← Retour au blog
tech 15 July 2026

Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel

Learn how an innovative approach using 20 Codex accounts in parallel tackles some of the world's hardest math problems.

Article inspired by the original source
Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel ↗ www.starfleetmath.com

Introduction

Erdős problems are among the most feared in the mathematics world. Named after Paul Erdős, these open problems have challenged mathematicians for decades. With the advent of artificial intelligence, a new approach is emerging. Imagine solving these puzzles with 20 Codex accounts running in parallel. Let's dive into this fascinating world where AI meets mathematics.

Why Erdős Problems?

Erdős problems are famous for their complexity and their ability to push the boundaries of mathematical research. They cover various fields, from number theory to combinatorics, and are often accompanied by prizes for anyone who manages to solve them.

The Power of Parallel Codex

Codex, the language model derived from GPT-3, has paved the way for complex natural language processing applications. By using 20 instances of Codex in parallel, each instance can focus on a specific problem. This allows for faster and deeper analysis, leveraging collective computing power to break down walls of complexity.

How It Works

Each Codex instance runs on a dedicated server with 60 vCPUs. In parallel, they leverage CPU bursts up to 2,000 vCPUs, coupled with H100 GPU bursts for massively parallel search programs. This robust infrastructure provides the exceptional computing power needed to tackle Erdős problems.

Success Stories

Take, for instance, Erdős Problem #123, which was solved using Lean 4. This problem, which had long resisted traditional approaches, was decomposed and solved thanks to the collective efficiency of Codex starships. The result was a proven theorem that paves the way for new perspectives in number theory.

Challenges Faced

While the parallel Codex approach is promising, it's not without challenges. One of the main hurdles is coordinating between different instances to avoid redundancy and maximize efficiency. Additionally, Erdős problems often require mathematical intuition that even the most advanced AI might struggle to replicate.

The Future of AI-Assisted Mathematics

Solving complex problems like Erdős problems with AI marks the beginning of a new era for mathematics. Researchers can now focus on more creative and innovative aspects, leaving complex calculations to machines.

Conclusion

The combination of AI and mathematics opens immense horizons. With approaches like using 20 Codex accounts in parallel, we are redefining what is possible. If you want to explore how this technology can benefit your project, let's discuss your project in 15 minutes.

Erdős Problems Codex AI Mathematics Parallel Computing Lean 4
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call