AI Math Breakthrough: OpenAI Disproves 80-Year Erdős Conjecture

In May 2026, OpenAI announced that an internal AI model had generated the core ideas and draft proof to disprove the Erdős unit distance conjecture, an 80-year-old problem in combinatorial geometry posed by Hungarian mathematician Paul Erdős in 1946. The conjecture asked how many pairs of dots on a flat surface could be placed exactly one unit apart, with mathematicians long assuming Erdős’s proposed upper limit was correct. A team of prominent researchers verified and refined the AI-generated proof, marking the first time an AI system independently contributed a major mathematical breakthrough. Google DeepMind followed days later, revealing its AlphaProof Nexus system had solved nine additional Erdős problems using formal proof verification through the Lean proof assistant. The two approaches differed fundamentally: OpenAI’s model relied on natural language reasoning with human oversight, while Google’s system combined language models with machine-checkable proofs. Both methods were endorsed by leading mathematicians, including Fields Medalist Tim Gowers, as valid contributions to the field. The Erdős unit distance problem had resisted solution for decades due to its deceptive simplicity—while small configurations could be solved manually, scaling to millions of points made it intractable for human mathematicians. OpenAI’s AI unexpectedly linked the geometric problem to algebraic number theory, a seemingly unrelated field, bypassing traditional grid-based approaches that had dominated research efforts. Experts caution that while these breakthroughs demonstrate AI’s potential to accelerate mathematical discovery, limitations remain. The technology still requires human verification to ensure accuracy, particularly for complex proofs. Nonetheless, the rapid succession of AI-driven results suggests a paradigm shift in how scientific research is conducted, with broader implications for fields beyond mathematics. The announcements came within a single week, underscoring the pace of AI progress in scientific reasoning. Fields Medalist Tim Gowers highlighted the significance, stating both sets of results represented genuine advancements, though he emphasized the need for caution in interpreting AI-generated proofs without rigorous human review. The developments signal AI’s evolving role from research assistant to potential co-discoverer in theoretical science.