Image Source: OpenAI
It took OpenAI’s artificial intelligence system just 88 hours to produce what the company describes as a solution to the 90-year-old Navier–Stokes problem. The claim has triggered excitement across the technology and mathematics communities, but experts are urging caution until the proposed proof is independently reviewed.
OpenAI, the company behind ChatGPT, said it used approximately 10,000 coordinating agents powered by an internal AI model. The agents worked in groups, communicated with one another, accessed a cached version of the internet and ran computer code while examining the complex fluid dynamics equation.
A Powerful AI Claim About the Navier–Stokes Problem
According to OpenAI, its research effort began on September 1. The company said the agents reached their resolution on September 5, roughly 88 hours after the first systems were launched.
Navier–Stokes describes how fluids move. Its equations are used in areas including aerodynamics, weather forecasting, oceanography and engineering. Despite their importance, mathematicians have not been able to prove whether smooth, physically meaningful solutions always exist in three dimensions.
The problem is one of seven Millennium Prize Problems established by the Clay Mathematics Institute in 2000. Each problem carries a $1 million prize for a valid solution, making the Navier–Stokes challenge one of the most prestigious unanswered questions in modern mathematics.
- OpenAI said about 10,000 AI agents worked on the challenge.
- The agents were organized into communicating groups.
- The reported resolution was reached on September 5.
- The Clay Mathematics Institute has not endorsed the proposed solution.
Critical Questions From a Leading Mathematician
The announcement quickly drew scrutiny from Tristan Buckmaster, a professor of mathematics at New York University. Buckmaster said he had been working with Levent Alpöge, a mathematician associated with OpenAI rival Anthropic, on mathematical problems that included Navier–Stokes.
Buckmaster said the approach described by OpenAI appeared similar to work he and Alpöge had been pursuing. He questioned whether a model could independently arrive at that direction in only a few days after receiving the problem statement.
However, Buckmaster emphasized that he had not seen OpenAI’s proof and was not claiming that the company had improperly accessed his research. He said he did not know exactly what the model had done or whether any of his data had been used.
His concerns focus partly on whether OpenAI systems may have been trained on, or exposed to, sessions conducted through the company’s Codex tools. The mathematicians had used several large language models during their own work, adding another layer to the debate over data privacy and AI-assisted research.
OpenAI Denies Accessing Private Research
OpenAI said its team began investigating the problem after hearing a rumor about progress that it later learned involved Buckmaster and Alpöge. The company stated that neither its researchers nor its agents saw the pair’s work before it was publicly released.
The company also said no specific user data was accessed to solve the problem. At the same time, OpenAI acknowledged that it could not completely rule out the possibility that de-identified information derived from product usage had helped improve its models.
That distinction is likely to remain important as AI systems become more capable of conducting research, writing code and coordinating complex tasks. A system can produce an apparently original result while still raising difficult questions about training data, data provenance and the boundaries of independent discovery.
Why Verification Will Be Historic
A genuine solution to Navier–Stokes would represent a historic achievement in mathematics. It could also demonstrate that multi-agent AI systems are capable of contributing to highly specialized research beyond routine calculations or language-based tasks.
Still, mathematical breakthroughs require more than a company announcement. Experts must inspect the complete proof, test every logical step and determine whether the argument satisfies the exact conditions set by the Clay Mathematics Institute.
The institute has not commented on OpenAI’s proposed solution. Until independent mathematicians review and accept the work, the claim should be viewed as a significant AI-generated proposal rather than a confirmed solution.
The development nevertheless highlights the rapid evolution of artificial intelligence. OpenAI’s use of thousands of coordinated agents suggests a future in which AI research teams can divide problems, compare approaches and refine technical arguments at a scale that would be difficult for human researchers working alone.
What is the Navier–Stokes problem?
It is a major mathematics challenge concerning the equations that describe fluid motion. The unresolved question involves whether smooth solutions always exist under specific three-dimensional conditions.
Did OpenAI officially solve the problem?
OpenAI said it was sharing a solution, but the claim has not been independently verified. The Clay Mathematics Institute has not endorsed the proposed proof.
How many AI agents did OpenAI use?
OpenAI said the group working on the Navier–Stokes problem involved approximately 10,000 concurrent coordinating agents.
Who questioned OpenAI’s claim?
Tristan Buckmaster, a New York University mathematics professor, raised questions about the proof’s development and whether research data could have influenced the result.
