OpenAI AI Disproves 80-Year-Old Math Conjecture

OpenAI claims its new reasoning model has produced an original proof disproving a geometry conjecture posed by Paul Erdős in 1946, marking what the company calls the first time AI has autonomously solved a prominent open math problem. The announcement is backed by respected mathematicians including Noga Alon, Melanie Wood, and Thomas Bloom. The claim comes seven months after OpenAI embarrassingly overstated GPT-5's math abilities, drawing ridicule from rivals. This time, the proof came from a general-purpose reasoning model, not a specialized system, suggesting AI can now handle complex, cross-disciplinary chains of reasoning with potential implications for biology, physics, and medicine.