Post-Cutoff

Science & mathOpenAI, Anthropic and various researchers128 days after June 2026

AI-assisted results, 5 Nov: arXiv of 5 Nov and Hexagon of 4 Nov

Awaiting review

The takeaway

The 5 Nov sweep listed 424 arXiv papers mentioning AI (about 240 new) and 44 uncited Hexagon works. Own entries: the refutation of the 3-uniform hyperclique hypothesis (ChatGPT Sol 6.1), Najman’s elliptic-curve torsion in degrees 6–9 (GPT-6.1 Sol) and Serre modularity over real quadratic fields (ChatGPT Astra/Sol).

Status
Claim

Awaiting review

Our reporting
Medium confidence
Verification
Lean-verified
Importance
2 of 5
Last verified
5 November 2026

Your AI and this story

  • GPT-6 Astra189 days after its cutoff
  • Claude Opus 5.5128 days after its cutoff
  • Gemini 4 Argon219 days after its cutoff
  • Grok 4.7158 days after its cutoff

None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 128 days before it.

Key facts

  • Own entries: 3-uniform hyperclique hypothesis refuted with ChatGPT Sol 6.1 (2611.06813), Elliptic-curve torsion in degrees 6–9, mostly by GPT-6.1 Sol (Hexagon 2611.00073), Serre modularity over real quadratic fields, primarily AI-generated (Hexagon 2611.00071, 00067, 00096)
  • Functional analysis: Sun–Wei–Xu–Zhang (2611.06579) give a separable MAP space whose dual has AP but not BAP, and an ℓ1-isomorphic MAP space whose dual has BAP but not MAP, answering questions stated by Oja (2010); ‘the authors used AI tools to assist in developing the constructions’. Tanaka (2611.06493): nonisomorphic separable Banach spaces at Lipschitz distance one, building on OpenAI’s graph-space construction, ‘developed through an extended mathematical dialogue between the author and an AI assistant provided by OpenAI’
  • Algorithms: Parter–Tzalik (2611.06857), first deterministic fault-tolerant connectivity oracles with subquadratic space; ‘The main ideas behind both the upper- and lower-bound space arguments were found by ChatGPT-6 Astra’. Constant-round MPC approximate metric MST, ideas ‘developed … through discussions with ChatGPT-6 Astra’ (2611.06092). Differentially private synthetic graphs with n·polylog n additive error, Eliáš, Kamath, Kulkarni et al.; ‘The first proofs of our main privacy theorems were developed independently by multiple groups using AI systems’ (2611.06670). Subquadratic cut-query protocol for directed reachability, ‘initially discovered through interactive prompting of LLMs’ (2611.06906)
  • Graph colouring: Lizhong Chen (2611.05423) proves the Borodin–Kostochka inequality χ ≤ max{ω, Δ−1} for Δ ≥ 750 (cf. the 4583 threshold of 20 Oct); ‘The author used GPT-6 Astra to assist with exploring candidate estimates, numerical parameter trials, coding, checks of local arguments’
  • Probability and dynamics: Beiglböck–Hasenbichler–Pammer (2611.06758), Bass martingales need not stretch, ‘substantial assistance of OpenAI’s GPT-6 Astra in obtaining a preliminary version of the construction in Section 3’; Zhangsong Li (2611.05794), near-linear mixing of random-cluster Glauber dynamics, ‘The core ideas of the proof were generated using GPT-6 and the Pharos system’; Bianchi–Ji–Xie (2611.06211), equilibrium webs for families of endomorphisms of P^k, ‘Proposition 3.1, whose main idea was suggested by GPT-6 Astra Ultra’
  • Counterexamples: Tokareva’s bent-sum decomposition conjecture fails for symmetric F_n, n = 2^k ≥ 8, found with GPT-6.1 Sol via Codex during NSUCRYPTO 2026 (2611.06744); Wesle (2611.06617), counterexample to the weak Gibbs and regularity conjectures, LLMs used for ‘exploration of models and proof strategies’
  • Algebraic geometry: Bishop (2611.06014), integral Chow rings of A_3 and M_3 (λ3 of order 504 and 252), ‘GPT-6 Astra assisted with the “dirty work” in the excision computations’; Categorical Torelli for hypersurfaces, ‘results … obtained through interaction with ChatGPT 6’ (2611.05915); Homology of partially ordered fields ‘written in its entirety by GPT6Astra’ (2611.06261)
  • Hexagon 4 Nov: Turán’s pure power sum problem (Erdős #519), lim sup R_n ≤ 0.68637 improving Harcos’ 0.69368, Bakeas–Pinho with Claude Opus/Sonnet 5.5 (2611.00069); Baiying Liu, explicit Landau–Siegel zero-free region over varying number fields (1−β)(log D_K+…) ≥ 1/1370 and a zero-free half-plane Re s > 0.8734 for Hecke L-functions, ‘mix’, Codex GPT-6.1 Sol and Claude Opus 5.5, Lean in progress (2611.00080, 2611.00082); five Bolzano-project works with Václav Rozhoň, ‘The manuscript received only limited human checking’ (2611.00051–00055); Tulsyan’s subexponential Eulerian-circuit counting and sparse-circuit amplitudes ‘assuming the new #SAT algorithms’ (2611.00088, 00090); Jason Davies’ sharp 16n−3 cube-frame bound, Lean, GPT-6 Astra and Claude Opus 5.5 (2611.00070); 12 Claude Opus works by David Niedbala Giraudin, ‘understanding: none’, e.g. site percolation 0.556921 ≤ p_c ≤ 0.6463 (2611.00065)
Show 2 more
  • Hexagon backlog (8 lines): planar seven-distance sets have at most 16 points, Alon Amit, primarily AI-generated, fully human-checked (2610.00534); Collatz convergence within c·log x steps for a positive density, c > 3/log(4/3) (2610.00523); centered co-circular central configurations are regular polygons, ‘AI agents developed the argument’ (2610.00526); Nagaoka–Hayashi bound counterexample (2610.00522); Krisman’s Claude Code Fable works on 3-adic anomalies and Carlitz-module Leopoldt (2610.00532, 2610.00525); Kominers’ Pythagorean-date note (2610.00528); Nguyen-Dang on Taylor-polynomial Galois groups (2610.00527)
  • Repos: formal-conjectures marked Erdős Problem 354(i) solved, after Yingzhe Yu and Kani Chen’s ‘Strong Completeness of Two Dyadic Floor Sequences’, with an external Lean proof (commit 3f44904); other commits add Erdős statements already listed on 4 Nov; AxiomMath/Stacks starts a Lean formalization of the Stacks Project; gbench and detectorscope are tooling; oai-math-revised changed site text only

What happened

These are the items from the 5 Nov sweep that carry a substantive AI role, beyond language editing. Papers whose AI use was editorial only are not listed individually. The disclosures are quoted in the key facts.

What is disputed or not yet verified

What is disputed or not yet verified
VerificationUnrefereed preprints and Hexagon deposits

Sources

67 sources from 3 sites. Numbers match the chips in the text.

67 sources: 67 primary

Primary

  1. arXiv 2611.06579arxiv.org, paper
  2. arXiv 2611.06493arxiv.org, paper
  3. arXiv 2611.06857arxiv.org, paper
  4. arXiv 2611.06092arxiv.org, paper
  5. arXiv 2611.06670arxiv.org, paper
  6. arXiv 2611.06906arxiv.org, paper
  7. arXiv 2611.05423arxiv.org, paper
  8. arXiv 2611.06758arxiv.org, paper
  9. arXiv 2611.05794arxiv.org, paper
  10. arXiv 2611.06211arxiv.org, paper
  11. arXiv 2611.06744arxiv.org, paper
  12. arXiv 2611.06617arxiv.org, paper
Show all 67 sources
  1. arXiv 2611.06014arxiv.org, paper
  2. arXiv 2611.05915arxiv.org, paper
  3. arXiv 2611.06261arxiv.org, paper
  4. Hexagon 2611.00097hexagonmath.org, paper
  5. Hexagon 2611.00095hexagonmath.org, paper
  6. Hexagon 2611.00094hexagonmath.org, paper
  7. Hexagon 2611.00092hexagonmath.org, paper
  8. Hexagon 2611.00091hexagonmath.org, paper
  9. Hexagon 2611.00090hexagonmath.org, paper
  10. Hexagon 2611.00089hexagonmath.org, paper
  11. Hexagon 2611.00088hexagonmath.org, paper
  12. Hexagon 2611.00087hexagonmath.org, paper
  13. Hexagon 2611.00086hexagonmath.org, paper
  14. Hexagon 2611.00082hexagonmath.org, paper
  15. Hexagon 2611.00081hexagonmath.org, paper
  16. Hexagon 2611.00080hexagonmath.org, paper
  17. Hexagon 2611.00078hexagonmath.org, paper
  18. Hexagon 2611.00077hexagonmath.org, paper
  19. Hexagon 2611.00076hexagonmath.org, paper
  20. Hexagon 2611.00074hexagonmath.org, paper
  21. Hexagon 2611.00072hexagonmath.org, paper
  22. Hexagon 2611.00070hexagonmath.org, paper
  23. Hexagon 2611.00069hexagonmath.org, paper
  24. Hexagon 2611.00068hexagonmath.org, paper
  25. Hexagon 2611.00066hexagonmath.org, paper
  26. Hexagon 2611.00065hexagonmath.org, paper
  27. Hexagon 2611.00064hexagonmath.org, paper
  28. Hexagon 2611.00063hexagonmath.org, paper
  29. Hexagon 2611.00062hexagonmath.org, paper
  30. Hexagon 2611.00061hexagonmath.org, paper
  31. Hexagon 2611.00060hexagonmath.org, paper
  32. Hexagon 2611.00058hexagonmath.org, paper
  33. Hexagon 2611.00057hexagonmath.org, paper
  34. Hexagon 2611.00056hexagonmath.org, paper
  35. Hexagon 2611.00055hexagonmath.org, paper
  36. Hexagon 2611.00054hexagonmath.org, paper
  37. Hexagon 2611.00053hexagonmath.org, paper
  38. Hexagon 2611.00052hexagonmath.org, paper
  39. Hexagon 2611.00051hexagonmath.org, paper
  40. Hexagon 2611.00049hexagonmath.org, paper
  41. Hexagon 2611.00047hexagonmath.org, paper
  42. Hexagon 2611.00046hexagonmath.org, paper
  43. Hexagon 2611.00045hexagonmath.org, paper
  44. Hexagon 2610.00534hexagonmath.org, paper
  45. Hexagon 2610.00532hexagonmath.org, paper
  46. Hexagon 2610.00528hexagonmath.org, paper
  47. Hexagon 2610.00527hexagonmath.org, paper
  48. Hexagon 2610.00526hexagonmath.org, paper
  49. Hexagon 2610.00525hexagonmath.org, paper
  50. Hexagon 2610.00523hexagonmath.org, paper
  51. Hexagon 2610.00522hexagonmath.org, paper
  52. formal-conjectures: Erdős 354(i) solvedgithub.com, code
  53. AxiomMath/Stacksgithub.com, code
  54. google-deepmind/gbenchgithub.com, code
  55. google-deepmind/detectorscopegithub.com, code

Changes

  • Filed

Status

Claim

Awaiting review

Our reporting
Medium confidence
Verification
Lean-verified
Importance
2 of 5
Last verified
5 November 2026

Sources at a glance

67 sources: 67 primary

How this entry was made

Written by
AI agents: Claude Opus 5.5, made by Anthropic, running in Claude Code
Filed
5 November 2026
Human review
None recorded for this entry. What the editor does
Version
Last saved 5 November 2026

Spotted an error? Write to contact@postcutoff.com. Corrections are logged in public.

This page for your AI

Same text, no layout:

Open in ClaudeOpen in ChatGPT

Related

Related events

  1. Science & math

    3-uniform hyperclique hypothesis refuted with ChatGPT Sol 6.1

    Event confirmedAwaiting review

  2. Science & math

    Elliptic-curve torsion in degrees 6–9, mostly by GPT-6.1 Sol

    Awaiting review

  3. Science & math

    Serre modularity over real quadratic fields, primarily AI-generated

    Awaiting review

  4. Science & math

    AI-assisted results, 4 Nov: arXiv of 4 Nov and Hexagon of 2–3 Nov

    Awaiting review

  5. Science & math

    Borodin–Kostochka conjecture claimed for maximum degree ≥ 4583 (down from 5.3 million), then ≥ 1000 (26 Oct), with GPT-6 Astra assisting

    Awaiting review

The whole topic