Each card below takes one factual claim from the manuscript, states it as written, and
tests it against the published statistics. The verdict comes from the data rather than from the draft,
so a claim that does not survive is marked and rewritten rather than quietly dropped. Every number is
read from a source file by scripts/derive_claims.py; nothing is typed in by hand. Each card
has its own permalink, so the manuscript can cite the exact evidence \u2014 card numbers follow
reading order and can change as cards are added, so cite the anchor, which does not.