OpenAI reports that a system of 10,000 artificial intelligence agents solved the Navier-Stokes existence and smoothness problem in 88 hours. This mathematical puzzle has remained one of the most difficult challenges in fluid dynamics for 90 years.
The Technical Approach
The company claims its agents worked in coordination to reach the result. These agents operated within a framework that granted them access to a cached version of the internet and provided the capacity to run code. OpenAI organized the agents into groups of varying sizes, with the largest coordination effort involving roughly 10,000 concurrent agents. The process began on September 1 and concluded on September 5.
OpenAI states that its researchers acted only as supervisors for the agent-driven effort. The agents communicated within their specific sub-groups to test proofs and iterate on potential solutions. This event marks a significant application of autonomous agents in theoretical mathematics, moving beyond standard chatbot responses to deep research-level problem solving.
Questions Regarding Methodology
Tristan Buckmaster, a professor at New York University, challenged the findings. Buckmaster notes that he had been working on the same problem with Levent Alpöge, a mathematician affiliated with Anthropic, in a personal collaboration. He received reports that information regarding their progress was shared with OpenAI researchers.
Buckmaster expressed concern that the tech company’s approach mirrored his own work too closely to be considered independent research. He raised the possibility that the model may have been trained on or gained access to the data from his sessions in OpenAI’s tools. While he admits he has not reviewed the specific proof provided by the company, he questions how the model arrived at the resolution in such a short window of time.
OpenAI countered these claims in its public statement. The company asserts that it began the attempt on September 1 after hearing rumors about progress on the puzzle. They maintain that no internal user data was accessed to generate the solution, though they acknowledged it is difficult to rule out that de-identified data from usage logs could have influenced the underlying models.
Broader Implications for Research
The Clay Mathematics Institute established the Millennium Prize Problems in 2000 to highlight the most difficult challenges in mathematics. Solving the Navier-Stokes problem carries a million-dollar prize. The Institute has not yet issued a comment on the validity of OpenAI’s work.
This development raises questions about how AI research should be verified when models arrive at solutions that human experts find puzzling or potentially derivative. If the math stands up to peer review, it will represent a major milestone for artificial intelligence in scientific discovery. If it fails, the episode will likely fuel debates about data sovereignty and the transparency of proprietary models in academic fields.