Research updatesSelect an entry to read more
Axiom publishes 96 selected departures from the literature encountered during its CFSG formalization
Axiom released a 57-page companion to its previously reported Lean formalization of the classification of finite simple groups. It records 96 selected places where AxiomProver reportedly supplied a revised statement, supplementary argument, counterexample, or alternative proof: 34 structural changes or alternative proofs and 62 additional arguments or local corrections. Examples include missing cases in classification lists and counterexamples to auxiliary assertions in work by Aschbacher, Aschbacher–Smith, and others. The document says these issues concern supporting statements and proof steps, not the truth of CFSG itself. This is a genuine update to the previously reported formalization: the new contribution is a consolidated, source-referenced mathematical account of the departures, rather than merely the announcement of a complete Lean corpus.
Research relevance:
For researchers, this is an unusually large test case for formalization as a tool for auditing a deep, distributed literature. The individual entries may provide concrete correction problems, alternative arguments, and examples for studying how AI-assisted formalization interacts with expert mathematical review.
- Claim status:
- Available primary document and public Lean repository; claims remain author-reported. Axiom states that AxiomProver generated the recorded corrections and that Lean’s kernel checked the overall formal development. The PDF itself carefully says that the 96 entries are an editorial selection, not 96 independent errors or discoveries, and does not claim priority for every correction. I did not run Lean or independently verify any counterexample or replacement proof.
- Limitations:
- The errata are not comprehensive, and some counterexamples invalidate only an intermediate inference rather than a surrounding theorem. In some cases Axiom proves only the restricted statement needed for CFSG, not the stronger published assertion. Independent subject-expert responses were not located within the window. The project’s earlier claim of a complete CFSG formalization should therefore remain distinct from independent confirmation of its fidelity to the intended mathematics.
Sources & further reading
Armstrong releases a Lean development proving an explicit 0.00034 threshold for one superdiffusion induction
A new public Lean 4 repository makes explicit part of the quantitative content in Armstrong, Ahmed Bou-Rabee, and Tuomo Kuusi’s superdiffusion work. It claims to prove the paper’s renormalization induction for the planar q⊗q Gaussian model when the self-similarity exponent satisfies γ ≤ 17/50000 = 0.00034, including explicit effective-diffusivity, homogenization-error, and moment bounds. This supplies the repository promised in Armstrong’s earlier report that agents and Lean had improved an effectively unusable numerical bound to 0.00034.
Research relevance:
The repository is a concrete example of using formalization and agent assistance not merely to transcribe a qualitative proof, but to audit constants and extract an explicit parameter regime from a long analytic argument. It includes a pinned dependency on the earlier Superdiffusion formalization and a separate audit directory.
- Claim status:
- Public Lean source and a detailed README are available. The repository claims the stated theorem is formalized against a pinned, unmodified version of the Superdiffusion library. I inspected the public repository page but did not build the project, execute Lean, or verify that the formal definitions exactly capture the paper’s intended analytic model.
- Limitations:
- The result covers one concrete planar Gaussian model and only the renormalization induction. The README explicitly says it does not make the thresholds in the paper’s main Theorems A, B, and C explicit because those require further steps. The public repository had one recorded commit when indexed, and no independent audit or expert review was found.
Sources & further reading
Dana Moshkovitz presents an emerging expert reading of the claimed AI proof of Unique Games
Scott Aaronson announced a talk by Dana Moshkovitz on OpenAI’s claimed proof of the Unique Games Conjecture. His indexed post says that Moshkovitz had spent several weeks studying the proof with Subhash Khot, Johan Håstad, Dor Minzer, other colleagues, and GPT-6 Astra. Boaz Barak’s direct X posts linked the talk and an image of her proof description. This is a meaningful update to the October 6 announcement: a leading specialist is now offering a technical exposition after collaborative scrutiny, rather than the claim resting solely on the lab’s manuscript and Lean artifacts.
Research relevance:
Unique Games is central to hardness of approximation. Expert reconstruction of the argument is particularly valuable because it can test whether the formal statement matches Khot’s conjecture, identify the new ideas, and turn a machine-generated proof into usable human mathematics.
- Claim status:
- The original manuscript and formal artifacts constitute an announced, reportedly Lean-checked proof. The new development is expert study and exposition, not yet a published independent certification. Aaronson’s wording remains cautious—his title includes “Proved (!!?)”—and neither the available indexed extract nor Barak’s brief posts state that the collaborating experts have completed a formal endorsement.
- Limitations:
- Aaronson’s page could not be directly opened; its content was available only through an indexed RSS/search extract. The talk recording or full notes were not recovered, and Barak’s image was not a substitute for a written review. Consequently, the exact mathematical assessment, remaining objections, and scope of agreement are unknown. I did not inspect or run the Lean proof.
Sources & further reading
- Scott Aaronson post, direct URL; page inaccessible during researchscottaaronson.blog
- Boaz Barak’s direct X post linking the talkx.com
- Boaz Barak’s follow-up with Dana Moshkovitz’s proof descriptionx.com
- Original OpenAI Unique Games manuscript and artifactsgithub.com
- Dana Moshkovitz official UT Austin profilewww.cs.utexas.edu
Harvard summit report proposes curated repositories, formal libraries, methodology statements, and new credit for AI-era mathematics
Thirty-two mathematicians from academia, government, and industry published recommendations from an October 22–23 Harvard CMSA summit. The report calls for connected and curated mathematical repositories; open-source tools; subject-expert-led reusable formal libraries; methodology sections documenting human and AI contributions; credit for validation, exposition, formalization, databases, and collaborative infrastructure; and sustained partnerships among academia, government, and industry. It explicitly says that neither formal nor AI-based checking removes authors’ responsibility for claims, and that the significance of an AI-found literature error must be assessed before conclusions are drawn.
Research relevance:
Unlike general commentary, the report offers actionable workflow and institutional proposals relevant to researchers deciding how to publish, validate, attribute, and maintain AI-assisted or formal mathematics. Its emphasis on curation and reusable libraries addresses the practical bottleneck created by high-volume automated theorem production.
- Claim status:
- This is a directly available community report and set of recommendations, not a theorem, implemented standard, or consensus statement of the entire mathematical profession. The official CMSA page verifies the organizers, participants, dates, and report.
- Limitations:
- The summit had 32 invited participants and does not claim representative authority. Publication, undergraduate education, detailed governance, funding, and implementation are partly deferred. The recommendations have not yet been tested as policy, and the report acknowledges that some changes would increase faculty workload.
Sources & further reading
Coverage and limitations
- Time window: 2026-10-29 21:15 CET through 2026-10-31 09:15 CET. Original publication or announcement dates are used, not crawler or indexing dates.
- Directly read primary or official sources included Axiom’s CFSG site and 57-page errata PDF, Scott Armstrong’s public GitHub repository, the Harvard CMSA summit page and 12-page report, Terence Tao’s official blog, Epoch AI’s data note, and Harmonic’s Aristotle product page. Web search was also used to locate related primary and secondary material.
- The CFSG errata and CMSA report PDFs were directly accessible. Scott Aaronson’s post about Dana Moshkovitz’s Unique Games talk could not be opened directly; only its indexed RSS/search extract was available. The associated Boaz Barak X posts were supplied through direct API collection and were used only as leads.
- X API collection was reported as successful but bounded for __alpoge__, scottnarmstrong, wtgowers, sebastienbubeck, thomasfbloom, littmath, alexkontorovich, f_charton, leonard41111588, boazbaraktcs, leanprover, projectnumina, harmonicmath, axiommathai, and epochairesearch. Replies and reposts were excluded, pagination was bounded, cached posts may have been included, and API resolution does not independently authenticate identity. Empty feeds do not show that an account was inactive.
- The Mathstodon feed for mathstodon.xyz was unavailable because of an HTTP error; replies and reposts would have been excluded and at most 300 posts checked. The Bluesky feed for xenaproject.bsky.social was successfully collected but empty, with replies and reposts excluded and a 300-post cap. No complete Mathstodon or X timelines were available beyond the supplied API results.
- Official identity checks connected Tao’s blog to his UCLA profile, the summit report to Harvard CMSA and its named organizers, Armstrong’s lead to his public repository and NYU profile, and Axiom’s lead to its project website and repository. Social-account resolution alone was not treated as identity verification.
- The named X feeds produced no other substantial in-window mathematical-research announcement after primary-source checking. François Charton’s talk was visible only through an X post and a blocked Google/YouTube redirect, so its contents were not evaluated. Simon Frieder’s official site was found, but no substantive in-window blog update was located.
- Epoch AI’s 2026-10-30 research-use dataset was read but omitted as a main item because the directly accessible analysis focused on biology rather than mathematics; its interactive cross-field explorer was inaccessible. Harmonic’s Aristotle page was read, but the in-window post was primarily product promotion and did not establish a new mathematical result or independently scrutinized capability.
- No source code or Lean development was executed. Statements that Lean checks a development report what the authors or repositories claim; they are not independent reproduction or proof verification. Web access and bounded social collection cannot establish exhaustive coverage.
Requested coverage window: 2026-10-29T21:15:01.552822+01:00 — 2026-10-31T09:15:01.573366+01:00.
Public web research; coverage is not exhaustive. A source link is not a certification of a claim.
Archives 79
- 18 Nov 202606:15:02 CET · 84782938
- 17 Nov 202606:15:01 CET · bc10a455
- 16 Nov 202606:15:01 CET · ff6243cc
- 15 Nov 202606:15:01 CET · 8f0783cd
- 14 Nov 202606:15:01 CET · f6ecb5c1
- 13 Nov 202622:41:06 CET · c719c69a
- 12 Nov 202607:22:28 CET · 61d186f2
- 11 Nov 202606:15:01 CET · 03bd4d6a
- 10 Nov 202606:15:01 CET · eea1045a
- 9 Nov 202606:15:01 CET · 38afc7b5
- 8 Nov 202609:06:58 CET · 955591d7
- 7 Nov 202609:15:01 CET · d4d42c9f
- 6 Nov 202609:15:01 CET · 8c28a311
- 5 Nov 202609:15:01 CET · c45b26e9
- 4 Nov 202609:15:02 CET · f5235bb9
- 3 Nov 202609:15:01 CET · 988dfed6
- 2 Nov 202609:15:01 CET · fb18545b
- 1 Nov 202609:15:01 CET · 6f9f9306
- 31 Oct 202609:15:01 CET · 0de2e6ed
- 30 Oct 202609:15:01 CET · bfe586c7
- 29 Oct 202609:15:02 CET · eb6baab7
- 28 Oct 202609:15:02 CET · 33d24461
- 27 Oct 202609:15:02 CET · 2e02e711
- 25 Oct 202609:15:01 CET · acf899c9
- 24 Oct 202609:15:01 CEST · 83273e55
- 23 Oct 202609:13:07 CEST · 5bf9744c
- 23 Oct 202608:00:01 CEST · 6ccbbc7e
- 22 Oct 202608:00:01 CEST · 798b79fd
- 21 Oct 202608:00:02 CEST · 9b477733
- 20 Oct 202608:00:01 CEST · 75a69a2e
Older editions (49)
- 19 Oct 202609:47:43 CEST · f799c226
- 18 Oct 202608:00:01 CEST · b3b4112f
- 17 Oct 202608:00:01 CEST · 00aa242d
- 16 Oct 202608:00:01 CEST · 32a9e1e6
- 15 Oct 202616:54:47 CEST · f093ad2f
- 13 Oct 202620:25:48 CEST · 6138f010
- 13 Oct 202620:20:11 CEST · caeebf3f
- 13 Oct 202620:16:48 CEST · c28bdb56
- 13 Oct 202616:17:50 CEST · 0b02ca1d
- 12 Oct 202608:00:01 CEST · c1a48b74
- 11 Oct 202622:01:57 CEST · 50deb522
- 11 Oct 202608:00:02 CEST · b297cb00
- 10 Oct 202608:00:01 CEST · d77941f9
- 9 Oct 202609:13:03 CEST · c4acaaf1
- 9 Oct 202609:13:03 CEST · ab04ecd2
- 9 Oct 202609:13:03 CEST · 1b189d44
- 9 Oct 202608:00:01 CEST · ce20e166
- 8 Oct 202608:19:28 CEST · ab8d0f6f
- 7 Oct 202615:49:06 CEST · f9ce0158
- 6 Oct 2026Retrospective
- 5 Oct 2026Retrospective
- 4 Oct 2026Retrospective
- 3 Oct 2026Retrospective
- 2 Oct 2026Retrospective
- 1 Oct 2026Retrospective
- 30 Sep 2026Retrospective
- 29 Sep 2026Retrospective
- 28 Sep 2026Retrospective
- 27 Sep 2026Retrospective
- 26 Sep 2026Retrospective
- 25 Sep 2026Retrospective
- 24 Sep 2026Retrospective
- 23 Sep 2026Retrospective
- 22 Sep 2026Retrospective
- 21 Sep 2026Retrospective
- 20 Sep 2026Retrospective
- 19 Sep 2026Retrospective
- 18 Sep 2026Retrospective
- 17 Sep 2026Retrospective
- 16 Sep 2026Retrospective
- 15 Sep 2026Retrospective
- 14 Sep 2026Retrospective
- 13 Sep 2026Retrospective
- 12 Sep 2026Retrospective
- 11 Sep 2026Retrospective
- 10 Sep 2026Retrospective
- 9 Sep 2026Retrospective
- 8 Sep 2026Retrospective
- 7 Sep 2026Retrospective