Sharp logarithmic Kakeya maximal and tube bounds in every dimension, registered as a Lean proof
Mark Lewko, with ChatGPT and Codex doing the mathematical development and Lean implementation, registered on Palomar a Lean proof that the Kakeya maximal operator is bounded from to with norm at most in every dimension , a logarithmic power the accompanying notes argue is optimal.
- Lab
- Independent
- Model
- OpenAI ChatGPT and Codex; versions not stated
- Field
- Mathematics
- Date
- 2026-11-15
- Human collaborators
- Mark Lewko
What was found
For the Kakeya maximal function takes the largest average of over radius- neighbourhoods of unit segments in each direction. The two Lean theorems are the maximal bound and a weighted tube bound: for -separated directions, arbitrary centres and nonnegative weights, . By Hölder's inequality the -neighbourhood of every Kakeya set then has volume at least , which matches the construction of Fernández-Delgado and de la Salle, and the estimates imply the Kakeya maximal function conjecture and full Minkowski dimension for every bounded Kakeya set. Those consequences and the matching lower bound are written arguments, not Lean theorems. The development reuses 1,682 files of OpenAI's Lean library at a pinned commit, which the record calls substantial reuse rather than an independent implementation of the analytic infrastructure.
Novelty check
arXiv, 2026-11-19, abs:Kakeya AND abs:logarithmic AND abs:maximal and abs:Kakeya AND abs:"maximal function" AND abs:"every dimension" both returned no records. Within this registry, 2026-10-26-local-smoothing-all-dimensions records OpenAI's claim, with its own unaudited Lean formalization, of the Kakeya maximal conjecture in every dimension as a consequence of local smoothing, and 2026-10-06-kakeya-3d-maximal-4d records the three-dimensional maximal and four-dimensional set claims; this registration is a different, sharper logarithmic statement by a different author, and it does not check either of those manuscripts. The 2026-11-20 watch run surfaced it through its Palomar registration.
Caveats and known objections
Held at claimed on a conservative reading. Palomar records trust level high and checks by the nanoda and con-ron kernels of KakeyaPalomar.sharpMaximalBound and KakeyaPalomar.sharpTubeBound against a 58-line Mathlib-only Challenge, which for other entries in this registry has supported formal. Against that: no human outside the author has audited the Challenge statements; the registry's review and the repository's own review were done by models; the repository README still says the candidate has not been registered and describes a failed preflight; and the geometric consequences that make the result notable are not formalized. There is no paper; the mathematics is in the repository's notes and a working manuscript. The Kakeya maximal conjecture in every dimension is a major claim, and nobody in harmonic analysis is on record as having read this. Autonomy is ai-assisted on the record's own wording, which credits Lewko with research direction and formulation and the models with 'AI-assisted mathematical development and Lean implementation'; collaborative or ai-led are arguable, and the weaker grade wins.
Machine-checked elsewhere
PalomarSharp logarithmic Kakeya estimates in every dimension
PALOMAR-2026-11-15-000001 · mrklwk/kakeya-sharp at 0ffa953d9275 · checked 2026-11-19
ProvedKakeyaPalomar.sharpMaximalBoundKakeyaPalomar.sharpTubeBound
A Lean proof that typechecks against the recorded statement at a pinned commit under a declared axiom set, checked mechanically rather than by human review. It does not certify that a result is new or of research interest: the only filter on that is a language-model screen, and Palomar states that it adds no human editorial step.
Trust level high; checked by the nanoda and con-ron kernels. Only the two upper estimates are Lean theorems; the geometric consequences and sharpness arguments are not formalized.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Claimed and AI-assisted.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded claimed for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Sharp logarithmic Kakeya maximal and tube bounds in every dimension, registered as a Lean proof. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-15-sharp-logarithmic-kakeya
BibTeX
@misc{whataifound-independent-2026-kakeya,
title = {Sharp logarithmic Kakeya maximal and tube bounds in every dimension, registered as a Lean proof},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Independent. Verification: Claimed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-11-15-sharp-logarithmic-kakeya}
}