OpenAI has once again made headlines with a bold claim, this time asserting that its new reasoning model has solved an 80-year-old math problem. However, this isn't the first time the company has faced scrutiny for such assertions. In the past, OpenAI's former VP Kevin Weil announced that GPT-5 had solved 10 previously unsolved Erdős problems, only to later retract the claim after it was revealed that the solutions already existed in the literature. This time, however, OpenAI seems to have taken a more cautious approach, publishing companion remarks alongside the announcement and seeking support from mathematicians like Noga Alon, Melanie Wood, and Thomas Bloom. The company claims that its new model has discovered an entirely new family of constructions that perform better than previously believed, marking the first time AI has autonomously solved a prominent open problem in mathematics. This development has significant implications for fields beyond mathematics, such as biology, physics, engineering, and medicine, as AI systems become more capable of holding together long, difficult chains of reasoning and connecting ideas across disciplines. However, as with any groundbreaking claim, it's essential to approach this development with a critical eye and consider the potential implications and limitations of AI in mathematics and beyond.
