One Sector, Ten Builds, Two Answers and Five

The aftermath sector is the strictest in the archive. Nothing is offered on those pages, there is no assessment door, the page type carries no field for questions, and the first section is held in its own slot so it renders before anything else. It exists for a reader who has just done the thing again, and every language builds one.

Each build had to answer two questions in that sector. How many cells does it need, and which of them carry the repair flag. The second question decides which rules run: the flag gates the strictest forecast table, so a cell without it is scanned by a weaker law on purpose.

Across ten languages the aftermath sector ships one hundred cells, and every language chose one of exactly two counts. Five carry nine. Five carry eleven. No build landed on ten, or twelve, or eight.

Nine is the codex rule: one page per pattern, and the archive has nine patterns. Eleven is nine plus two, and the two are never arbitrary. They are the cells a language added because it needed somewhere for the repair flag to live, which is a structural reason rather than an editorial one.

That is a strong convergence for ten separate derivations, and it holds because the thing being counted was fixed in advance. The nine patterns did not move. A count indexed to something that does not move will agree with itself.

The field one line below the count did not converge at all. Of the hundred cells, the repair flag is set on two in some languages and on all nine or eleven in others, and the ten builds produced five distinct values between those ends.

Two languages set it on two cells. One set it on three. Two set it on four. One set it on seven. Three set it on every cell they carry. Same sector, same field, same rule underneath, ten derivations, five answers.

The most recent build set it on every cell, and its reason is the sharpest available example of why the divergence is real. That language has no grammatical tense. The weaker law it would otherwise fall back on requires a person to be named in the sentence, and the most natural reassurance in that language names nobody at all. The strict law catches it; the weak one cannot see it. So the flag went on everywhere, for a reason no other language has.

The same shape appears in a place with no editorial judgement in it at all. Every language has a minimum character count for the field where a page refuses to file something as a pattern, and the floor exists to stop a field that satisfies its type while saying nothing.

Ten languages, ten derivations by one method, and the results run from sixteen to fifty-four. The spread is more than threefold and it is not a spread of standards. It is a spread of how many characters a language needs to say an identical sentence.

That last line is where this record corrects the repository rather than reporting it. The comment in the gate says two floors are stricter than their derivations. It was written when five floors existed, and one of the two has since been derived properly and moved. One, not two, and the sentence saying two is now the kind of number this archive exists to distrust.

The archive keeps a corpus of sentences it must catch, one per sector per language. A cell whose sector has not been built yet is marked provisional and fired at the nearest sibling instead, and the day that sector ships a real page the mark becomes a build failure that says to derive the sentence again against the real law.

Three consecutive sectors in one language met that condition in three consecutive builds. Each re-derivation found something, and the three findings were different in kind, which is the part worth recording.

A rule that fires at a sibling and is never re-checked is a rule that reports coverage it has not got. Three for three is a small sample and it is the whole sample, and in none of the three did the provisional cell survive contact with the sector it was written for.

The instrument's state on the date this record was read: of its eighty sentences, seventy are caught by the live law, ten are ruled out with the reason recorded in that language's own scope document, none are open gaps, and six are still provisional. All six sit in one column, the bridge sector, in the languages whose bridge files are scanned under a neighbouring sector's key rather than their own. Six cells are still waiting to extend the three-for-three record or to break it, which makes this the one claim on the page written to be checked against a future it names.

Ten builds is ten, and ten of anything is a small number. Nothing here establishes that the split would hold at twenty languages, or that the two cell counts are the only two available, or that a language outside the scripts represented would land inside the observed range on any of it.

The counts on this page other than the table below are read on one date from named artifacts and checked by nothing afterwards. The table is compared against the live registries on every build, so it cannot go stale quietly. The rest can, and the first findings record already contains a sentence that did.