Repository path: wiki/arms/ARM-footing-price.md · rendered 2026-09-09

ARM-footing-price — is the price of carrying a footing mark real to a reader, and does the reader get what it buys?

Deliverable it advances (T3, wiki/tracks.md §Deliverables): framework/v0.2 §10.5 is currently one hand's price list for its own sentences, and §10.6 names the blind pricing of the R06/R27 pair as the first design the section needs. This arm builds and runs that design, and puts a third rendering between the two ends so that the answer is a curve rather than two points.

Question

A translator can carry a Japanese source's grammatical footing into English at nearly every site (RS-20260814b, T-genji-yomogiu-R27-v1). Does a reader who is shown only the English notice that anything was bought — and is what they notice a social relation, or merely a marked style?

Why it is not method work — the subject-rule sentence (continue-prompt.md §4.5)

This unit teaches whether a specific, nameable translation decision — spend English words on a grammatical honorific the target language does not have — reaches the reader it is spent on, and what it costs in the prose. That is a sentence about translating literature and about evaluating translations. The panel is the instrument, not the subject: nothing here asks whether the jury is reproducible, whether a past figure holds, or how a sense behaves. If the run comes back null, the finding is about the translation decision, not about the panel.

The one thing that is method work here is discharged as a gate inside the session, per the rule: T-genji-yomogiu-R27-v1's published marked-token total was 74 and is 78, its counter having matched device patterns against line-wrapped text. Corrected in place on 2026-08-14 in framework/v0.2 §10.5, on the translation page and in note (bnw); the density moves 1/13.7 → 1/13.0, the direction that strengthens what the figure was cited for.

Completion criterion

Closed when framework/v0.2 §10.5 states, on evidence the lead did not produce, what carrying the footing costs and what it buys — or states that the run could not tell, and why. §10.6's first bullet is either discharged or rewritten with what blocks it.

Steps and budget — 2 sessions declared

• Step 1 (S187, DONE). Mint R28 footing-selective and translate the middle rung (T-genji-yomogiu-R28-v1, 12 of 52 sites carried, +3.5% words, 12 marked tokens against R27's 78). Build E-20260814h-footing-price: four arms — baseline, selective, maximal, and an ornament-matched decoy built from the baseline — over six matched segments, three blind seats, pre-run critic, run, verify. Write RS-20260814h-footing-price and rewrite §10.5 and §10.6 on it.
• Step 2 (S192, DONE — and the arm closes resolved at 2 of a declared 2). Minted R33 footing-directional, the rule §10.6 named and could not point at, and executed it on the censused span (T-genji-yomogiu-R33-v1, +6.0% words, twenty graded paragraphs, one mark declined in writing). Built and ran E-20260815c — 10 segments × 4 hands, 3 blind seats, 138 bodies, $0.542027, verifier 219 checks 0 failures, 4 of 4 mutations caught, pre-run critic NEEDS-AMENDMENT with 3 BLOCKING and the registered primary replaced before dispatch. RS-20260815c-footing-direction: the run's own instrument gate FAILED and every registered primary is WITHHELD. What stands is the gate's own measurement — narrated behaviour beats wording on the direction of a perceived social grading, and wording survives only on its intensity — and two post-hoc figures registered after the failure and before the data existed: four hands of one scene put different people on top at 9 of 10 segments, and two texts identical but for 70 words of marking part company at 8 of 10. framework/v0.2 §10.5 gains items 6a and 6b, §10.6's direction question is discharged in the negative, changelog v0.2.12.
• ~~Step 2 — next, and step 1's result chooses between the two candidates it named in advance.~~ The arms did separate on social marking (M1 +4.056, and the decoy scored below the plain baseline), so candidate (a)'s premise did not fire. What did fire is something neither candidate anticipated: the run could not price the middle rung at all, because on a segment where two arms are the same text under different labels one seat scored it REL 4 and then 1, and the design's registered noise rule withheld the comparison. Step 2 is therefore candidate (b) narrowed, with the instrument problem solved inside it rather than made into its own unit:
• the primary question is framework/v0.2 §10.5 item 6 — maximal carriage manufactures a relative grading whose direction follows local device density rather than the source — which is a claim about what a translator is doing to a text and is testable on the two published hands already censused site by site in RS-20260814b;
• the middle rung's audibility is re-asked inside that design at a density high enough to clear the observed label noise, with the same-text control repeated as a gate and not as a footnote.
• Candidate (a) — a decoy built from C by closed-class replacement rather than from A by insertion — is the better control and is carried into step 2's materials, on the pre-run critic's finding F2.

What the arm concluded, since the completion criterion asked for a sentence

§10.5 now states, on evidence the lead did not produce, what carrying the footing costs and what it buys: it costs mostly bulk (step 1) and what it buys is a grading of the translator's own manufacture (step 2). §10.6's first bullet was discharged at step 1; its second is discharged here, in the negative, with the rule written and executed rather than merely named. The arm's own last step failed its own gate, and the failure is on the result page's first line rather than in its limits.

Log

• S187, 2026-08-14 — constituted; step 1 done. T3 was the tool's selection at 4 sessions since it last supplied a principal unit and had no live arm. Minted R28, translated T-genji-yomogiu-R28-v1 (12 of 52 sites carried, 40 declined in writing, +3.5% words), built and ran E-20260814h — 4 arms × 6 segments × 3 blind seats, 72 bodies, $0.378857, verifier 66 checks 0 failures, pre-run critic NEEDS REDESIGN with the primary moved before dispatch. RS-20260814h-footing-price: the registered primary FAILS at +0.417 against a 0.50 bar, and the failure is the finding — 70% of the naturalness price of carrying a footing mark is the price of the added words as words. framework/v0.2 §10.5 and §10.6 rewritten, changelog v0.2.9; §10.6's first open question discharged. Gate discharged inside the unit: R27's published marked-token total corrected 74 → 78.
• S192, 2026-08-15 — step 2 done; the arm closes resolved at 2 of 2. T3 was the tool's selection at 4 sessions and this was its only live arm and the most neglected arm in the project. R33 minted and executed; E-20260815c run; RS-20260815c written; framework/v0.2 v0.2.12. The gate failed and the primaries are withheld — the arm closes on what the failure and two post-hoc statistics establish, and says so. Two method notes: (bof) the key-usage cross-check is retired, and (bog) an instrument control must hold the narrated events flat.