OpenAI disproves 80-year-old hypothesis — and mathematicians back it

OpenAI claims that its reasoning model has refuted one of the classic geometric hypotheses that remained unresolved for over 80 years. This time, the result received support from the very mathematicians who previously debunked the company's previous controversial claims. This is a significant step in proving that AI can generate genuine new mathematical discoveries.

Which hypothesis did the reasoning model solve?

This time, we're talking about a geometric hypothesis that has remained open since 1946 — one of those classic questions in topology and geometry to which mathematicians have returned for many decades without success. This is not an entertaining puzzle, but a serious problem of interest to fundamental science. OpenAI used its reasoning model to analyze the problem. The company did not disclose the details of the mathematical aspects of the solution, but indicated that AI was able to find a logical and demonstrative path leading to the refutation of the hypothesis. For AI, this is non-trivial: it is necessary not simply to enumerate options, but to construct a chain of reasoning that is understandable and verifiable by humans.

Support from authoritative experts

Notably, the result received validation from leading mathematicians — the very same scientists who a year ago criticized OpenAI for making unwarranted bold claims. Then the company announced the solution to another problem, but upon careful examination, it turned out that the proof contained an error. OpenAI was forced to acknowledge the inaccuracy. Now the situation is different. Experts independently verified the logic and calculations of the reasoning model and confirmed their correctness. This is confirmation of the result by the scientific community, not just a PR move. For OpenAI, this reduces reputational risk and demonstrates the increased reliability of AI systems for solving abstract, creative mathematical problems.

What does this mean for science?

This achievement is important for several reasons:

• For the first time, an AI system independently solved a multi-decade-old open problem, which is recognized by leading mathematicians
• The reasoning model can be applied to other unsolved hypotheses in geometry, topology, and pure mathematics
• It is proven that AI is capable of generating genuinely new mathematical knowledge, not just copying known patterns
• Computer assistance transitions from the category of auxiliary tool to the status of active participant in discoveries

If the results are confirmed and published in leading journals, it will redefine the role of AI in fundamental science. From a tool for verifying already known ideas, AI transitions to a generator of new discoveries. Even if the reasoning model does not become a fully independent researcher, it can already now accelerate the process of mathematical search by proposing new paths of proof.