OpenAI Reasoning Model Disproves 80-Year-Old Mathematical Conjecture
An OpenAI reasoning model disproved a planar unit distance conjecture mathematicians accepted for nearly 80 years.
An OpenAI reasoning model has disproved a longstanding belief about the planar unit distance problem, a question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians assumed that optimal configurations of unit-distance points in a plane looked roughly like square grids. OpenAI announced that one of its models found a configuration that invalidates that assumption.
The result carries weight beyond the mathematical milestone itself because it came from a general-purpose reasoning model — not a system purpose-built to tackle geometry or this problem specifically. For builders and practitioners, that distinction matters: it suggests current AI systems are developing the capacity to sustain long, complex chains of reasoning across disparate domains, connecting ideas that human researchers might not have linked. OpenAI framed this as an early indicator of the kind of reasoning capability it expects to accelerate scientific and technical work more broadly.