Preprint credits the ‘Odin’ AI research agent with a positively curved metric on S³×S³, which would disprove Hopf’s Euler-characteristic sign conjecture (unverified)
Awaiting review
Importance: major (4 of 5)The takeaway
If correct, a 1930s conjecture of Hopf falls in dimension 6, with the construction and proof credited to an AI agent; no geometer has confirmed it yet.
Status
- Claim
Awaiting review
- Our reporting
- Medium confidence
- Verification
- Preprint only
- Importance
- Major (4 of 5)
- Last verified
- 16 October 2026
Your AI and this story
- GPT-6 Astra129 days after its cutoff
- Claude Opus 5.568 days after its cutoff
- Gemini 4 Argon159 days after its cutoff
- Grok 4.798 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 68 days before it.
Key facts
- Paper: ‘A Metric with Positive Sectional Curvature on S³×S³’, Shengtao Guo, Ethan X. Fang, Junwei Lu (Department of Biostatistics, Harvard T.H. Chan School of Public Health), arXiv 2609.11980 [math.DG], v1 6 Sept 2026, still v1 on 16 Oct
- AI statement (abstract): ‘The metric and the proof were discovered by the Odin Automatic AI Research Agent.’ Introduction: ‘Odin Automatic AI Research Agent was used to construct the metric and the proof.’ No model behind Odin is named
- Claim: since χ(S³×S³) = 0, the result ‘disproves the positive-curvature case of Hopf’s sign conjecture’ (closed even-dimensional manifolds of positive sectional curvature have χ > 0)
- Method (paper’s outline): perturbations of a nonnegatively curved metric with height functions selecting a torus of zero-curvature planes, a smooth global correction tensor and a ‘conformal Poisson correction’; finite algebraic identities checked exactly over Q(z) by code at github.com/shengtaoguo/s3xs3-curvature-calculations (‘python3 verification/verify.py’ prints ‘VERIFIED:’), while the passage to positive curvature on the whole Grassmann bundle is a written proof (Sections 4–7), not a formal one
- Earlier step by the same authors: ‘A Metric with Positive Sectional Curvature on S²×S³’ (arXiv 2608.22133, 22 Aug 2026), whose PDF also says the results ‘are discovered by the Odin Automatic AI Research Agent’
- Human context: Simon Brendle and P. K. Hung, ‘A metric on S²×S² with positive sectional curvature’ (arXiv 2608.19068, 19 Aug 2026), disproving Hopf’s product conjecture with a third-order perturbation of a Cheeger–Müter metric and Mathematica calculations; no AI credit in the abstract
- Specialist reaction: Nienhaus and Wulle’s Hopf T³ paper (arXiv 2610.22173, 15 Oct) footnotes ‘an announcement of a counterexample to the Hopf conjecture in dimension 6, but to the knowledge of the authors this work is entirely AI and not verified by geometers’ (no citation given; this preprint matches the description)
- The same agent is credited in Fang, Lu et al.’s forest-unimodality paper (arXiv 2609.20961): ‘The Odin Automatic AI Research Agent was used to find the initial proof’ (see Erdős Problem #993 (1987))
Show 1 more
- No Hacker News thread, press coverage or expert confirmation found as of 16 Oct 2026
What happened
Heinz Hopf asked two famous questions about positive sectional curvature: whether S²×S² admits such a metric (the product conjecture) and whether every even-dimensional closed positively curved manifold has positive Euler characteristic (the sign conjecture). In August 2026 Simon Brendle and P. K. Hung posted a metric of positive curvature on S²×S², disproving the product conjecture. Within days a team at Harvard’s Department of Biostatistics posted a positively curved metric on S²×S³ (22 Aug), and on 6 Sept a metric on S³×S³. Because χ(S³×S³) = 0, the second would be a counterexample to the sign conjecture in dimension 6, where Geroch’s 1976 algebraic example had shown that the four-dimensional Gauss–Bonnet argument fails.
Both papers credit the “Odin Automatic AI Research Agent” with discovering the metric and the proof. The S³×S³ paper runs to 30 pages, ships Python code that checks its finite algebraic identities exactly, and leaves the global positivity argument as a written proof.
Why it matters
If it holds, one of the central conjectures of Riemannian geometry fails, and an AI agent is credited with the full construction. The claim is unverified: on 15 Oct the geometers Nienhaus and Wulle, whose own AI-assisted paper proves the conjecture under T³ symmetry, wrote that the dimension-6 announcement is “entirely AI and not verified by geometers”. The case shows how AI-produced claims about famous problems can sit unexamined for weeks.
What is disputed or not yet verified
| Verification | Unrefereed preprint; exact computer checks of finite identities in a public repository; not verified by geometers per Nienhaus–Wulle |
|---|
Sources
6 sources from 3 sites. Numbers match the chips in the text.
6 sources: 6 primary
Primary
- arXiv 2609.11980: A Metric with Positive Sectional Curvature on S³×S³ (Guo, Fang, Lu)arxiv.org, paper
- Verification code: shengtaoguo/s3xs3-curvature-calculationsgithub.com, code
- arXiv 2608.22133: A Metric with Positive Sectional Curvature on S²×S³ (Guo, Fang, Lu)arxiv.org, paper
- arXiv 2608.19068: Brendle and Hung, A metric on S²×S² with positive sectional curvaturearxiv.org, paper
- arXiv 2610.22173: Nienhaus and Wulle, The Hopf conjecture with T³-symmetry (footnote on the dimension-6 claim)arxiv.org, paper
- MathWorld: Hopf Conjecture (lists the Brendle–Hung announcement)mathworld.wolfram.com, docs
Changes
- Filed (arXiv abstracts and both Odin PDFs read; Brendle–Hung abstract read)