OpenAI says it solved an 80-year-old math problem, for real
▼ Summary
– OpenAI claims its new reasoning model produced an original mathematical proof disproving a famous unsolved geometry conjecture posed by Paul Erdős in 1946.
– Seven months ago, OpenAI’s former VP claimed GPT-5 solved 10 previously unsolved Erdős problems, but those solutions already existed in the literature, leading to criticism and retraction.
– OpenAI now published companion remarks from mathematicians supporting the new disproof, avoiding a repeat of the earlier misrepresentation.
– The company says this is the first time AI has autonomously solved a prominent open problem central to a field of mathematics.
– OpenAI states this demonstrates AI’s ability to hold long reasoning chains and connect ideas across fields, with implications for biology, physics, engineering, and medicine.
OpenAI has announced that its latest reasoning model successfully generated an original mathematical proof, disproving a long-standing unsolved conjecture in geometry first introduced by mathematician Paul Erdős in 1946.
If this declaration sounds familiar, that is because it echoes a previous claim that did not hold up. Seven months ago, Kevin Weil, then an OpenAI vice president, posted on X that “GPT-5 found solutions to 10 (!) previously unsolved Erdős problems and made progress on 11 others.” In reality, GPT-5 had not solved those problems at all; it had simply rediscovered solutions already documented in academic literature.
The backlash was swift. Rivals including Yann LeCun and Google DeepMind CEO Demis Hassabis publicly mocked the claim, and Weil quickly deleted his post. This time, however, OpenAI appears to have taken greater care. Alongside the latest announcement, the company published supporting statements from mathematicians such as Noga Alon, Melanie Wood, and Thomas Bloom. Bloom, who runs the Erdos Problems website, had previously described Weil’s earlier post as “a dramatic misrepresentation.”
“For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids,” OpenAI stated on X. “An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.”
The company further asserted that this represents “the first time AI has autonomously solved a prominent open problem central to a field of mathematics.” Notably, the proof came from a general-purpose reasoning model, not a system custom-built for mathematical problem-solving or even for this specific challenge.
According to OpenAI, this breakthrough is significant because it demonstrates that AI systems can now sustain long, complex chains of reasoning and connect ideas across different disciplines in ways researchers may not have previously considered. That capability has potential implications for fields like biology, physics, engineering, and medicine.
“AI is helping us to more fully explore the cathedral of mathematics we have built over the centuries,” Bloom said in a statement. “What other unseen wonders are waiting in the wings?”
(Source: TechCrunch)
