TL;DR
OpenAI's GPT 5.2 and other AI models have solved 15 previously open Erdős problems since Christmas, with 11 crediting AI involvement, demonstrating new capability in autonomous mathematical reasoning.
Key Points
- 15 Erdős problems moved from 'open' to 'solved' since Christmas 2025; 11 solutions credited AI models
- Terence Tao documented 8 problems with autonomous AI progress, 6 with AI-assisted research
- GPT 5.2 described as 'anecdotally more skilled at mathematical reasoning' than previous iterations
- Lean proof assistant and tools like Harmonic's Aristotle accelerating formalization and verification of proofs
Why It Matters
This demonstrates AI's growing capability to tackle frontier mathematical problems autonomously, with implications for research acceleration and problem-solving at scale. For developers and researchers, it validates AI tools like Lean and proof assistants as legitimate infrastructure for mathematical work—when mathematicians with reputations to protect actively adopt these tools, it signals genuine technical merit beyond hype.
Source: techcrunch.com