Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel
starfleetmath.com
July 15, 2026
94 min read
🔥🔥🔥🔥🔥
52/100
Summary
Star Fleet is an AI system designed to solve complex open mathematics problems using Lean 4. It operates as a Mac desktop app, utilizing up to 20 custom agentic harnesses called "starships," each running a GPT-5.6 instance on a dedicated 60-vCPU server.
Key Takeaways
- Star Fleet Math is an AI system designed to solve complex open mathematics problems using Lean 4, operating on a Mac desktop app with up to 20 parallel GPT-5.6 instances.
- The system utilizes a dedicated 60-vCPU server with advanced computational resources, including H100 GPU bursts and a large corpus of Lean 4 premises.
- The first solved problem is Erdős Problem #123, demonstrating that every sufficiently large integer can be expressed as a sum of distinct terms from a pairwise-coprime triple of integers without any term dividing another.
- The solution to Erdős Problem #123 was achieved through a combination of advanced mathematical techniques and computational checks, addressing the challenges posed by the divisibility condition.
Community Sentiment
Positives
- The sheer computational power harnessed for solving these problems is a game changer, with thousands of vCPUs leading to unprecedented exploration in mathematics.
- The ability to build advanced AI systems for tackling novel math is thrilling, and it's inspiring to see others making real progress alongside.
- Fable's referee system for Lean 4 proofs is fascinating, even if the writeups still need to improve in readability.
Concerns
- The proof summaries from Fable are struggling with human readability, making it tough for non-experts to engage with the results.
- Some commenters are skeptical about the funding and sustainability of such a massive computational project.
