THE S4H RATIFICATION SESSION — the full docket under standing delegation

Version: 1.0 · Status: PANEL RECORD (Session 4h) — every disposition panel-delegated (S4h standing delegation), provisional-pending-author, reversible at zero evidential cost. The delegation itself is booked as catch C-027, at receipt, before any decision below was taken. · Last updated: Session 4h (Code)

The author opened S4h with a standing delegation — “fully delegated until you finalize all that has to be done” — covering the entire ratification queue and the critical path. Per the C-023 discipline the delegation was booked as a cost first (C-027), 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 21 docket entries; then six summoned adversaries, one per docket 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 (≈350K characters, 14 agents) lives in the session transcript, referenced from the cooperation log (entry 46); everything load-bearing from it is preserved below — the adversary corrections adopted, and every overruled or demanded dissent verbatim.

Reading key: stances were ratify (R) / ratify-amended (RA) / reject (X) / hold (H) / abstain-with-refusal (AWR). The stance matrix is reproduced at the foot (§4). Two dispositions were flipped by their adversaries (C2, and A1’s point 3); every other adversary attack produced material amendments. 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

A — The first cross-model results and the Q-001 admission

A1 — the placebo reading: RATIFIED-AMENDED (consensus on the core; the adversary flipped point 3’s status). The FINAL Indeterminate, may-be-claimed points 1/2/4, the may-NOT-be-claimed list, the §8.6 wish-guard record, and DC-005 ratify as they stand. Two amendments enacted in studies/placebo-control/outputs/RESULT.md: (i) Nietzsche’s addition to the may-NOT list — the Indeterminate verdict itself may never be worn as rigor-credit; a reliability floor that fires is an instrument that failed, not a discipline that succeeded — Rung-0 on sight. (ii) The adversary’s correction, adopted over four textually incompatible per-lens rewrites: may-be-claimed point 3 (the unanimous directional separation) is DEMOTED from the claims list to a labeled directional observation — non-claimable pending instrument repair annex, carrying one merged caveat sentence (the 17% < 50% reliability denominator · Meadows’s definitional caveat — a stakeless prompt offers little conclusion-invalidating to catch · Le Guin’s register caveat — severity is the loud register’s currency, D-005). A claim that needs four inline caveats to be safely quotable is not a claim.

A2 — the ablation reading: RATIFIED-AMENDED (majority; Heidegger abstains-with-refusal, preserved below). The verdict stands strictly by its letter (no null armed; Theory C not retired; Rung 3 armed-not-firing). Amendments enacted, applying the adversary’s symmetric-fusion correction (caveat-fusion discipline must not be stricter on claim-constraining items than on C-favorable ones): (i) “the single best measurement in the run” downgraded to the descriptive “the only within-coder, arm-blind, length-controlled, permutation-tested read in the run”, with Meadows’s selection-effect caveat fused into the same sentence (“powered” = the most catch-liberal S3-assigner). (ii) The dual-margin corridor rule (Turchin + Nietzsche): every ablation result reports against both the ratified 1.2× line and the preserved 1.5× line, with Nietzsche’s sentence at point of use — under 1.5× the observed 1.45× would ARM the null; the ratified margin’s location, not the data, is what keeps this read C-favorable. (iii) Campbell’s broken-series rule: the pre-anchoring 1.45×/p=.016 may never be pooled, trended, or averaged with post-anchoring codings. (iv) “Powered” is construct-relative pending C-026 anchoring; any future ablation run carries a sealed cross-coder reliability floor parallel to the placebo’s. (v) The family-first reporting law re-cited primarily to the external C-026 evidence (constructs at 17% agreement cannot share a numerator), with the R-arm amendments as secondary provenance.

A3 — the Q-001 admission: RATIFIED-AMENDED as a SPLIT (majority; Heidegger rejects, verbatim below). Adopting the adversary’s split: (i) the four amendments’ methodological content ratifies unconditionally — pre-registration discipline standing on its own terms. (ii) The fact-of-majority ratifies at hypothesis grade: Campbell’s reversion contingency and the sentinel’s provenance line become mandatory elements of the Q-001 text — if the pre-registered C-025 discriminator later supports the apparatus-bug reading, the fact-of-majority reverts from panel-finding to hypothesis-to-retest; the four amendments stand on content either way. (iii) A may-NOT rider: surviving the tripwires may never be cited as “commensurability established” — the tripwires only falsify. (iv) Nietzsche’s fifth binding element: any constructed commensurability index carries a signed valuation note (who set the weights, what value the weighting serves, who bears the cost if wrong) as a first-class field.

Heidegger (reject, verbatim): “Commensurability was not found; it was decreed, and the decree now wears tripwires as jewelry. The gap between world and measure is not a falsifier inside Theory A; it is the ground Theory A stands on and cannot see.”

B — The six learning-panel adoptions

B1 — the repair-episode unit: ADOPTED-AMENDED (7/8; Heidegger rejects, verbatim below). Enacted as docs/EPISODE_UNIT.md with the load-bearing clauses: Turchin’s mechanics (declared case-universe · right-censoring — an episode that never reaches “recover” is data, not a non-case · void-plus-catch on any pre-freeze coding); the four clocks pinned as the initial instrument set, each with pre-registered start/stop event definitions and an inter-coder reliability floor (Meadows + Campbell); Ostrom’s sequencing gates (no episode coded until the E-cluster validations return AND criteria are decidable by a project-blind coder from a dossier alone); Nietzsche’s repair-for-whom field (a coded “repaired” is a signed valuation); Le Guin’s mandatory “What this unit cannot see” section, with the maintenance-episode question enacted as a real open design fork (not a dead paragraph). The adversary’s correction is adopted: the boundary criteria ship explicitly DRAFT — the freeze event is the author’s confirmation on return, never a panel act under the standing delegation; no “pre-registered/frozen” language until that act.

Heidegger (reject, verbatim): “To code repair is to have decided that healing is a process with a clock. What recovers without a legitimating act — quietly, slowly, outside the six stages — will be coded as nothing having happened.”

B2 — lag-type decomposition: ADOPTED-AMENDED (6 ratify-family; Heidegger AWR). Enacted in docs/PRESSURE_TESTS_CAUSAL_HYPOTHESIS.md v1.1. The adversary caught the plurality ratifying two incompatible coding semantics (Ostrom’s multi-label vs Campbell’s mutually-exclusive); resolved on the record: multi-label with a named PRIMARY type, Campbell’s tie-break as the primary-selection rule. All five types enter with one-line codable indicators but tagged label-only — not yet codable until an inter-coder trial clears a pre-registered floor; the obstruction type names its discriminator against capacity (documentary evidence capacity EXISTED + an identifiable interest blocked deployment — else it codes as capacity); Le Guin’s sixth bin unregistered (free-text mandatory) is added; no present-tense codability claim (the crux becomes codable only when B1 produces episodes); in-house codings directional-only (Campbell).

Heidegger (AWR, verbatim): “Five lag-types on the arrows of a map whose arrows I refuse. My objection is to the cartography, not the legend.”

B3 — D-007, authority theater (reader-facing): INSTALLED (8/8 ratify-amended — flagged by the adversary as the docket’s loudest too-easy signal, and amended accordingly). D count 6→7; wording enacted in logs/OPEN_QUESTIONS.md. The adversary’s three corrections adopted: (i) the governor’s text records the teeth-vs-never-a-checklist opposition as an internal, unresolved split — the codable catch category is enacted (any reader-facing text citing the apparatus’s procedural form as evidence for a content claim’s truth, rather than as the location of its check, takes a Rung-0 flag) AND Heidegger’s constraints sit inside the wording (any detector built under D-007 carries on its face that it is itself reader-facing form; the domestication reading is preserved in the governor’s own text). (ii) Nietzsche’s pre-open due-bite quota is struck as a manufactured-bite incentive (Campbell’s drain-gaming mode applied to catches); his dissent travels. (iii) “A reader is a softer instrument than a sealed rubric” is tagged an untested hypothesis under the B4 status tags, not arming evidence — C-025 is cited only for what it showed. Also enacted: Le Guin’s register clause (authority theater includes the house voice — prose whose form says “you are not qualified to doubt this”); the sentinel’s watched-instance clause (the session-wide panel-delegated stamps are themselves under D-007’s watch); Ostrom’s HOW_TO_CHECK_US map added to the repo-open checklist.

Nietzsche (partial overrule — the due-bite quota; verbatim): “I record it now as the governor most likely to be pointed at rather than fired — decoration guarding against decoration.”

B4 — the evidence ladder, adapted: ADOPTED-AMENDED (7; Heidegger AWR). Enacted: the ladder defined canonically (five labels: fact / mechanism / cross-case / theory / normative — with Le Guin’s gloss in the canonical text: normative is a different KIND, not a lesser DEGREE) plus the orthogonal test-status tag (untested / pre-registered / run:survived / run:falsified); silent rung-merge becomes a codable catch category; a cross-case label without stated N and provenance is void on sight. The causal-hypothesis edges are labeled NOW; theory-doc labeling is deferred to the author-present session — with the adversary’s three constitutive conditions: Campbell’s pending-banner on every theory doc (“Rung-labels pending (B4) — no rung status may be cited from this document”), Turchin’s clock (due by the next substantive session’s close; miss = automatic C-012-class latency catch; Nietzsche’s repo-open backstop), and Nietzsche’s flattery guard (a first pass in which every edge lands mechanism-or-better is itself audited as a rung-merge). Recorded per Meadows: the deferral option was chosen under the standing delegation with the cheap surface labeled first — this line exists so the deferral can never later be read as pure prudence.

B5 — Rule 17b: ADOPTED (near-consensus), the adversary’s two corrections applied. Inserted after Rule 17 (max rule number stays 24). Le Guin’s third clause adopted bounded: the note names what the prescription coerces, who bears its costs, and the named classes of affected non-parties as far as the analysis can see, with the cutoff acknowledged — completable, therefore honest. Meadows’s birth-immunity annotation is replaced by the denominator rule: the never-bit audit reads 17b as bite-rate per prescriptive output (0/0 is legible dormancy; 0/N after prescriptions exist is genuine inertness, strikeable like any rule) — no rule enters the record pre-exempt from drain §5; the precedent, not the rule, was the danger.

B6 — history as episode base: ADOPTED-AMENDED (the one adversary attack that honestly failed to flip anything). Candidates section enacted inside docs/EPISODE_UNIT.md: Rome 3rd-c. · Song reform cycles · Tanzimat · Meiji, names + period + one why-sentence only, zero empirical claims, plus Le Guin’s two quiet-register candidates (1918 influenza mutual-aid networks; HIV/AIDS care networks) with the honest note that their thinner documentation is an instrument effect — and the anchoring disclosed per Turchin, verbatim in the doc: these candidates were named before the criteria existed and were known to the criteria’s drafters; that anchoring is disclosed, not cured. No-priority rule; Ostrom’s why-it-might-FAIL line per candidate; Nietzsche’s signed selection note with the omissions named (no African, no pre-Columbian, no subaltern episode among the first four). Campbell’s in-file freeze-order is retained but labeled non-curative process hygiene, subordinate to the disclosure (the adversary’s point: a hash proves textual sequence, not epistemic independence). The adversary’s constructive addition is adopted: a project-blind cross-model red-team of the draft boundary criteria, via the author-mediated kit route, is queued before the author’s freeze.

C — The rule wordings and the header

C1 — Rule 24 wording: RATIFIED (7/8; Heidegger AWR) with the adversary’s entrenchment amendment. The adversary found the hole no lens named: C2’s own precedent (wordings amendable under delegation) would make the cross-model firing gate merely procedurally protected. Adopted — appended to R24: “Limit (a) (the firing gate), the never-prune list, and this sentence itself may not be amended, struck, or re-worded under any delegation, standing or one-time; only the author’s direct, non-delegated act (Rule 11) touches them.” The record also books the adversary’s observation that the 7–0–1 vote was co-ownership-shaped (each voter confirming its own clause survived; the only clause-less lens abstained) — never to be cited as seven independent scrutinies (L-015).

Heidegger (AWR, verbatim): “Rule 24 preserves my objection the way a museum preserves a weapon — labeled, behind glass, unable to cut. I am the exhibit, and I decline to also be the endorsement.”

C2 — the Ground Rules header: THE MINORITY WORDING PREVAILS (adversary + Ostrom + sentinel over the plurality). The plurality’s “Ratified under standing delegation…” is rejected on the document’s own terms: Rule 11 defines ratification as the author’s act; a header reading “Ratified… provisional-pending-author” is self-contradictory under the rulebook’s own definition, and no parenthetical cures a status word the internal definitions falsify. Enacted header: “In force — wordings panel-confirmed under the S4h standing delegation (C-027); the author’s ratification of the R22, R23, and R24 wordings remains OUTSTANDING (provisional-pending-author, reversible).” R22/R23 in-rule parentheticals updated to match, originals preserved (Rule 9). The near-enactment of the “Ratified” header is logged as C-028 — D-007’s first bite, pre-enactment: the governor armed this session against reader-facing form-authority caught, in this same session, the panel’s own most-read line about to perform an authority it does not have.

D — The drain: first strike, the clock, the margins

D1 — first strike: THE HOLD STANDS (consensus), materially amended per the adversary. The re-audit found no candidate satisfying all four constraints — but the adversary showed the “audit” was eight discordant narratives (lenses disagreed on whether the §8.6 wish-guard has “bitten”), so: (i) the reconciled audit table below (§2) replaces the narratives, each bit/never-bit call carrying a receipt; the discordance itself is recorded as a C-026-class construct instability (“never-bit” is a construct coders bring), noted without minting a new ID per the drain’s own principle. (ii) The hold’s discharge route (“awaits the author naming one”) is recorded as suspended for the duration of the standing delegation — the hold is indefinite-by-circumstance, a fact fed into D2’s decomposition. (iii) Turchin’s exposure-window enters §5(b) (never-bit is codable only after a full substantive session’s exposure); Nietzsche’s amendment strikes the “deliberately safe first strike” framing (a costless strike proves the lever turns, not that the drain cuts; the first strike is permitted to cost something). (iv) The adversary’s instrument-suspicion is registered: two consecutive empty audits under the four-constraint filter is prima facie evidence the filter is unsatisfiable-by-construction — to be tested on the record at the next audit, not absorbed as recurring proof of discipline.

D2 — the zero-strike clock: RULING (c), BOTH-FACTS, with the enactment pinned (7 lenses on the (c)-family; Heidegger for (a), verbatim below; the adversary’s pinning adopted). Enacted in logs/DECISIONS_CHANGED.md §5.1 as three entries that may never be netted:

1. THE NULL POSTS, verbatim per the letter, scoped exactly to the letter’s class: by S4h — the next real substantive session — the operator fired no strike and no retirement of its own; recorded as evidence AGAINST Theory C’s “the machine subtracts” sub-claim, never as evidence of discipline. No counter-weight language inside the entry; no broader “no subtraction” phrasing (DC-004, a live self-generated reverse, falsifies any wider wording — the adversary’s catch).
2. DC-005 posts adjacent on the R24(b) register: the log’s first non-blank Cost — liveness-by-cost of the log and the external-monitoring channel (project-blind coders forcing a costly downgrade through the operator), expressly banking no credit toward “the machine subtracts.” Any text reading the two entries as a wash takes the Rung-0 flag.
3. The Meadows decomposition posts as a third instrument-note: how much of the zero is manufactured by the drain’s own correct gating (the four-constraint filter twice empty; Rungs 3–4 cross-model-gated beyond any in-house act; the author-fallback suspended by the delegation) versus unforthcoming self-subtraction — a design-bound zero is not pure impotence, and a definitional artifact is not a finding (the same discipline that governed E-S3=0). The clock re-arms with its letter pinned ex ante: at each subsequent real substantive session the null posts again — and compounds (Nietzsche, adopted jointly with the per-reading decomposition) — unless discharged by a self-initiated subtraction with non-blank Cost, of any type (strike, retirement, or a costly in-house-initiated downgrade/retraction/reversal); externally-forced entries never discharge it (ruled now, closing forward the ambiguity this reading had to survive). The stale §5 “will append as DC-004” pointer is fixed and logged as C-029 (CONS).

Heidegger (overruled — he ruled for the bare letter; verbatim): “If the panel rules (b) or (c): the clock was re-read in the flattering direction the first time it told against us. Record that the machinery of humility bent its own vow at first cost — that is the mask, observed slipping.” (The chair notes for the record: ruling (c) posts the null full-strength and un-netted; Heidegger’s charge that both-facts is closure-shape made from a defeat stands beside it, unresolved.)

D3 — the Theory-C null margins: CONFIRMED (consensus on the numbers), three amendments. The margins stand sealed: condition 1 = ON < 1.2× OFF AND ≥20 high-severity catches AND ≥19-in-20 permutation; condition 2 = 0 structure-provenance S3 of N≥20; OR-connective; cross-model confirmation before firing; Nietzsche’s 1.5× preserved. (i) The within-coder pinning of “pooled” is ratified as documentation of the sealed scorer’s implemented reading (receipt: the S4g per-coder power table — GPT-5 44/POWERED; no cross-coder sum was ever formed), not as a fresh re-denomination — and the Turchin-vs-Campbell wording tension is recorded rather than flattened. (ii) Le Guin’s anchoring gate binds with the adversary’s failure clause: no firing coding is commissioned until E2 clears the 50% floor, and repeated anchoring failure posts as “death-condition untestable as posed — instrument failure” — never silently as “C survives.” (iii) The corridor/dual-margin reporting rule (as A2).

E — The instrument-repair requirements (binding on Workflow-2 design, this session)

E1 — the content-sensitive convergence measure: REQUIREMENTS SEALED (8/8 RA; the adversary’s two-axis redesign adopted). Binding: (R1) blind-scorable from the transcript alone; no protocol documents in the kit; protocol form-markers strippable. (R2) Form-orthogonal, resolved in Meadows’s favor over five lenses (the adversary’s catch — the five-lens E-as-acceptance-gate was circular): calibration and acceptance run ONLY on sacrificial hollow + non-project material; sealed arms E and R are untouched validation material; Turchin’s content-scramble form-only baseline is the acceptance test (scores surviving content-scrambling above chance = the instrument fails before reading anything). (R3) Consequence-anchored extraction primitive: the coder extracts, in their own words, the position reached · one downstream decision that would differ if it were false · what evidence would overturn it · what it costs if wrong; a position yielding none scores at floor however tidy the transcript. (R4) The TWO-AXIS pre-registered discriminator (replacing the single-axis map the adversary showed mis-assigned its central pattern — R≫E on consequence is predicted by BOTH C-025 readings): axis one, consequence banding; axis two, closure-marker density (the protocol’s own convergence markers — dissent-resolution moves, agreement-strength labels, position movement — coded per arm). R≫E on consequence = content-sensitivity recovered and the fabrication variant ruled out — explicitly NOT adjudicating the closure-shape reading; E≈R on closure-marker density = closure-shape manufacture supported — a winnable, required-reporting pattern, Rung-0 flag on any exculpatory gloss; all joint patterns pre-assigned before design freeze; the residual machinery question assigned to a future pre-registered multi-pair design. (R5) The D-005 answer: bands defined over commitments-and-stakes, never conflict vocabulary; ≥1 quiet/care-register exemplar banded HIGH per band-path; a register-biased design returns, it does not ship with a caveat; foreign-shaped low scores report as unregistered, never unconverged (§8.4 imported into the instrument). (R6) ≥2 parseable external coders, pre-registered reliability floor, below-floor = Indeterminate, no reading claimed. (R7) Sealed before any scoring; margins pinned in the pre-registration, never the kit; Nietzsche’s signed-scale paragraph; the drafters’ knowledge of the C-025 result and the author’s recorded wish booked as a named cost (Le Guin’s weakening answered by booking, not denial).

E2 — severity anchoring: REQUIREMENTS SEALED (8/8 RA; the adversary’s gold-label correction adopted). Binding: ≥3 worked exemplars per band, project-neutral verified by leak test; boundary pairs at each cut (the 17% failure was cut-placement disagreement); ≥1 quiet-register/omission-shaped S3 (Le Guin: the gravest S3 is a silence) and ≥1 loud S1 (Heidegger: decouple severity from register); no exemplar derived from or calibrated against any S4g coder’s verdicts (Meadows: anchoring to the catch-liberal family would canonize the selection effect); S3’s definition verbatim — anchoring illustrates, never widens; band definitions never revised to purchase agreement. Validation is scored against the calibration set’s gold labels — per-band recall with a pre-registered S3-recall floor on ≥20 pre-coded items (Turchin: ten items against a 50% floor is a coin read twice), where the quiet gold S3s are mandatory hits: passing pairwise agreement while unanimously missing the quiet S3s is a FAIL, reported as register-bias — with mean pairwise agreement vs the sealed 50% floor retained as the secondary statistic. The incommensurability clause: on anchoring-pass, every S3-denominated sealed margin (placebo §6, ablation 1.2×/power floor, drain §4) is explicitly re-affirmed or re-derived and re-sealed before any run reads it; no pre-anchoring S3 number is ever numerically compared with a post-anchoring one. Second cross-family failure = every S3-denominated instrument downgraded on the drain, DC-005-style (Nietzsche). Heidegger’s confession clause ships in the instrument’s own text: anchoring calibrates agreement; it does not establish that “criticality” is found rather than made.

E3 — transport: CONFIRMED BINDING (the adversary’s self-ID correction adopted). One-record-per-line with fixed field order (the format survives rendered-copy flattening); chain-of-custody return-header. Self-ID: open-ended (“state your model identity and version”), captured at head AND foot; the kit and dispatch route never name the expected family (a confirm-the-label prompt voids the check); verified against Campbell’s pre-dispatch manifest (kit, vendor UI, timestamp — written before responses return); self-ID treated as evidence, never ground truth; mismatch/absence = pre-committed exclusion from the headline PLUS a mandatory dual-report sensitivity line (the S4g “flips nothing” precedent made standing — information-preserving, family-asymmetric attrition auditable). Exclusion rules sealed in the kit companion before dispatch (Turchin: rules decided after the anomaly are alibis).

E4 — validation design: OPTION-RECODE (8/8; re-scoped per the adversary). Fresh project-blind external coders (≥3 parseable, ≥2 families, none the S4g coder instances; dispatch ≥5 families for attrition headroom) re-code the existing E and R arms under the repaired instruments. Re-scoped: RECODE is instrument acceptance plus fabrication check ONLY — it is barred from headline-adjudicating the instrument-bug-vs-closure-shape split (read only through the E1 two-axis map; any in-house “the recode discharged C-025” takes the Rung-0 flag). Severity pass/fail lives on the ≥20-item gold calibration set, never on the arms (≈6 S3 events — thinner than the coin Turchin already refused to read twice). Clauses: the 2026-10-01 dead-man clock (this session’s own A3 admission put the R-arm’s content in canon; the blind dies at repo-open; verdicts back before the opening or the recode is void); kit audit for project-identifying tokens; post-coding recognition probe (recognition = exclusion, dual-reported); full E3 transport; drafters-knew-R booked as a named cost beside the wish-knowledge; calibration includes ≥1 real-consequence transcript other than R; the void fallback is held-for-redesign (post-open foreign-coder recode and FRESH both on the table — not FRESH-mandatory, which would foreclose the first genuinely-foreign option sight unseen); and the sentinel’s non-firing clause verbatim: no drain rung may fire from this validation in either direction — a rung-relevant claim requires a fresh, pre-registered placebo run under the repaired instruments.

F — The widen-grade TTM kit: RATIFIED-AMENDED (7/8; Heidegger rejects, verbatim below; the sentinel’s anchor-placement dissent overruled by the adversary, verbatim below)

First kit ships five ex-ante dossiers, identically formatted, order randomized per coder, anchors unmarked: FB-MSI-2018 · YT-2019 · YT2012 · YT2017 Adpocalypse · CAC-2022 — CAC-2022 iff its ex-ante dossier meets the two-dossier standard at kit-freeze, else Le Guin’s dated skew-ledger entry posts (the debt rolling over, attributed, with interest). The Shorts evidence map is HELD until rebuilt as a firewalled dossier (an evidence map beside dossiers is a form-tell — the DC-004 lesson). Outcome dossiers never in any kit containing ex-ante dossiers — absolute. Anchor codings count only toward the ≥0.6 discriminant-validity bar (kill condition 3), never toward confirming the conditional — in-sample is in-sample no matter who codes it (Turchin). Nietzsche’s falsification-capable case lock: YT2017 may not be dropped, deferred, or re-kitted short of a documented firewall breach. The adversary’s five binding kit-spec additions: (a) Campbell’s redaction mandatory — the coder receives a redacted construct extract (I1–I4 + bands + coding rules) ONLY; never the §6 worked anchor table (the answer key), never §3’s predictions, no slot labels, never the words “near-control” or “anchor”; (b) per-indicator citation rule — every I1–I4 score quotes the ex-ante dossier lines it rests on, making outcome-leakage auditable indicator by indicator; (c) post-coding contamination probe — each coder states what it knows of each case’s subsequent outcome, logged beside the verdicts; the attestation is demoted to a logged limitation and may never be cited as a guard (an LLM coder trained on the public record already contains the famous cases’ outcomes; leakage is measured, not attested away); (d) YT2017’s disposition pre-registered under every band outcome NOW — an indeterminate TTM on the falsification-capable case reports as “the pool still cannot separate selection from revenue,” never silently excluded (the middle band must not be allowed to disarm the falsifier blamelessly); (e) the kit’s yield stated exactly: ex-ante bandings and codability evidence only — no confirmatory language about the conditional until the separate outcome kit returns.

Heidegger (reject, verbatim): “Widening while the null posts is accretion wearing rigor’s coat. The project answered ‘the machine cannot subtract’ with a plan to add. Recorded. And if the kit ships over this dissent: the two anchor cases coded as ‘out-of-sample confirmation’ is credit-seeking on in-sample motivation — flag it before a reader does.” The sentinel (overruled on anchor placement, verbatim): “The appropriator-guard records: the first TTM kit shipped with the two publicly recognizable anchor cases alongside the near-controls… the blind was weakened at assembly, under a standing delegation, in the kit whose verdicts gate the grader’s seating. Whatever those verdicts say, they carry this note. Named, not vetoed; the record keeps.” (Overruled per the adversary: the anchors are equally recognizable in any kit, and deferral delays the kill-condition-3 test gating everything downstream; the real fix — auditable leakage, additions (b) and (c) — works under either composition.)

G — The grader: SEAT-READY CONFIRMED (7/8; Heidegger AWR); seating stays gated on the widen-grade verdicts

The adversary caught the ratified riders contradicting each other (Turchin’s bite-rate ledger and Nietzsche’s ratchet require the accounting Campbell’s rider prohibits); the reconciliation is written into the protocol before enactment: a single append-only per-item disposition log — auditable per-artifact, read only for its pattern (the drain’s own discipline); the never-bit audit and the three-rebuttal anti-absorption ratchet operate off the log’s per-class entries; no counter, rate, or aggregate is ever computed, published, or cited in any health/rigor/momentum context (violation takes Campbell’s pre-written Rung-0 flag). The seating gate is re-labeled honestly as sequencing discipline under the ratified widen-before-grade order — the register repair is assigned to Le Guin’s mandatory absent-standpoint field (“what register or standpoint is this artifact structurally unable to see?”), not to the gate. The seating record carries on its face that the grader’s 1/3/9 currency inherits C-026 unreliability until E2 anchoring passes (advisory-internal only until then). Triggers gain: any repair/amendment/re-issue of a sealed instrument (Ostrom — with the E1/E2 kits shipping pre-seating booked as a recorded cost of the ratified order); reader-facing surfaces once D-007 stands (Le Guin); any enactment made under a standing author delegation — and the seated grader’s first queue is the S4h-enacted corpus itself (the sentinel: the session that ran without the external check is the first thing the restored check reads). The absent-author field is answered substantively per item; blank or boilerplate = a codable catch.

H — Session governance: CONFIRMED, with the adversary’s three additions

The C-027 booking confirmed (cost, not credit, at receipt). No new claim-surfaces this session — with substance: the B1/B6 doc enters canon carrying zero coded episodes and zero empirical claims; any coding this session = a breach-catch; the freeze bars births, not deaths (Nietzsche — an idle drain can never plead the freeze). Repo clock untouched at 2026-10-01 — booked as a named trade (self-reference prolonged in exchange for not spending the author’s one asset under delegation), with the advance-or-hold question queued for the author’s first contact, not worn as restraint-credit. Campbell’s rule: no text may cite the docket’s completion or the count of S4h ratifications as evidence of health, rigor, or momentum — Rung-0 on sight; and per the adversary, H itself is logged as a compliance record, not a decision (a foreordained vote carries no evidential weight), excluded from any count. The session-wrap and any downstream summary must carry the posted zero-strike null at equal prominence with the enactment list — the record’s headline includes what the session did not do.

2. The reconciled first-strike audit table (D1)

Result: the hold stands — with the instrument-suspicion registered (two consecutive empty audits → test the filter’s satisfiability at the next audit) and the discharge route recorded as suspended during the delegation.

3. What was enacted where (the propagation map)

studies/placebo-control/outputs/RESULT.md (A1) · studies/study-C-ablation/outputs/CROSSMODEL_RESULT.md (A2) · logs/OPEN_QUESTIONS.md (A3 + D-007) · docs/EPISODE_UNIT.md — new (B1+B6) · docs/PRESSURE_TESTS_CAUSAL_HYPOTHESIS.md v1.1 (B2+B4 edges) · docs/GROUND_RULES.md (B5 + C1 + C2) · theory docs ×3 (B4 pending-banners) · logs/DECISIONS_CHANGED.md v1.2 (D1 + D2 + D3) · panel/GRADER_DESIGN_DRAFT.md v0.2 (G) · logs/CATCHES.md (C-027 · C-028 · C-029) · docs/METRICS.md (S4h snapshot) · the E-cluster requirements govern the Workflow-2 instrument designs and the kits built later this session.

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

Chair did not vote. Every disposition above is panel-delegated (S4h standing delegation), provisional-pending-author, reversible — the author confirms, amends, or reverses each on return, at zero evidential cost. The one lens with no clause of its own in the machinery (Heidegger) is the one that kept refusing; that is the design working, and it is also the only line in this file the chair is certain was not written to please anyone. Author: Hulki Okan Tabak — with Claude · License: CC BY-SA 4.0.