The Navier Stokes existence and smoothness problem was addressed by an internal OpenAI model through 88 hours of coordinated computation. This effort utilized 10,000 AI agents that exchanged 2.7 million messages and generated 130 billion output tokens to create a 166 page construction. GPT-6 Astra then spent 17 hours verifying the resulting argument in Lean, a proof assistant that checks logical steps against a formal set of axioms.
These equations, which describe fluid motion for aircraft design and weather forecasting, have remained unsolved for 90 years. A proposed solution to the Millennium Prize problem requires a qualifying publication and two years of scrutiny before the Clay Mathematics Institute can grant a prize OpenAI does not intend to claim. Technical analysis of company posts suggests the model may have been a resumed reinforcement learning run started August 28 rather than a fresh training effort.
Key sources
- SOURCE@corvusxbt“roughly 10,000 AI agents spent 88 hours producing a proposed solution to the Navier–Stokes existence and smoothness problem”x.com
- SUPPORT@deredleritt3r“on August 28th, we restarted the large frontier RL run that was previously paused”x.com
- SUPPORT@thestalwart“didn’t even take OpenAI two weeks to train a model that’s a step change above Astra”x.com
- SUPPORT@teortaxestex“resumed RL on a base that's been ready for a while”x.com
- SUPPORT@teortaxestex“final RL can take arbitrarily little time in the current MOPD regime”x.com
- SOURCEmarketbrief.now
- SOURCE@a16z“new knowledge created by AI and it unlocking a whole wave of scientific discovery, medicines, all of those things”x.com
- SUPPORT@ctrlsecint“produced a proof resolving the 90-year-old Navier–Stokes problem, one of mathematics’ seven Millennium Prize Problems”x.com