“Erdos problem #728 was solved more or less autonomously by AI”
TL;DR Highlight
Terence Tao used AI tools (Aristotle + LLM) to formally prove unsolved math problem Erdős #728 in Lean.
Who Should Read
Mathematicians and CS researchers interested in AI-assisted formal verification, and ML researchers tracking AI's progress on math reasoning.
Core Mechanics
- Terence Tao (Fields Medal winner) used an AI tool called Aristotle in combination with an LLM to construct a formal Lean proof for Erdős conjecture #728 — a long-standing open problem.
- This is significant because it's not a toy problem — Erdős conjectures are real mathematical research-level challenges, and #728 had resisted proof for decades.
- The proof process was collaborative: Tao provided the high-level mathematical insight and strategy, while Aristotle/LLM handled the Lean proof formalization.
- The result was machine-verified — not just plausible but formally checked by Lean's type system, guaranteeing correctness.
- This demonstrates the emerging pattern: world-class mathematicians can now prove harder theorems faster with AI assistance than without.
- Lean proof assistants become a trust anchor — even if the AI makes reasoning mistakes, the proof checker catches them before they're accepted.
Evidence
- The Lean proof was made publicly available and verified by the community — not just a claim but checkable artifact.
- Tao himself described the experience as significantly faster than manual formalization, even with his deep mathematical expertise.
- HN discussion noted the importance of the 'Aristotle' tool specifically — it bridges between natural language mathematical reasoning and Lean's formal syntax.
- Researchers in the Lean community confirmed this is a meaningful result — Erdős #728 is a legitimate research-level theorem, not a textbook exercise.
- Comparison was made to AlphaProof's work on IMO problems — a pattern of AI enabling mathematical progress rather than replacing mathematicians.
How to Apply
- If you work in software verification, track Aristotle and similar tools — the ability to bridge natural language specs to formal proofs is the missing piece for wider adoption.
- For researchers: AI-assisted Lean proof writing is now at the stage where it's worth attempting on your own open problems, not just educational exercises.
- The Tao collaboration model (human provides insight, AI handles formalization) suggests the near-term workflow: focus your effort on the mathematical ideas, delegate the proof mechanics.
- Monitor the Lean/Mathlib ecosystem for AI tooling improvements — this is moving fast and the tools available in 6 months will be significantly better.
Terminology
Related Papers
Show HN: OpenKnowledge – open source AI-first alternative to Obsidian/Notion
Git 기반 동기화와 Claude/Codex/Cursor 연동을 내장한 로컬 우선 마크다운 에디터로, AI 에이전트의 두 번째 뇌(LLM Wiki)로 활용할 수 있는 오픈소스 도구다.
The Unfireable Safety Kernel: Execution-Time AI Alignment for AI Agents and Other Escapable AI Systems
AI 에이전트가 자신의 안전장치를 우회할 수 없도록, 에이전트 프로세스 바깥에 수학적으로 증명된 강제 통제 게이트를 배치하는 아키텍처
RubyLLM: A Ruby framework for all major AI providers
OpenAI, Claude, Gemini 등 주요 AI 프로바이더를 단일 인터페이스로 통합한 Ruby 프레임워크로, Rails 통합과 에이전트 기능까지 지원해 Ruby 개발자가 AI 기능을 빠르게 붙일 수 있다.
Qwen-AgentWorld: Language World Models for General Agents
Alibaba Qwen 팀이 AI 에이전트가 행동 결과를 미리 시뮬레이션할 수 있는 'Language World Model'을 공개했다. 에이전트 훈련과 실행 경로 검증에 새로운 패러다임을 제시하는 연구다.
SHERLOC: Structured Diagnostic Localization for Code Repair Agents
버그 위치만 알려주는 게 아니라 '왜, 어떻게 고쳐야 하는지'까지 진단 리포트를 생성해서 코드 수정 에이전트의 성능을 높이는 training-free 프레임워크
Show HN: peerd – AI agent harness that runs entirely in your browser
백엔드 서버 없이 Chrome/Firefox 확장 프로그램으로만 동작하는 AI 에이전트 실행 환경으로, 브라우저 탭을 직접 조작하고 WASM Linux VM까지 구동할 수 있어 프라이버시와 보안을 동시에 챙길 수 있다.