OpenAI of training Astra on his unpublished work; OpenAI denies(2026-09-10, 9 posts)
- Episode 9: Navier–Stokes Authorship Dispute Fuels Debate Over AI Labs' Scooping Culture(2026-09-10, 6 posts)
- Episode 10: OpenAI denies Codex prompts influenced training as "stolen chat data" claims unravel(2026-09-11, 8 posts)
- Episode 11: OpenAI's Millennium Prize Claim Mired in Authorship and Scooping Dispute(2026-09-11, 10 posts)
- Episode 12: AI Circles Speculate Millennium Prize Problems May Fall to AI(2026-09-11, 2 posts)
- Episode 13: Math Community Splits Over AI's Role After OpenAI's Millennium Prize Claim(2026-09-11, 8 posts)
- Episode 14: Rumor: OpenAI Aims Internal Model at Riemann Hypothesis and P vs NP(2026-09-11, 7 posts)
- Episode 15: OpenAI's Navier-Stokes Proof Passes Rechecks but Faces Scrutiny(2026-09-11, 2 posts)
- Episode 16: OpenAI's 10,000-Agent Navier-Stokes Claim Sparks Math Community Credit War(2026-09-12, 18 posts)
- Episode 17: Misread as Anti-AI, Terence Tao Sparks Debate Over the Value of Mathematics(2026-09-12, 18 posts)
- Episode 18: Fields Medalists' Open Letter on AI Sparks Cross-Disciplinary Debate(2026-09-12, 6 posts)
- Episode 19: OpenAI Claims AI Agent Solved the Navier–Stokes Blow-up Problem(2026-09-13, 8 posts)
- Episode 20: OpenAI Claims 10,000 Agents Solved Navier-Stokes Problem in 88 Hours(2026-09-14, 11 posts)
Primary sources
- Grok on AI cracking Navier-Stokes: not surprising, AI stood on giants' shoulders — robleclerc · 2026-09-09
- [source] OpenAI's Navier-Stokes run used a model just 4 days into training, still unfinished — haider1 · 2026-09-09
- OpenAI's new model reportedly went from RL restart to Navier–Stokes solution in two weeks — haider1 · 2026-09-10
- OpenAI's 10,000 parallel agents crack the 90-year Navier-Stokes smoothness problem in 88 hours — r0ck3t23 · 2026-09-10
- Report: OpenAI internal model claims to crack 200-year-old Navier-Stokes problem — AlexTensor · 2026-09-10
- [source] OpenAI internal model solves Navier-Stokes in 88 hours using ~10,000 coordinating agents — TheMoonMidas · 2026-09-10
- Report: OpenAI's 10,000 Agents Closed Navier-Stokes in 88 Hours, Verified in Lean — r0ck3t23 · 2026-09-10
- OpenAI Publishes a Navier-Stokes Solution, Teaming Humans with AI on a Millennium Prize Problem — Two Minute Papers · 2026-09-10
- OpenAI's Navier-Stokes proof push and ZrTe5 quantum limit experiments meet at electron hydrodynamics — kacoef · 2026-09-10
- Goertzel on OpenAI's Navier-Stokes solution: $22M compute for a $1M prize — bengoertzel · 2026-09-10
- [source] OpenAI ran 10,000 agents for 88 hours to prove Navier-Stokes — and it cost millions — AGI Hunt · 2026-09-10
- OpenAI Internal Model Ran 10K Agents for 88 Hours, Claims Partial Navier-Stokes Proof — eyishazyer · 2026-09-11
- 10,000 OpenAI agents crack Navier-Stokes, resolving a Millennium Prize Problem — fortnow · 2026-09-11