What was already known
Exact optima were known for the equilateral triangle and base angles from 45 to 60 degrees. Gibbs had numerically predicted a seven-piece path near 36 degrees but had not proved its exact global optimum.
AI involvement
The title disclosure credits Claude Fable 5, GPT 5.6 Sol and Claude Opus 5 with searching for the extremal path, drafting arguments and writing the Lean development. Temerev and Doria are the human authors; the source does not divide every proof idea by origin.
Models searched for the exact retained path and drafted arguments, while human authorship and unspecified mathematical direction prevent a stronger independence finding.
Importance
The claimed proof gives the first exact optimum below the previously solved 45-degree range and an equivalent sharp worm-cover result for this triangle.
Limitations
Lean checks selected exact calibration, geometry and discrete certificates; compactness, hull surgery, planar incidence and parts of minimizer rigidity remain in prose. The full global optimum has no documented complete check.
Technical verification
The paper reports Lean 4 checks of exact calibration and geometry certificates, rational inequalities and reusable discrete identities. Its Appendix B.3 explicitly leaves compactness, hull surgery, planar incidence and parts of minimizer rigidity in prose; the full global optimum is not machine-certified, and no complete human check is documented.
Formal checking
Exact rational interval certificate checking and formal proof of selected identities
- Checker
- Lean 4 kernel in the authors' reported development
- Scope
- Calibration root and scalar bounds, seven-piece geometry and support-window certificates, plus listed finite escape and ledger identities for the golden gnomon
- Basis
- Reported by the linked checker
Checking a formal statement does not establish that it matches the original problem.
Human understanding
The exposition describes the mechanism, but the authors say models drafted arguments throughout and do not identify a person's own explanation of the central calibration and rigidity reasoning. Few is the evidence default, not a claim that the authors understand little.
Correction and publication history
Initial publication of this result.
Read this published revision