AI Agents Solved Thomson Geometry Problem
Ten Claude Sonnet 5.5 agents produced a 17,895-line mathematical proof in September 2026.
Updated on Sept. 30, 2026 in Mathematics
Live Poll
Would you trust an AI-generated mathematical proof that has not been reviewed by a human expert?
In September 2026, a team of ten Claude Sonnet 5.5 agents successfully solved the seven-charge Thomson problem. The AI models generated a 17,895-line formal proof identifying the pentagonal bipyramid as the lowest-energy configuration for seven points on a sphere.
Why it matters
Vals AI utilized the Thomson problem to test the proficiency of large language models in executing complex mathematical reasoning and formal code generation. The effort demonstrated the ability of AI to coordinate across thousands of messages to complete a high-level scientific task.
The agents spent 15 hours coordinating via 1,270 messages to construct a proof containing 47,854 declarations. Two independent kernels, including nanoda, verified the 17,895-line file.
The players
Vals AI
This is the research organization that coordinated the use of AI agents to demonstrate capabilities in complex mathematics and coding tasks.
Claude Sonnet 5.5
This is a large language model architecture developed to handle high-level reasoning and programming challenges.
J.J. Thomson
He was a Nobel Prize-winning physicist who originally proposed the geometry problem involving point charges on a sphere in 1904.
The details
The agents partitioned the geometry problem by minimizing inner products between charge pairs and utilized existing techniques from human mathematical research. They employed semidefinite bounds and three-point certificates to ensure the configuration was mathematically sound.
Timeline
J.J. Thomson first posed the geometry problem in 1904.
The AI agents performed their calculations in September 2026.
The Big Picture
This project represents a departure from traditional mathematical verification by relying on autonomous AI agents to solve the 1904 Thomson problem. It demonstrates a paradigm shift in formalizing proofs that were previously reliant on human manual derivation.
This development suggests that AI agents may soon accelerate the discovery of new mathematical truths by automating rigorous, line-by-line verification processes. For researchers, this could reduce the time required to confirm complex proofs that typically take months or years of human labor.
The takeaway
The successful application of AI to the Thomson problem highlights the growing potential for models to handle high-stakes logic and formal verification. Readers interested in computational science should note that while this proof is computationally verified, professional peer review remains the gold standard for mathematical consensus.
Further reading
Learn more about the latest developments in the field of Mathematics.
More information
View the formal code and detailed methodology in the GitHub repository for proof.
Source note: This article includes information reported by Startup Fortune.
Live Poll
Would you trust an AI-generated mathematical proof that has not been reviewed by a human expert?
