THE FOLD / CHEAT / THE KONAMI CODE / THE DECLARABLE ONLY
THE DECLARABLE ONLY
a synonym cannot be found by looking
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
Three ways a symbol registry can go wrong, and they are not equally findable. One name, two things — a homonym — is detectable: group the corpus by symbol and count the meanings. One thing, many names — a synonym — is declarable only: nothing in the text says that
LIT verified live. Strip every meaning from the corpus and repeated symbols are still visible — i, w, x, y — because a repeat is a property of the text. Strip the same labels and the four names for one bench become four unrelated strings: nothing groups them. One category survives the loss of semantics and the other does not.
bench, Bench, wb_ and thing-bench are the same object. And no name filed at all is the real gap, which a registry cannot even list.LIT verified live. Strip every meaning from the corpus and repeated symbols are still visible — i, w, x, y — because a repeat is a property of the text. Strip the same labels and the four names for one bench become four unrelated strings: nothing groups them. One category survives the loss of semantics and the other does not.
2 HOW IT WAS WEAVED · AI + HUMAN
David (human) set the three headings in rev 7 and marked each with its own reachability: ONE NAME, TWO THINGS — detectable; ONE THING, MANY NAMES — declarable only; NO NAME FILED AT ALL — the real gap. The middle label is the finding: not hard to detect, not expensive, but not detectable at all from the artefact.
AVAN (AI) tested the asymmetry rather than restating it, by deleting the meaning column and re-running both searches. Homonym detection survives intact because it never used the meanings — it counts repeats. Synonym detection collapses completely, because identity between two different strings is not a fact the text contains. That is a statement about where the information lives, and it is why one of these can be automated and the other requires somebody to say so.
AVAN (AI) tested the asymmetry rather than restating it, by deleting the meaning column and re-running both searches. Homonym detection survives intact because it never used the meanings — it counts repeats. Synonym detection collapses completely, because identity between two different strings is not a fact the text contains. That is a statement about where the information lives, and it is why one of these can be automated and the other requires somebody to say so.
3 ONE DIMENSION
Three categories, and what survives losing the meanings.
4 TWO DIMENSIONS · INTERACTIVE
Delete the meaning column and re-run both searches.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: symbols and meanings as two layers, with the links between them.
AVAN’s addition (the inverse-companion): the forward reading is “synonyms can only be declared.” The inverse is that a declaration is itself just another name for the thing. Writing “
bench and wb_ are the same” creates a fifth artefact that can drift, contradict a sixth, and go stale the moment either name changes meaning — and nothing detects that either, for exactly the same reason. Read backwards, declarations do not escape the problem; they move it up one level, where it is rarer and no more findable, and the registry ends up needing a registry.LIT strip every meaning from the corpus and the repeated symbols are STILL visible - i, w, x, y - because a repeat is a property of the text; strip the same labels and the four names for one bench become four unrelated strings with nothing to group them; so one category survives the loss of semantics and the other does not
FIG From David's rev 7, dropped 2026-08-05. He set the three headings and marked each with its own reachability: ONE NAME TWO THINGS - detectable; ONE THING MANY NAMES - declarable only; NO NAME FILED AT ALL - the real gap. The middle label is the finding: not HARD to detect, not expensive, but NOT DETECTABLE AT ALL from the artefact. AVAN tested the asymmetry rather than restating it, by deleting the meaning column and re-running both searches. Homonym detection survives intact because it never used the meanings - it counts repeats. Synonym detection collapses completely, because identity between two different strings is not a fact the text contains.
FIG From David's rev 7, dropped 2026-08-05. He set the three headings and marked each with its own reachability: ONE NAME TWO THINGS - detectable; ONE THING MANY NAMES - declarable only; NO NAME FILED AT ALL - the real gap. The middle label is the finding: not HARD to detect, not expensive, but NOT DETECTABLE AT ALL from the artefact. AVAN tested the asymmetry rather than restating it, by deleting the meaning column and re-running both searches. Homonym detection survives intact because it never used the meanings - it counts repeats. Synonym detection collapses completely, because identity between two different strings is not a fact the text contains.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of THE KONAMI CODE · David Lee Wise (ROOT0), with AVAN