😸 OpenAI solved an 80-year math problem by... disproving it

2026-05-12 · Source: The Neuron · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Software Development & Engineering · Depth: Novice, long

Summary

OpenAI's internal reasoning model reportedly disproved the 80-year-old Erdős unit distance conjecture, a discrete geometry problem from 1946. The model found a new infinite family of point arrangements creating more unit-distance pairs than previously believed, with Princeton mathematician Will Sawin sharpening the proof to show over n1.014 pairs. This achievement, verified by external mathematicians, highlights AI's advanced reasoning capabilities beyond standard benchmarks. Concurrently, the AI landscape saw Qwen 3.7 Max complete a 35-hour agent run, Cohere release its open enterprise model Command A+, and Codex receive upgrades including "/goal mode". Other developments include OpenAI's Q1 revenue of \$5.7B, California's AI worker preparation order, Intuit's 3,000+ AI-focused layoffs, and Samsung's \$26.6B chip worker bonuses.

Key takeaway

For AI scientists and developers evaluating advanced reasoning systems, OpenAI's disproof of the Erdős conjecture signals a critical shift towards verifiable AI-driven mathematical discovery. You should explore how similar AI models could accelerate research in fields with clear verification mechanisms, potentially uncovering novel solutions that human effort alone might miss. Additionally, integrate tools like Codex's "/goal mode" to enhance agent reliability for complex, multi-step projects.

Key insights

An OpenAI model disproved an 80-year math conjecture, showcasing AI's advanced reasoning and verifiable problem-solving capabilities.

Principles

• AI proofs offer a robust test of reasoning.
• Mathematical problems are uniquely AI-friendly due to verifiable proofs.
• AI can uncover solutions that human experts might overlook.

Method

Codex's "/goal mode" enables persistent objectives for long agent tasks, defined by outcome, constraints, and tests.

In practice

• Utilize Codex "/goal mode" for complex, multi-step coding tasks.
• Evaluate Qwen 3.7 Max for extended autonomous agent operations.
• Consider Cohere Command A+ for private, multilingual enterprise AI deployments.

Topics

• AI Reasoning
• Discrete Geometry
• Mathematical Proof
• Large Language Models
• AI Agents
• Codex
• Enterprise AI

Best for: Research Scientist, AI Scientist, Director of AI/ML, AI Student

Related on AIssential

• An OpenAI model solved a famous math problem that stumped humans for 80 years — AI disproved an 80-year-old math conjecture by combining existing techniques, showcasing its ability to resolve major open problems.
• OpenAI reasoning model disproves 80-year-old Erdős geometry conjecture — A general-purpose AI model has autonomously disproved an 80-year-old mathematical conjecture, demonstrating advanced reasoning capabilities.
• OpenAI Says Its AI Just Disproved a Famous 80-Year-Old Math Problem – And This Time, Mathematicians Agree — A general-purpose AI autonomously disproved an 80-year-old math conjecture by making a cross-domain mathematical leap.
• OpenAI’s math breakthrough played to AI’s strengths — AI disproved an 80-year-old math conjecture by applying broad knowledge and persistent strategy exploration, marking a significant autonomous contribution.
• OpenAI makes breakthrough on 80-year-old maths problem — OpenAI's AI disproved an 80-year-old math conjecture, showing AI can find novel solutions in complex problems.
• OpenAI claims it solved an 80-year-old math problem — for real this time — OpenAI's new model disproved an 80-year-old math conjecture, demonstrating advanced autonomous reasoning in AI.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Neuron.