{"schema":"postcutoff/event@1","as_of":"2026-11-19T22:11:00+01:00","url":"https://postcutoff.com/e/2026-11-16-ai-assisted-results-november-16-listing/","md":"https://postcutoff.com/e/2026-11-16-ai-assisted-results-november-16-listing/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-11-16-ai-assisted-results-november-16-listing","date":"2026-11-16","date_precision":"day","short_title":"AI-assisted results, 16 Nov","deck":null,"takeaway":"The sweep listed 260 arXiv papers with AI statements, 52 conjecture papers without one, 62 new Hexagon works and 16 result-repo commits. Every PDF was read at its AI statement. Eight results got their own entries; 54 arXiv papers and 45 Hexagon works where AI supplied proofs, constructions, code or Lean are listed here. All are unrefereed.","category":"science","category_label":"Science & math","importance":2,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":[{"n":1,"title":"arXiv 2611.18541: No free lattice contains an infinite ascending chain of copies of FL(3) (DeMeo, Freese)","url":"https://arxiv.org/abs/2611.18541","type":"paper","group":"primary","domain":"arxiv.org"},{"n":2,"title":"arXiv 2611.17375: The four-dimensional Borsuk number is five or six (Chalmers)","url":"https://arxiv.org/abs/2611.17375","type":"paper","group":"primary","domain":"arxiv.org"},{"n":3,"title":"arXiv 2611.17682: Agent-written heuristics versus commercial solvers on unsolved job-shop instances","url":"https://arxiv.org/abs/2611.17682","type":"paper","group":"primary","domain":"arxiv.org"},{"n":4,"title":"arXiv 2611.17542: On the Hofer–Zehnder conjecture for Reeb flows without symmetry","url":"https://arxiv.org/abs/2611.17542","type":"paper","group":"primary","domain":"arxiv.org"},{"n":5,"title":"Hexagon 2611.00443: Automorphisms of the modular curves X_Δ(N) (Najman, Claude Opus 5.5)","url":"https://hexagonmath.org/abs/2611.00443","type":"paper","group":"primary","domain":"hexagonmath.org"},{"n":6,"title":"Hexagon 2611.00442: A counterexample to the Barker–Larman problem in dimension three","url":"https://hexagonmath.org/abs/2611.00442","type":"paper","group":"primary","domain":"hexagonmath.org"},{"n":7,"title":"Hexagon 2611.00473: The sum of the reciprocal cubes of the Fibonacci numbers is irrational","url":"https://hexagonmath.org/abs/2611.00473","type":"paper","group":"primary","domain":"hexagonmath.org"},{"n":8,"title":"Hexagon 2611.00465: Essential enhancements of site percolation in every dimension (Tomar)","url":"https://hexagonmath.org/abs/2611.00465","type":"paper","group":"primary","domain":"hexagonmath.org"},{"n":9,"title":"formal-conjectures #7571: WOWII conjecture 19 solved (Gemini-assisted)","url":"https://github.com/google-deepmind/formal-conjectures/commit/a3aa3ca653a3ed3d9dc52ad9fe1b2c4ed54fa4f3","type":"code","group":"primary","domain":"github.com"},{"n":10,"title":"formal-conjectures #8015: AI audit tracker fixes","url":"https://github.com/google-deepmind/formal-conjectures/commit/3fa745850e48108a793e2fd06e1e5560650a1a50","type":"code","group":"primary","domain":"github.com"}],"official":10,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI","Anthropic","Google","various researchers"],"title":"AI-assisted results in the 16 Nov sweep: quartic hypersurfaces not stably rational, Pixton's 3-spin relations incomplete, Kim's dimension inequality, the Ulam sequence, Jamneshan–Tao, Petty's problem, matrix ultraproducts, additive cubes, plus 62 Hexagon works and 10 backlog lines","summary":"Catalogue for the 16 Nov 2026 sweep. Own entries: very general quartic hypersurfaces are not stably rational in any dimension (constructions by ChatGPT), a cohomological relation on M̄₂,₁₄ outside Pixton's 3-spin relations (Codex agent team), Kim's dimension inequality for curves over ℚ (GPT-6 Astra), a subexponential bound for the Ulam sequence (proofs by ChatGPT 6), the Jamneshan–Tao conjecture (GPT-Astra), Petty's polar projection problem (GPT-6), trivial fundamental group of matrix ultraproducts (ideas by ChatGPT 6 Astra) and additive cubes over {0,1,2,3} (ChatGPT's morphism). Qiyue Tang's 15 Nov Hexagon batch (HLRV, inductive Alperin–McKay) was added to his entry and a Yang–Mills compilation paper to the Yang–Mills entry.","key_facts":["Method: all 260 lines of 'arXiv papers that disclose or mention AI use' were uncited (98 of them dated 8–13 Nov and probably re-listed from the 15 Nov sweep, which dropped them; all were read again). Every PDF was downloaded and the AI statement read in the text, not only the sweep's quote; 7 own entries; 54 listed below; 199 dropped as editorial: roughly 105 disclose only editing, literature search, proofreading, figures, routine code or proof-checking, about 15 state that no AI was used, and about 80 (mostly cs, stat, physics and q-bio) study, benchmark or cite language models rather than use them in a result","Conjecture papers without a disclosure (52): PDFs scanned; one disclosure the sweep missed (Jamet–Rosenfeld, own entry); Smale's mean value constant = 4 (2611.17373, 'done overwhelmingly by AI models') was already an entry and gets its arXiv link; the rest show no AI use. Bläser–Czerner's 'Go under Chinese rules … is complete for exponential space' (2611.18201) has no AI involvement and goes to the off-scope file","Algebra, geometry and number theory: free lattices contain no infinite ascending chain of copies of FL(3), answering R. A. Dean's question, proofs and first draft 'produced by large language models working as agents' (Claude Fable 5.1, Claude Opus 5.5, GPT-6-Astra in Codex, Kimi K3) under William DeMeo with Ralph Freese ([2611.18541](https://arxiv.org/abs/2611.18541)); moduli of canonical threefolds of minimum volume, 'OpenAI's GPT models were used throughout the mathematical development' ([2611.18357](https://arxiv.org/abs/2611.18357)); quadratic APN functions of F₂¹⁰, Claude wrote and ran the code and drafted all sections, 298 types settled by certificates ([2611.18358](https://arxiv.org/abs/2611.18358)); Rollenske on irrational Seshadri constants, 'Claude pointed out the connection that leads to Theorem A' ([2611.18289](https://arxiv.org/abs/2611.18289)); Helly theorem for the Bruhat–Tits building, 'the statement and proof of Theorem 3.5 were proposed by … GPT-5.6 Sol' (Kaveh et al.; [2611.17844](https://arxiv.org/abs/2611.17844)); nuclear dimension of uniform Roe algebras, first proof by GPT-6.1 Sol via the Lovász Local Lemma, made deterministic by the authors ([2611.18311](https://arxiv.org/abs/2611.18311)); counterexample to Arveson's finiteness question for subdiagonal algebras, arguments refined with GPT-6 ([2611.17636](https://arxiv.org/abs/2611.17636)); rigidity of pivotal braided categories of moderate growth via 'a ChatGPT-generated proof, prompted by Bin Gui' ([2611.17732](https://arxiv.org/abs/2611.17732)); extreme central L-values of quadratic twists, 'ChatGPT introduced Richert's logarithmic weighted sieve to reduce the bound to five' prime factors (Shenghao Hua; [2611.17848](https://arxiv.org/abs/2611.17848)); Chow stability of K3 surfaces with ADE singularities, 'an LLM to assist in developing the proof' ([2611.17901](https://arxiv.org/abs/2611.17901)); character conductors and Sylow subgroups, ChatGPT proved the abelian case and found Example 4.1 ([2611.17993](https://arxiv.org/abs/2611.17993)); universal quadratic forms, Claude Opus 5.5 proved Lemma 2.6 and Sections 4.1–4.2 ([2611.17965](https://arxiv.org/abs/2611.17965)); Wasserstein geometry of cycle spaces, ChatGPT 5.6 supplied Lemma 2.16 and the lower bound of Theorem 3.17 ([2611.17496](https://arxiv.org/abs/2611.17496)); exceptional sequences on Fl(1,ℓ;ℓ+1), GPT-6 found configurations ([2611.17782](https://arxiv.org/abs/2611.17782)); Hodge–Riemann intersection inequalities discovered with the Danus multi-agent system (GPT-5.5, GPT-5.6 Sol, GPT-6 Astra) ([2611.17841](https://arxiv.org/abs/2611.17841)); motivic invariants and the plane Cremona group, AI found the sextic candidate ([2611.18222](https://arxiv.org/abs/2611.18222)); the Hofer–Zehnder conjecture for Reeb flows on S⁵ (exactly three prime closed orbits), Claude Opus 5.5 and Gemini 3.1 Pro 'assist in generating proofs' (Çineli, Ginzburg, Gürel; [2611.17542](https://arxiv.org/abs/2611.17542)); quasicentral modulus for almost normal operators with Danus and Pharos agents ([2611.15604](https://arxiv.org/abs/2611.15604)); Kourovka Notebook problems 18.62(a), 18.63, 18.65 and 20.120, 'An AI language model assisted in preparing the initial manuscript' ([2611.18465](https://arxiv.org/abs/2611.18465))","Combinatorics and discrete maths: four-dimensional Borsuk number is 5 or 6, Lean-checked, 'OpenAI's Astra assisted with exploring proofs, preparing the Lean formalisation' (Logan Chalmers; the independent announcement mentioned in the 15 Nov listing; [2611.17375](https://arxiv.org/abs/2611.17375)); Goodman-type bounds for homomorphism densities, initial proofs by GPT-5.5 and GPT-6 Pro, 117 cases by GPT-5.6 Codex, Lean by Claude ([2611.17885](https://arxiv.org/abs/2611.17885)); real separated subset sums f(7) = 44, f(8) = 84, computations by Claude and a second checker by Astra (Esther Notik; [2611.17360](https://arxiv.org/abs/2611.17360)); higher Bruhat orders, 'ChatGPT Astra was used at all stages of the project, except for asking the questions' (Daria Poliakova; [2611.17465](https://arxiv.org/abs/2611.17465)); no quasi-symmetric 2-(33,15,35) design, GPT-6 Astra helped develop the proofs (Masaaki Harada; [2611.17678](https://arxiv.org/abs/2611.17678)); Schreier sets and Fibonacci×Catalan numbers, ChatGPT drafted proofs, answering a problem in the negative ([2611.17657](https://arxiv.org/abs/2611.17657)); nearly generalised polygons, ChatGPT 'gave us the main idea of Proposition 2.12' ([2611.17855](https://arxiv.org/abs/2611.17855)); n-dimensional recurrences over p-groups in an 'AI-assisted workflow' ([2611.17358](https://arxiv.org/abs/2611.17358)); zero-sum partitions with ChatGPT-drafted proofs ([2611.18059](https://arxiv.org/abs/2611.18059)); TSP instances needing exponentially many subtour constraints, constructions developed 'in concert with Claude' ([2611.17386](https://arxiv.org/abs/2611.17386)); color refinement on outerplanar graphs, proof shortened and Rocq proof built with Claude ([2611.18391](https://arxiv.org/abs/2611.18391)); interchange-process rigidity, ChatGPT 6 Sol discovered Theorem 2.3 and several lemmas ([2611.18483](https://arxiv.org/abs/2611.18483))","Analysis, probability and physics: harmonic-map counterexamples to Eells' conjecture, Claude built the candidate function and drafted the computer-assisted proof, Codex reviewed ([2611.17697](https://arxiv.org/abs/2611.17697)); counterexamples to the clamped-plate eigenvalue ratio conjecture, ChatGPT 6.1 Sol helped find examples ([2611.17606](https://arxiv.org/abs/2611.17606)); mean value sets for elliptic operators, AI 'especially' developed the counterexample in Section 7 (Armstrong, Blank; [2611.15988](https://arxiv.org/abs/2611.15988)); sharp finite-time blow-up for unforced axisymmetric Euler without swirl, data in all C^α, α < 1/3, ChatGPT 'used to optimize arguments' (Bang, Cheskidov; [2611.18151](https://arxiv.org/abs/2611.18151)); Fer expansion converges almost twice as far as Magnus for qubits, Claude Code suggested the tools ([2611.18249](https://arxiv.org/abs/2611.18249)); fast mixing from strong decay of correlations, ChatGPT helped develop the proofs ([2611.17433](https://arxiv.org/abs/2611.17433)); tight quantum query bounds for collision finding, 'GPT-6 Astra developed the detailed proofs' ([2611.17804](https://arxiv.org/abs/2611.17804)); cone of states over time, 'LLM assistance has been used to derive the results' ([2611.18001](https://arxiv.org/abs/2611.18001)); Müller density-matrix functional asymptotics with ChatGPT and Gemini (Heinz Siedentop; [2611.18018](https://arxiv.org/abs/2611.18018)); mean-field Ising propagation of chaos, ChatGPT 6 gave a less technical proof ([2611.17911](https://arxiv.org/abs/2611.17911)); high-dimensional Gaussian bootstrap, 'Artificial intelligence was used … to develop proofs' ([2611.17923](https://arxiv.org/abs/2611.17923)); Borges's question on bilinear spherical averages, ChatGPT and Codex drafted parts ([2611.18061](https://arxiv.org/abs/2611.18061)); DLA on random planar maps, ChatGPT 5.6 and 6 Pro expanded the proof sketch ([2611.18540](https://arxiv.org/abs/2611.18540)); Painlevé quadrics for dKP, Claude Opus 4.8 in the proof of Theorem 6 ([2611.18209](https://arxiv.org/abs/2611.18209)); link invariants in nilpotent gauge theories, Claude assisted derivations ([2611.18237](https://arxiv.org/abs/2611.18237)); causal tail coefficients, Opus 5.5 and ChatGPT 6 refined Lemmas 4.4–4.5 ([2611.18225](https://arxiv.org/abs/2611.18225)); a large ultrafilter number at ℵ_ω, 'no human intelligence was involved in writing this paper. All ideas are due to GPT 6 Astra' (Fanxin Wu; [2611.17643](https://arxiv.org/abs/2611.17643))","Algorithms and formal methods: agent-written job-shop heuristics beat OptalCP, Hexaly and CP Optimizer on tai_100_1000_1 (525,515), improved all ten instances of the family and closed two ([2611.17682](https://arxiv.org/abs/2611.17682)); smoothed analysis to backward error, ChatGPT 5.6–6.1 and Gemini 4 Argon assisted the proofs ([2611.18545](https://arxiv.org/abs/2611.18545)); tight bounds for k-median++, ChatGPT gave initial proofs, and a ChatGPT-claimed constant ≈ 4.162 for k-means++ is left out because its certificate was not verified ([2611.18544](https://arxiv.org/abs/2611.18544)); cost-preserving rounding for laminar thin trees with ChatGPT-6 ([2611.18399](https://arxiv.org/abs/2611.18399)); long rational and Gaussian-integer multiplication, AI-assisted search for evaluation sequences ([2611.18028](https://arxiv.org/abs/2611.18028)); HOL in HOL soundness and completeness, Isabelle scripts drafted by Claude (Benzmüller, Kirchner; [2611.17884](https://arxiv.org/abs/2611.17884)); anisotropic eikonal narrow-band method, Claude found the main gap during Lean formalization (Mirebeau et al.; [2611.17947](https://arxiv.org/abs/2611.17947)); agentic evolutionary search for fault-tolerant Clifford circuits ([2611.18504](https://arxiv.org/abs/2611.18504))","Hexagon, 62 new works (15 Nov): 8 by Qiyue Tang (HLRV conjecture, inductive Alperin–McKay condition, a counterexample to the Huang–Huh–Michałek–Wang–Wang realizability conjecture, Cuenca's Schur-nonnegativity conjecture, Markoff–Hurwitz results) added to [Tang's entry](/e/2026-11-13-qiyue-tang-gpt-6-pro-representation-theory-conjectures/); Zeng Chunping's Kimi-assisted compilation of the Yang–Mills construction added to the Yang–Mills entry; 21 by Tejasvi Singh Tomar ('mixed', all understood, no tools named), among them Conjecture 1 of Balister–Bollobás–Riordan on essential enhancements for site percolation in every d ≥ 2 ([2611.00465](https://hexagonmath.org/abs/2611.00465)), three-copy undistillability of Werner states ([2611.00464](https://hexagonmath.org/abs/2611.00464)), connective constant of ℤ² ≤ 2.6586513 ([2611.00474](https://hexagonmath.org/abs/2611.00474)), Chvátal–Sankoff γ₂ ≤ 0.821963 ([2611.00467](https://hexagonmath.org/abs/2611.00467)), Romanoff's constant ≥ 0.110825 ([2611.00472](https://hexagonmath.org/abs/2611.00472)), the beat-the-average game C ≤ 0.40069634, less than 10⁻⁶ above the conjectured value ([2611.00476](https://hexagonmath.org/abs/2611.00476)), and bounds for Moser's worm, the snake-corridor sofa, the Bohr radius of the bidisc, Mahler-measure limit points and the Schur–Siegel–Smyth trace problem","Hexagon, other works: Filip Najman, automorphisms of the modular curves X_Δ(N), 'Claude Opus 5.5 produced almost all of the mathematical content of the first versions of the proofs' and the Lean files; finds a level-63 curve with previously unnoticed non-modular automorphisms ([2611.00443](https://hexagonmath.org/abs/2611.00443)); Sihan Chen, a counterexample to the Barker–Larman problem in dimension three, 'a result that the AI unintentionally produced' (ChatGPT GPT 6 Astra) ([2611.00442](https://hexagonmath.org/abs/2611.00442)); Shawn Krisman, Σ 1/F_n³ is irrational and the same for Lucas sequences of quadratic units, Claude Opus 5.5, 'Primarily AI-generated', none checked by a human, Lean claimed ([2611.00473](https://hexagonmath.org/abs/2611.00473), [2611.00491](https://hexagonmath.org/abs/2611.00491)); András Pongrácz refutes the existential Thomas conjecture even under the Ramsey hypothesis, human-written with GPT-6 Astra listed ([2611.00444](https://hexagonmath.org/abs/2611.00444)); Alberto Gandolfi (NYU Abu Dhabi), bulk stress tensor and Ward identities for critical percolation, '2-300 interactions' with GPT 6 Pro, checked by Fable 5.2 Max, 105 pages ([2611.00497](https://hexagonmath.org/abs/2611.00497)); Chen–Wang, point–unit-circle incidences, GPT-6 Astra, 178 pages ([2611.00452](https://hexagonmath.org/abs/2611.00452)); Takloo-Bighash–Chimni, subrings of ℤ[x]/(x⁵) with ChatGPT 6 Astra ([2611.00436](https://hexagonmath.org/abs/2611.00436)); Samir Hamad (Kiel), an MLUR Banach space with the Daugavet property, GPT-6 Astra ([2611.00453](https://hexagonmath.org/abs/2611.00453)); Niedbala Giraudin, average sensitivity of PTFs ≤ (6/√(2π)) d√n, extending 'OpenAI's weighted-grading argument', Claude ([2611.00439](https://hexagonmath.org/abs/2611.00439)); Teo Banica's two Claude/GPT-written papers ([2611.00456](https://hexagonmath.org/abs/2611.00456), [2611.00446](https://hexagonmath.org/abs/2611.00446)); Codex-written works by Zhen Gao (Ekeland–Hofer capacities, [2611.00454](https://hexagonmath.org/abs/2611.00454)), Daniel Wang (Miller–Monticone OQ14, [2611.00438](https://hexagonmath.org/abs/2611.00438)), Brice Pouly ([2611.00447](https://hexagonmath.org/abs/2611.00447)), Sunay Venkat ([2611.00462](https://hexagonmath.org/abs/2611.00462), [2611.00463](https://hexagonmath.org/abs/2611.00463)) and twinheterodox ([2611.00440](https://hexagonmath.org/abs/2611.00440)); also Lavi (vector packing, GPT Sol/Astra, Fable; [2611.00449](https://hexagonmath.org/abs/2611.00449)), Rui Chen ([2611.00487](https://hexagonmath.org/abs/2611.00487)), Besevic ([2611.00495](https://hexagonmath.org/abs/2611.00495)), Heffern ([2611.00434](https://hexagonmath.org/abs/2611.00434)), Sivarajan with Claude Opus 5.5 ([2611.00433](https://hexagonmath.org/abs/2611.00433)), Frank Vega (Claude-written check of Nicolas' criterion; [2611.00455](https://hexagonmath.org/abs/2611.00455)). Dropped as editorial: 8 (human-written works without AI tools, an expository essay, two Riemann-hypothesis reformulations by one author, a Collatz note)","Hexagon backlog (10 lines of the 29 Oct file): Michael Mislan, productive autocatalytic cores NP-complete and D-stability coNP-hard (a question of Chen, Fan and Yu, 1995), 'primarily AI-generated' with Lean ([2610.00390](https://hexagonmath.org/abs/2610.00390), [2610.00368](https://hexagonmath.org/abs/2610.00368)); Djalil Chafaï, Gaussian weights on fractals and Edelman's ghosts and shadows, ChatGPT ([2610.00347](https://hexagonmath.org/abs/2610.00347)); Shawn Krisman, a Lean formalization of Roy's strong six exponentials theorem in a single Claude Code session ([2610.00338](https://hexagonmath.org/abs/2610.00338)); Sabrina Hackfort, optimal test functions for analytic ranks, Claude ([2610.00346](https://hexagonmath.org/abs/2610.00346)); four Brice Pouly works on multiplicative sequences and power-residue runs with GPT-6 Astra ([2610.00361](https://hexagonmath.org/abs/2610.00361), [2610.00362](https://hexagonmath.org/abs/2610.00362), [2610.00332](https://hexagonmath.org/abs/2610.00332), [2610.00336](https://hexagonmath.org/abs/2610.00336)); Anwesh Ray's Drinfeld-module paper declares no AI and is dropped","Result repos (16 google-deepmind/formal-conjectures commits, 14–16 Nov): mostly Lean links for known results (R(5,5) ≥ 43, Banach–Mazur in finite dimension, AME(12,5), OEIS congruences, Erdős 935 and 1194 variants, Green 27 and 35). New: Graffiti 'Written on the Wall II' conjecture 19 marked solved with a Palomar-registered Lean proof, 'Developed with AI assistance (Gemini)' (#7571); a positive-lower-density variant of Erdős 338 proved by its contributor with Claude as assistant (#7779); six misformalized statements fixed from an 'AI audit tracker' (#8015). No openai/math or anthropics/formal-math commits","Reactions to last week's results (searched 16 Nov: Shtetl-Optimized, Kalai, Tao, Not Even Wrong, Mathstodon tags and accounts, HN Algolia): none found for k-Clique, ℤ in ℚ, Dvir's LCCs, the groupoid models or the Gödel workbooks beyond what their entries hold. The Xena Project blog post '…and the proof is checked in Lean' (12 Nov, HN 16 Nov, 42 points) was added to the 722-manuscripts entry. Related, written by the news agent: OpenAI's Standard Model lattice result with Nathaniel Craig (2026-11-15-openai-standard-model-lattice-craig-gpt-6-astra)"],"key_numbers":[],"tags":["math","computer-science","physics","arxiv","hexagon","catalogue","gpt-6-astra","gpt-6-pro","chatgpt","claude-opus-5-5","claude-fable","codex","gemini","lean"],"science":null,"body_md":"## What happened\n\nThe 16 Nov sweep listed the first full arXiv day after the weekend plus 62 Hexagon works from 15 Nov. Each arXiv PDF was read at\nits AI statement. Eight results got their own entries; the rest of the substantive ones are grouped above by field. Statements\nare quoted from the papers or from Hexagon's provenance fields. Hexagon does not judge correctness, and many deposits declare\nthat no human has checked them.\n\n## Why it matters\n\nThe day's papers include several long-open problems where authors credit models with the main construction or proof (quartic\nhypersurfaces, Pixton's relations, the Ulam sequence, matrix ultraproducts, additive cubes, free lattices). The same day's\nHexagon batch shows the other end: dozens of deposits that no human has checked.","disputed":[],"related":[{"id":"2026-11-15-ai-assisted-results-november-15-listing","url":"https://postcutoff.com/e/2026-11-15-ai-assisted-results-november-15-listing/","date":"2026-11-15","date_precision":"day","short_title":"AI-assisted results, 15 Nov","deck":null,"takeaway":"arXiv posts no listings at weekends and Hexagon had no new works, so this sweep read the 366 arXiv papers of 7–13 Nov with AI statements that no file cited, line by line in the PDFs, and 59 conjecture papers without a statement. Five got their own entries; 115 with a substantive AI role are listed here. All are unrefereed.","category":"science","category_label":"Science & math","importance":2,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":9,"official":9,"filed":"2026-11-15","updated":"2026-11-15","orgs":["OpenAI","Anthropic","Google","Axiom Math","various researchers"]},{"id":"2026-11-13-chabauty-kim-dimension-inequality-zehavi-gpt-6-astra","url":"https://postcutoff.com/e/2026-11-13-chabauty-kim-dimension-inequality-zehavi-gpt-6-astra/","date":"2026-11-13","date_precision":"day","short_title":"Kim's dimension inequality claimed (Zehavi, GPT-6 Astra)","deck":null,"takeaway":"Minhyong Kim's non-abelian Chabauty method proves finiteness of rational points once a dimension inequality holds, which was known only in special cases such as CM Jacobians. A single-author preprint claims it for every curve of genus ≥ 2 over ℚ with a rational point, and says the main bound came from a 'lengthy back-and-forth interaction with GPT 6 Astra'. Unrefereed.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":2,"official":2,"filed":"2026-11-16","updated":"2026-11-18","orgs":["OpenAI","Max Planck Institute for Mathematics"]},{"id":"2026-11-13-petty-polar-projection-problem-solved-gpt-6","url":"https://postcutoff.com/e/2026-11-13-petty-polar-projection-problem-solved-gpt-6/","date":"2026-11-13","date_precision":"day","short_title":"Petty's polar projection problem answered (GPT-6 assisted)","deck":null,"takeaway":"Petty asked in 1971 whether a convex body in ℝⁿ, n ≥ 3, must be an ellipsoid if its projection body is a dilate of its polar. A single-author preprint says yes; 'GPT-6 was used to assist with proof development'.","category":"science","category_label":"Science & math","importance":3,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI","Tongji University"]},{"id":"2026-11-13-matrix-ultraproducts-trivial-fundamental-group-popa-question-astra","url":"https://postcutoff.com/e/2026-11-13-matrix-ultraproducts-trivial-fundamental-group-popa-question-astra/","date":"2026-11-13","date_precision":"day","short_title":"Matrix ultraproducts: trivial fundamental group (Astra's ideas)","deck":null,"takeaway":"Two operator algebraists prove that every tracial ultraproduct of matrix algebras has trivial fundamental group, answering Sorin Popa's question whether there are uncountably many non-isomorphic matrix ultraproducts. They attribute the main ideas to ChatGPT 6 Astra and say they rewrote its hard-to-read proof entirely.","category":"science","category_label":"Science & math","importance":3,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI","Chiba University","UC Irvine"]},{"id":"2026-11-13-additive-cubes-avoidable-0123-jamet-rosenfeld-chatgpt","url":"https://postcutoff.com/e/2026-11-13-additive-cubes-avoidable-0123-jamet-rosenfeld-chatgpt/","date":"2026-11-13","date_precision":"day","short_title":"Additive cubes avoidable over {0,1,2,3}","deck":null,"takeaway":"Whether an infinite word over {0,1,2,3} can avoid three consecutive blocks with equal length and equal sum was open since 1994. Jamet and Rosenfeld prove it can, with eight explicit morphisms. They write that ChatGPT (GPT-6) 'played a notable role in initiating this work' by proposing the first candidate morphism, without proof.","category":"science","category_label":"Science & math","importance":3,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI","Université de Lorraine","Université de la Polynésie française"]},{"id":"2026-11-13-qiyue-tang-gpt-6-pro-representation-theory-conjectures","url":"https://postcutoff.com/e/2026-11-13-qiyue-tang-gpt-6-pro-representation-theory-conjectures/","date":"2026-11-13","date_precision":"day","short_title":"Six unchecked GPT-6-Pro proofs of named conjectures (Qiyue Tang)","deck":null,"takeaway":"In one day a Tsinghua contributor posted AI-written proofs of several well-known conjectures in representation theory and geometry, each produced in GPT-6-Pro sessions of 2 to 110 hours. Every deposit declares that no human has checked or understood it; reviews by the Rethlas system are promised later.","category":"science","category_label":"Science & math","importance":3,"confidence":"low","status":{"key":"pending","labels":["Awaiting review"]},"sources":7,"official":7,"filed":"2026-11-14","updated":"2026-11-16","orgs":["OpenAI","Tsinghua University"]},{"id":"2026-11-12-pixton-3-spin-relations-incomplete-m2-14-codex-agents","url":"https://postcutoff.com/e/2026-11-12-pixton-3-spin-relations-incomplete-m2-14-codex-agents/","date":"2026-11-12","date_precision":"day","short_title":"Pixton's 3-spin relations incomplete in cohomology (M̄₂,₁₄)","deck":null,"takeaway":"Pixton's conjecture says the 3-spin relations give all relations in the tautological ring of the moduli space of curves. A paper by six specialists, including Pixton, constructs a relation in cohomology on M̄₂,₁₄ that is not among them. It began when Bochao Kong noticed a Codex agent writing a file named potential_counterexample.md while it was trying to prove the opposite.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI"]},{"id":"2026-11-12-ulam-sequence-grows-subexponentially-kravitz-chatgpt","url":"https://postcutoff.com/e/2026-11-12-ulam-sequence-grows-subexponentially-kravitz-chatgpt/","date":"2026-11-12","date_precision":"day","short_title":"Ulam sequence grows subexponentially (ChatGPT 6)","deck":null,"takeaway":"Ulam's 1964 sequence had only exponential upper bounds (best growth rate 1.454ⁿ, 2025). A preprint proves u_n ≤ e^{C√n}. The author states that ChatGPT 6 produced the successive bounds 'with minimal substantive human input'.","category":"science","category_label":"Science & math","importance":3,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI"]},{"id":"2026-11-12-jamneshan-tao-conjecture-nilspaces-gpt-astra","url":"https://postcutoff.com/e/2026-11-12-jamneshan-tao-conjecture-nilspaces-gpt-astra/","date":"2026-11-12","date_precision":"day","short_title":"Jamneshan–Tao conjecture proved (nilspaces, GPT-Astra)","deck":null,"takeaway":"Three nilspace theorists claim the Jamneshan–Tao conjecture and an inverse theorem for Gowers norms on all second-countable compact abelian groups. They write that, after their own U⁴-norm results, collaboration with GPT-Astra 'produced an initial version of the proof', which they refined over many iterations.","category":"science","category_label":"Science & math","importance":3,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI"]},{"id":"2026-11-11-quartic-hypersurfaces-not-stably-rational-all-dimensions-chatgpt","url":"https://postcutoff.com/e/2026-11-11-quartic-hypersurfaces-not-stably-rational-all-dimensions-chatgpt/","date":"2026-11-11","date_precision":"day","short_title":"Quartic hypersurfaces not stably rational in every dimension","deck":null,"takeaway":"Whether a very general quartic hypersurface can be stably rational was settled only up to dimension 6. A preprint of 11 Nov 2026 claims every dimension, reducing the rationality problem for very general hypersurfaces to cubics. The authors write that 'the final constructions presented in Sections 3 and 4 were found by ChatGPT', while the theoretical framework is human.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-16","updated":"2026-11-16","orgs":["OpenAI","Universität Hamburg","New York University"]},{"id":"2026-11-11-smale-mean-value-constant-equals-4-goel-lean","url":"https://postcutoff.com/e/2026-11-11-smale-mean-value-constant-equals-4-goel-lean/","date":"2026-11-11","date_precision":"day","short_title":"Smale's mean value constant = 4 (claimed, Lean-checked)","deck":null,"takeaway":"Two months after GPT-6 Astra refuted the constant-1 form, a Hexagon deposit says the best constant in Smale's mean value theorem is exactly 4, so Smale's 1981 bound cannot be improved at all; a Lean proof accompanies it, experts have not yet commented.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":6,"official":6,"filed":"2026-11-12","updated":"2026-11-16","orgs":["Anthropic","OpenAI"]},{"id":"2026-10-26-openai-yang-mills-qcd-existence-claim","url":"https://postcutoff.com/e/2026-10-26-openai-yang-mills-qcd-existence-claim/","date":"2026-10-26","date_precision":"day","short_title":"OpenAI claims 4D Yang–Mills and QCD existence","deck":null,"takeaway":"The physics third of OpenAI's 26 Oct announcement claims that certain interacting 4D quantum field theories exist mathematically, among them pure Yang–Mills and quantum chromodynamics. That is the existence half of the Yang–Mills Millennium problem, not the mass gap. A group led by David Gross, with Michael Douglas, Arthur Jaffe, Igor Klebanov and Nikita Nekrasov, is reviewing it. At the group's request the AI-produced material is not public while they write their own account. Nothing can be checked yet.","category":"science","category_label":"Science & math","importance":5,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":22,"official":5,"filed":"2026-10-27","updated":"2026-11-16","orgs":["OpenAI"]}],"people":[{"id":"nathaniel-craig","name":"Nathaniel Craig","url":"https://postcutoff.com/person/nathaniel-craig/"},{"id":"zeev-dvir","name":"Zeev Dvir","url":"https://postcutoff.com/person/zeev-dvir/"}],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-11-16","type":"filed","text":"Created (science agent; 260 + 52 arXiv PDFs, 62 Hexagon provenance records, 10 backlog lines, 16 commits)"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-11-16","run":null,"sources_read":null,"updated":"2026-11-16","human_review":null,"version":{"date":"2026-11-16"}},"gaps":[{"model_id":"gpt-6-2-astra","name":"GPT-6.2 Astra","cutoff":"2026-05","days_after":169,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":139,"in_training_data":false},{"model_id":"gemini-4-hydrogen","name":"Gemini 4 Hydrogen","cutoff":"2026-03","days_after":230,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":169,"in_training_data":false}],"short_url":null}