Curtis's conjecture: the mod-2 stable Hurewicz image of the sphere is spanned by the Hopf- and Kervaire-invariant-one classes
For let be the mod-2 Hurewicz homomorphism into the homology of the basepoint component of . The obvious elements with nonzero image are the Hopf-invariant-one classes in degrees 1, 3, 7 and the Kervaire-invariant-one classes in degrees , when they exist. Curtis (1975), in his study of the Dyer-Lashof and lambda algebras, conjectured that nothing else is detected; May (1977) recorded a gap, found by Wellington, in the proposed proof. The conjecture implies Eccles's conjecture for spheres. Is the image of spanned over by the Hurewicz images of and of the Kervaire-invariant-one classes that exist?
- Result
- Proved(see note)
- Status
- Candidate (review pending)
- AI contribution
- AI-discovered
- Method
- Argument
- Field
- Stable homotopy theory
- Posed by
- Edward B. Curtis
- Year posed
- 1975
- Years open
- 51y
- Solved
- 2026-09-25
- Model
- Unreleased internal OpenAI model
- Vendor
- OpenAI
- Collaborators
- —
- Verification
- Unreviewed
- Publication
- Announced
- Collection
- OpenAI math release (October 2026), version adc7f12
- Significance
- 48 / 100
- Disclosed cost
- —
- Wikipedia
- No dedicated article
What was actually shown
Theorem 1.1: the image of is spanned over by the Hurewicz images of and of the Kervaire-invariant-one classes (degree ) whenever they exist; no existence is posited. Corollary 1.2: Eccles's conjecture holds for every sphere , : each nonzero spherical class in is the bottom class or a suspended Hopf-invariant-one image. Corollary 1.3 (with Hill-Hopkins-Ravenel): for outside , so the positive image vanishes above degree 126. Not shown: an exact dimension of the image or nonvanishing in each listed degree, odd primes, or Eccles's conjecture for with other than a sphere.
What the AI did
The release README says the results were produced by an unreleased internal OpenAI model with a fixed procedure, on average about three hours of ChatGPT Pro thinking compute per result, and that some outputs build on earlier model results. This result is not among the README's exceptions (the Hodge conjecture for CM abelian varieties and the Re(s) > 11/12 zero-free region). The manuscripts are authored 'OpenAI' and name no human author. The single manuscript (September 25, 2026) is the whole family.
Verification
No independent mathematician has checked this yet. Checked here: the abstract, introduction and Theorem 1.1 with Corollaries 1.2 and 1.3 were read against Curtis's conjecture as the manuscript cites it; Theorem 1.1 states the conjecture in every positive degree. The proof (last nonzero homology suspension, weight decomposition through Dickson-algebra modules, a Steenrod-module obstruction, Kuhn's lifting theorem and Browder's identification of Kervaire classes) was not refereed. No Lean formalization accompanies this manuscript. The paper says its exclusion of fourth powers (Section 4) is already present in Eccles-Zare, and Corollary 1.3 uses the Hill-Hopkins-Ravenel nonexistence theorem as an input.