Formally verified AI-led

Large triangles in planar sets of positive upper density, resolving Castro-Silva's conjecture, with a Lean proof

Polona Durcik directed GPT-6 Astra Pro to a proof, formalized in Lean by Claude Opus 5.5 and registered on Palomar, that every measurable planar set of positive upper Banach density contains a rotated and translated copy of every sufficiently large dilate of any fixed non-degenerate triangle, which gives the planar conjecture of Davi Castro-Silva.

Model
GPT-6 Astra Pro via Codex (proof and manuscript); Claude Opus 5.5 at medium effort (Lean formalization)
Field
Mathematics
Date
2026-11-10
Human collaborators
Polona Durcik
Problem posed
2021 · open 5 yrs

What was found

Furstenberg, Katznelson and Weiss showed that every positive-radius neighbourhood of a set of positive upper density contains congruent copies of all large dilates of a triangle, Bourgain obtained exact copies of non-degenerate simplices in dimension three and higher, and Falconer, Kovac and Yavicoli reached the planar statement for density above one half. Castro-Silva conjectured it for every positive upper density, measured in squares centred at the origin. The note proves it under upper Banach density, which dominates that centred density, and with orientation-preserving copies, so it implies the conjecture. The argument combines the density Hales-Jewett theorem with an operator-norm continuity estimate for the triangle-counting form, derived from a trilinear smoothing theorem. Its provenance statement says the manuscript came from three conversations with GPT-6 Astra Pro on 4 and 5 November 2026, in which Durcik directed the model toward trilinear oscillatory integral methods from papers by Christ and by Christ, Durcik and Roos that she supplied, with otherwise generic prompts.

Novelty check

arXiv, 2026-11-16, abs:triangle AND abs:"upper density" AND abs:plane returned no records. The note's own literature review stops at Falconer, Kovac and Yavicoli's 2022 result for density above one half. Castro-Silva's broader conjecture for d+1d+1 points in ℝd\mathbb{R}^d in other dimensions is not addressed. Nothing in this registry covered it; the nearest entry, 2026-10-19-erdos-173-monochromatic-triangles, concerns colourings of the plane.

Caveats and known objections

Graded formal on the Palomar registration, trust level high, which vibemathed reports as checked with the nanoda and con-ron kernels against a Challenge.lean whose headline statement, including the definition of upper Banach density, it read and found to match Theorem 1.1. That match was made by the author and by an aggregator, not by an independent specialist, and both the Lean statement and its proof were machine-generated. The note is nine pages, unrefereed, and labelled on Hexagon as primarily AI-generated with human understanding of some parts. No outside mathematician is on record as having read it. Autonomy is ai-led rather than autonomous on the strictest defensible reading: the human posed the problem, chose the family of methods and supplied the papers that carry them, and the model produced the argument.

Machine-checked elsewhere

PalomarTriangles in sets of positive upper density in the plane

PALOMAR-2026-11-10-000004 · durcikp/lean-trianglesinplane at 3fb29cdc7ad9 · checked 2026-11-16

ProvedAuto.triangles_in_the_plane

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.

Also recorded at

vibemathedthe-positive-density-configuration-problem

Nothing mechanically. It is a parallel listing of the same result, carrying its own verification label rather than an independent check.

Labels it Lean-checked with the statement unaudited, and records reading Castro-Silva's Conjecture 1 in arXiv 2102.10018 against the Lean headline statement.

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

Entry history (1 event)
  1. AddedEntered the registry graded Formally verified and AI-led.

Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.

Graded formally verified for verification and ai-led for autonomy. What these mean.

Cite this entry

Plain text
whataifound.org. (2026). Large triangles in planar sets of positive upper density, resolving Castro-Silva's conjecture, with a Lean proof. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-10-castro-silva-triangles-plane
BibTeX
@misc{whataifound-independent-2026-plane,
  title        = {Large triangles in planar sets of positive upper density, resolving Castro-Silva's conjecture, with a Lean proof},
  author       = {{whataifound.org}},
  year         = {2026},
  howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
  note         = {Result by Independent. Verification: Formally verified. Autonomy: AI-led.},
  url          = {https://whataifound.org/finding/2026-11-10-castro-silva-triangles-plane}
}

Related findings

← All mathematics findings in the registry