AI Solves 80-Year-Old Paul Erdős Math Problem

An internal OpenAI model recently discovered a counterexample to Paul Erdős' 1946 planar unit distance problem ^[1]. This breakthrough proves that traditional grid-like point arrangements are not the most efficient configurations ^[1]. The discovery marks a significant step in the ability of artificial intelligence to perform autonomous mathematical research ^[1].

Reports also indicate that Anthropic's Claude Mythos model has produced a simplified proof for the same landmark problem ^[2]. This development follows the initial findings by the OpenAI system, showcasing intense competition in the AI sector ^[2]. Experts expressed surprise at the simple and "cute" nature of the proof provided by the Claude model ^[2].

The Erdős problem had remained unsolved for 80 years before these recent AI advancements ^[1]. Researchers believe this indicates a shift toward AI models becoming partners in solving complex theoretical challenges ^[1][2]. The findings suggest that AI can move beyond pattern recognition into genuine logical discovery ^[2].