Cut the circle at the origin. An Evidence Press briefing. Look at the eigenvalues of a large random matrix, on the scale of one gap between neighbours, and the same random pattern appears for a very wide range of models. It is called the sine-beta process. The number beta measures how strongly neighbouring points repel. Exact formulas for this pattern are rare, so shape properties that hold for every beta are valuable. In twenty twenty-four, Johnston and Prochno conjectured that if you number the points outward from a fixed origin, any selection of them has a log-concave joint law: a single-peaked, well-concentrated shape. Near the edge of the spectrum this was known, because the k-th point from the edge is a fixed coordinate. In the middle it was not, because how many points lie to the left of the origin is random, and mixing log-concave laws can spoil log-concavity. The idea of this release is to put the particles on a circle and cut the circle at the origin. Then the first point to the right of the origin is simply the first coordinate, with nothing random about the labelling. On the cut circle, log-concavity and a strong form of positive dependence hold, and both survive the limit. So the conjecture is proved, for every beta. A stronger version keeps the repulsion between points. As a consequence, the typical spacing is less spread out than a Gamma variable of shape beta plus one, so its variance is at most one over beta plus one. That shape cannot be improved. The same bounds hold at every size for finite matrix ensembles with a convex potential. There is also an exact threshold when a charge sits at the origin, and a new range at the hard edge. Now the limits. At beta equal to two, the variance bound is one third, against a true value of about nought point one eight. For very large gaps the bounds are weak: the bound on a gap three times the mean is about seven in a hundred, while the truth is about five in a hundred thousand. Nothing here concerns zeros of the Riemann zeta function, and the exact hard-edge threshold remains open. The proofs rely on strong known limit theorems. This is an anonymous, AI-assisted, unrefereed candidate, published on the fourteenth of November twenty twenty-six. Its review was internal and AI-assisted, not peer review. The paper, the checking scripts and the evidence are linked from the Evidence Press release page. This briefing uses an AI-generated voice and is not additional scientific evidence.