Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Putterman 2026 infinite sets no line
Moe Putterman, Mehtaab Sawhney, Gregory Valiant, On infinite sets with no on a line. arXiv:2602.21275 (2026). The arXiv record (https://arxiv.org/abs/2602.21275, read 2026-10-02) names the Creative Commons CC0 1.0 Universal public domain dedication. The folder-name PDF is arXiv v1 of 24 February 2026.
Theorem 1.1 gives an infinite A in R^2 such that every n-point subset P of A contains a subset of size at least n/2 with no three points collinear, yet A admits no partition into finitely many pieces each free of collinear triples. The construction indexes points by the 2-element subsets {i,j} of N, i < j, via P_{i,j} = (t_i + t_j, t_i^2 + t_i t_j + t_j^2) for an algebraically independent sequence (t_i). Claim 2.1 classifies collinear triples: P_{i,j}, P_{j,k}, P_{i,k} are collinear and every other pattern is not, verified by showing that each of the other determinants has a monomial with nonzero coefficient, so that algebraic independence keeps it nonzero. The two properties then follow from graph theory: identifying a point set with a graph, a bipartite subgraph keeps at least half the edges (first property), while a finite partition would give a triangle-free m-coloring of K_N, contradicting Ramsey's theorem (second property). The finitary version yields a lower bound of Omega(log k/log log k) parts for k points against an O(log k) upper bound. This resolves the question of Erdos, Nesetril and Rodl behind Erdos problem 846 with a short proof, avoiding density Hales-Jewett; Rodl noted it also follows from Reiher-Rodl-Sales. On the use of AI the paper says (Section 1.1, p. 1): "The construction and proof were generated by a model internal to OpenAI." The human authors then worked through the proof and rewrote it for human readers.
Source: https://arxiv.org/abs/2602.21275.
Bears on. #846
Results to transcribe.
- Theorem 1.1: There is an infinite A in R^2 where every n-point subset has a subset of size at least n/2 with no 3 collinear, yet A is not a finite union of sets with no 3 collinear.
- Construction: A = { (t_i + t_j, t_i^2 + t_i t_j + t_j^2) : i < j in N } for an algebraically independent sequence (t_i).
- Claim 2.1: Among distinct indices, P_{i,j}, P_{j,k}, P_{i,k} are collinear and the four other index patterns are not, so collinear triples correspond exactly to triangles in K_N.
- Finitary bound: For k points the construction forces Omega(log k/log log k) parts in such a partition, with an O(log k) upper bound from iterating the first property.