OpenAI model disproves Erdős unit distance conjecture

openai's system solved a unit distance problem posed by mathematician paul erdős in 1946 regarding the optimal arrangement of points on a plane. contrary to the established conjecture that the square grid was nearly optimal, the model identified a superior construction using class field theory and complex number systems. princeton mathematician will sawin noted the solution yields roughly one percent more unit-distance pairs per doubling of point count, a gain previously thought impossible. while a theoretical upper bound remains unbroken, the discovery demonstrates ai's capacity to synthesize specialized mathematical knowledge to bypass human biases against certain approaches.