{"schema":"postcutoff/event@1","as_of":"2026-11-19T22:11:00+01:00","url":"https://postcutoff.com/e/2026-11-19-ai-assisted-results-november-19-listing/","md":"https://postcutoff.com/e/2026-11-19-ai-assisted-results-november-19-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":"2026-11-19-ai-assisted-results-november-19-listing","date":"2026-11-19","date_precision":"day","short_title":"AI-assisted results, 19 Nov","deck":null,"takeaway":"The sweep listed 379 arXiv papers with AI statements (253 new on 19 Nov), 72 conjecture papers without one (20 new), 71 Hexagon works, 6 result-repo commits and 1 watched repo. Every new line was read and its AI statement checked in the PDF or the Hexagon provenance record. Four results got their own entries, a fifth (Graph Minors in Lean) came from Mathstodon, seven existing entries were extended, and 112 arXiv papers and about 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 2611.24441: An additive-square-free word over an alphabet of size 6 (Shallit)","url":"https://arxiv.org/abs/2611.24441","type":"paper","group":"primary","domain":"arxiv.org"},{"n":2,"title":"arXiv 2611.23641: Almost all vertex-transitive graphs and digraphs are Cayley (Su, Xia)","url":"https://arxiv.org/abs/2611.23641","type":"paper","group":"primary","domain":"arxiv.org"},{"n":3,"title":"arXiv 2611.24386: Even faster algorithms for bounded integer programming (Chen, Mao, Zhang)","url":"https://arxiv.org/abs/2611.24386","type":"paper","group":"primary","domain":"arxiv.org"},{"n":4,"title":"arXiv 2611.23693: Separating QMA from the QCMA hierarchy (Yiming Wang)","url":"https://arxiv.org/abs/2611.23693","type":"paper","group":"primary","domain":"arxiv.org"},{"n":5,"title":"arXiv 2611.24027: Boundary C^{2,γ} regularity for optimal transport potentials (Hong, Wang)","url":"https://arxiv.org/abs/2611.24027","type":"paper","group":"primary","domain":"arxiv.org"},{"n":6,"title":"Hexagon 2611.00653: Ehrhart positivity for polypositroids (Ferroni)","url":"https://hexagonmath.org/abs/2611.00653","type":"paper","group":"primary","domain":"hexagonmath.org"},{"n":7,"title":"Hexagon 2611.00666: Non-existence of non-Eulerian irregular Diophantine quadruples (Dujella, Slijepčević)","url":"https://hexagonmath.org/abs/2611.00666","type":"paper","group":"primary","domain":"hexagonmath.org"},{"n":8,"title":"Hexagon 2611.00652: Deterministic almost-linear time for general matching and network flow (Liu)","url":"https://hexagonmath.org/abs/2611.00652","type":"paper","group":"primary","domain":"hexagonmath.org"}],"official":8,"filed":"2026-11-19","updated":"2026-11-19","orgs":["OpenAI","Anthropic","Google DeepMind","various researchers"],"title":"AI-assisted results in the 19 Nov sweep: additive squares avoided on six integers, the McKay–Praeger conjecture, Graph Minors in Lean, Brakensiek–Guruswami claimed, GPT-6.2 Astra's first research wave","summary":"Catalogue for the 19 Nov 2026 sweep. Own entries: Jeffrey Shallit's additive-square-free word on six integers (GPT-6.2 Astra wrote almost all of it, with Lean), Su and Xia's proof of the McKay–Praeger conjecture (key proposition from GPT-6.1), Yuwen Zhou's Hexagon claim of both Brakensiek–Guruswami promise-colouring conjectures (details checked only by AI), and Hajiaghayi and JafariRaviz's Lean formalization of Graph Minors I–XXII. Extended: Hadamard matrices below 2000 (written version), AKLT (free-boundary follow-up), Xi Chen's NLS cascades (arXiv version), non-split Cartan (level 25), Dave Goel (Černý), Anthropic's black-hole preprints (research post) and the GPT-6.2 Astra amateur wave.","key_facts":["Method: all 253 lines of 'arXiv papers that disclose or mention AI use' dated 19 Nov were read and each PDF scanned for its AI statement (about 75 lines quoted only a heading or unrelated text); the 126 lines dated 11–18 Nov are re-listings read in earlier runs and were not re-read. 2 own entries and 1 extension from this list; 44 listed by field and 68 with smaller roles below; 138 dropped as editorial: about 74 disclose only editing, proofreading, literature search, figures, routine code or proof checking, about 29 state that no AI was used or that results predate it, and about 35 mention AI only in references, funding, the topic or not at all","Conjecture papers without a sweep disclosure: the 20 new PDFs (19 Nov) were scanned in full; none credits AI. Human results of note: a counterexample to the Harper-norm silver-mean conjecture with an exact certificate (Shchukin, van Loock; 2611.23265) and new cases of Beauville's conjecture characterizing Jacobians by theta singularities (Meng, Popa; 2611.24361). The other 52 lines are re-listings","Algebra, number theory and geometry: all F_q-linear relations among positive-characteristic multiple zeta values re-derived, 'it generates all mathematical results in this paper', Lean by GPT-6 Astra (Jinyuan Hu; [2611.23413](https://arxiv.org/abs/2611.23413)); partial flags of tropical linear spaces may not complete, 'This example was found entirely by ChatGPT-6 Astra' (Christopher Eur; [2611.23785](https://arxiv.org/abs/2611.23785)); a Weinstein manifold without any spherical Fukaya category, example found with GPT-6 Astra ([2611.24930](https://arxiv.org/abs/2611.24930)); self-extensions of mod p Jacquet–Langlands, 'Some key ideas of the paper are given by GPT-6 Astra' ([2611.24405](https://arxiv.org/abs/2611.24405)); Bernstein–Zelevinsky duality on Drinfeld spaces, AI's 'main mathematical contribution' is key Lemma 4.14 ([2611.23831](https://arxiv.org/abs/2611.23831)); smooth minimal toric fourfolds not connected by smooth elementary flops, examples developed with GPT-6.1 ([2611.24292](https://arxiv.org/abs/2611.24292)); tropical geometry of G₂, Claude Opus 5.5 drafted results and proofs throughout ([2611.24456](https://arxiv.org/abs/2611.24456)); convergence of the big J-function of GKM manifolds, Claude found the resummation 'with substantial assistance' ([2611.23650](https://arxiv.org/abs/2611.23650)); free products of vector lattices, GPT-5.5/5.6 proofs and Lean ([2611.24528](https://arxiv.org/abs/2611.24528)); regularity-free bisymmetric means, uniqueness proof from ChatGPT 6 Astra ([2611.23300](https://arxiv.org/abs/2611.23300)); ultralogarithm branches, 'Gemini Sector-Based Conjugation Criterion' ([2611.23317](https://arxiv.org/abs/2611.23317)); large entropy implies Hölder regularity, ChatGPT 6 proposed the strategy ([2611.23490](https://arxiv.org/abs/2611.23490)); heights of polarized fibrations, adapted Bismut–Vasserot torsion with ChatGPT ([2611.23739](https://arxiv.org/abs/2611.23739))","Analysis, PDE, probability and physics: boundary C^{2,γ} regularity for optimal transport potentials under A3s, 'The initial proof was generated entirely by OpenAI's GPT-6 Astra through the Danus project' after four days (Hong, Wang; [2611.24027](https://arxiv.org/abs/2611.24027)); rough vector fields with vanishing Lie brackets, 'we were very surprised when GPT-6 Astra suggested (in a single response) the counterexample in the regime p < 2' after the authors had been stuck since 2025 (Bianchini, Talamini; [2611.24175](https://arxiv.org/abs/2611.24175)); sharpness of the Osgood threshold for anomalous dissipation, key ideas with ChatGPT 6 Astra ([2611.24784](https://arxiv.org/abs/2611.24784)); quasimonomial Monge–Ampère for big classes, AI proved Proposition 3.8 and simplified Theorems 4.1 and 4.9 ([2611.24589](https://arxiv.org/abs/2611.24589)); Riesz transforms on homogeneous trees, Theorem 1.1 with ChatGPT-6 ([2611.24683](https://arxiv.org/abs/2611.24683)); orthonormal Strichartz estimates in all remaining cases ([2611.23797](https://arxiv.org/abs/2611.23797)); hydrodynamic limit of the Toda lattice, GPT-5.6 Pro drafted the lower-lying results ([2611.23805](https://arxiv.org/abs/2611.23805)); majorisation extremes, 'ChatGPT generated a complete proof' of the authors' conjectured Theorem 4.12 ([2611.23612](https://arxiv.org/abs/2611.23612)); heterogeneous mean-field steady states, 'Most proofs were generated … through iterative interactions between ChatGPT and the author' ([2611.24123](https://arxiv.org/abs/2611.24123)); SRB measures for Rössler systems, Claude Opus 5.5 computations and Lean ([2611.23279](https://arxiv.org/abs/2611.23279)); long-lived quasinormal modes for multiple black holes, Proposition 4.1 first proved by ChatGPT, its numerical experiments 'not independently verified by the authors' ([2611.24063](https://arxiv.org/abs/2611.24063)); free bosonic CFT integrals, an identification found by Claude review ([2611.23237](https://arxiv.org/abs/2611.23237)); ellipsoid fitting, dual result found with GPT-5.6 ([2611.23431](https://arxiv.org/abs/2611.23431)); camera calibration from right-angle corners, Claude proposed the main proofs ([2611.23934](https://arxiv.org/abs/2611.23934))","Combinatorics, algorithms and complexity: even faster bounded integer programming (Õ(n + N^{1.9602d}) randomized), 'We fed these two papers into GPT-6.2 Astra, and with almost no human input and after several runs, it obtained the result in this paper. The authors take no intellectual credit' (Lin Chen, Yuchen Mao, Guochuan Zhang; [2611.24386](https://arxiv.org/abs/2611.24386)); a classical oracle relative to which QMA is not in the QCMA hierarchy, GPT-6 Astra 'produced nearly the entire proof in a single attempt, which took approximately 10 hours', later simplified by the author (Yiming Wang; [2611.23693](https://arxiv.org/abs/2611.23693)); optimal universal sets for submodular functions, construction 'found using extensive interactions with ChatGPT-6 Astra' ([2611.23637](https://arxiv.org/abs/2611.23637)); (12/11+ε)-approximation for nonpreemptive throughput, ChatGPT 6 Pro ([2611.24201](https://arxiv.org/abs/2611.24201)); sharp quadratic bound for selector advice complexity, Claude and Codex ([2611.24913](https://arxiv.org/abs/2611.24913)); depth hierarchies for symmetric algebraic circuits, Claude Code and Lean (Dawar et al.; [2611.23541](https://arxiv.org/abs/2611.23541)); ordinal Hegselmann–Krause convergence bounds refined by ChatGPT 5.6 Sol ([2611.24449](https://arxiv.org/abs/2611.24449)); subexponential advantage for vanilla adiabatic optimization, 'GPT-6 Luna suggested the main proof strategy' ([2611.24035](https://arxiv.org/abs/2611.24035)); forbidden 0-1 patterns classified with Codex GPT-6 Astra Ultra ([2611.23333](https://arxiv.org/abs/2611.23333)); MSTD set families ([2611.23338](https://arxiv.org/abs/2611.23338)); intersections of odd polygons, ChatGPT found a gap in a published proof and lowered the constant (first to 11) ([2611.24343](https://arxiv.org/abs/2611.24343)); Bapat–Sunder eigenvalue conjecture fails over the reals, search built with GPT-5.6 Sol and Claude Fable 5 ([2611.23269](https://arxiv.org/abs/2611.23269)); degree-three Euclidean spanning trees within 1.514 ([2611.23703](https://arxiv.org/abs/2611.23703)); 'anatomy' of the AI-found Dinitz–Garg–Goemans counterexample, ChatGPT then built counterexamples on 93 other arborescence shapes (Ge et al.; [2611.23244](https://arxiv.org/abs/2611.23244)); downstream citations of AI-generated proofs: removable forests use 'the solution to Erdős-Sós in full by GPT-6 Astra' ([2611.23647](https://arxiv.org/abs/2611.23647)) and proper thinness cites Shitov's GPT-6 Astra Pro NP-completeness manuscript ([2611.23576](https://arxiv.org/abs/2611.23576)); the Borel Ruziewicz problem, now on arXiv, 'The core mathematical ideas and the proofs in this paper were found by AI models' (Konstantin Slutsky; [2611.24066](https://arxiv.org/abs/2611.24066), Hexagon 2611.00309 in the 12 Nov listing)","Smaller or less specific AI roles (a lemma, an example, exploration, code): [2611.23238](https://arxiv.org/abs/2611.23238), [2611.23280](https://arxiv.org/abs/2611.23280), [2611.23295](https://arxiv.org/abs/2611.23295), [2611.23325](https://arxiv.org/abs/2611.23325), [2611.23430](https://arxiv.org/abs/2611.23430), [2611.23437](https://arxiv.org/abs/2611.23437), [2611.23478](https://arxiv.org/abs/2611.23478), [2611.23488](https://arxiv.org/abs/2611.23488), [2611.23579](https://arxiv.org/abs/2611.23579), [2611.23656](https://arxiv.org/abs/2611.23656), [2611.23662](https://arxiv.org/abs/2611.23662), [2611.23681](https://arxiv.org/abs/2611.23681), [2611.23694](https://arxiv.org/abs/2611.23694), [2611.23721](https://arxiv.org/abs/2611.23721), [2611.23733](https://arxiv.org/abs/2611.23733), [2611.23743](https://arxiv.org/abs/2611.23743), [2611.23778](https://arxiv.org/abs/2611.23778), [2611.23814](https://arxiv.org/abs/2611.23814), [2611.23868](https://arxiv.org/abs/2611.23868), [2611.23902](https://arxiv.org/abs/2611.23902), [2611.24040](https://arxiv.org/abs/2611.24040), [2611.24087](https://arxiv.org/abs/2611.24087), [2611.24128](https://arxiv.org/abs/2611.24128), [2611.24131](https://arxiv.org/abs/2611.24131), [2611.24132](https://arxiv.org/abs/2611.24132), [2611.24195](https://arxiv.org/abs/2611.24195), [2611.24237](https://arxiv.org/abs/2611.24237), [2611.24238](https://arxiv.org/abs/2611.24238), [2611.24282](https://arxiv.org/abs/2611.24282), [2611.24319](https://arxiv.org/abs/2611.24319), [2611.24322](https://arxiv.org/abs/2611.24322), [2611.24332](https://arxiv.org/abs/2611.24332), [2611.24341](https://arxiv.org/abs/2611.24341), [2611.24519](https://arxiv.org/abs/2611.24519), [2611.24532](https://arxiv.org/abs/2611.24532), [2611.24563](https://arxiv.org/abs/2611.24563), [2611.24564](https://arxiv.org/abs/2611.24564), [2611.24570](https://arxiv.org/abs/2611.24570), [2611.24575](https://arxiv.org/abs/2611.24575), [2611.24595](https://arxiv.org/abs/2611.24595), [2611.24645](https://arxiv.org/abs/2611.24645), [2611.24652](https://arxiv.org/abs/2611.24652), [2611.24677](https://arxiv.org/abs/2611.24677), [2611.24700](https://arxiv.org/abs/2611.24700), [2611.24712](https://arxiv.org/abs/2611.24712), [2611.24745](https://arxiv.org/abs/2611.24745), [2611.24751](https://arxiv.org/abs/2611.24751), [2611.24760](https://arxiv.org/abs/2611.24760), [2611.24808](https://arxiv.org/abs/2611.24808), [2611.24819](https://arxiv.org/abs/2611.24819), [2611.24820](https://arxiv.org/abs/2611.24820), [2611.24828](https://arxiv.org/abs/2611.24828), [2611.24835](https://arxiv.org/abs/2611.24835), [2611.24842](https://arxiv.org/abs/2611.24842), [2611.24865](https://arxiv.org/abs/2611.24865), [2611.24922](https://arxiv.org/abs/2611.24922), [2611.24934](https://arxiv.org/abs/2611.24934), [2611.23547](https://arxiv.org/abs/2611.23547), [2611.23827](https://arxiv.org/abs/2611.23827), [2611.24227](https://arxiv.org/abs/2611.24227), [2611.24753](https://arxiv.org/abs/2611.24753), [2611.24821](https://arxiv.org/abs/2611.24821), [2611.24866](https://arxiv.org/abs/2611.24866), [2611.24236](https://arxiv.org/abs/2611.24236), [2611.23387](https://arxiv.org/abs/2611.23387), [2611.24167](https://arxiv.org/abs/2611.24167), [2611.24931](https://arxiv.org/abs/2611.24931), [2611.24526](https://arxiv.org/abs/2611.24526)","A reported failure: the product-structure paper of the MATRIX workshop group (no AI in its proofs) writes that 'it seems silly to propose a conjecture without first checking if it can be resolved quickly by an LLM'; GPT-6 Astra and Gemini 4 Argon were asked to settle its Conjecture 23 and 'Neither model succeeded' ([2611.24544](https://arxiv.org/abs/2611.24544))","Hexagon, 71 works in the sweep (plus 21 published after it, read ahead): Zhou's Brakensiek–Guruswami claim has its own entry; Zhong (AKLT), Jaswal (non-split Cartan level 25), Hasani–Nellore–Ward (Hadamard), Dave Goel (Černý 2611.00635 and the four 'miscellanea' records 2611.00513/00527/00546/00548) join existing entries. Listed: Luis Ferroni, Ehrhart positivity for integral polypositroids, 'this proves the positroid conjecture of Ferroni, Jochemko and Schröter', a 117-page first proof by ChatGPT-6 Astra in October with 'quite low' human input, simplified by GPT-6.2 Astra and checked by Claude Opus 5.5 ([2611.00653](https://hexagonmath.org/abs/2611.00653)); Andrej Dujella and Siniša Slijepčević with the AI assistant Aleph, non-Eulerian irregular Diophantine quadruples reduced to nine families with c < 10^99 for every irregular quadruple, final computation 'in progress' ([2611.00666](https://hexagonmath.org/abs/2611.00666)); Yilin Liu, deterministic almost-linear time for general matching and min-cost flow, with ROABP identity testing and an answer to Dey and Kavitha on arborescences, all GPT-6 Astra Pro, 'None of this contribution has been checked and understood by any human' ([2611.00652](https://hexagonmath.org/abs/2611.00652), [2611.00634](https://hexagonmath.org/abs/2611.00634), [2611.00665](https://hexagonmath.org/abs/2611.00665)); Jason Davies with Claude Opus 5.5 and Codex GPT-6.2 Astra, no human check: the shortest word containing every permutation of ten symbols has length 83 ([2611.00630](https://hexagonmath.org/abs/2611.00630)), the equilateral triangle maximises σ₁L among triangles, the n = 3 case of the Steklov analogue of Pólya–Szegő ([2611.00629](https://hexagonmath.org/abs/2611.00629)), odd-cycle homomorphism complexity with Lean ([2611.00633](https://hexagonmath.org/abs/2611.00633)); David Jordan, Christol's conjecture for non-resonant globally bounded ₃F₂ ([2611.00628](https://hexagonmath.org/abs/2611.00628)); Danny Plymesser's Jocelyn harness, consecutive powerful numbers O(x^{1/4+ε}) improving Reuss's 29/100, Lean with six literature hypotheses ([2611.00602](https://hexagonmath.org/abs/2611.00602)); Igor Kortchemski, BKT asymptotics for hierarchical pinning, central argument by ChatGPT ([2611.00654](https://hexagonmath.org/abs/2611.00654)); Fabrice Baudoin, gradient structures on metric measure spaces, 'all the proofs were produced by the model Astra 6.1' ([2611.00669](https://hexagonmath.org/abs/2611.00669)); Gerald Höhn, an extremal self-dual VOA of central charge 64 with GPT 6 pro ([2611.00631](https://hexagonmath.org/abs/2611.00631)); Rob Sullivan, a 'potentially LLM-generated' argument proving the main conjecture of his paper with Winkel ([2611.00610](https://hexagonmath.org/abs/2611.00610)); Hernandez Juan with GPT, Conjecture 8.11 of Gross–Hacking–Keel–Kontsevich disproved ([2611.00649](https://hexagonmath.org/abs/2611.00649)) and Somos-k non-integrality for k ≥ 8 ([2611.00597](https://hexagonmath.org/abs/2611.00597)); Sunay Venkat, GPT-6.2 Astra 'carried out the mathematical research', Erdős's 1976 equal-support predictions disproved (problem 931 stays open; [2611.00599](https://hexagonmath.org/abs/2611.00599)) and Ramanujan primes under abc ([2611.00598](https://hexagonmath.org/abs/2611.00598)); twinheterodox with Codex GPT-6.2-Astra max, irrationality of the alternating reciprocal Fibonacci cube sum ([2611.00613](https://hexagonmath.org/abs/2611.00613)) and the reciprocal Mersenne cube sum ([2611.00643](https://hexagonmath.org/abs/2611.00643)), plus eight related deposits (2611.00600, 00601, 00605, 00615, 00623, 00624, 00638, 00639); Sebastien Palcoux, Grothendieck rings of fusion categories of ranks 2, 3 and 4 with GPT-6/6.2 Astra ([2611.00641](https://hexagonmath.org/abs/2611.00641), [2611.00642](https://hexagonmath.org/abs/2611.00642), [2611.00645](https://hexagonmath.org/abs/2611.00645)); Banica with ChatGPT, Grok, Codex and Claude Fable (2611.00603, 00616, 00632); Antoine Olivier, formal duals for Bierbrauer–Edel configurations, Claude agents, author 'did not check the proofs' ([2611.00626](https://hexagonmath.org/abs/2611.00626)); Marc Sperzel, Claude-produced exclusion of Gaussian covariant classical–quantum gravity by XENONnT X-ray data, likelihood fit made for the paper ([2611.00594](https://hexagonmath.org/abs/2611.00594)); MengHan Tsai, a Claude-produced no-shortcut theorem for moving wormhole mouths, author 'not a professional physicist' ([2611.00646](https://hexagonmath.org/abs/2611.00646)); Yair Lavi, Penrose polynomial coefficient conjectures refuted ([2611.00612](https://hexagonmath.org/abs/2611.00612)); Rui Chen, a non-stable unramified multiplicity-one symmetric pair over Q₁₇ ([2611.00664](https://hexagonmath.org/abs/2611.00664)); Jiarui Fei, Schubert Cox rings developed by ChatGPT 6.0 Pro ([2611.00668](https://hexagonmath.org/abs/2611.00668)); Lüwei Dong and Hongyi Wang, the regular pentagon maximises the first Neumann eigenvalue among simple pentagons ([2611.00671](https://hexagonmath.org/abs/2611.00671)); Niels Feld, Gersten exactness for excellent regular local surfaces, audited 'line by line' ([2611.00625](https://hexagonmath.org/abs/2611.00625)); Luis Dieulefait with Claude as assistant, a converse to Ramakrishnan's Rankin–Selberg automorphy ([2611.00622](https://hexagonmath.org/abs/2611.00622)); Paul Pollack, a variant of Daboussi's elementary PNT proof ([2611.00636](https://hexagonmath.org/abs/2611.00636)); Alberto Gandolfi, CFT of critical percolation II ([2611.00651](https://hexagonmath.org/abs/2611.00651)); Riesen and Evans, Flohr's W-algebras ([2611.00640](https://hexagonmath.org/abs/2611.00640)); Samuil Petkov, eight deposits with GPT-5.6 Sol and Claude Opus 5.5 (2611.00656–00663, among them a four-state Markovian Mpemba effect refuting a triplet-test conjecture); Roy van Rijn, five no-human-check deposits (2611.00593, 00595, 00596, 00672, 00673); Babanskyy (2611.00590), Niedbala Giraudin (2611.00607, 00617, 00618), Turner (2611.00606), Diakin (2611.00604), Fradj (2611.00627), MacGregor (2611.00620), Arhancet (2611.00592), Zeng Chunping (2611.00572, 00608, 00621), Hackbarth (2611.00647), Vilela Jato (2611.00655), Kruer's group (2611.00670). Dropped as editorial: 9 (six human-written with no AI, three with AI for figures, review or a single illustration)","Zenodo (lines in the sweep's Stories list, not edited there): Zihuan Xu, about twelve deposits on 19 Nov with Codex, among them strict rank log-concavity for lattice path matroid quotient intervals, answering a question of Benedetti-Velásquez and Knauer ([24365209](https://zenodo.org/records/24365209)); Yitzchak Shmalo, the logarithmic correction to the relaxation time of the 2D Fredrickson–Andersen model, Conjecture 4.3 of Hartarsky and Toninelli (no AI statement in the record; [24363035](https://zenodo.org/records/24363035)); Elias Botsford, cycle-spectra lower bound for Erdős problem 84 raised to 37.18 × 2^(n/2) ([24356011](https://zenodo.org/records/24356011)); Felipe Santibáñez-Leal's two follow-ups to Alpöge's Jacobian counterexample ([24358311](https://zenodo.org/records/24358311), [24358296](https://zenodo.org/records/24358296)); Maria Smith's 'Fold' deposits are crank material and dropped","GitHub: google-deepmind/efficient_evals is evaluation tooling (no result). Result repos: of the 6 formal-conjectures commits, 4 (16–18 Nov) were read on 18 Nov; the 2 new ones (#8033 Kourovka 20.51, #7598 Erdős 790 with known bounds) add statements, no results. No openai/math or anthropics/formal-math commits. Off-sweep: mahdi-jfri/GraphMinor and StrongPerfectGraphTheorem (own entry); both accounts were added to the sweep's GH_WATCH","GPT-6.2 Astra (released 17 Nov), research uses in its first two days: Shallit (additive squares), Chen–Mao–Zhang (integer programming, no human input), Xi Chen (NLS cascades), Zhou (Brakensiek–Guruswami), Venkat, twinheterodox, Ferroni and Palcoux's revisions, Davis's repo (see the amateur-wave entry)","Reactions (searched 19 Nov: HN Algolia, Mathstodon tag timelines, Terence Tao's blog, Shallit's blog, Erdős Problems pages, web): none found yet for Hayden's quantum memory, AKLT, Reinhardt, Green's conjecture counterexamples, k-Clique, ℤ in ℚ or the new results above. The only expert comment found was Ross Kang's Mathstodon post on the Graph Minors formalization. Anthropic's black-hole post reached 218 points on HN (see its entry)"],"key_numbers":[],"tags":["math","physics","computer-science","arxiv","hexagon","zenodo","catalogue","gpt-6-astra","gpt-6-1-sol","gpt-6-2-astra","chatgpt","claude-opus-5-5","claude-fable-5-2","codex","gemini","lean"],"science":null,"body_md":"## What happened\n\nThe 19 Nov sweep listed 253 new arXiv papers with AI statements and 71 Hexagon works. Each new arXiv line was read and the AI\nstatement found in the PDF; each Hexagon record's provenance was read through the API. Four results got their own entries and\nthe substantive rest are grouped above by field. Statements are quoted from the papers or from Hexagon's provenance fields.\nHexagon does not judge correctness, and many deposits declare that no human has checked them.\n\n## Why it matters\n\nGPT-6.2 Astra's first research outputs arrived within two days of release, including two that settle named problems (additive\nsquares, an integer-programming exponent) where the authors credit the model with almost all of the work. In parallel, a few\nsenior groups (Shallit, Su–Xia, Ferroni, Dujella) now describe the model's share in plain terms, while a growing number of\ndeposits on hard problems (matching in almost-linear time, Brakensiek–Guruswami) have been checked only by AI.","disputed":[],"related":[{"id":"2026-11-18-additive-square-free-word-six-letters-shallit-gpt-6-2-astra","url":"https://postcutoff.com/e/2026-11-18-additive-square-free-word-six-letters-shallit-gpt-6-2-astra/","date":"2026-11-18","date_precision":"day","short_title":"Additive-square-free word on six integers","deck":null,"takeaway":"Can an infinite word over a finite set of integers avoid two adjacent blocks with equal length and equal sum? The question goes back to Justin (1972) and Brown and Freedman (1987), who conjectured it cannot. Shallit gives an explicit word on six positive integers that does, says the paper was 'almost entirely written by GPT 6.2 Astra', and links a Lean verification the model also built. Unrefereed.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":2,"official":2,"filed":"2026-11-19","updated":"2026-11-19","orgs":["OpenAI","University of Waterloo"]},{"id":"2026-11-18-mckay-praeger-conjecture-almost-all-vertex-transitive-cayley-su-xia-gpt-6-1","url":"https://postcutoff.com/e/2026-11-18-mckay-praeger-conjecture-almost-all-vertex-transitive-cayley-su-xia-gpt-6-1/","date":"2026-11-18","date_precision":"day","short_title":"McKay–Praeger conjecture claimed","deck":"Almost all vertex-transitive graphs are Cayley graphs, with the key proposition found by GPT-6.1","takeaway":"McKay and Praeger's tables from the 1990s suggested that almost every vertex-transitive graph is a Cayley graph. Ning Su and Binzhou Xia claim a proof, with exponentially few exceptions, and the digraph analogue. Their disclosure credits GPT-6.1 with 'the discovery of Proposition 4.1 and the supporting arguments developed in Sections 3 and 4'; the rest builds on Xia's earlier unpublished work. Unrefereed.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-19","updated":"2026-11-19","orgs":["OpenAI"]},{"id":"2026-11-18-aklt-square-lattice-spectral-gap-aronow-lopatto-machine-written","url":"https://postcutoff.com/e/2026-11-18-aklt-square-lattice-spectral-gap-aronow-lopatto-machine-written/","date":"2026-11-18","date_precision":"day","short_title":"AKLT square-lattice spectral gap claimed (machine-written)","deck":null,"takeaway":"Affleck, Kennedy, Lieb and Tasaki proposed in 1987–88 that their two-dimensional valence-bond-solid models should be gapped, like the spin-1 chain. The hexagonal (spin-3/2) case was settled by computer-aided and later analytic work; the spin-2 square lattice stayed open. On 18 Nov P. M. Aronow and Patrick Lopatto posted a 147-page proof of a uniform gap for large square-lattice tori. The authors say the whole document is machine-written; Hexagon records that no human has checked it.","category":"science","category_label":"Science & math","importance":4,"confidence":"low","status":{"key":"pending","labels":["Awaiting review"]},"sources":3,"official":3,"filed":"2026-11-18","updated":"2026-11-19","orgs":["Anthropic","OpenAI","Yale University","University of North Carolina at Chapel Hill"]},{"id":"2026-11-18-bourgain-sobolev-growth-cubic-nls-torus-xi-chen-gpt-6-2-astra","url":"https://postcutoff.com/e/2026-11-18-bourgain-sobolev-growth-cubic-nls-torus-xi-chen-gpt-6-2-astra/","date":"2026-11-18","date_precision":"day","short_title":"Unbounded Sobolev growth for cubic NLS on T² claimed (ChatGPT 6.2 Astra)","deck":null,"takeaway":"Bourgain asked whether higher Sobolev norms of solutions to the periodic defocusing cubic Schrödinger equation can grow without bound. Colliander–Keel–Staffilani–Takaoka–Tao (2010) showed arbitrarily large but finite growth, and unbounded orbits were known only for related or waveguide models. On 18 Nov Xi Chen of the University of Basel posted a 40-page claim of a global solution on T² whose H^{6/5} norm tends to infinity, written with ChatGPT 6.2 Astra. Unrefereed; Hexagon says only some parts were checked by a human.","category":"science","category_label":"Science & math","importance":4,"confidence":"low","status":{"key":"pending","labels":["Awaiting review"]},"sources":3,"official":3,"filed":"2026-11-18","updated":"2026-11-19","orgs":["OpenAI","University of Basel"]},{"id":"2026-11-18-brakensiek-guruswami-promise-coloring-conjectures-claim-yuwen-zhou","url":"https://postcutoff.com/e/2026-11-18-brakensiek-guruswami-promise-coloring-conjectures-claim-yuwen-zhou/","date":"2026-11-18","date_precision":"day","short_title":"Brakensiek–Guruswami conjectures claim","deck":null,"takeaway":"Yuwen Zhou (UBC Okanagan) deposited a 25-page claim that promise graph homomorphism is NP-hard for every non-bipartite source (Brakensiek–Guruswami 2017), which would include hardness of colouring 3-colourable graphs with any constant number of colours, plus their 2016 hypergraph conjecture. The record says the AI generated the initial draft and that 'the detailed proofs in the remaining sections were checked only by AI systems'. Unrefereed, no expert reaction yet.","category":"science","category_label":"Science & math","importance":3,"confidence":"low","status":{"key":"pending","labels":["Awaiting review"]},"sources":1,"official":1,"filed":"2026-11-19","updated":"2026-11-19","orgs":["OpenAI","Anthropic","University of British Columbia Okanagan"]},{"id":"2026-11-18-gpt-6-2-astra-first-48-hours-amateur-maths-wave","url":"https://postcutoff.com/e/2026-11-18-gpt-6-2-astra-first-48-hours-amateur-maths-wave/","date":"2026-11-18","date_precision":"day","short_title":"First 48 hours of GPT-6.2 Astra","deck":"A high-school student posts 25 claimed results, half Lean-checked","takeaway":"It shows how fast a new frontier model turns into a stream of formal-looking results from outside academia.","category":"science","category_label":"Science & math","importance":3,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":9,"official":3,"filed":"2026-11-18","updated":"2026-11-19","orgs":["OpenAI"]},{"id":"2026-11-18-ai-assisted-results-november-18-listing","url":"https://postcutoff.com/e/2026-11-18-ai-assisted-results-november-18-listing/","date":"2026-11-18","date_precision":"day","short_title":"AI-assisted results, 18 Nov","deck":null,"takeaway":"The sweep listed 323 arXiv papers with AI statements (208 new on 18 Nov), 68 conjecture papers without one (16 new), 54 Hexagon works, 18 result-repo commits and 1 watched repo. Every new line was read and its AI statement checked in the PDF. Five results got their own entries, six existing entries were extended, and 96 arXiv papers and 31 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":8,"official":8,"filed":"2026-11-18","updated":"2026-11-18","orgs":["OpenAI","Anthropic","Google DeepMind","various researchers"]},{"id":"2026-11-17-openai-gpt-6-2-astra-sol","url":"https://postcutoff.com/e/2026-11-17-openai-gpt-6-2-astra-sol/","date":"2026-11-17","date_precision":"day","short_title":"OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol","deck":null,"takeaway":"The three US frontier labs now ship within eight days of each other (Fable 5.2 on Nov 9, Gemini 4 Carbon generally available on Nov 10, GPT-6.2 on Nov 17).","category":"model-release","category_label":"Model releases","importance":4,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":21,"official":7,"filed":"2026-11-17","updated":"2026-11-18","orgs":["OpenAI"]},{"id":"2026-11-17-osserman-conjecture-matsaev-conjecture-dave-goel-claude-chatgpt","url":"https://postcutoff.com/e/2026-11-17-osserman-conjecture-matsaev-conjecture-dave-goel-claude-chatgpt/","date":"2026-11-17","date_precision":"day","short_title":"Osserman, Matsaev and more: Goel's AI-written claims","deck":null,"takeaway":"In three days Dave Goel deposited sixteen Hexagon works (most results also in an expanded “and other miscellanea” version) stating that they were 'done overwhelmingly by AI models' (Claude and ChatGPT) and only partly checked by a human. The headline claims: the Osserman conjecture holds in every dimension, and Matsaev's conjecture fails for every p ≠ 2 already for a 2×2 matrix. None is refereed.","category":"science","category_label":"Science & math","importance":3,"confidence":"low","status":{"key":"pending","labels":["Awaiting review"]},"sources":10,"official":10,"filed":"2026-11-17","updated":"2026-11-19","orgs":["Anthropic","OpenAI","independent researchers"]},{"id":"2026-11-13-graph-minors-series-lean-formalization-hajiaghayi-jafariraviz","url":"https://postcutoff.com/e/2026-11-13-graph-minors-series-lean-formalization-hajiaghayi-jafariraviz/","date":"2026-11-13","date_precision":"day","short_title":"Graph Minors series in Lean","deck":null,"takeaway":"A public repo claims a sorry-free Lean 4 proof of every numbered result (838) of Robertson and Seymour's Graph Minors I–XXII, including the graph minor theorem, the structure theorem and the O(n³) disjoint-paths algorithm. The formalization found 30 statement mistakes, 60 proof gaps and 106 places needing a missing argument, none affecting a main result. A sibling repo by the same group (Strong Perfect Graph Theorem) says its Lean code 'was written with AI agents'; the Graph Minors README does not describe its method. Not yet reviewed by others.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":5,"official":4,"filed":"2026-11-19","updated":"2026-11-19","orgs":["University of Maryland"]},{"id":"2026-11-12-ai-assisted-results-november-12-listing","url":"https://postcutoff.com/e/2026-11-12-ai-assisted-results-november-12-listing/","date":"2026-11-12","date_precision":"day","short_title":"AI-assisted results, 12 Nov: Hexagon","deck":null,"takeaway":"The largest Hexagon day yet: 94 uncited deposits on 12 Nov, a quarter of them Tejasvi Singh Tomar's Lean-checked bounds on classical constants. Several claim named conjectures (Petty, Marcus–Minc, Kim–Roush, Humphreys, quantum Kazhdan–Lusztig) with little or no human checking; read each as a claim.","category":"science","category_label":"Science & math","importance":2,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":102,"official":102,"filed":"2026-11-12","updated":"2026-11-12","orgs":["OpenAI","Anthropic","various researchers"]},{"id":"2026-11-09-non-split-cartan-121-two-ai-proofs","url":"https://postcutoff.com/e/2026-11-09-non-split-cartan-121-two-ai-proofs/","date":"2026-11-09","date_precision":"day","short_title":"X_ns+(121): two AI-generated proofs in two days","deck":null,"takeaway":"Arango-Piñeros, Najman and Zureick-Brown (Claude Opus 5.5, GPT-6.1 Sol, GPT-6 Astra) and, separately, Manvir Jaswal (an agent system on Claude Opus 5.5) posted proofs on Hexagon that X_ns+(121) has exactly 7 rational points, all CM. Neither is fully human-checked.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":5,"official":5,"filed":"2026-11-10","updated":"2026-11-19","orgs":["Anthropic","OpenAI","University of Zagreb","UMass Amherst","Amherst College"]},{"id":"2026-11-07-anthropic-claude-hawking-page-modular-bootstrap-hexagon","url":"https://postcutoff.com/e/2026-11-07-anthropic-claude-hawking-page-modular-bootstrap-hexagon/","date":"2026-11-07","date_precision":"day","short_title":"Anthropic's Claude physics papers: Hawking–Page and modular bootstrap","deck":null,"takeaway":"Anthropic's own Hexagon account posted two 'primarily AI-generated' hep-th papers on 7 Nov: an internal Claude research model computed a three-loop coefficient unknown for two decades (b = −1.48×10⁻³ λ²), and found the optimal slope of the spinless modular bootstrap, later reproduced by public Claude Opus 5.5. A physicist checked the results.","category":"science","category_label":"Science & math","importance":4,"confidence":"medium","status":{"key":"pending","labels":["Awaiting review"]},"sources":5,"official":4,"filed":"2026-11-08","updated":"2026-11-19","orgs":["Anthropic"]},{"id":"2026-08-12-claude-hadamard-matrices-below-2000","url":"https://postcutoff.com/e/2026-08-12-claude-hadamard-matrices-below-2000/","date":"2026-08-12","date_precision":"day","short_title":"Claude-assisted constructions complete Hadamard matrices for every order below 2000","deck":"Including 668","takeaway":"It closed a famous \"smallest unknown case\" that had stood for two decades, with an easily verified result.","category":"science","category_label":"Science & math","importance":3,"confidence":"medium","status":{"key":"confirmed","labels":["Result confirmed"]},"sources":4,"official":3,"filed":"2026-09-29","updated":"2026-11-19","orgs":["Anthropic"]}],"people":[{"id":"terence-tao","name":"Terence Tao","url":"https://postcutoff.com/person/terence-tao/"},{"id":"levent-alpoge","name":"Levent Alpöge","url":"https://postcutoff.com/person/levent-alpoge/"}],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-11-19","type":"filed","text":"Created (science agent; 273 arXiv PDFs, 92 Hexagon provenance records, Zenodo metadata, 6 commits, 1 repo; Graph Minors found via Mathstodon)"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-11-19","run":null,"sources_read":null,"updated":"2026-11-19","human_review":null,"version":null},"gaps":[{"model_id":"gpt-6-2-astra","name":"GPT-6.2 Astra","cutoff":"2026-05","days_after":172,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":142,"in_training_data":false},{"model_id":"gemini-4-hydrogen","name":"Gemini 4 Hydrogen","cutoff":"2026-03","days_after":233,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":172,"in_training_data":false}],"short_url":null}