How a study is chosen. Each Monday the pipeline scores every fully-verified init of the trailing week for instructiveness: how far its Day-5 skill sat from the season mean (busts and nails both teach), how widely the models disagreed (a winner and a loser to name), and whether the issue-time Fork call — the calibrated bust probability — turned out right. The standout becomes that week's study. Everything is traceable: each figure comes from the same frozen artifacts served on the Verification board and each study links to the full per-run post-mortem, so any claim can be checked against the record it cites.