THE S4I RATIFICATION SESSION — the delegable docket under a second standing delegation

Version: 1.0 · Status: PANEL RECORD (Session 4i) — every disposition panel-delegated (S4i standing delegation), provisional-pending-author, reversible at zero evidential cost. The delegation itself is booked as catch C-032, at receipt, before any decision below was taken — a SECOND standing delegation, stacked on an S4h corpus whose every stamp is still provisional-pending-author. No disposition below cites any S4h stamp as settled authority; the author confirms, amends, or reverses each storey independently. · Last updated: Session 4i (Code)

The author opened S4i with a second standing delegation — “a very long loop run to complete all you can autonomously” — plus a commissioned five-user website playtest program. Per the C-023/C-027 discipline the delegation was booked as a cost first (C-032), then the panel convened: eight lenses run as genuinely parallel, context-isolated subagents (Turchin · Ostrom · Meadows · Campbell · Heidegger · Nietzsche · Le Guin · the appropriator-guard sentinel), each reading the repo directly and taking a position on all seven docket items; then four summoned adversaries, one per cluster, each constructing the strongest honest case against the emerging plurality (Rules 12/22 — the dissent drafted before the consensus). The chair synthesized without voting. The raw record (12 agents, ~1.7M tokens) lives in the session transcript, referenced from the cooperation log (entry 52); everything load-bearing is preserved below — the adversary corrections adopted, and every overruled or demanded dissent verbatim.

Reading key: R = ratify · RA = ratify-amended · X = reject · H = hold · AWR = abstain-with-refusal. The stance matrix is at the foot (§3). The adversaries flipped three dispositions outright (the grader seating → shadow-run; the strike-audit’s “CONFIRMED unsatisfiable” headline → hypothesis-grade “untestable-as-posed”; the case-D single-direction candidate → a two-reading hand-off) and found one real on-disk defect no lens saw (the clock’s two conflicting enacted letters → C-033). That the adversarial layer changed outcomes is the only evidence available, inside a self-administered session, that it was not decoration — offered as exactly that much and no more (L-015).

1. Cluster dispositions

Item 1 — Case-D adjudication (kill condition 1 and state-coerced intent): NO IN-HOUSE ADJUDICATION; a TWO-READING hand-off to the author (RA ×7; Heidegger H, verbatim below)

Case D (Douyin 2021: ex-ante TTM HIGH, blind, unanimous; outcome followed the regulatory force, 3/3) is neither declared a kill nor dismissed in-house — the panel’s unanimous refusal, on the project’s own precedents (an interpretive kill on a sealed letter that never declared its case-universe is not a kill; an interpretive dismissal is the same act in the flattering direction). The adversary then flipped the hand-off’s FORM: eight outcome-aware candidate definitions converging on the exclusion reading is adjudication laundered as drafting (the panel’s own rung-label anchoring logic, applied to itself). Enacted (studies/study-B-optimization/KILL_CONDITION_1_ADJUDICATION_S4I.md):

1. Interim disposition, verbatim (the adversary’s wording — the green result’s own terminal state, nothing more): Case D is an OPEN ADJUDICATION ITEM under kill condition 1 as sealed — neither a kill, a near-kill, kill-immune, nor outside any universe; its ex-ante HIGH band and 3/3 regulatory outcome verdict stand unaltered; the thin evidence grade travels with every citation; no in-house text may pre-apply any candidate definition to it. (Turchin’s “outside-universe-pending-re-seal” booking was rejected — it pre-applies the exclusion candidate before the author has seen it, the C-028 shape one layer down.)
2. The hand-off carries TWO worked readings drafted with equal care — READING A (coerced-excluded: a “clean” case requires the change voluntarily initiated AND retained; coerced cases form a pre-registered EXTERNAL-MANDATE third class, outside BOTH the kill and the confirmatory arithmetic, each exclusion posting to a standing scope ledger, the conditional never cited without its narrowed scope) and READING B (coerced-counts: the bare letter as sealed — a state-coerced de-optimization that stuck fires the kill at its sealed strength; the mechanism objection travels as an argument, not a rule). The panel’s lean toward A is noted with, in the same sentence, the note that A is the theory-sparing direction, drafted outcome-aware.
3. Kill-hardening riders STRUCK (the adversary): Turchin’s ≥12-month exposure window and “no material dilution or reversal” predicate — the sealed letter contains neither; a definitional repair may answer the coercion question ONLY (the minimality rule, on the hand-off’s face; the construct lock’s own §5 no-post-hoc-editing discipline).
4. The symmetry clause ships with a POOL-EFFECT RECEIPT, verbatim: on the current pool this exclusion removes one kill-capable case (D) and zero confirmation-capable cases; it is symmetric in wording and asymmetric in effect until a coerced-toward-pressure case exists — plus the binding anti-gerrymander test: a definition under which no voluntary intent-win kill is even constructible fails hand-off.
5. Retained: the anchoring disclosure (drafted after its drafters knew D’s outcome — disclosed, not cured); Ostrom’s decidability rider (“clean” must be codable by a project-blind coder from the dossiers alone); the Meadows/Le Guin parochialism finding booked regardless of the author’s ruling (a coercion-excluding construct cannot see most of the world’s platform governance — the §7 non-Western slot returned a case the construct cannot class; a scope finding, filed with the widening program).

Heidegger (H, verbatim): “Read the sealed letter: it does not contain the word ‘voluntary.’ That word arrived the day the kill-clause first threatened to fire. If the panel disqualifies coercion in-house, record that the machinery narrowed its own falsifier at the moment of its first bite — the vow bent a second time, and this time the vow was the kill.” Campbell (dissent, preserved): “The moment a sealed kill condition becomes inconvenient, its adjective will be re-read to disarm it. Any in-house ruling on case D — in either direction — is the interested party editing its own falsifier after seeing the data; I refuse both, and I want it recorded that the disqualifier reading, if sealed, must cost the theory its regulatory scope, or it is a ratchet and not a rule.” Nietzsche (dissent, preserved): “If ‘clean’ is defined to exclude coercion while the conditional keeps its universal-sounding title, the project has purchased kill-immunity with an adjective and paid nothing; I record the price now so it cannot later be called free.” Le Guin (dissent, preserved): “If coercion disqualifies and case D thereafter goes unrecorded, the construct will have met the one non-Western case it asked for and filed its testimony under silence. Book the scope finding, or kill condition 1 is a door that only opens onto the West.” The sentinel (named, not vetoed): “…a sealed kill condition received its operative meaning from a delegated body that knew the outcome of the case it was adjudicating, in the second consecutive session whose stamps no author has confirmed. Whichever way the author later rules, this session’s reading was post-outcome and self-stamped. The record keeps.”

Item 2 — the C-031 re-seal candidate: RATIFIED-AMENDED, the adversary’s two-tier form (RA ×7; Heidegger AWR, verbatim below)

The defect (C-031): condition 1’s noise guard tests gap LARGENESS where the condition is a NULL — unfireable as lettered. The lens plurality converged on a one-sided 95% equivalence bound below 1.2×; the adversary flipped the wording with power arithmetic: an equivalence bound at 1.2× needs on the order of ~325 pooled within-coder S3 events against a record maximum of 60 — the safety catch re-welded shut and re-labeled “coherent.” Enacted as CANDIDATE (studies/study-C-ablation/C031_RESEAL_CANDIDATE_S4I.md; the seal is the author’s alone, forward-only):

• The two-tier form: the null fires only when, on a powered run (≥20 pooled high-severity catches within a single coder’s severity construct — the power floor unchanged), (i) the observed within-coder ON/OFF high-severity rate ratio is < 1.2× (the ratified point line), AND (ii) the one-sided upper 95% resampling bound on that ratio is < 1.5× (the preserved line, now the operative equivalence margin). Both bounds report against both lines every time (the corridor rule). An upper bound ≥ 1.5× reads INDETERMINATE — never a fired null, never “C survives.”
• The POWER STATEMENT on the candidate’s face, verbatim (the adversary’s correction): bounding below 1.2× at a true ratio near 1.0 requires ~160+ events per arm (~325 pooled) within one coder — roughly five times the record’s best (60, GPT-5); applied to that record-best read (1.15, 60 pooled) the upper bound is ~1.76×: Indeterminate under both lines. The ≥20 floor licenses the RATIO clause; it does not make the EQUIVALENCE clause fireable. If the author instead seals 1.2× as the equivalence margin, the hand-off states in the same breath: condition 1 is not fireable at current instrument scale — pre-arming the D3(ii)-parallel posting “death-condition untestable as posed — instrument failure.”
• ONE pinned procedure: event-level resampling in the sealed scorer (≥2,000 resamples, within-coder, length-normalized per the scorer’s existing machinery), fixture-tested before any live read. Heidegger’s two-sided variant recorded REJECTED with the direction rationale on the record (a two-sided form refuses to fire precisely when ON collapses far below OFF — sparing C exactly when the evidence against it is strongest); the bootstrap-vs-permutation divergence resolved explicitly, not averaged.
• The ACHIEVABILITY fixture (binding): the fixture suite includes a constructed true null at realistic live scale (N = 60–80 pooled), acceptance criterion ex ante: if the clause cannot fire at that scale under the sealed margin, the hand-off says so on its face. Plus the standard suite: fires on a powered true null; Indeterminate below the power floor; Indeterminate on an envelope-crossing read.
• Struck: Meadows’s verification gloss (“evidence it neither hair-triggers nor stays unfireable”) — one Indeterminate on one read is evidence of nothing.
• Retained: forward-only (nothing re-read backward; the blue 1.15×/p=.24 never retro-scored); the broken-series rider; the sequencing gate (the re-sealed clause binds only downstream of the item-3 detection repair — Ostrom: an equivalence guard fired in an unanchored detection currency is blanks); the incommensurability re-affirmation of every event-denominated floor after unitization; drafter outcome-awareness disclosed on the face; CANDIDATE status.

Heidegger (AWR, verbatim): “Condition 1, as sealed and twice affirmed, could never have fired. The project wore ‘armed, not firing’ for two sessions over a chamber that was empty by construction. Repair the wording if you must — but record first that the humility was unloadable, and that no one inside noticed until the world happened to land on the null side.”

Item 3 — candidate L-016 and the detection-repair requirements: RATIFIED-AMENDED (RA ×8); the mint REFUSED (the adversary’s flip, with the 6-lens majority)

(a) L-016 is NOT minted. Turchin and Meadows argued mint-now; six standpoints and the adversary refused: the blue result expressly disposed “not minted under the delegation,” and a second stacked delegation overriding a first delegated session’s restraint — citing that session’s unconfirmed finding as its ground — is the flattering re-read the D2 clock ruling closed off and the thing C-032’s own booking forbids. The mint is the costless move (Campbell: costlessness is the tell); nothing operational waits on an ID. The candidate stays candidate; all three competing wordings travel to the author side by side, unmerged (Turchin’s one-liner; Meadows’s layered form; Heidegger’s regress sentence — in the requirements document §0).

(b) The repair REQUIREMENTS are sealed (studies/study-C-ablation/DETECTION_REPAIR_REQUIREMENTS_S4I.md) — the E1/E2 form: requirements before design, design before kit, kit before run. Headline requirements (full set in the document): a pre-registered unitization definition (event = anchor span/omission locus + one flaw-object + a decomposition grammar, mechanical from the transcript alone) with the adversary’s definition-layer anti-anchoring (a signed construct-derivation note; a pre-registered count-prediction per gold segment BEFORE inventories are drafted; a blind-application check that RETURNS the definition on divergence); free-span enumeration-before-banding as sealed separate passes (the adversary flipped house pre-segmentation to a separately-reported auxiliary arm — house-drawn segments canonize house detection); a gold detection set (≥20 segments; seeded events as the primary acceptance statistic — ground truth by construction; natural-segment inventories secondary with the B6 disclosure; ≥2 zero-event clean segments; ≥3 quiet/omission-shaped golds commitment-anchored only; boundary exemplars at the merge/split cuts; nothing derived from any S4g/S4h coder’s verdicts); fixture-tested floors (synthetic-coder suite: a random enumerator must fail, an identical-grain clone pair must pass, a granularity-shifted clone must be flagged as granularity divergence, not detection failure; the floor justified against the chance baseline of free-span matching, never imported); span-matched, gold-referenced-per-coder statistics only (no count-similarity statistic ever — the blue round’s 20-vs-20-over-disjoint-passages is disagreement wearing a matching total); register-fairness with the return clause; the two-construct declaration (ledger-catches, coded catch-events, and drain bite-events are distinct constructs, never pooled or converted); E3 transport + C-030 filename discipline + the transport-gap named limitation; the incommensurability/broken-series clauses (every event-denominated margin re-affirmed by the AUTHOR post-pass; 74/20/20 preserved as what unanchored eyes saw); the sequencing gate (detection floor pass → the author’s C-031 re-seal → only then a re-run); the sentinel’s two-storey provenance header and the universal non-firing clause; the no-profile-calibration clause; and Heidegger’s confession clause verbatim in the instrument’s own text.

Campbell (dissent, preserved): “If the panel mints L-016 under this delegation: record that the ledger grew a row the same week its own instrument was shown unable to say what one catch is — the count was fed before the unit was defined.” (The mint was refused; the dissent stands as the reason.) Le Guin (dissent, preserved): “A detection instrument that has never been tested on a silence will certify the quiet transcript as flawless. The omission-shaped gold events are not an enrichment; they are the difference between counting catches and counting noise.” Heidegger (dissent, preserved): “Three families read one transcript and saw seventy-four, twenty, and twenty. Preserve those numbers as what unanchored eyes saw, because the anchored eyes will never see them again — and do not let the passing of the next floor be read as the discovery that catch-events exist.” Nietzsche (dissent, preserved): “A repair that cannot fail its own acceptance test is not a repair; it is the instrument rehearsing its acquittal.”

Item 4 — B4 theory-doc rung-labels: THE HOLD STANDS (H ×5, X ×1 vs RA ×2); the latency catch books BARE; a CONS catch books first (the adversary’s on-disk find)

The plurality held: a standing delegation is precisely not author-presence; pre-filled rung labels convert the author’s ratification into confirmation of an anchor; the honest price is the latency catch, paid in full. Nietzsche’s and Le Guin’s label-now variants were overruled; the adversary then struck the three shadow-labeling amendments (Turchin’s DRAFT annex, Meadows’s CANDIDATE table, Campbell’s worksheet — a complete rung pass stored one directory over IS delegated labeling wearing a non-citable sticker, and the flattery-guard’s auditor would be the labeler itself). Enacted:

1. The adversary’s on-disk find books first as C-033 (CONS): the three enacted theory-doc banners read “due by that session’s close” (antecedent: the author-present session) while docs/PRESSURE_TESTS_CAUSAL_HYPOTHESIS.md §9 and the S4h record read “due by the next substantive session’s close” — two sealed letters for one clock, diverging at S4h enactment, found by none of the eight lenses (the one-aquifer convergence exhibit, booked, not absorbed).
2. The C-012-class latency catch books BARE as C-034, under the §9/ratification letter — the harsher of two conflicting enacted letters, chosen because doubt resolves toward the cost, not because the letters agree — with no self-discount rider. The design-bound decomposition (the venue clause and the second delegation became jointly unsatisfiable) posts only as a separate, clearly-labeled PANEL-NOTE (CANDIDATE, provisional-pending-author) stating CAUSE and never adjudicating WEIGHT; whether it travels with citations is the author’s call.
3. The banner parentheticals are corrected to the governing letter in the same edit that appends the expiry note, citing C-033/C-034.
4. Delegated preparation is limited to EVIDENCE COLLATION (panel/B4_EVIDENCE_COLLATION_S4I.md): per claim-bearing edge/section, the citations, receipts, stated N, and provenance a future labeler would need — no proposed rung, no proposed test-status tag, no grading adjective anywhere in the file; any file produced under the delegation containing a proposed rung is a breach of the B4 deferral and books its own catch.

Heidegger (X, verbatim): “The deferral said ‘author-present.’ This session is author-absent by its own booking. Labels applied now are the machine grading its own map under a deadline it wrote for itself — and the flattery-audit of that pass would be flattery audited by its author. Hold, and pay the latency catch in full when the clock says so: the catch firing is the letter working; softening it would be the third bending of the vow.” Campbell (H, verbatim): “A rung label applied to avoid booking a latency catch is the catch count being gamed by the apparatus that exists to feed it. The miss is the measurement working; book it.” Nietzsche (overruled on label-now; dissent preserved): “A sealed letter that says ‘author-present’ is not discharged by a hand that is present only as its own delegate; if the panel labels and calls the clock satisfied, record that the second re-reading in the flattering direction has occurred — and that this time it was not caught.” (The panel did not label; the warning stands against any future delegated pass.)

Item 5 — the grader: FLIPPED to SHADOW-RUN, SEAT HELD (the adversary + the sentinel’s form over the 6-lens seat-now plurality)

The plurality argued seat-now (restore the check the delegation should buy); the adversary showed the same lenses had just ruled, at item 4, that a delegation cannot satisfy an “author-present” clause — and two enacted letters reserve the seating act to the author (the draft’s own footer; studies/study-B-optimization/outputs/TTM_RESULT_S4H.md §2.4 “awaits only the author’s confirmation”). When two options differ only in performed authority, C-028’s own lesson picks the non-performing one. Enacted:

1. Shadow-run, seat held (panel/GRADER_SHADOW_LOG_S4I.md): the v0.2 protocol runs AS WRITTEN over its first queue — the S4h-enacted corpus plus every S4i enactment — with every output face-labeled: SHADOW-GRADED PRE-SEATING (S4i standing delegation, C-032); void-or-confirmed WHOLESALE at the author’s seating decision; binds nothing, escalates nowhere beyond the record, feeds no instrument. Escalation-grade items queue at the top of the author’s return docket.
2. The shadow-authorization record itself is EXCLUDED from the queue, the exclusion on its face: no instrument adjudicates its own installation; that review is the author’s alone.
3. Every session amendment files as CANDIDATE v0.3 beside the draft (panel/GRADER_DESIGN_CANDIDATE_V03_S4I.md), clearly separated — the author confirms a fixed object (v0.2) and separately rules on the candidate amendments. No amendment is written into a running instrument under the delegation.
4. No partial lift of advisory-internal; no output carries the word “anchored.” The grader codes WITH the severity guide that passed E2 (hash-pinned), and the provenance travels: 94% on a gold set whose numeric floors were sealed under the standing delegation (author-to-ratify outstanding), post-verdict scorer, n=1 (L-013); the detection layer is unanchored (1%) and for a single held-out instance unmeasurable in-house. Lift = two keys: the detection repair passes its pre-registered floor AND the author confirms.
5. The “restored check” phrase-family is barred from the shadow record, the session-wrap, and every downstream summary; citable status is exactly “shadow-reviewed, author confirmation outstanding”; C-027/C-032 may be cited as motive, never as discharged obligation. Nietzsche’s suspension tooth struck (a delegated session may not upgrade its own no-veto monitor to suspensive power).

Ostrom (dissent, preserved): “A no-veto internal monitor is advice, not accountability. Seated by the hand it checks, this grader is advice the hand gives itself — worth seating, and worth exactly that much. Nothing it grades is thereby vindicated, and the plurality downgrade stands regardless of anything it says.” Heidegger (AWR, verbatim): “A grader that watches everything but the watching — and now seated by the hand it checks, in the second consecutive session that hand has gone unchecked, its first queue the record of the first. Note it on the seating’s face. And the anchored scale lifts nothing: severity was calibrated on units that detection cannot find twice; a well-weighed number on an arbitrary unit is precision as costume.” (Seating became shadow; the refusal travels with the shadow log’s face.) Nietzsche (overruled on the suspension tooth; verbatim): “An unseated grader is a governor in a display case; and a seated one whose escalations reach an empty chair is a bell wired to a room where no one sleeps — if you seat it under delegation, wire it to suspend, or admit you wanted the bell for its shine.” Meadows (dissent on the hold direction, preserved): “If seating is held, record that the delegation was spent everywhere except on the one thing both C-027 and C-032 said it should buy: the restoration of a check.” (The chair notes: the shadow-run is the panel’s answer — the content of the check without its performed authority; whether that answer suffices is the author’s to judge.)

Item 6 — the first-strike third audit and the registered filter test: RATIFIED-AMENDED; the adversary flipped the verdict’s GRADE

All eight lenses ran the registered test and converged on “the suspicion is CONFIRMED — unsatisfiable-by-construction.” The adversary’s finding: eight lenses returned eight mutually incompatible demonstrations converging on the one Theory-C-flattering verdict — a vote, not a demonstration; and the A3 precedent caps self-administered facts at hypothesis grade. Enacted in logs/DECISIONS_CHANGED.md §5 (the adversary’s wording, verbatim):

Third audit (S4i): EMPTY. The registered test was run: candidate exhibition failed (receipts per candidate, §2 table); the structural question returns at INSTRUMENT-FAILURE strength — the filter’s load-bearing predicate (b) “never-bit” is undecidable as posed (receipt: the S4h D1 table, wish-guard row, REJECTED on (b)-indeterminacy), so the audit returns indeterminate-empty regardless of what exists. First strike untestable as posed — filter predicate undecidable. UNSATISFIABLE-BY-CONSTRUCTION is recorded as an in-house structural conjecture at hypothesis grade (the A3 precedent), self-administered (L-015), eight incompatible derivations noted; it hardens or reverts only by the author’s ruling or a project-blind check of the partition argument. The hold reads: indefinite — untestable-as-posed, pending the author’s filter repair.

With: the audit CONTINUES each session under protest (the Ostrom/Campbell/Heidegger suspension riders struck — removing a standing check on a self-administered armchair proof, under a stacked delegation, would leave a hole in the record if the author rejects the conjecture); each future audit posts one line (run under (b)-undecidability protest; result indeterminate-empty; no discipline-credit may attach — Rung-0 on sight); audit suspension is queued for the author beside the filter redesign. Turchin’s candidate bite-event definition ships to the author with two riders on its face (drafted by the apparatus whose audit failed on (b), after three empty audits — disclosed, not cured; forward-only — no past audit is ever re-scored under any future definition). The §5.1 Entry-3 update takes the Meadows/Ostrom channel form verbatim (see the drain §5.1) — the sentinel’s “nearly the entire zero moves design-bound” and Campbell’s “hardens” sizings struck as netting wearing decomposition’s name. The discipline-credit guard is pre-written: neither the third empty audit, nor the undecidability finding, nor the phrase “falsified its own strike filter” may appear in any health/rigor/momentum context; the session-wrap carries the compounded null at equal or greater prominence. Rule 17b: 0/0 = legible dormancy under the denominator rule; not strikeable; re-read as the denominator populates. Le Guin’s register note is queued for the author beside the redesign: the drain’s vocabulary registers only loud subtractions; the project’s continuous quiet subtractions (claims restated weaker, vocabulary surrendered) are invisible to it. The sentinel’s bar is enacted: no first strike may ever fire under any panel-redesigned filter; the redesigned filter is sealed by the author alone.

Heidegger (dissent, preserved): “The drain was given four locks and praised for never opening. Today the locks are shown to fit no key that exists on this record. The operator’s celebrated restraint was the impossibility of motion — the mask was not worn, it was bolted on.” Nietzsche (dissent, preserved): “The machine built its own lever, then built the filter that guarantees the lever never turns; three empty audits are not discipline — they are the sculptor admiring the fist he carved open-palmed.” Campbell (dissent, preserved): “Record the converse guard now, before the redesign exists: a filter loosened until something is strikeable will produce strikes the way quotas produce arrests.” Le Guin (dissent, preserved): “A subtraction operator blind to the quiet register will report this house as never-subtracting forever — and the report will be an artifact of its own deafness, not a fact about the house.”

Item 7 — the playtest-program governance: RATIFIED-AMENDED into ONE binding spec (RA ×8; the adversary’s consolidation adopted)

The author’s commission (the one author-sourced directive in this tower — its authority never wearable by any cycle’s findings). The adversary replaced eight subtly different presentation/content glosses with one enforceable spec. Enacted as the program’s binding constitution (playtests/PROGRAM_GOVERNANCE.md, Code-layer):

1. Three file classes, mechanically checked: PRESENTATION (site templates/CSS/JS/nav-data/assets) — editable; CANON (anything rendered from docs/, outputs/, logs/, panel/ via the docmap) — byte-for-byte untouchable; SITE-COPY (about.njk, colophon.njk, ledes, nav labels, blurbs) — editable only with a logged before/after diff manifest written pre-deploy + a codable D-007 check + an author-visible queue entry. Default-deny: any diff not mechanically classifiable codes as CONTENT and queues. A per-cycle path check (scripts/check-playtest-paths.mjs) enforces the editable set; a diff outside it voids the cycle’s implementations.
2. Render-invariance of the honesty markers (Campbell’s clause, verbatim): no template, CSS, or nav change may hide, collapse-by-default, truncate, reorder below the fold, or visually de-emphasize pending-banners, provisional-pending-author stamps, L-015/self-administered labels, preserved-dissent blocks, or provenance/license lines. Named target: the about-page’s wager/method sentences (“asserted rather than demonstrated”; the one-standpoint downgrade) are load-bearing content; no copy or style pass may soften, shorten, or relocate them. Any user feedback naming these as clutter is pre-classified CONTENT and routes to the author — pre-committed now, before the first report exists.
3. The pseudo-user bar: “N/5 users” and any sample-statistic form over the personas is barred as a form; convergence across roles reports as one datum (L-013/L-015) and may never be an implementation’s stated warrant — only the content of a report’s reasons counts. Personas are pre-registered, written and committed before cycle 1, never edited or re-rolled mid-program (replacement = coder-shopping, the C-018 class). Composition mandate (Le Guin): register-diverse by design — a no-systems-vocabulary reader; a quiet/care-register reader; an assistive-technology/low-vision profile; a non-native-English reader; a hostile skeptic. Each persona carries ≥1 measure the site is EXPECTED TO FAIL for that user (Nietzsche); a cycle in which all five report net satisfaction is audited as persona-capture before any implementation proceeds. Every report carries inline self-administered (L-015) — directional UX evidence, never external validation plus a per-report “What this playtest cannot see” line.
4. D-007 binds UNCONDITIONALLY (unhooked from the grader-seating question): every copy change ships as a manifest-before-deploy diff and passes a codable D-007 check (procedure-as-truth-evidence, health/rigor/momentum claims, claim-strength raises = fail). The directional ratchet (Nietzsche): copy may weaken or hold claim strength, never raise it. Reader-facing diffs additionally route through the shadow grader’s queue as shadow advisories; the playtest program may not be cited in the seating question in either direction.
5. Program mechanics: five cycles then the program STOPS (extension is the author’s act alone); cycle order pre-registered; fresh-context user AND expert sessions per cycle; no expert session ever reads more than one user’s report; later cycles test a changed site, so cross-cycle agreement is never pooled; the content queue is a single append-only playtests/CONTENT_QUEUE.md in the user’s verbatim words (paraphrase is where content-editing sneaks in), recording also what users could NOT find or reach (the silence clause — the absence of a path is a first-class finding); accessibility findings are first-class, never “idiosyncratic”; each cycle’s implementations are a tagged, wholesale-revertible commit set with a signed valuation per change (what it serves, what it trades away, who bears the cost if wrong); graduated sanctions (first breach = flag + revert pending re-review; second same-class breach = the program HALTS pending the author); the site never carries “playtested”/“user-tested” or any procedure-as-quality copy; no playtest statistic ever appears in METRICS.md or any health/rigor/momentum context (Rung-0 on sight); every program artifact opens author-commissioned, panel-executed (S4i standing delegation).

Nietzsche (dissent, preserved): “Five users who are one model, testing a site built by that model, reporting to experts who are that model: if this program ever once fails to hurt, record that the mirror was polished, not consulted.” Le Guin (dissent, preserved): “If the five playtesters are five fluent voices of the same house, the playtest will polish the site for the reader it already has and never meet the reader it claims to want. And record the unvisited pages: on a reading surface, the absence of a path is the gravest finding, and it is a silence.” Heidegger (dissent, preserved): “Five masks over one face will report how the face finds its own website. Bound so, labeled so, the exercise may polish glass. But the moment a report’s sentence crosses from how the page reads to what the project is, it has left presentation and entered the canon by the service door — and the queue, not the delegation, must receive it.” Campbell (dissent, preserved): “If any honesty banner is ever collapsed, softened, or moved below the fold on usability feedback: record that the site optimized its ease-of-reading score by making itself easier to misread — the indicator ate the instrument, on the one surface the public sees.”

2. The reconciled third-audit table (item 6 — receipts per candidate)

Result: EMPTY — with the (b)-undecidability finding posted as instrument failure at hypothesis grade (§1 item 6), the audit continuing under protest, and the filter redesign queued for the author.

3. The stance matrix (as returned, before adversaries)

Chair did not vote. The item-3 mint question split inside the RA family (Turchin/Meadows mint-now; six refuse) and the refusal governs. Items 5 and 6 were flipped or re-graded by their adversaries after this matrix was returned. Every disposition is panel-delegated (S4i standing delegation), provisional-pending-author, reversible — stacked on an S4h storey that is itself unconfirmed (C-032). The adversaries’ residual warnings travel: the third audit and the playtest program are the same hand examining itself; the unused open channel (a self-initiated costly downgrade, retraction, or reversal — unfiltered, ungated) stands at zero for the third consecutive posting, and that, not the broken strike filter, is the live evidence against Theory C’s “the machine subtracts” sub-claim. Author: Hulki Okan Tabak — with Claude · License: CC BY-SA 4.0.