OpenAI previously released solutions to 10 open math problems on GitHub

OpenAI is preparing to publish hundreds of AI-generated mathematical solutions to GitHub in a single release, capping off what has become a remarkably consequential stretch for the company’s work in formal mathematics.

The planned upload follows an announcement on August 28, 2026, in which OpenAI disclosed that it had begun training an AI model capable of solving the Navier-Stokes Millennium Prize problem alongside more than 100 other longstanding open mathematical problems. That is not a minor footnote: the Navier-Stokes problem is one of the seven Millennium Prize Problems designated by the Clay Mathematics Institute, each carrying a prize for a correct solution and representing some of the deepest unsolved questions in mathematics.

What OpenAI is releasing

The forthcoming GitHub release is expected to include formal proofs written in Lean 4, a proof assistant language that allows mathematical arguments to be verified by computer.

This follows an earlier milestone in August 2026, when OpenAI published solutions to 10 open mathematical problems on GitHub, also accompanied by formal Lean 4 proofs. The upcoming release would represent a much larger batch, described as a significant dump of results.

OpenAI spokesperson Lindsay McCallum confirmed that the company is consulting its Advisory Group on Mathematics and Artificial Intelligence to handle disclosure responsibly. No confirmed release time has been set as of this writing.

The math community has concerns

Not everyone in the mathematical world is celebrating. Some mathematicians have raised pointed concerns about the pace at which AI firms are generating and publishing proofs, with critics drawing sharp comparisons to competitive behavior among leading AI companies. One comparison that has circulated describes the dynamic as resembling mobster behavior, a phrase that captures the anxiety around who gets credit, how quickly results are verified, and whether the traditional peer-review process can keep up.

The broader question the mathematical community is now grappling with is what credit attribution looks like when a machine produces the proof. If an AI solves the Navier-Stokes problem, does anyone collect the Millennium Prize? Does the team that trained the model? The researchers who formalized the problem? These are not hypothetical edge cases anymore.

Why this release is worth watching

OpenAI’s decision to route these results through GitHub rather than a traditional academic journal is itself a signal. GitHub publication is fast, public, and version-controlled, but it bypasses the editorial and peer-review infrastructure that mathematics has relied on for validation.

Whether the AI’s solution is complete, partial, or conditional on assumptions that narrow its scope is precisely the kind of detail the forthcoming GitHub release will need to make clear. OpenAI’s engagement with its advisory group suggests the company understands that, even if the timeline for release remains open.