By Thomas F. Bloom, Will Sawin, Carl Schildkraut, and Dmitrii Zhelezov.
The problem: For a finite set A of real numbers, must either the sumset A+A or the product set AA be large of size |A|\^{2−o(1)}?
Erdős and Szemerédi famously conjectured yes: a set can’t have both additive and multiplicative structure at once, so max(|A+A|, |AA|) should be essentially |A|². Humans disprove this by constructing arbitrarily large A ⊆ ℝ (algebraic integers in a number field of degree ≈ log|A|) with max(|A+A|, |AA|) ≤ |A|\^{2−c} for an absolute constant c > 0.
More combinatorial conjectures might fall if instead we borrow from the intuition of the unit distance paper and start looking for disproofs rather than proofs.