OpenAI claims its new reasoning model has produced an original mathematical proof disproving a famous unsolved conjecture in geometry, first posed by Paul Erdős in 1946. This announcement follows a previous assertion by OpenAI’s former VP Kevin Weil, who seven months ago claimed that GPT-5 found solutions to 10 previously unsolved Erdős problems and made progress on 11 others. It was later revealed that GPT-5 did not actually solve these problems but instead identified existing solutions in the literature.
After facing criticism from rivals like Yann LeCun and Google DeepMind CEO Demis Hassabis, Weil removed his initial post. OpenAI has now published supporting remarks from mathematicians Noga Alon, Melanie Wood, and Thomas Bloom, the latter of whom described Weil’s earlier statement as “a dramatic misrepresentation.”
Today is my last day at OpenAI, as OpenAI for Science is being decentralized into other research teams. It’s been a mind-expanding two years, from Chief Product Officer to joining the research team and starting OpenAI for Science. Accelerating science will be one of the most…
— Kevin Weil 🇺🇸 (@kevinweil) April 17, 2026
According to OpenAI, mathematicians have believed for nearly 80 years that the best solutions to these problems resembled square grids. The company stated, “An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better.” This proof marks the first time AI has autonomously solved a significant open problem in mathematics.
The proof was generated by a new general-purpose reasoning model, rather than a system specifically designed for mathematical tasks. OpenAI asserts that this development indicates AI systems are becoming more adept at connecting ideas across different fields, which could have implications for various disciplines including biology, physics, engineering, and medicine.
Thomas Bloom remarked on AI’s potential to explore the depths of mathematics, questioning, “What other unseen wonders are waiting in the wings?”
