Open; the site labels it OPEN. Erdős and Graham state the bound with $c/K^2$ in place of $e^{-cK}$ (the monograph, without proof or reference); the best bound found is Korsky's $\varepsilon(A)\le\exp(-c\sqrt{K\log K})$ for $R(A)>K\ge K_0$, an arXiv preprint of July 2026 whose acknowledgment declares extensive assistance from GPT-5.5 Pro, with no journal record and, by its author's own statement, no journal submission planned; a construction (the site's thread; the preprint's display (1.4)) shows that no bound better than $\exp(-(1+o(1))K\log K)$ holds for all multisets. Korsky's bound settles no instance of the question, so it has no claim page. The conjectured exponential rate is open, and no proof, disproof or proof claim for it was found in the search whose scope the Current assessment records; this is a bounded negative finding.