The Sunflower Conjecture

Let \(f_{r}\)(w) be the largest family of sets of size ≤ w containing no r-sunflower (r sets whose pairwise intersections all coincide). Erdős–Rado conjectured \(f_{r}\)(w) ≤ \(c_{r}^{w}\) with \(c_{r}\) depending only on r.

The sunflower lemma powered circuit-complexity lower bounds for decades. Alweiss–Lovett–Wu–Zhang's 2019 bombshell introduced robust sunflowers and dropped the base to (r log w loglog w)-ish, winning Frankston–Kahn–Narayanan–Park a refinement that PROVED the Kahn–Kalai expectation conjecture and Park–Pham took Talagrand's selector process. Bell–Chueluecha–Warnke trimmed to O(r log w); Fukuyama's 2025 note shaved further. The logarithm refuses to die — and the surrounding machinery now permeates probabilistic combinatorics.