VibeMathedMath problems solved by AI

Rectangles versus Isosceles Triangles in Lattice Sets

Can a finite set of lattice points determine many rectangles but few isosceles triangles? Both parts of the governing question have negative answers, quantified by explicit blowup rates, and the resulting configurations give obstructions in the Mizohata-Takeuchi circle of problems.

Result
Disproved
Status
Resolved
AI contribution
AI co-developed
Method
Construction
Field
Combinatorial geometry
Posed by
Jonathan Bennett and coauthors
Year posed
Years open
Solved
2026-06-29
Model
ChatGPT 5.5 Pro
Vendor
OpenAI
Collaborators
Jonathan Bennett, Vjekoslav Kovac, Shohei Nakamura, Itamar Oliveira
Verification
Unreviewed
Publication
Preprint
Significance
15 / 100
Disclosed cost
Wikipedia
No dedicated article

What the AI did

The paper says in its introduction, and again in a dedicated AI usage declaration, that the combinatorial construction of the finite lattice sets achieving the stated bounds was found by ChatGPT 5.5 Pro; the authors ran it repeatedly and developed the surrounding estimates. Gemini 3.1 Pro produced the TikZ for the figures.

Verification

arXiv preprint; not yet peer-reviewed.

Source

arXiv:2606.30178 - Rectangles, triangles and Schrodinger waves

Discussion