EqualifyEverything / equalify-iris

Closes #442. A copula has two ways of denying its subject and the reader knew one of them. `Handwriting is absent.` declares the page blank; `The heading is empty.` says the heading holds nothing — the same news about text — but nothing denied `heading`, so the sentence read as an affirmation and the blank page was reported as a hole, with the word that denied it quoted inside the evidence (`affirmed: "heading is empty"`). `ABSENCE_COMPLEMENT` is the six wordings that do it: `empty`, `blank`, `unmarked`, `unfilled`, `featureless`, and `void` only with its preposition — a stamp that "is void" is a mark ON the paper. `is illegible` stays out: marks that cannot be read are not an absence of marks. 6 complements x 4 text-naming subjects go from 0 of 24 declared to 24 of 24; the same complements against 4 subjects naming none were 24 of 24 and are unchanged. Four page-LOSING defects came off it, none of them in the issue and two of them mine: - `The heading is not empty.` declared the page blank on base, all 24 grid rows: the `not` denied a clause and nothing read what it denied. - `The heading isn't empty.` was the same sentence one apostrophe over and the losing side of it — a contraction is in none of this file's verb lists on purpose, so it found no verb and went out blank while `is not empty` was reported. - Six exceptive wordings (`Blank apart from a caption.`) had no negator in them and shipped a caption out as nothing; the denial-anchored scan reads these complements now. - That scan took the FIRST denial position, and a complement stands BEHIND the negator of its own sentence, so `The page is empty, nothing on it except handwriting.` anchored on `empty` and lost the handwriting. Every position is tried now, which can only add an affirmation. Which nouns survive an exceptive is not decided again here: each row answers as the negator wording that says the same thing already answered — 8 of 8, generalized to 120 of 120 cells over 5 complements x 8 objects x 3 orderings, base-independently. The complement is read at the word right after the verb and after a LINKING verb only, since half of `AFFIRMING_VERB` takes an object and `empty` is the ordinary adjective for a cell or a row. What it costs is stated rather than left to be found: eight wordings of the shape `The heading is empty, handwriting only.` go out blank, at the comma bound that already loses `The heading is absent, handwriting only.` — so the repair belongs at the comma. Of 3,747 page replies with a parseable log, 153 write one of these complements and 135 sit inside a blank declaration, 0 negated and 0 contracted, and 0 verdicts move; 1,583 logs carry more than one negator and none of those moves either. The corpus could not pick the arm, and the wide read rests on the constructed rows. Three review rounds, each finding what the round before it created. Co-authored-by: bbertucc <bbertucc@users.noreply.github.com> Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>