What happened
OpenAI's new reasoning model produced an original mathematical proof disproving the 1946 planar unit distance conjecture by Paul Erdős. This follows a previous, retracted claim regarding GPT-5's supposed solutions to other Erdős problems. This time, mathematicians including Noga Alon, Melanie Wood, and Thomas Bloom supported the disproof, with Bloom previously calling the earlier claim "a dramatic misrepresentation". OpenAI stated this marks the first autonomous AI solution to a prominent open mathematical problem, achieved by a new general-purpose reasoning model, not a system specifically designed for mathematics.
Why it matters
This demonstrates AI systems' enhanced capability for long, complex reasoning chains and cross-field idea connection, moving beyond literature retrieval. For research teams and architects, this mechanism suggests new avenues for AI application in biology, physics, engineering, and medicine, potentially accelerating discovery by exploring previously unconsidered constructions. The metric is the disproof of an 80-year-old problem, indicating a shift in AI's reasoning capacity. This follows OpenAI's earlier retracted claim about GPT-5's math capabilities, underscoring the current validation's significance.
