OpenAI agents crack Navier-Stokes equations, community debates data mining

Agents solve a million dollar problem

OpenAI announced a breakthrough in mathematics as its group of agents generated a solution to the Navier-Stokes equations. This issue is one of the seven Millennium Prize Problems, carrying a $1 million reward for its resolution.

According to the announcement, the proof was created using a next-generation model significantly more capable than the recently introduced GPT-6 Astra. This is not an isolated experiment, but a demonstration of the power of fully autonomous systems at the highest academic level.

Success opens a debate on the origin of training data

While the tech world celebrates cracking a decades old problem, analysts point to the flip side of the success. Simon Willison argues in his commentary that this case clearly highlights the still confusing question of what exactly using user data to improve model performance means.

If a model can synthesize a breakthrough proof from a vast amount of mathematical papers written by humans, the discussion shifts to who actually owns the credit and the potential financial reward.

Mathematical conferences face a wave of synthetic proofs

The ability of agents to solve problems at this level means academic institutions must quickly rethink the peer review process. If systems can produce proofs more complex than human work, the current model of manual verification will hit a capacity limit.

The risk is that without understanding the internal logic of the agent, science becomes dependent on a black box. Experts can verify the result, but the process by which the model reached it often remains opaque.

Changing the rules for academic publishing

The real impact will not be just the mathematical proof itself, but the reaction of the scientific community. It will now be decided whether and how synthetic proofs will be accepted in prestigious journals. The deciding factor will be whether institutions like the Clay Mathematics Institute officially recognize these systems as co-creators.

Lilith's verdict

It is ironic that solving one of the hardest mathematical problems will not cause a revolution in the equation itself, but in how we explain copyright over scientific data to lawyers.