The method
Every dossier on this site is scored by a machine on six dimensions, out of a hundred, from facts already stored and already checked: which citations name a dated document, whether an argument is declared and mirrored, how much prose there is, whether the arithmetic is shown in both languages. Nothing on this page is a judgement anyone here typed about their own work.
Measured across the live library on 2026-09-19 · These numbers are re-measured against the live library on every release, and the release is refused if the page and the library disagree.
Evidence, argument and depth must each reach the floor on their own. A dossier that averages well while failing one of the three does not meet the standard, because an unarguable dossier with good pictures is not a good dossier.
Where a measure cannot be read, the dossier is recorded as unscored. It is never counted as a zero and never dropped from the denominator, so the number of dossiers meeting the standard can never be improved by failing to measure one.
Every published dossier is scored. Only a dossier long enough to carry an argument and cited enough to support a figure is charged against the standard; a shorter note is scored and left uncharged rather than quietly excluded.
What it measures. Citations that name one dated document, graded by the same rules the citation registers enforce.
What it measures. A declared argument: the question, the tension, the mechanism, and what would prove it wrong, with a French version to dispute it in.
What it measures. Words of prose, against the two thousand word target the house sets for an analytical dossier.
What it measures. Exhibits that actually render to a reader, counted in both languages by the engine that draws them.
What it measures. The length of the French document against the English one.
What it measures. Whether every citation carried can be traced to an edition, not only most of them.
