Solving 20 Erdős Problems with 20 Codex Accounts Running in Parallel (www.starfleetmath.com)

A groundbreaking AI system called Star Fleet has successfully tackled 20 problems from Paul Erdős's renowned list of open mathematics challenges. This innovative desktop application utilizes Lean 4 and orchestrates up to 20 parallel instances of GPT-5.6, each operating on distinct mathematical problems. The architecture is built from the ground up using TypeScript and Bun, leveraging advanced computational resources, including a dedicated 60-vCPU server topology for intensive mathematical searches and GPU acceleration for massively parallel processing. This achievement is significant for the AI and machine learning community as it demonstrates the potential of AI to solve complex mathematical problems that have historically resisted traditional methods. The system merges powerful computational techniques with dedicated natural language processing capabilities, creating a comprehensive platform for exploring difficult mathematical theories, such as those involving pairwise-coprime integers and Ramsey-type problems. The formal results, verified in Lean 4, mark an important milestone in the intersection of formal verification, AI, and open mathematics, paving the way for enhanced computational proofs and further exploration of unsolved problems in the field.