What was already known
Smale proved the upper bound of 4 in 1981 and asked whether 1 suffices. Adamczewski’s model-generated counterexample ruled out 1. Later lower bounds exceeded 1.1018 and then 1.1559.
AI involvement
Goel credits Claude and ChatGPT with most of the mathematical work. He steered their work through an ongoing discussion.
- Dave Goel, Smale's mean value constant (1981) equals 4
- Authorship and correspondence with the paperAutomation notes and paper’s AI disclosure
Importance
This determines the optimal universal constant in Smale’s inequality. The limit is approached as degree grows, while every fixed-degree constant is below 4.
Technical verification
The project reports Lean 4 Comparator acceptance of the sharp constant and the degree limit on 11 November. It also checked the joint statement in the paper’s normalization. The Lean proof uses a disc version of the Jatar–Ng reduction and Fekete points in place of two steps in the written route.
Formal checking
The project reports Lean 4 Comparator acceptance of the sharp constant and the degree limit on 11 November. It also checked the joint statement in the paper’s normalization. The Lean proof uses a disc version of the Jatar–Ng reduction and Fekete points in place of two steps in the written route.
- Checker
- Lean 4 via the project’s Comparator run
- Scope
- For each degree , let be the best constant in Smale’s mean value inequality for complex polynomials of that degree. Goel proves that tends to 4 and that the best constant valid for all degrees is 4.
- Basis
- Reported by the linked checker
- Lean statements and Comparator checkingProof and documented checking scope
Human understanding
Goel directed the project and reports checking the formalization. The available sources give no separately attributed human explanation of the central construction.
Correction and publication history
Published a scoped research advance with primary evidence of central AI work.
Read this published revision