OpenAI has announced that its latest reasoning model has produced an original mathematical proof, purportedly disproving an unsolved conjecture first posed by Paul Erdős back in 1946. This claim marks what the company says is ‘the first time AI has autonomously solved a prominent open problem central to a field of mathematics.’

However, echoes of past controversies linger. Just last year, OpenAI’s former VP Kevin Weil prematurely claimed that GPT-5 had found solutions to 10 previously unsolved Erdős problems, only for the company to backtrack when it emerged these 'solutions' were already in circulation.

On this occasion, the company has gathered support from notable mathematicians like Noga Alon and Thomas Bloom, who have verified the new proof. This breakthrough could signal a significant leap forward in AI’s ability to engage in complex reasoning across various disciplines – potentially revolutionising fields as diverse as biology, physics, engineering and medicine.

‘AI is helping us to more fully explore the cathedral of mathematics we have built over the centuries,’ Bloom stated, hinting at the untapped potential that AI might yet reveal. Only now, with this new claim, does it seem OpenAI has truly learned from its mistakes.