--- id: "2026-11-22-ai-assisted-results-november-22-listing" url: "https://postcutoff.com/e/2026-11-22-ai-assisted-results-november-22-listing/" as_of: "2026-11-22T13:32:00+01:00" date: "2026-11-22" date_precision: day category: science importance: 2 confidence: medium status: [Partly confirmed] sources: 39 editor: human editor human_review: null version: null --- As of: 2026-11-22 13:32 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-11-22-ai-assisted-results-november-22-listing/ # AI-assisted results, 22 Nov Full title: AI-assisted results in the 22 Nov sweep: a polynomial sample compression bound, sharp topological overlap, El-Zahar–Erdős and Stafford settled, a 2D Boussinesq blow-up claim, and 127 arXiv lines earlier runs had skipped Catalogue for the 22 Nov 2026 sweep. Own entries: Cai, Moran, Warmuth et al. (Google) on an O(d² log² d) sample compression scheme found autonomously by Cogentic with Gemini 4 Carbon; Norin's sharp topological overlap constants found by GPT-6 Astra; Liu and Meng's disproof of the El-Zahar–Erdős conjecture; Albert, Moser and Philipp's agent-found, Lean-checked proof of Stafford's conjecture; Chen, Hou and Xiang's claim of smooth blow-up for 2D Boussinesq on the torus; and a Lean-checked proof of Brück's conjecture. Extended: biased random walks (Kortchemski, Schwarzfischer), Peter Graham's deposits, the k-Clique refutation (6-cliques, Max Cut), AlphaProof Nexus (Krenn's GHZ no-go) and the elliptic-curve rank record (rank 32, Quanta). ## Key facts - Method: arXiv's newest announcement was still Friday 20 Nov (checked on arxiv.org/list/math/new), so no line was new. But none of the 127 uncited lines dated 14–19 Nov was cited or mentioned anywhere, and the strongest of them (sample compression, El-Zahar–Erdős, topological overlap, 6-cliques) could not have been dropped as editorial: runs since 19 Nov read only lines with the newest date and treated older-dated lines as re-listings, while the sweep's date is the submission date, a day before the first sweep that sees a paper. All 131 PDFs (those lines plus four cs.GT/cs.AI lines that the cs.GT scan added) were read today, the 125 lines dated 20 Nov were re-read (30 heading-only statements in the PDF), and every URL read and dropped is now in data/_arxiv-read-log.md so the sweep stops re-listing them - Fixes found that way: 25952 (20 Nov, Xiyu Hu) had been dropped although its statement credits a GPT-6.1 Pro and Hunyuan 4 multi-agent system with the three-dimensional argument; listed below - arXiv, combinatorics and TCS: Sam Spiro, small and large quasi-kernel conjectures equivalent for sublinear bounds and best bounds n − Ω(n^{3/4}) and Ω(n^{2/3}), 'Every proof in this paper is due to ChatGPT's Astra 6' ([2611.22881](https://arxiv.org/abs/2611.22881)); Kozma and Tan, polynomial-space poset algorithms in O(2.7693^n), 'The AI tool then generated the majority of the technical work' with GPT Astra 6.0 and Claude Opus 5.5 ([2611.23653](https://arxiv.org/abs/2611.23653)); Teng and Yang, hardness of online stochastic matching, 'main hardness proofs were obtained from extensive interactions with GPT 6 Astra and Gemini 4 Argon' ([2611.23157](https://arxiv.org/abs/2611.23157)); exact degree of symmetric Boolean functions at least n − O(n^{1/3}(log n)^{2/3}), beating the prime-gap barrier toward von zur Gathen–Roche, with Claude Code, Codex and Lean ([2611.26446](https://arxiv.org/abs/2611.26446)); Hopkins et al., optimal robust community detection, GPT-5.6 Sol upgrade and Lean certificate ([2611.25157](https://arxiv.org/abs/2611.25157)); linear-size WSPD for unit-disk graphs with ChatGPT 6 and Claude 5.5 ([2611.24295](https://arxiv.org/abs/2611.24295)); polylog packing of convex polyhedra, where a failed ChatGPT 5.5 repair's reasoning trace pointed the authors to Wang–Zahl and Kakeya ([2611.26352](https://arxiv.org/abs/2611.26352)); 74 toroidal obstructions with no K5 minor, AI scripts and ideas ([2611.22089](https://arxiv.org/abs/2611.22089)); Jooken, an infinite family of bipartite hypotraceable graphs from a Fable 5.2 example ([2611.23018](https://arxiv.org/abs/2611.23018)); Corless's Bohemian-matrix conjecture with Claude writing the Lean ([2611.25266](https://arxiv.org/abs/2611.25266)); Santonastaso, Cooperstein conjecture, counting theorem 4.6 'provided' by ChatGPT 6-Astra ([2611.22622](https://arxiv.org/abs/2611.22622)); EGZ for Z_n^3 ([2611.25050](https://arxiv.org/abs/2611.25050)); Stanley's conjecture on leading Ehrhart coefficients of the Birkhoff polytope ([2611.25973](https://arxiv.org/abs/2611.25973)); Zaks–Perles–Wills in dimension three ([2611.22337](https://arxiv.org/abs/2611.22337)); FunSearch on weak Schur numbers ([2611.22032](https://arxiv.org/abs/2611.22032)); Yuster's transversal factor, AI offered a stronger unused argument ([2611.26259](https://arxiv.org/abs/2611.26259)); correspondence packing, GPT-6.1 Sol chose the explicit k(D) ([2611.20504](https://arxiv.org/abs/2611.20504)); pattern-avoiding BSTs, LLM-suggested constants and k-finger BSTs ([2611.20088](https://arxiv.org/abs/2611.20088)) - arXiv, game theory (cs.GT, scanned for the first time): proportional representation for almost-solid coalitions, GPT-6.1 Sol 'extensively used … including in the generation of many of the proofs' ([2611.25677](https://arxiv.org/abs/2611.25677)); Feldman, Gkatzelis et al., budgeted crowdsourcing contracts, 'Some of the proofs and some counterexamples … developed with the assistance of' ChatGPT and Claude ([2611.25988](https://arxiv.org/abs/2611.25988)) - arXiv, probability, analysis, PDE: Novikov, counterexample to Molchan's orthant persistence conjecture, 'developed by GPT-6 Astra Ultra' and first drafted by GPT-6 Astra Pro ([2611.24520](https://arxiv.org/abs/2611.24520)); counterexample to the Hytönen–Lacey one-supremum conjecture with ChatGPT and Codex ([2611.22097](https://arxiv.org/abs/2611.22097)); Xiyu Hu, one-sided Liouville question of Buhovsky–Logunov–Malinnikova–Sodin, three-dimensional argument from a GPT-6.1 Pro/Hunyuan 4 agent system ([2611.25952](https://arxiv.org/abs/2611.25952)); higher-derivative Sendov conjectures, ChatGPT drafted some proofs ([2611.23039](https://arxiv.org/abs/2611.23039)); Bogosel, Bucur and Fragalà prove the polygonal Saint-Venant inequality of Pólya–Szegő by hand and report that an LLM-assisted adaptation also proves polygonal Faber–Krahn, which they did not publish ([2611.24573](https://arxiv.org/abs/2611.24573)); Hedong Wang's single-author proof of both Pólya–Szegő polygon inequalities for n ≥ 5 has no AI statement ([2611.22305](https://arxiv.org/abs/2611.22305)); Haraux, dry-friction string with Claude numerics ([2611.21213](https://arxiv.org/abs/2611.21213)); Eliashberg condensation energy, LLM agents chose the conjectures and drafted proofs and Lean ([2611.21773](https://arxiv.org/abs/2611.21773)); Stratonovich waves from the 'She' research environment ([2611.23014](https://arxiv.org/abs/2611.23014)); Strichartz counterexamples ([2611.19555](https://arxiv.org/abs/2611.19555)); SU(3) Toda degree ([2611.20013](https://arxiv.org/abs/2611.20013)); ellipsoids by intrinsic volumes ([2611.24923](https://arxiv.org/abs/2611.24923)) - arXiv, algebra, geometry, number theory: Shuoxing Zhou, a new proof of Connes' bicentralizer conjecture and Marrakchi's relative version, where GPT-6 Astra 'helped identify a solvable problem' for the author's Poisson-boundary techniques ([2611.24787](https://arxiv.org/abs/2611.24787)); failure of the AJ conjecture for Whitehead doubles of torus knots, GPT-6 Astra 'contributing the principal derivations', Lean ([2611.24826](https://arxiv.org/abs/2611.24826)); Lin and Zhang, functorial height, with applications to Orlov's rationality conjecture 'found with AI assistance' ([2611.23707](https://arxiv.org/abs/2611.23707)); Geiger and Kretschmer, Green's conjecture on anticanonical surfaces, 'a significant part of the paper relies on the AI contribution' ([2611.22997](https://arxiv.org/abs/2611.22997)); BMM symmetrising trace conjecture for all exceptional complex reflection groups, programs built with an LLM ([2611.21912](https://arxiv.org/abs/2611.21912)); Păunescu, Rădulescu and Thom, Bessis–Moussa–Villani in finite von Neumann algebras with OpenAI models ([2611.23122](https://arxiv.org/abs/2611.23122)); Pasten and Vidaux, abc over K implies Hilbert's tenth problem fails over quadratic fields, AI-assisted ([2611.24748](https://arxiv.org/abs/2611.24748)); follow-ups of OpenAI's Hilbert-tenth and Uchida manuscripts without AI statements, Zhi-Wei Sun over Q(√2) ([2611.25002](https://arxiv.org/abs/2611.25002)) and Yu Mao ([2611.24495](https://arxiv.org/abs/2611.24495)); Sherman, Coulembier–Flake saturation found with ChatGPT Astra 6.1 ([2611.24471](https://arxiv.org/abs/2611.24471)); fence conjecture with ChatGPT 6 Astra ([2611.24022](https://arxiv.org/abs/2611.24022)); envelope property for unibranch varieties, initial proof by the Danus agent and GPT-5.6 Sol ([2611.26301](https://arxiv.org/abs/2611.26301)); Brown–Wemyss type E with Danus and Rethlas ([2611.25467](https://arxiv.org/abs/2611.25467)); Casselman's comparison conjecture, corollaries with AI, and Hang Xue's separate AI-assisted proofs ([2611.25367](https://arxiv.org/abs/2611.25367)); Kobayashi volume Schwarz lemma ([2611.26524](https://arxiv.org/abs/2611.26524)); Petrescu Hadamard subfactors ([2611.25638](https://arxiv.org/abs/2611.25638)); bounded functions on character varieties ([2611.25162](https://arxiv.org/abs/2611.25162)); algebraic expanders ([2611.24034](https://arxiv.org/abs/2611.24034)); Jack b-positivity ([2611.26439](https://arxiv.org/abs/2611.26439)); Terao for 17–18 lines ([2611.21506](https://arxiv.org/abs/2611.21506)); Küchle fourfolds ([2611.21576](https://arxiv.org/abs/2611.21576)); Riemann-zero windows with GPT-6 and Fable 5.2 ([2611.26149](https://arxiv.org/abs/2611.26149)) - arXiv, formal methods and physics: Claude Fable 5.1 wrote the full Verus proof of an SDP solver ([2611.25174](https://arxiv.org/abs/2611.25174)); Aristotle filled qlin's Lean proofs ([2611.19618](https://arxiv.org/abs/2611.19618)); Claude-written Lean for Hubbard spectral moments ([2611.19929](https://arxiv.org/abs/2611.19929)); Lean for a hypersequent calculus ([2611.19940](https://arxiv.org/abs/2611.19940)); formal Vlasov–Maxwell proofs reused for plasma codes ([2611.23319](https://arxiv.org/abs/2611.23319)); coding agents training optical neural networks in situ ([2611.23875](https://arxiv.org/abs/2611.23875)) - Hexagon, 33 works read through the API: own entries for Hexagon 2611.00792 (Boussinesq) and 00826 (Brück); added to existing entries 00820 (Kortchemski) and 00793–00797 (Graham); Alpöge follow-ups for the RH entry: twinheterodox's three simplifications with Codex GPT-6.2-Astra max (00828), Slijepčević's Aleph outline from the Lean (00812), Robin Zhang's GPT 6 Pro provenance note (00809), Jaswal's reader's guide, unchecked by any human (00799), and Leo Pintar's conditional K_{4k}(ℤ) = 0 for k ≤ 26 from Alpöge's Step 17 with Codex (00829) - Hexagon, other claims: Yang and Wang, smooth isometric embedding of closed n-manifolds into R^{s_n+5} for n ≥ 10^62 + 4, 'detailed technical arguments have not yet been fully verified' (00824); Hong and Yu, exceptional lattice nonzero only up to rank 26, answering Grabbel et al., human-checked (00821); Enomoto, an extriangulated category with no triangulated embedding, constructed by GPT-6.2-Astra (00805); Sihan Chen, sharp Holmes–Thompson systolic inequality for non-reversible Finsler RP² (00817); Fomin–Pylyavskyy Conjecture 8.12 and surface Conjecture 5 (Hernandez Juan, 00815); Christopher D. Long, Fang–Chang optimal-package conjecture and factorial moments (00814, 00813); Kazalicki and Slijepčević, Atkin–Swinnerton-Dyer congruences with Aleph (00822); Vega, 83/84 approximation for a Horn class (00827); Henriksson, semiprimes in short intervals under RH, no human check (00808); Krisman, log-product irrationality, no human check (00807); Aquila, Hooley-type sieve limits, no human check (00798); Banica with Claude Fable (00818); Sivarajan with Claude Opus 5.5 (00791); Huang with GPT-6.1 Sol and Grok 4.7 (00802); Noori Jihad (00800). Dropped: 00825 and 00790 (human-written), 00816 (no AI role stated), 00819 (physics essay) - Zenodo (sweep's search lines, maths records read): Yitzchak Shmalo's project 'on proving open problems with AI systems', Claude Opus 5.5 agents advised by Fable 5.2 or Codex and Claude agents: the sharp within-level Fourier entropy constant 2/ln 2 for symmetric functions (O'Donnell–Wright–Zhou had 9.04) ([Zenodo 24435309](https://zenodo.org/records/24435309)), two questions of Jerrum and Patel ([24434504](https://zenodo.org/records/24434504)), Alon–Puder KMP eigenvalue chains ([24433750](https://zenodo.org/records/24433750)), Paul's greedy-sequence conjecture ([24435134](https://zenodo.org/records/24435134)) and moment-constrained alignment ([24435174](https://zenodo.org/records/24435174)); Allahyar and Yianni, prime gaps ≤ 180 under Sharifi's exponential-sum hypothesis (182 before), Fastino agent harness ([24430219](https://zenodo.org/records/24430219)); Lupini, answering Harbourne–Migliore–Nagel–Teitler on free extensions, ChatGPT, Codex and Claude for proof development ([24438399](https://zenodo.org/records/24438399)); Botsford, Dean's conjecture for k = 5 shortened with GPT-5.6 Sol, Opus 5, Danus, Fable 5.2 and GPT-6.2 Astra ([24434347](https://zenodo.org/records/24434347)); rainbow C5-saturation with Codex ([24433034](https://zenodo.org/records/24433034)); Huynh, Kowitz–Zhao Conjecture 26, no AI statement ([24432860](https://zenodo.org/records/24432860)). Dropped: the 'Unconditional Proof of the Inescapable Form' series on the Millennium problems, the 'Postmodern Physics of Hamzah Information' series and non-maths records - Hexagon backlog (oldest 8 lines, 18 Oct): Caceres's Euler–Caceres bound, human-checked ([Hexagon 2610.00343](https://hexagonmath.org/abs/2610.00343)); Yuran Xiu, three AI-only answers on Stirling lucky sets and acyclic polynomials, negative answers to Barton–Brown–Pike, no human check ([2610.00324](https://hexagonmath.org/abs/2610.00324), [00325](https://hexagonmath.org/abs/2610.00325), [00326](https://hexagonmath.org/abs/2610.00326)); Machado Manhães de Castro, counterexamples to the Barát–Gyárfás–Tóth nonnested matching conjecture with ChatGPT, Codex and Claude, human-checked ([2610.00329](https://hexagonmath.org/abs/2610.00329)); Hao-Yu Sun, Heegaard boundary modules for the Z-hat invariant ([2610.00339](https://hexagonmath.org/abs/2610.00339)); Brice Pouly, two Poisson-limit papers with Fable 5, Opus 4.8 and GPT-6 Astra, no human check ([2610.00333](https://hexagonmath.org/abs/2610.00333), [2610.00334](https://hexagonmath.org/abs/2610.00334)) - GitHub, watched accounts (3): AxiomMath/LogNarayana and AxiomMath/HigherDyson3, Lean formalizations of log-concavity for stack-sorting fibers and Narayana transforms and of Alfes–Ono–Swaminathan's higher Dyson ranks; google-deepmind/robotics-antigravity-plugin (★207), robotics agent skills (tooling). Result-repo commits (2): formal-conjectures added the multiplicative persistence conjecture and clarified that solved statements are out of scope - Reactions (task: envy-free cake cutting cascade, tight five-cycle): none found on blogs, HN or Google News by 22 Nov; a cs.GT paper (arXiv 2610.30370) says the cascade began with an X thread by Nisarg Shah and argues the field needs a protocol people can check and teach; Quanta's weekly 'Transformation' column (20 Nov) covered the RH and Navier–Stokes claims, Alessandro Masullo's 940-page human GRH manuscript and Radziwiłł's criticism (news agent's entries) ## What happened No new arXiv announcement reached the 22 Nov sweep, so its lists looked like weekend re-listings. A check of which lines had ever been cited showed that the 127 uncited lines dated 14–19 November had not been read in any run: since 19 November the runs had read only the lines carrying the newest date and treated the rest as re-listings, while the sweep's date is a paper's submission date, one day before the first sweep that lists it. Those lines contained four results strong enough for their own entries and several extensions. Every PDF was read for its AI statement; the 20 Nov lines were re-checked, which turned up one more paper dropped by mistake. Hexagon added 33 works on 21 November, many of them responses to Alpöge's Riemann Hypothesis claim (listed here; the RH entry is maintained separately). Two Hexagon works have their own entries: a claimed smooth blow-up for 2D Boussinesq on the torus and a Lean-checked proof of Brück's conjecture. ## Why it matters The strongest items of the week were sitting in the sweep's output for several days, unread, because of a dating rule in how runs triaged the list. They include an autonomous agent proof (sample compression, co-signed by the conjecture's originator) and a sharp constant in every dimension (topological overlap). The read log introduced today keeps read lines out of future sweeps, so each run sees only lines nobody has read. ## Your AI and this story - GPT-6.2 Astra (training cutoff May 2026): 175 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 145 days after its cutoff - Gemini 4 Hydrogen (training cutoff March 2026): 236 days after its cutoff - Grok 4.7 (training cutoff May 2026): 175 days after its cutoff ## Sources 1. [arXiv 2611.22881: Equivalence of small and large quasi-kernels (Spiro)](https://arxiv.org/abs/2611.22881) (arxiv.org, paper) 2. [arXiv 2611.24520: A counterexample to Molchan's orthant persistence conjecture (Novikov)](https://arxiv.org/abs/2611.24520) (arxiv.org, paper) 3. [arXiv 2611.24787: A Poisson boundary solution to Connes' bicentralizer conjecture (Zhou)](https://arxiv.org/abs/2611.24787) (arxiv.org, paper) 4. [arXiv 2611.26446: Degree lower bounds for symmetric Boolean functions](https://arxiv.org/abs/2611.26446) (arxiv.org, paper) 5. [arXiv 2611.25677: Proportional representation for almost-solid coalitions](https://arxiv.org/abs/2611.25677) (arxiv.org, paper) 6. [arXiv 2611.24573: Polygonal Saint-Venant inequality (Bogosel, Bucur, Fragalà)](https://arxiv.org/abs/2611.24573) (arxiv.org, paper) 7. [arXiv 2611.22305: The Pólya–Szegő inequalities for polygons (Hedong Wang)](https://arxiv.org/abs/2611.22305) (arxiv.org, paper) 8. [Hexagon 2611.00828: Three simplifications in the Alpöge–Claude argument](https://hexagonmath.org/abs/2611.00828) (hexagonmath.org, paper) 9. [Hexagon 2611.00812: Outline of Alpöge's proof of GRH (Slijepčević)](https://hexagonmath.org/abs/2611.00812) (hexagonmath.org, paper) 10. [Hexagon 2611.00809: Preliminary note on Alpöge's proposed proof (Robin Zhang)](https://hexagonmath.org/abs/2611.00809) (hexagonmath.org, paper) 11. [Hexagon 2611.00799: RH for Dirichlet L-functions, after Alpöge (Jaswal)](https://hexagonmath.org/abs/2611.00799) (hexagonmath.org, paper) 12. [Hexagon 2611.00829: Hecke calculations for conditional vanishing of K_{4k}(Z) (Pintar)](https://hexagonmath.org/abs/2611.00829) (hexagonmath.org, paper) 13. [Hexagon 2611.00824: Smooth isometric embeddings into R^{s_n+5}](https://hexagonmath.org/abs/2611.00824) (hexagonmath.org, paper) 14. [Hexagon 2611.00821: Maximal rank of an even hyperbolic lattice with nonzero exceptional lattice](https://hexagonmath.org/abs/2611.00821) (hexagonmath.org, paper) 15. [Hexagon 2611.00805: An extriangulated category with no triangulated embedding (Enomoto)](https://hexagonmath.org/abs/2611.00805) (hexagonmath.org, paper) 16. [Hexagon 2611.00817: Pu's inequality for non-reversible Finsler metrics](https://hexagonmath.org/abs/2611.00817) (hexagonmath.org, paper) 17. [Hexagon 2611.00815: Proofs of two conjectures of Fomin and Pylyavskyy](https://hexagonmath.org/abs/2611.00815) (hexagonmath.org, paper) 18. [Hexagon 2611.00814: Optimal coupon packages (Long)](https://hexagonmath.org/abs/2611.00814) (hexagonmath.org, paper) 19. [Hexagon 2611.00813: Factorial moments in two variables (Long)](https://hexagonmath.org/abs/2611.00813) (hexagonmath.org, paper) 20. [Hexagon 2611.00822: Atkin–Swinnerton-Dyer congruences for Kummer twists](https://hexagonmath.org/abs/2611.00822) (hexagonmath.org, paper) 21. [Hexagon 2611.00827: An 83/84 approximation for a cubic Horn class (Vega)](https://hexagonmath.org/abs/2611.00827) (hexagonmath.org, paper) 22. [Hexagon 2611.00808: Balanced semiprimes in short intervals under RH](https://hexagonmath.org/abs/2611.00808) (hexagonmath.org, paper) 23. [Hexagon 2611.00807: The product log(1+1/N)log(1+2/N) (Krisman)](https://hexagonmath.org/abs/2611.00807) (hexagonmath.org, paper) 24. [Hexagon 2611.00798: Hooley's twisted cancellation below level 4/15](https://hexagonmath.org/abs/2611.00798) (hexagonmath.org, paper) 25. [Hexagon 2611.00818: Noncommutative curves and surfaces in small degree (Banica)](https://hexagonmath.org/abs/2611.00818) (hexagonmath.org, paper) 26. [Hexagon 2611.00791: Rational points on Baxter's curve (Sivarajan)](https://hexagonmath.org/abs/2611.00791) (hexagonmath.org, paper) 27. [Hexagon 2611.00802: Separate fits and joint compatibility (Huang)](https://hexagonmath.org/abs/2611.00802) (hexagonmath.org, paper) 28. [Hexagon 2611.00800: Holes and islands of visible points (Noori Jihad)](https://hexagonmath.org/abs/2611.00800) (hexagonmath.org, paper) 29. [Hexagon 2611.00825: Reconstructing the Simulmatics 1960 simulation (dropped, human-written)](https://hexagonmath.org/abs/2611.00825) (hexagonmath.org, paper) 30. [Hexagon 2611.00790: Crystal structures on theta bases (dropped, human-written)](https://hexagonmath.org/abs/2611.00790) (hexagonmath.org, paper) 31. [Hexagon 2611.00816: Cluster-crystal realization of KR Cartan algebras (dropped)](https://hexagonmath.org/abs/2611.00816) (hexagonmath.org, paper) 32. [Hexagon 2611.00819: A stochastic molecular mechanism for the arrow of time (dropped)](https://hexagonmath.org/abs/2611.00819) (hexagonmath.org, paper) 33. [GitHub AxiomMath/LogNarayana](https://github.com/AxiomMath/LogNarayana) (github.com, code) 34. [GitHub AxiomMath/HigherDyson3](https://github.com/AxiomMath/HigherDyson3) (github.com, code) 35. [GitHub google-deepmind/robotics-antigravity-plugin](https://github.com/google-deepmind/robotics-antigravity-plugin) (github.com, code) 36. [formal-conjectures commit: multiplicative persistence conjecture](https://github.com/google-deepmind/formal-conjectures/commit/24371fbe346f01704b4aa18a6b60b41aac4d9622) (github.com, code) 37. [formal-conjectures commit: solved statements out of scope](https://github.com/google-deepmind/formal-conjectures/commit/2207c54d0dbe960de591d280f997b195bf3f6e28) (github.com, code) 38. [arXiv 2610.30370: Cutting a cake for human consumption](https://arxiv.org/abs/2610.30370) (arxiv.org, paper) 39. [Quanta Magazine: As mathematicians race to verify AI Riemann proof, a human proof emerges (20 Nov)](https://www.quantamagazine.org/update/169569/) (quantamagazine.org, press) ## Changes - 2026-11-22 (filed): Created (science agent; 131 arXiv PDFs plus 30 re-checked, 33 Hexagon API records, 12 Zenodo records, 8 backlog works) ## Related - 2026-11-21: [2D Boussinesq blow-up on the torus (claim)](https://postcutoff.com/e/2026-11-21-boussinesq-2d-torus-smooth-blowup-chen-hou-xiang-chatgpt/index.md) - 2026-11-21: [Tight 5-cycle Turán density (AI, Lean)](https://postcutoff.com/e/2026-11-21-tight-five-cycle-turan-density-peter-graham-ai-lean/index.md) - 2026-11-21: [Brück conjecture (ChatGPT, Lean)](https://postcutoff.com/e/2026-11-21-bruck-conjecture-proof-tang-chatgpt-codex-lean/index.md) - 2026-11-21: [AI-assisted results, 21 Nov](https://postcutoff.com/e/2026-11-21-ai-assisted-results-november-21-listing/index.md) - 2026-11-20: [Anthropic mathematician Levent Alpöge posts a Lean proof of the Riemann Hypothesis found with Claude; the Lean and Mathlib communities certify it](https://postcutoff.com/e/2026-11-20-alpoge-claude-riemann-hypothesis-proof-claim/index.md) - 2026-11-20: [Biased random walk speed monotonicity (GPT-6 Astra)](https://postcutoff.com/e/2026-11-20-biased-random-walk-speed-monotonicity-heng-ma-gpt-6-astra/index.md) - 2026-11-19: [Sharp topological overlap (GPT-6 Astra)](https://postcutoff.com/e/2026-11-19-sharp-topological-overlap-every-dimension-norin-gpt-6-astra/index.md) - 2026-11-18: [Sample compression O(d² log² d) (Gemini 4 Carbon)](https://postcutoff.com/e/2026-11-18-sample-compression-conjecture-polynomial-cogentic-gemini/index.md) - 2026-11-18: [El-Zahar–Erdős disproved (GPT-6.2 Astra)](https://postcutoff.com/e/2026-11-18-el-zahar-erdos-conjecture-disproof-gpt-6-2-astra/index.md) - 2026-11-18: [Stafford's conjecture (AI agents, Lean)](https://postcutoff.com/e/2026-11-18-stafford-conjecture-weyl-algebra-ai-agents-lean/index.md) - 2026-11-06: [k-Clique hypothesis refuted (Vassilevska Williams, Zamir, ChatGPT-6)](https://postcutoff.com/e/2026-11-06-k-clique-hypothesis-refuted-vassilevska-williams-zamir/index.md) - 2026-10-08: [Google DeepMind's AlphaProof Nexus paper published in Science](https://postcutoff.com/e/2026-10-08-alphaproof-nexus-science-paper/index.md) - 2026-08-23: [Claude-assisted search breaks the elliptic curve rank record](https://postcutoff.com/e/2026-08-23-elliptic-curve-rank-record-31/index.md) - People: [Levent Alpöge](https://postcutoff.com/person/levent-alpoge/)