🐯虎嗅•Stalecollected in 17m
OpenAI model solves 80-year-old math conjecture
💡AI autonomously solved a classic math problem, proving its capability for original scientific research.
⚡ 30-Second TL;DR
What Changed
The model solved the 'unit distance problem' by finding a new construction that exceeds previous limits.
Why It Matters
This achievement demonstrates that AI is moving beyond simple text generation into the upstream of scientific research, capable of original mathematical discovery.
What To Do Next
Explore the potential of reasoning models for your domain-specific research tasks by testing complex logical reasoning prompts.
Who should care:Researchers & Academics
Key Points
- •The model solved the 'unit distance problem' by finding a new construction that exceeds previous limits.
- •The proof utilized advanced algebraic number theory, demonstrating the model's ability to connect cross-disciplinary knowledge.
- •Fields experts, including Fields Medalists, consider this a milestone for AI in autonomous scientific discovery.
🧠 Deep Insight
Web-grounded analysis with 13 cited sources.
🔑 Enhanced Key Takeaways
- •The 'unit distance problem' was originally posed by the renowned Hungarian mathematician Paul Erdős in 1946, who even offered a monetary prize for its resolution, highlighting its long-standing difficulty and significance in combinatorial geometry.
- •Contrary to the prevailing belief among mathematicians for nearly 80 years that optimal solutions for the problem resembled square grids with a growth rate of roughly n^(1+o(1)), OpenAI's model discovered an entirely new family of constructions that achieve a polynomial improvement, specifically n^(1+δ) for some fixed positive constant δ (e.g., δ=0.014).
- •The proof was generated autonomously by an internal OpenAI general-purpose reasoning model, not a specialized system, and was subsequently rigorously checked by a group of external mathematicians, including Fields Medalists, who also co-authored a companion paper explaining the argument and its context.
- •This achievement is considered a qualitative shift, marking the first time AI has autonomously solved a prominent open problem central to a field of mathematics, thereby transitioning AI's role from a research tool to a research agent.
- •The model's novel solution involved applying unexpected and sophisticated concepts from algebraic number theory, such as class field towers and Golod-Shafarevich theory, to an elementary geometric question, demonstrating its ability to connect disparate mathematical domains.
🛠️ Technical Deep Dive
- The solution was produced by an internal, general-purpose reasoning model from OpenAI, rather than a system specifically trained for geometry or theorem proving.
- These reasoning models, such as OpenAI's o-series (e.g., o1, o3, GPT-5.2), utilize 'reasoning tokens' to internally 'think,' breaking down prompts and exploring multiple approaches to problem-solving.
- The models employ 'chain-of-thought reasoning,' maintaining an internal dialogue or 'thinking block' to methodically work through problems, refine solutions, and self-correct before generating a final answer.
- The mathematical proof itself leverages advanced concepts from algebraic number theory, including infinite class field towers and Golod-Shafarevich theory.
- The AI-generated proof was formally verified using Lean, a rigorous proof assistant tool, to ensure its correctness and reliability.
🔮 Future ImplicationsAI analysis grounded in cited sources
AI will increasingly become an autonomous research agent in scientific discovery.
This breakthrough demonstrates AI's capacity to independently generate novel mathematical proofs for long-standing open problems, shifting its role from a tool to a co-discoverer.
The demand for verifiable computation infrastructure will grow significantly.
As AI systems generate complex proofs, human researchers will need robust methods like blockchain-based proof verification and zero-knowledge systems to trust and audit these machine-generated results.
AI will accelerate progress on other complex, open mathematical problems.
OpenAI's models already claim to have solved over 10 research-level problems in combinatorics, including several by Erdős, indicating a broader capability to tackle difficult conjectures.
⏳ Timeline
1946
Paul Erdős poses the planar unit distance problem.
1984
Spencer, Szemerédi, and Trotter establish the best known upper bound (O(n^(4/3))) for the problem.
2025-01
OpenAI's GPT-5.2 answers Erdős problems #397 and #728, open for about thirty years.
2025-02
An internal OpenAI model solves at least five of ten 'First Proof' research-level math problems.
2025-12
OpenAI highlights GPT-5.2 Pro and GPT-5.2 Thinking as its strongest models for scientific and mathematical work.
2026-05
OpenAI's general reasoning model autonomously disproves the planar unit distance conjecture.
📎 Sources (13)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 虎嗅 ↗