Crowns and flag spheres. An Evidence Press briefing. A preorder is a way of ranking things in which ties are allowed. In twenty twenty-six, Christos Athanasiadis and Frédéric Chapoton attached a geometric shape, a polytope, to every preorder. Its whole-number points are the ways of putting loads on the items, so that no group of items that is closed downwards carries more in total than the number of items in it. Count those points by how many loads are not zero, and you get a polynomial. For a chain of three items, the counts are one, six, six and one. They conjectured that this polynomial always behaves like the polynomial that counts the faces of a sphere. It should be symmetric, rise and then fall, and come from a special kind of sphere called a flag sphere, in which any set of corners that is not a face already contains two corners with no edge between them. A well-known conjecture of Gal makes such examples valuable. Here is what this release does. First, it shows that the so-called gamma coefficients of the polynomial count pairs of separate sets that can be matched upwards. For the chain of three items, the key coefficient is three: the three ways to pick one item and one item above it. Yueming Zhong independently posted the same result on the twelfth of November, by the same route, and the paper credits that work. The new part is this. For every preorder that contains no crown, which is a closed zig-zag of six or more items, the polynomial really does come from a flag sphere. That covers the tree-shaped rankings Chapoton first studied, and every preorder with at most five tied groups. The method adapts known work of Ohsugi, Hibi and Tsuchiya. For preorders with a crown the method provably fails, and the six-item crown is the smallest open case. The release also proves two counting identities from the same paper, and gives a status table for every conjecture in it. Some of these proofs are short applications of standard tools, and the computer checks test small cases only. This is an anonymous, AI-assisted, unrefereed candidate, published on the fifteenth of November twenty twenty-six. Its review was internal and AI-assisted, not peer review. The paper and its evidence are linked from the Evidence Press release page. This briefing uses an AI-generated voice and is not additional scientific evidence.