OPEN LINK ↗
// 1h agoRESEARCH PAPER
OpenAI o1 disproves 80-year-old Erdős geometry conjecture
OpenAI's general-purpose reasoning model has autonomously disproven the 1946 Erdős planar unit distance conjecture, a central problem in discrete geometry that had resisted human efforts for eight decades. By utilizing advanced algebraic number theory and Golod–Shafarevich theory to construct new point-set families, the model demonstrated that long-held square-grid assumptions were suboptimal, marking a watershed moment for AI in pure mathematics.
// ANALYSIS
This is a "Stockfish moment" for mathematics, signaling that AI has graduated from a proof assistant to an autonomous researcher capable of genuine discovery.
- –The model bypassed human cognitive biases by discovering point-set constructions that outperformed the "intuitively optimal" square grids.
- –By producing fully formalized proofs in Lean 4, the system eliminates hallucination risks and allows immediate machine verification by the mathematical community.
- –The discovery relied on connecting disparate fields—algebraic geometry and discrete combinatorics—a feat typically reserved for top-tier human mathematicians like Terence Tao or Timothy Gowers.
- –Success with a general-purpose model (o1) suggests that scaled "Thinking" capabilities are now powerful enough to solve frontier problems without domain-specific engineering.
// TAGS
openaiopenai-o1leandiscrete-geometryresearchreasoningmathformal-verification
DISCOVERED
1h ago
2026-05-20
PUBLISHED
3h ago
2026-05-20
RELEVANCE
10/ 10
AUTHOR
tedsanders