Noam Brown defends OpenAI's million-dollar Millennium Prize sprint as test-time compute costs collapse
anshulkundaje · x · 2026-09-09
The controversy over OpenAI's Navier–Stokes sprint deepens. @tylerjohnston alleges OpenAI heard rumors of a Millennium Prize breakthrough (reportedly from Anthropic, under external review), spent millions in a week-long sprint that may have scooped the authors, then considered offering Levant/Tristan co-authorship before learning they'd only solved the forced Euler problem — a key step, not the full milestone.
Noam Brown responds that while the result cost millions, o3 once needed $500,000 to score 87.5% on ARC-AGI 1 while Astra now scores higher for $20; IMO gold took enormous compute in 2025, but a $20/month ChatGPT subscription could do it for the 2026 IMO. He believes everyone will have an AI of this caliber within a year.
Related event: OpenAI Claims Navier-Stokes Breakthrough as Credit Dispute Simmers(160 posts)→
More from Companies & People
- FLORA opens applications for 5-week AI design Immersion Program, deadline Sept 18 — round · 2026-09-09
- OpenAI burned $22.5M in compute to beat NYU mathematician to Navier–Stokes proof, he says — SumitGup · 2026-09-09
- Joke: Anthropic reportedly planned its Navier-Stokes reveal right before IPO — IgorCarron · 2026-09-09
- GitLab CEO Sid Sijbrandij, 16 months cancer-free, launches 15 biotech startups in 18 months — juanbenet · 2026-09-09
- OpenAI claims internal AI with 10,000 agents solved 90-year-old Navier-Stokes problem — The Verge AI · 2026-09-09
- Documents show OpenAI and Google advising Pentagon on military AI strategy and risks — ambaonadventure · 2026-09-09