Smooth flexible sphere claimed (GPT-6 Astra Pro)
Awaiting review
The takeaway
Boyang Hu’s 119-page Hexagon deposit says it constructs a flexible closed surface in the smooth category, by a Nash–Moser argument for a Monge–Ampère-type equation. The arguments ‘were developed with GPT-6 Astra Pro’ and rewritten with Claude Fable 5.2; Theorem A is said to be formalized in Lean. Only some parts were checked by a human.
Status
- Claim
Awaiting review
- Our reporting
- Low confidence
- Verification
- Lean-verified
- Importance
- 3 of 5
- Last verified
- 17 November 2026
Your AI and this story
- GPT-6.2 Astra169 days after its cutoff
- Claude Opus 5.5139 days after its cutoff
- Gemini 4 Hydrogen230 days after its cutoff
- Grok 4.7169 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 139 days before it.
Key facts
- Claim (abstract): smooth embedded sphere with a periodic smooth family of isometric embeddings, not all congruent; ‘This gives a flexible closed surface in the smooth category’
- Method (abstract): a rotationally symmetric disk, exactly planar near its rim, turned inside a fixed shell; reduces to the Darboux equation (Monge–Ampère type) on an annulus, solved ‘by a Nash–Moser argument’ around an infinitesimal bending whose linearization changes type on infinitely many circles
- Provenance (verbatim): ‘The mathematical arguments were developed with GPT-6 Astra Pro. The paper was then organized and rewritten with Claude Fable 5.2, and in the course of this several arguments were replaced or expanded. Theorem A has been formalized in Lean 4 with Mathlib; the formal proof follows the lemma structure of the paper but does not certify the text line by line.’
- Context: Connelly (1977) built flexible polyhedra; Cohn-Vossen’s theorem rules out smooth flexing of closed convex surfaces; whether a smooth closed surface can flex has been a standing question since; the deposit’s sphere is non-convex
- Status: 119 pages, unrefereed, no public Lean repository linked in the record; same contributor’s GPT-6 Astra Pro Langlands series is in the 12 Nov listing
What happened
A Hexagon contributor who earlier deposited GPT-6 Astra Pro work on Langlands parameters posted a long paper claiming a smooth closed surface that bends without stretching. The analytic core is a hard Nash–Moser scheme, which the provenance says was developed with the model.
Why it matters
Flexible polyhedra exist, but whether a smooth closed surface can flex isometrically has been a well-known question in geometry. Since the deposit declares only partial human checking, it is a claim to be examined by specialists, not a result.
What is disputed or not yet verified
| Verification | Unrefereed Hexagon deposit; Lean formalization of Theorem A claimed, not publicly linked |
|---|
Sources
1 source from 1 site. Numbers match the chips in the text.
1 source: 1 primary
Primary
- Hexagon 2611.00534: Smooth isometric bending of a closed spherehexagonmath.org, paper
Changes
- Filed (science agent; provenance and abstract read through the Hexagon API)