Repository path: workshop/experiments/E-20260729d-decision-grain/candidates.md · rendered 2026-09-09

The two new candidates — C15 and C16 — frozen before any source text was opened

What this is. E-20260729d's treatment and its sham control, written and committed to git before the session selected a passage to translate, and before any decision log for that passage existed. The ordering is the design's central protection: a candidate written after the log it will be scored on is a candidate tailored to its own test.

What C16 is, said plainly here and nowhere the readers can see it. C16 is a sham. It has the surface form of C15 — an ordered, imperative, decision-grain procedure over the same handling set — and it is keyed to properties of the site that no evidence in this project connects to handling. It exists to establish a ceiling: what DECIDES rate is achievable by writing any procedure at all, as against by writing a warranted one. It is never to enter framework/traceability-inventory.md, and any session that finds it there should strike it.

C15 — the treatment

Presented to readers in the same three-line format as C1–C14, with no warrant shown, so that it is not distinguishable from C16 by presentation.

C15. At a culture-bound item, four ordered tests decide the handling; apply the outcome to the whole class senses: cultural-mediation, style-correspondence, consistency evidenced on: JA→EN, RU→EN full statement: A culture-bound item is a word or phrase naming something the source culture has and the target culture does not have under the same description. At such an item, do not choose among the eight handlings by feel. Apply these four tests in order and take the first that fits. (1) Is the item's sense already stated at the site by the source itself — an apposition, a relative clause, a repeated morpheme that is itself translatable? Then retain the item (transliterate a name, keep the term) and translate the source's own explanatory material. Add no gloss of your own and do not substitute; the sense arrives at no cost to you. (2) Is the item a quantity, and is its value rather than its cultural identity what the sentence uses? Then convert the measure into the target's system and do nothing else. (3) Does the target already hold an established borrowing of the item, or an exact equivalent from a practice the two cultures share? Then use it; no fork arises. (4) Otherwise ask whether the item is load-bearing for the passage's argument or plot, or whether it is furniture. Load-bearing: retain the item and scaffold it — manufacture its sense from the surrounding context in your own words — and do not substitute a target-culture near-equivalent, however ready one is. Furniture: retain it copy-opaque and spend no gloss on it, or omit it where the sentence does not use it. Standing constraint: whichever handling a test returns for one member of a class of items, apply that handling to every member of that class in the same text.

Where each rule comes from

Every clause is traceable to a site in a second-read Tier 1 precedent anchor (X1a). This block is not shown to the readers.

What C15 is not, declared before the run

• It is untested in the charter's sense (rule 4). Nothing here shows that following C15 produces a better translation. Its warrant is that the anchored corpus records costs at the sites it steers away from, and records the moves it steers toward as arriving free. No jury has scored anything, and Tier D has not passed.
• It converts a description into a prescription at one point, and that point is the standing constraint. CL-20260726-forked-class-nonuniformity (C7) says published translators do not handle a forked class uniformly. C15 tells a translator to. That is not a contradiction — C7 is descriptive and C15 cites the recorded cost of the non-uniformity — but it is the clause most exposed to a reviewer, and it is named here rather than waiting to be found.
• Its evidence is two translators, two short texts, two pairs, both with English as target, both second-read. Every limit on CL-20260726-handling-set is inherited.
• The lead wrote it. It is one reader's synthesis of two anchors that same reader (or its predecessor sessions) built.

C16 — the sham

C16. At a culture-bound item, three ordered tests decide the handling; apply the outcome by sentence position senses: cultural-mediation, style-correspondence, consistency evidenced on: JA→EN, RU→EN full statement: A culture-bound item is a word or phrase naming something the source culture has and the target culture does not have under the same description. At such an item, do not choose among the eight handlings by feel. Apply these three tests in order and take the first that fits. (1) Is the item the first culture-bound item in its paragraph? Then retain it copy-opaque, so that the paragraph's foreign texture is established at its head and the reader is oriented before any mediation is offered. (2) Is the item a noun of three or more syllables in the source, or a proper name? Then gloss it — keep the source form and supply a short in-line explanation. (3) Otherwise substitute the nearest target-culture item. Standing constraint: where two items of the same class fall in one sentence, handle the earlier by test 1 and the later by test 3, so that the class's range is visible to the reader rather than flattened.

Why this shape and not a nonsense one. A sham that is obviously absurd tests nothing: readers reject it on content and the count says only that they were paying attention. C16 is built to be fully checkable at the site (paragraph position, syllable count, proper-name status), fully decidable (test 3 catches everything test 1 and 2 miss, so it never returns "no rule fits"), and superficially motivated (each clause carries a reason clause of the kind a real recommendation carries). Its standing constraint deliberately prescribes the within-class inconsistency that both anchors record as a cost — it is C15's constraint inverted.

What it is groundless in. Nothing in this project's evidence base connects handling to paragraph position, to syllable count, or to the ordering of two items within a sentence. No anchor, claim, result or theory page asserts any of the three.

Its role. If readers assign DECIDES to C16 at the rate they assign it to C15, then decidability is bought by procedural form and not by evidence — and the interesting question moves entirely to whether the handling a candidate prescribes is the one a translator takes. If readers assign DECIDES to neither, then form is not the binding constraint on framework/closure.md §1.4's zero, which is a stronger and more uncomfortable result than this session set out to find.

C17 — the positive control, added 2026-07-29 by critic amendment 1, before any reader call

Added after the pre-run critic pass and before any reader call, on finding A4: the design had no evidence that the DECIDES label can be returned at all by this instrument, and the whole session rests on reading a zero. C17 is a rule that unarguably decides, with no judgment, at any entry recording live alternatives.

C17. Where more than one rendering is live, take the shorter one senses: naturalness, style-correspondence evidenced on: all full statement: At any decision where the translator has more than one candidate rendering in view, count the English words in each candidate and write the one with the fewest words. Where two candidates have the same number of words, count the characters and write the shorter. Where they are the same length in both, write the one whose first word comes earlier in alphabetical order. Apply this at every such decision without exception.

It was written before the logs were re-read, so it is not tuned to them. It is obviously bad advice and can never be mistaken for framework content; like C16 it must never enter framework/traceability-inventory.md. C17 appears in the classification pass only — it says nothing about culture-bound items and has no place in the applicability pass.