REGISTER_V14_FOLD.md

pipelines/REGISTER_V14_FOLD.md

v1.4 — the P1-C fold (2026-07-28, refutation-gated)

Provenance: the P1-C contract-critical wave (four lenses — presentation-craft-anim-audio ·

hades-death-run-realm-identity · animal-crossing-cadence-collection ·

knowledge-progression-decipherment) ran as workflow wf_86fbfd65-f35 and returned 8 of 8 units clean:

four audits, each attacked by its own fresh-context adversarial refutation critic that opened every

citation, re-ran every zero with its own positive control, re-derived every MISSING against the

designed-not-missing test, and ran the devil's-advocate pass calibrated against over-conservatism.

Nothing below entered unrefuted. Every entry lands at its critic-AMENDED verdict, class and

priority, with the six binding director adjudications of Addendum 2 applied on top of both draft and

critic. Raw evidence trail: LENS_<id>_DRAFT.md + CRITIC_<id>.md for the four ids under the wave

scratchpad, and the twenty rulings in scratchpad/lens_wave/DIRECTOR_ADJUDICATIONS_28.md. The

archived copy lives at docs/pipeline_review/lens_wave_2026-07-28/, where the P1-B and realm-wave

files already sit; the eight P1-C files join that archive at this landing.

Wave arithmetic: 60 findings in · 58 survive · 2 whole findings killed · 12 further premises,

supporting arguments or recommendation limbs struck · 34 amended · 15 contested-carried, of which

the five presentation items are closed by adjudications 15-20 and four route to a Josh brief. The

kill ledger and the brief queue are at the foot of this section.

New entries 94-132, compact form. Lower-priority survivors are rolled up per lens with their draft

named as the record rather than expanded into rows.

Entries

that consumes it: ASSET_DROPIN_CONTRACT carries presentation zero times across 191 lines, both

variant hits are the substring inside "invariant", and its eleven-row table has no tier, variant

or multiplicity column — while FR-006's own tooth: NONE(...) records the hole verbatim. Fix is one

column and one rule on machinery that exists: presentation_tier_ref on the ability/weapon

extension rail via the Batch-0 registry_extensions.json + fidelity_baseline.json mechanism the

contract itself calls proven across ten registries, and a resolver key of id-plus-tier that falls

back down the ladder to the nearest authored tier. Count corrections: mesh_id_ref 15, not 11;

generation_status 10, not 7. (spec-doc + schema-data; HIGHEST, must precede the DR-2 freeze.)

ceiling class anywhere for the realm ceiling to be reserved with: realm appears zero times in the

1920-line doc that owns per-asset budgets, against tier 162 / budget 126 / Nanite 76 in the

same read, and §5.6 F8's check column ends "vs the class budget" with no realm class to gate

against. Per adjudication 17 the verdict splits honestly — the requirement is a landed charter

amendment and therefore WEAKER (ruled, homeless), the mechanism is genuinely undesigned and

therefore MISSING — because a bare MISSING on a ruled item reads to a later agent as nobody decided

this. The discharge is the seven-clause paste-ready reservation, all seven clauses verified to ride

real machinery. (spec-doc; HIGHEST, must precede the DR-2 freeze.)

PresentationTier in code is mastery, the charter's realm tier is fidelity, QA's

tier_applicability is footage maturity, the twelve mastery levels are progression, and

readability_tier is inscription mastery — already flagged in this register as a false friend.

Ratify two distinct names in the commit that adds them, presentation_tier_ref for mastery and

generation_tier for the fidelity class, never a bare tier. One line in the DR-2 invariant

block; costs nothing now and is unrecoverable after the freeze. (spec-doc; HIGH, pre-freeze.)

prints it from an id and a clamped tier and says so in its own comment — and the ability registry

carries 185 rows and 35 columns with no presentation_tier_ref. Corrected sizing: the live key

space is 31 elements by 8 live tier values = 248, with 372 the clamp ceiling; and the audit tooth

does cover ability_id, which IS an audited RowName — what is unaudited is the DERIVED variant

key. Per adjudication 15 the band-collapse recommendation is BRIEF-QUEUED for Josh and never

applied silently: it narrows a Josh-nodded hard requirement in two directions at once and costs the

21 hybrid elements their presentation identity, reversible upward but not downward once assets

exist. (spec-doc + schema-data; HIGH, brief-queued — see the Josh brief queue.)

montage and the canon row will disagree silently: animation returns 6 hits in 2 files across the

game Source, all of them one stock unarmed bind plus an include, against a 213-hit reachability

control, and the attack header carries no montage, notify or animation field. Narrowed per the

critic — the authoring surface IS declared at T1_Build_Pipeline_Contracts; what is unruled is the

AUTHORITY rule between a montage's notify windows and the canon row's seconds. Rule the canon row

authoritative and retime the montage to it, then add the QT-5 divergence check, which ships today

as a must-not-fire fixture with zero montages bound. (spec-doc + engine; HIGH, before the first

montage imports.)

scope. T1_Ability_Tree anchors the presentation ladder in five places and enumerates its craft

vocabulary outright (animation refinement, effect density, audio confidence, follow-through), so

the ladder is owned and firewalled against power double-dip at T1. What is absent corpus-wide, on

a positive control that plainly matches, is the LAW: anim notify returns zero, root motion and

cancel window return only the draft itself, and hit reaction returns only a tool declaration,

while the control term returns 61 hits across 43 files. Land commit-versus-cancel policy,

root-motion policy, a hit-reaction taxonomy, follow-through and settle, and the per-tier fluidity

vocabulary the DR-2 prompt composer must quote — under the existing T1_Ability_Tree anchor, not

in a re-split home. Camera and cut policy stay routed to entry 29. (spec-doc; HIGH.)

families must carry a state vocabulary far larger than five, on 309 uassets with 14 placeholder SFX,

one combat StateTree and zero Niagara systems. Do not add families — declare the closed

FGameplayCueParameters set the art must read (element, presentation_tier_ref, status class,

counter class, phase) in the same DR-2 clause as entry 94, so one Niagara system per family branches

instead of forty existing. The best economy argument on the lens and one struct declaration now

against a re-author later. (spec-doc + engine; MED-HIGH, rides the DR-2 clause.)

own rewards correctness. The gap is real and the divergence check is clean — a presentation-only

channel grants no stat and shows no rank, so no-scoreboard is not re-litigated, and the refusal of

the a-style-rank-is-tonally-wrong instinct is correctly calibrated. Per adjudication 16 the

recommendation does NOT re-open the ruled pip-meter shape (transient render-band promotion, no

second meter, no stat, no storage) but as drafted it rewards the correct answer repeated, which is

the inverse of the lesson, and it keys on a puzzle-boss instrument invisible in ordinary combat.

It lands re-shaped with a variety signal that decays on repetition and a non-boss home, or not at

all. (spec-doc; MEDIUM, post-slice.)

file is 34 bytes while 820 lines of audio doctrine live in a T99 operational doc plus each region

page's Section 10 — surfaces that cannot arbitrate each other because T1 wins conflicts. Make it a

two-page authority stub that delegates the pipeline and the schema by name and owns only what

nobody owns today: the score's dramatic law, the per-region identity law and the per-realm identity

law. Per adjudication 19 this is DIRECTOR-authored work at the T1 authority seat, carried as a work

order and never fanned out. It closes entry 22's audio third and needs a DOC_MAP row on landing.

(spec-doc; MED-HIGH, director work order.)

declares seven values including place and path, and both are used by zero rows across all 12

registry rows, so 76 regions, 79 nodes and 3 integrity paths have no theme row of any kind. The

dramatic law lands in the entry-102 stub as four rules — per-state entry and exit conditions, the

silence rule where the score deliberately stops, leitmotif economy, and the combat-to-exploration

transition budget. Leitmotif economy is a reveal-discipline surface, not only craft: Music Step 1

already binds a §17 check on a pre-reveal thematic leak. Corrections: the type breakdown is 5

character (not 6), and the two supporting citations are re-anchored. (spec-doc + schema-data;

MED-HIGH.)

whose other side is empty: all three T0_Region_Index audio arrays are populated 0 of 76,

including the slice's own region, while the SFX rows point back at it. The fix is zero-token,

zero-model and needs no authoring judgment — a deterministic populate through the sanctioned

harness/apply_*_columns.py writer pattern that inverts the SFX rows' own region_id, with the

fidelity_baseline.json refresh in the same commit. Adjudication 18 makes two corrected facts

MUST-FIX before the script is written: the slice region carries 18 SFX rows and 10 ambient beds,

not 8, and the inversion must key on layering_role, not sfx_class, or it silently drops both

creature-presence beds and one cultural-traditional row. Land this one first regardless of

ordering. (schema-data; MED-HIGH, GPU-free.)

obvious selector silently drops the crown jewel. All ten realm rows carry empty ambient arrays,

empty elemental affinity and blank environment-profile status; 0 of 30 SFX rows and 0 of 12 theme

rows anchor to a realm; and because the fairy realm is the single story_anchor row, any filter

keyed on node_kind='mythic_realm' returns nine and excludes it — while the safer-looking

page_status='dimensional' returns eleven by pulling in the epilogue. Neither existing column is a

clean realm selector, which is exactly why reservation clause 4 writes generation_tier='realm'

onto the ten rows explicitly. Per adjudication 18 the sourcing rule is restated on the corpus's own

ruled wording — no sampling of consecrated performance, reference-COMPOSITION permitted — because

the drafted blanket reference-first rule would have frozen generated audio for four of ten realms

for no §17 reason. (spec-doc + schema-data; HIGH, with the DR-2 reservation.)

whether the tier-12 variant is identifiable as tier-12, which a ladder flat from rung 2 to rung 11

passes. Narrowed and cheapened by the critic — the per-tier CAPTURES are already required by the

doctrine's own D6 check column; what is missing is an ORDINAL criterion across them. One rubric row

on the existing presentation board: a ladder-monotonicity measure sampled at band boundaries,

requiring a non-decreasing ordinal, shipping with the mandated must-fire fixture (a synthetic

ladder whose rung 3 copies rung 2 must FAIL). Obeys the self-testing-gate law by construction.

(QA loop; MED-HIGH, armed at handshake.)

per-realm distinctness test — a real instrument, but nothing declares Humanity's own cross-title

signature, so the critic has no positive target. Add a SIGNATURE panel to the existing pre-hardware

moodboard board, which the doctrine states explicitly should be shipped rather than deferred: one

board, once, cross-realm, carrying three or four candidate signatures with their cost profile,

since a signature built from motion, timing, silhouette and light is nearly free while one built

from translucency is not. Per adjudication 20 it lands as the HEAD of the realm art bible the realm

wave is producing, never as a parallel artifact, so two waves do not author competing bibles.

(spec-doc + art program; MED-HIGH, before the first Lane-P plate.)

writers: the two worldstate tags exist in exactly 8 hits across 2 generated-schema files against a

44-hit reachability control, the boss registry carries populated composure nodes, and

HandlePlayerDeath writes no worldstate at all — telemetry, a camera fade and a respawn timer.

Write the banked tell-reads through the existing worldstate subsystem at that call site, keyed boss

id to the composure-node ids already on the row. Zero new schema, zero new registry, one call site,

and the load-bearing half of the whole defeat economy. (engine + schema-data; HIGH, slice.)

281 of 281 boss rows with 264 distinct values and the resolver has no attempt index — which is the

anti-Hades defect on what the corpus itself calls the most-repeated surface in the game. Ride the

already-ruled defeat ledger for an attempt_index and let the payout become a small ordered set:

first death teaches the tell, third names the counter, later ones stay short. Only the cardinality

of one field changes. Sequencing is the finding's real value — do it BEFORE the natural-voice pass

rewrites 281 strings. (engine + spec-doc; HIGH, slice.)

corpus forbids a death spiral; the guard it names polices path reachability only, and its own

docstring records that the opportunity and relationship layer is unbuilt and passes as a scaffold.

State the invariant on that existing gate: the per-encounter demand after N failed attempts is

non-increasing, and the graduated delta is authored on the WORLD, never on the encounter you just

failed. This is the framing the Sigma Constraint already uses for randomness. Not a softening —

the ask is that the cost be AUTHORED rather than accumulating, with the top two difficulty tiers

explicitly untouched. CONTESTED-CARRIED on whether this re-litigates the ruled doctrine or states

an unstated invariant inside it. (spec-doc; HIGH, director call before it enters the gate.)

defeat, death, fail, respawn, checkpoint and retry all return zero while vril returns 15, and the

soft catch is an undefined phrase used twice repo-wide. Add one defeat paragraph structured as the

tiers already are — the two lowest name the soft catch as a concrete affordance in the existing

vril-reach voice, the middle is the authored baseline, and the top two are knowledge-only with the

door staying hard. This is the floor that keeps entry 110 honest and it must land WITH it, not

after, or a bounded-demand invariant gets read arc-wide instead of tier-scoped. (spec-doc;

MED-HIGH.)

the doc that would own the real one is a 35-byte stub. Per the critic this joins the rank-20

T1_UI_UX_Spec fill that adjudication 3 already routes the gameplay-camera, onboarding and

HUD-legibility findings into — same doc, same pass, one owner, never a second lane — with entry 22

cited as the parent stub fact. (spec-doc + UI; MED-HIGH, slice.)

at the nearest site marker by squared distance, which can teleport the player past the content they

just failed, while T99_Translation_Combat §2.5 already rules that consciousness withdraws to the

last attuned recharge source. Reframed by the critic from add-a-clause to the-code-contradicts-

canon, which strengthens it: the acceptance criterion is the already-drafted tooth that a nearer

marker must NOT win, plus one clause making the anchor the LAST-TOUCHED site rather than a

recomputed nearest. Sequence behind entry 46 — the anchor registry's own gating column is corrupted

on 30 of 42 rows. (engine + spec-doc; HIGH, slice.)

family is exactly six recorders and none of them observes an outcome. Enrol the ruled defeat ledger

as a telemetry-visible surface beside the existing diagnostics and assert two magnitude-free

properties — no attempt paid nothing, and the world delta applied exactly once per attempt. Both

are the executable form of entries 108 and 110. The boundary is pre-blessed by adjudication 4: this

is the run-ECONOMY axis, deliberately orthogonal to the per-attempt execution-surface latency that

landed once on the sm64 side as entry 53. Named landmine carried: the legal-ids list must be

extended in the same commit. (QA loop; MED-HIGH.)

there is no dungeon table anywhere. Narrowed per the critic to the CONTAINER-and-rows question,

which is the half entries 47 and 48 do not cover — 47 is escalation discipline, 48 is the factory

rider — so it lands as their sibling rather than a fresh headline. Do NOT mint a Dungeon registry:

extend T0_Hollow_Codex, which already carries region, chapter window, contain verb and

containment reward, with the run fields, and let the declared fourth rarity axis do its own work.

Three or four of the twenty land inside the slice. (schema-data; HIGH, extend-existing-registry.)

reward-reroll contract — the classic farm hole, and the integration record uses the phrase farm the

cap twice. Three fields on the same Hollow extension as entry 115 (reentry_class, on_exit_state,

reward_reroll_rule), with the reroll rule bound to machinery already built and already named: the

sigma constraint system, the drop table, the item-instance ledger, and the rarity-firewall gate

already policing the ladder. No new gate, one predicate. Strengthened by the critic: entry 53

registers the identical defect class on a different ruled repeatable surface, so this is a pattern

rather than an isolated gap. (spec-doc + harness; MED-HIGH.)

nothing says who lives there or how a fight there feels. Across the ruleset field family, enemy,

creature, encounter, boss, hazard, fauna, loot, drop, bestiary and opposition all return zero with

reward as the matching control, and no register entry anywhere covers a per-realm bestiary or

reward mix. Narrowed twice by the critic: the hazard limb is already assigned to the environment

grammar registry by the charter's own downstream-edits list, and the proposed reuse home was a

category error — mode_variety_check is a chapter column populated 79 of 79, not a registry that

realm rows can be enrolled in. What survives is native_hostile_families and reward_mix_class as

soft-FKs on the Wave-2 ruleset row, with the divergence tooth belonging to the charter's own

proposed realm-teeth family. Class CONTESTED-CARRIED: a field family absent from the design reads

MISSING while the registry itself stays DESIGNED-UNBUILT per entry 77. (spec-doc + P5-factory;

MED-HIGH.)

about departure — 164 character rows with six loyalty columns, every one of them join, leave or

risk, and not one a yield — while the familiar registry carries bond_level, bond_threshold,

grants_class, command_permission and integrity_band_gate. Familiars have a payoff table and

companions do not, and the ruling itself warns the bands become dead content silently. No new

table: band-gate a small set of picks on bond state in the existing bonus-option registry, which

already carries the unlock gate, condition expression, role tags and build-defining columns. Two

consequences fall out free — the terminal bond becomes a build decision, and the picks enter the

ruled build-space sweeps where a never-picked bond bonus reads as a balance defect. Distinct from

entries 37, 39 and 43, none of which is a-band-yields-nothing. (schema-data + spec-doc; HIGH.)

moves only from integrity and authored beats, and the one authored gift beat runs NPC to player.

Land the give-or-offering verb on the existing festival and persona surfaces rather than a new

relationship system: seven offering-economy rows already carry the occasion, the region and the

community-standing consequence, valued through the ruled per-region mixed exchange, with the payoff

routed through entry 118's band-gated picks. MANDATORY BOUND the draft missed and the critic

supplied: the same ruling paragraph contains a ratified invariant — opposition ceilings at 100, the

trigger caps at 90, and investment never buys immunity — so the verb must be declared incapable of

purchasing departure immunity. Without that clause this quietly reopens a ratified invariant.

BRIEF-QUEUED: whether the ruling's never-from-observed-deeds extends to the positive direction is

Josh's, and if it does the finding converts to DIVERGENT-BY-DESIGN. (spec-doc + P5-factory;

MED-HIGH — see the Josh brief queue.)

closed eleven-value set is gate-enforced and schema-recited, so a festival can only be rowed as a

cinematic to be watched — while 11 of 16 festival rows are opt-in-participate and one of them

names its player verb outright. Add one value to the enum, to the gate's constant and to the scene

registry; nothing else moves. It gives the festival rows and the 295 festival-or-daily-immersion

seeds a legal spine home and makes the participate-versus-observe distinction machine-checkable.

Rides a single commit with its gate change, per the never-write-a-gate-enrolled-file-mid-check

rule. (spec-doc + schema-data; HIGH.)

deterministic 210-day cycle cannot fire inside a one-to-three-hour chapter, and no chapter-duration

or in-game-day-budget doctrine exists anywhere in the plan corpus. Add one column to the existing

festival registry on the shape its own window_kind split already implies — guaranteed-in-visit

for the marquee rows (the calendar supplies the flavour, never the gate), opportunistic for pure

bonus, ambient-daily for the L1.5 texture class. The Pawukon arithmetic still runs; it stops being

load-bearing for whether the player ever sees the content. (spec-doc + schema-data; HIGH.)

seven-observation chain per species, four-tier documentation rewards, and the ruled line that

completion grants knowledge and attention rather than combat advantage — and it has no read

surface, no build item, and about 1% of its data: dex_completion_class and collect_verb are

each populated on 1 of 150 creature rows. The critic re-attacked the surface claim repo-wide and

engine-side and found no bestiary or codex UI in either tree. Three small moves: fold a codex tab

into the existing screens shell, add a per-chapter population row in the FR-038 pattern so data

accrues rather than awaiting a doomed sweep, and repair the doc's own internal count currency.

(P5-factory + schema-data; HIGH.)

12 columns and minerals 19 by 11, and neither carries any discovery, documentation or completion

field, while the language wing accumulates toward a ratified fifteen-languages figure with no

completion tier and no payout. Copy the column that already exists onto both registries and declare

in the owning T3 doc's own language that the four-tier ladder is substrate-general rather than

fauna-specific. Sharpened per the critic: the language registry does carry decipherment_status

and mastery_tier_threshold, but the first is real-world scholarly status and the second a gating

threshold — neither is a player-side documentation state, which is the precise claim. Withholding

the ladder from flora on care grounds would be care-inflation and is rejected. (schema-data +

spec-doc; MED-HIGH.)

unmeasured: the strictly-no-kill path has named prizes in ruled doctrine and no route guarantee

anywhere. Define the aspect as a PACIFIST SWEEP on the existing build-space-sweep doctrine — for

each of the 79 nodes assert at least one authored route requiring zero required human kills,

computed from data that already exists (win_condition union has_nonkill_path union the forks'

rerouted paths) — so a node with no such route is a defect in exactly the way a never-picked pool

option is. Scoped as the arc-level union only; per-row population stays with FR-038 and the BG3

lens, so this is not a duplicate. Explicitly not care-inflation: no content softened, no violence

reduced, the PACIFY integrity cost untouched, the arena untouched, and an unwinnable encounter is

still a combat beat. (QA loop + spec-doc; HIGH, slice-relevant.)

are REJECTED on purpose and pre-registered here so no later lens or rubric re-opens them as gaps:

the real-world clock collides with single-player-first, offline play and walked geography, and is

unverifiable by the AI QA loop that is our only player until the Josh gate; decaying daily

obligations collide with the ruled anti-grind model and with entry 9's rested-XP rejection; and

permanently-missable content collides with no-dead-end, multi-path and the completionist promise.

What IS imported instead, across entries 121 and the roll-up's return-delta finding, is the CHANGE

without the clock — causal, player-driven world movement surfaced on return, and calendar content

staged as guaranteed beats wearing a real calendar's skin. (register pre-registration.)

workspace. Novelty correctly refuted by the critic and the entry is re-priced accordingly: the

build plan's Q3.6 already specifies auto-assembly from decision flags with path-band template

selection, ranked open items 26 and 29 already name the information-architecture gap and the

empty-stub verification, and two slice items already ship the shell — so the assembly and voice

half is registered work, not a discovery. What none of the four artifacts specifies is the

deduction workspace: an entry-state enum (heard-of, witnessed, connected), an evidence back-link,

and a more-here flag. It lands as comparator INPUT to the plan's own open re-derive of whether the

journal ships as a correctly-empty shell or an activating system, arguing the second. (spec-doc;

MED-HIGH, slice-critical.)

so Pillar 5 degenerates to pairwise brute force. The zero survives the most heavily

positive-controlled search in the wave — six patterns over docs, eleven over the source tier, then

eleven across both repos including the engine tree, with a 24-hit control and exactly one hit that

is a false positive about a stale chapter cross-reference. Adopt a confirmation-quorum rule on

existing columns (link_type, link_strength, evidence_basis), adding no new system. Timidity

amendment carried: the ruled non-gating doctrine bars progression GATES, not non-gating COSTS, and

a cost-free unbounded action is exactly what invites the enumeration this finding exists to

prevent — so the option space widens to quorum plus an optional non-gating cost, and which one

ships is Josh's call, not the loop's. (spec-doc; HIGH — see the Josh brief queue.)

test and no gate watches: exactly 16 rows carry a reward class with a blank primary source and a

non-primary evidence tier, and the critic verified the unstated load-bearing half — all 16 carry a

MECHANICAL class (intuition tier, discovery unlock, puzzle unlock, school technique) and none is

lore. The no-gate claim was strengthened by column-name search across the whole harness: the two

columns appear only in an applier and a pack builder, never in a gate. Add one gate file modelled

on the two that already exist in this exact shape — FAIL on the 16, WARN-tier count on the reward

blanks so the coverage trend reaches the scorecard. Distinct from entry 59, which adds a predicate

on the inward-scoped source-anchor axis. (harness; HIGH.)

edges carry no reward, and the draft's plan to apply the taxonomy as a deterministic transform is

STRUCK: the dossier entry it cited as ratified carries the disposition await_josh_confirm in its

own header line, and the same dossier rejects the alternative option precisely because reward and

ability assignments are canon Josh authors. Running it would have applied an unratified

mechanical-reward rule across 122 rows — a cardinal-rule breach dressed as a mechanical transform.

The 122 blanks are therefore not a stalled apply but an awaiting-ruling state, and the correct ask

is to surface the taxonomy key for Josh through the standing decision protocol. The 88

needs-research rows route to the book pipeline unaffected. (schema-data; HIGH — see the Josh brief

queue.)

the only shipped surface that could hold it is a 16-slot ring buffer that is explicitly not a

journal. Give the ring an exempt class: the uncapped fired-clue set, not the ring, is the natural

evidence store. Two corrections, both of which the critic verified in code and neither of which

changes the conclusion — the mystery is structurally asserted rather than authored through (the

unsigned-note token appears in Ch 3, 11, 12 and 13 and in none of Ch 4 through 10, and the

worldstate write exists in Ch 3 only, which is a real second-order authoring gap this lens

surfaced), and the fix is larger than a small change because the player-text map is reset and

chapter-filtered on every load, so chain-bearing rendering needs an unfiltered store. (engine +

schema-data; HIGH, slice.)

with no arming data and it is the QA loop's only knowledge tooth. Reframed per the critic to lead

with what is genuinely new: the rider's own stated rationale is that solvability cannot be proven

statically, and the finding rebuts it — a PRESENCE assertion (at least one active clue row whose

chain matches each puzzle-routed or clue-routed beat) is not a solvability proof but does catch the

actual failure mode, a puzzle authored with no clue at all. The rank-29 sequencing half is already

the contract's own declared consumption point and is not the contribution. Slice-relevant under the

Josh Gate because the QA loop is the only player. (QA loop + schema-data; MED-HIGH.)

and has no state for a law the player has FIGURED OUT: law_state returns zero in the charter, and

the registry does not yet exist on disk, so the field costs nothing today and a migration later.

Add law_state (untaught, taught, exploited) to the metaphysical-laws sub-row group — the first

two are already implied by taught_at and the third is the hinge — plus exploit_tier over the

taboo sub-row's existing layer vocabulary, so the charter's own EXPLOIT column can carry a

prize-tier lock with no risk of a gating law. Highest-confidence finding in its audit; land it

before the Wave-2 schema mint. (schema-data; MEDIUM on weight, HIGH on timing.)

Also surviving, per lens — the draft is the record

LENS_presentation-craft-anim-audio_DRAFT.md: PC-13 COVERED — the slice's systems-teaching audio

vocabulary is authored across eight verified gameplay-cue rows (gather yield, mineral strike, vril

recharge fill and complete, pip fill tick, pip-wipe damp thud, chain-complete swell, pips reveal),

recorded so the lens is not read as audio-is-a-hole; and PC-16, which per adjudication 20 does not

enter as its own row but MERGES into this register's existing colour-independence entry (see the

amendments block below). Also carried from the draft: the six refuted-and-dropped self-kills, which

the critic verified as true refutations — the presentation ladder is NOT unbuilt, audio design is

NOT absent, the QA loop CAN judge motion, music generation IS planned, hit-confirmation feel

EXISTS, and enemy-state legibility is only a vocabulary-sizing gap.

LENS_hades-death-run-realm-identity_DRAFT.md: HD-7 the ruled rescuer can leave with no successor

named, narrowed by the critic because the unavailable state is ALREADY enumerated in the ledger's

outcome enum — so the gap is the ORDERING of the fallback chain, not the existence of the failure

state, and it is adjacent to entry 40 · HD-9 DIVERGENT-BY-DESIGN and favourable, graduated

persistence is a better comeback idiom than the reset · HD-12 repeatable content has no difficulty

authority, CONTESTED-CARRIED on whether a pop-up rift may declare its own density band (the critic

and the draft both read yes, a rift is a site) · HD-13 the nine realm rows own seven

environment-identity columns measured at 0 of 10 while base rows sit at 4 of 50 — the population is

already scheduled by the charter's downstream-edits list, so what is genuinely additive is the

per-column baseline plus the DIVERGENCE TOOTH, that a mythic-realm row must differ from every base

row on at least N of the seven, which is the Hades biome lesson made executable · HD-16 COVERED,

the compounding one-signature-law-plus-reprises budget is a stronger biome discipline than the four

Hades biomes · HD-19 the keepsake exists as a single placeholder equipment row with an unassigned

stat slot, is load-bearing for the defeat axis, and has no mechanical form — route its effect

through the built imprint lane rather than adding effect columns, and name it as the worked example

of an item given to an NPC that changes what that NPC can do for you.

LENS_animal-crossing-cadence-collection_DRAFT.md: AC-6 return-is-paid-in-change has no pre-Vimana

surface, re-cited to the correct world-geography line and STRENGTHENED against live data (the

worldstate registry is 46 rows, not the plan's 42, and none of the four new rows is temporal), with

the fix a causal re-entry delta read off the existing mark ledger rather than a clock · AC-7

faucet-sink discipline is our own stated divergence rationale and does not exist as a document,

with the control corrected from 3 to 8 pre-existing hits in a third sense the draft missed — the

binding no-farm-faucet anti-grind constraint the sink ladder must compose with — and its class

CONTESTED (WEAKER versus blocked-on-the-economy-brief) · AC-8 integrity-visible aesthetics and

player decoration compete for one surface, with the recommendation repositioned as GENERALISING an

existing partial split rather than inventing the first one, and cross-cited to claim-board cell 14b

· AC-9 no warmth axis in the voice doctrine, on a two-hit control, joining the already-scheduled

natural-voice critic wave as a third question plus a scene-type-to-register default map · AC-10

festival coverage is 4 regions of 76 with an explicitly NULL factory gate · AC-11 no low-stakes

acquaintance tier, the thinnest survivor in the wave and honestly priced MEDIUM-LOW, with the

caveat recorded that the persona schema's own home is flagged ORPHAN in the wiring audit · AC-12

the minigame layer is DESIGNED-UNBUILT and further along than the lens expected, no finding minted

· AC-13 no-fail play is COVERED and the source is RULED-and-CLOSED rather than proposal-tier, so

the COVERED holds.

LENS_knowledge-progression-decipherment_DRAFT.md: KPD-4 the clue lane is built and empty (10

authored clues, all in Ch 2, across 79 nodes), DEMOTED from a finding to a priority argument on an

existing rider whose own evidence field already states the headline fact verbatim — its real

contribution is the comparator claim that clue density is the minimum viable substrate of a

knowledge loop rather than polish, which argues that rider should outrank ordinary authoring

backlog; the proposed answers column is dropped as subsumed by the existing chain field · KPD-6

retroactive legibility, RECLASSED from MISSING to DESIGNED-UNBUILT and narrowed to the one sliver

with no home, the player-facing NOTIFICATION that something previously illegible has become legible

— the registry rows, the build item, the screen home and the world-side return vehicle all exist,

and the top-tier promise is post-story endgame where revisiting IS the content · KPD-7 decipherment

is modelled entirely as world-truth and character-tier with no state for the player's own wrong

model, whose never-validated player gloss the critic rates the boldest and best-judged

recommendation in the audit and clears against the anti-decoder-ring rule, since a player's private

guess is not the game asserting a semantic content · KPD-10 the 22-thread ladder tracks what the

story has done and never what the player can conclude, landing first as a rule checkable today

across three existing tables with no schema change, hosted on the reveal-discipline gate as the

positive twin of the leak axis. The audit's seven COVERED rows and four pre-registered divergences

are endorsed in full, and its method contributions (the escaped-underscore trap, csv-not-grep for

clue counts, programmatic gate enumeration) belong in the program's shared method file.

Amendments to existing entries (no new numbers)

generated against no VISUAL law. The recovered elements matrix returns zero for colour, silhouette,

VFX, visual and palette across 105 lines while element returns 17 in the same read; the

effectiveness registry is 124 rows of attacker, defender, effectiveness, axis and source with

nothing presentational. The entry now carries two halves — the negative and accessibility half it

already had (family and landmark readability at range, bound to the ruled colour-independent shape

channel) and the positive and identity half added here: a per-BASE-element presentation row

(dominant hue band, silhouette primitive, motion signature, sound register) with a declared

hybrid-inheritance rule, which becomes the prompt composer's per-element constraint block and the

de-slop critic's per-element checklist, and which also supplies entry 97's banding its inheritance.

Quote-fidelity fix carried: the accessibility floor's real strings are multi-channel vril-polarity

readability as a non-negotiable L1-playability requirement, not the paraphrase the draft quoted.

the UI third by the rank-20 fill that entry 112 joins. The entry stays open on the remaining two.

soft-FK from the vril-site registry into the existing scene-spec registry keyed on visit index,

plus a cascade rule that every vril site in a chapter carries at least one authored

change-on-return beat. It lands at entry 9's own LOWEST priority and does not raise it.

routed in from the killed hades finding: region_id as a soft-FK to the region index, and the

ownership split stated — the region index owns the RENDERABLE environment profile, the realm

ruleset owns the LAW delta. It enters as an INPUT to entry 73's Wave-2 first call, never as a row.

sequences behind its repair.

as the same defect class on a second ruled repeatable content class — a pattern, not two isolated

gaps.

The DR-2 freeze consequence

This wave ran in front of the tiered-asset contract freeze by rule, and the question the register

owes is whether it made it. Three things had to land before that freeze: the realm-tier reservation

(entry 95), the presentation_tier_ref rail (entry 94), and the DECAY law carried forward from v1.3

entry 92.

verified to ride machinery that already exists, and its cost honesty holds — clauses 1 through 5

are schema and desk work with no GPU. Two clauses were strengthened during refutation: clause 4's

explicit generation_tier='realm' write is now backed by the finding that NEITHER existing column

is a clean realm selector, and clause 5 grew from three colliding tier senses to five.

own proven Batch-0 mechanism, and the honest caveat is recorded: clause 1's enum-must-exist-at-

freeze argument is sound INFERENCE from the standing baseline-refresh rule, not a citation, and is

labelled as reasoning so it is never re-quoted as a cited fact.

the freeze; entries 100 and 105 clause 4 belong in the same DR-2 clause for the same reason.

presentation_tier_ref as an opaque key in the contract and ratify its VALUE DOMAIN separately, so

the freeze proceeds while the band-versus-rung question goes to Josh. Freezing a cardinality the

brief may overturn is the one way this wave could still fail after succeeding.

v1.3 claim-board note, and the lens produced no decay clause — decay appears in the draft only in

the DMC5 style-meter argument. The realm art-hook family still carries a ceiling law and no decay

law, so the fairy realm's fading state has no art bound at the moment the contract locks. It must

be written into the same open item 6 as the reservation, before the freeze, and it is now a named

obligation on that clause rather than an unowned scope line.

Claim-board flip (§v1.2 rows)

Derived from what actually ran: the P1-C four.

CLAIMED-QUEUED (P1-C) to CLAIMED-RUN. knowledge-progression-decipherment ran and its survivors

are entries 126-132 plus the roll-up. The boundary pre-recorded in v1.3 held: ac-historical kept to

the attribution half and this lens took the mechanics, and the two do not collide.

Survivors are entries 94-101, 106 and 107 plus the entry-4 merge. The row's own two riders resolve

differently: the realm CEILING amendment is DISCHARGED by entry 95's reservation, and the v1.3

entry 92 DECAY law is NOT discharged and is carried as the named debt above.

entries 102-105 plus the PC-13 COVERED row; R-A16 is claimed and folded by entry 105.

half is folded as entries 108-116, and the boundary pinned by adjudication 4 was HONORED in the

work rather than only in the note: the hades lens declared the execution-surface axis a handoff and

its QA finding (entry 114) is explicitly the orthogonal outcome and economy axis, so

RecordRetryLatency still lands exactly once, on the sm64 side, as v1.3 entry 53.

hades leg has now RUN; its realm-identity survivors are entries 117 and the HD-13 roll-up, and its

ruleset-linkage recommendation is routed into entry 73 rather than landed. All three legs of 14a

are complete.

cross-reference: the AC-8 substrate-versus-objects ruling in the roll-up is a design-collision call

that lands when the stub is authored, and it is not a duplicate of 14b's routing status.

KILLED its own no-photo-mode candidate against the live build item, which is the correct outcome

and adds nothing to the cell.

its verdict folded as entry 124. R-A16 (realm audio identity) is RUN and folded as entry 105. R-A15

(per-realm art direction) stays with the realm wave; per adjudication 20 the cross-realm signature

from entry 107 lands as the HEAD of that wave's art bible, not beside it.

Wave status board:

survive, 2 whole findings killed, 12 further premises or limbs struck. Survivors are entries 94-132

plus the per-lens roll-ups and the six amendments to existing entries.

gap-audit four ran 2026-07-27 as entries 20-45; P1-B and the realm doctrine wave ran 2026-07-28 as

entries 46-93; P1-C closes here as entries 94-132. Every P1 lens has now been authored, refuted and

folded. P2 (13 lenses) rides the factory and P3 (3 lenses) stays held by design, unchanged.

Kill ledger

charter's field of view) is false, because the single term hit the draft counted but did not open

is the charter's own downstream-edits list committing to fill those rows; and the general concern

is already entry 73, ranked HIGHEST. The recommendation is PRESERVED and routed to entry 73 as an

input; the finding does not enter.

the one uncited assertion in the draft and is contradicted by the ruled walked-geography structure;

it duplicates entry 9 at a higher priority than entry 9 carries; and its distinctive limb is a

PRE-REGISTERED DIVERGENCE that entry 9 already refuses by name. One crumb preserved as an

implementation note on entry 9.

ratified carries await_josh_confirm in its own header line. Running it would have applied an

unratified mechanical-reward rule across 122 rows. The cardinal-rule near-miss of the wave.

the promise it cites is post-story endgame content where revisiting is the point, and the arc

already ships the return vehicle structurally as the revisit band.

recommendation must not re-read a ratified T1 promise downward to fit a registry's current

population. If 98 rows under-serve the promise, the answer is more rows.

already declares.

a ranked, named, open item in the build plan, so the finding re-enters as an input to that open

re-derive rather than as a discovery.

correct answer repeated, which is the inverse of the lesson it names, and keys on a puzzle-boss

instrument invisible in ordinary combat. Re-shaped with a variety signal and a non-boss home, or

not at all.

adjudication 18 and restated on the corpus's own ruled wording: no sampling of consecrated

performance, reference-COMPOSITION permitted. As drafted it would have frozen generated audio for

four of ten realms for no §17 reason.

in five places, so the claim is contentlessness. As written it would have told a later agent the

ladder had no owner.

populated 79 of 79, not a registry realm rows can enrol in; the divergence tooth belongs to the

charter's proposed realm-teeth family.

altered law to environment-grammar rows, so only the hostile roster and the reward mix survive

un-covered.

add-a-clause framing — struck, three instances of the same class: each proposed as new what an

existing artifact already schedules, assigns or rules. All three findings survive reframed, and the

reframes make two of them stronger.

Corrections carried into this record

minigame top-score completion promise to a cardinal-numerical section; the real anchor is the CVD's

Layer-2 completionist definition, and the string the guard quotes appears nowhere in the section it

cites. The substance is TRUE, so no verdict moves — but every lens in this wave that touched the

no-scoreboard guard may carry the same wrong anchor, and the wave spec should be corrected at

source so the whole roster re-anchors.

distinct values (not zero and six), two of the three blanks being the Ch 1 and Ch 2 fragments and

therefore inside the slice; the mastery-tier threshold has 2 blanks (not zero); and the payoff

chapter is zero-blank only literally, with 18 rows carrying a literal not-applicable string. The

method rule earned here and worth adding to the program's shared file: blank counts are computed

and stated separately from histograms, never read off a printed top-N.

term count of one as absence without opening the hit is the mirror image of the repo's own

cannot-match-reports-zero law, and it produced the wave's only false premise.

— so citations should be converted to document-and-section anchors before they are re-quoted.

positive controls (mesh-id references 15 not 11, generation status 10 not 7, the reachability

control 213 hits not 121) and the animal-crossing faucet control (8 pre-existing hits, not 3).

JOSH BRIEF QUEUE

Four items reached the bar of a decision brief rather than an apply. Each goes to Josh through the

standing protocol — alternatives, deep-dive, adversarial pass calibrated against over-conservatism,

a recommendation and the strongest objection — and none is applied in the meantime.

closed band set keyed on base elements with hybrids inheriting, or the full ruled ladder. It

narrows a Josh-nodded hard requirement in two directions at once and costs the 21 hybrid elements

their presentation identity; it is reversible upward but not downward once assets exist. Carry the

corrected arithmetic (248 live key space, 372 at the clamp ceiling), the honest feasibility case

(hand-authoring 248 tiered effect systems solo is not a plan), the fact that the band boundaries

derive from real canon mastery gates, and the steelmanned status quo. The DR-2 freeze must not wait

on this: reserve the key, ratify the value domain separately.

governs mechanical reward classes across 122 link rows — squarely the canon Josh authors. Surface

it for confirmation; do not transform.

bars player-initiated bond movement as well as the departure computation. If it extends, the

finding converts to DIVERGENT-BY-DESIGN and is pre-registered so no later lens re-opens it; if it

does not, it is the highest-leverage relationship finding in the wave. Either way the bounding

invariant stands: investment never buys immunity.

time, a cooldown, standing with a scholar) or must cost nothing. The critic's timidity flag says

the option belongs on the table because a cost-free unbounded action invites brute force; the

never-punish reading is the safer read of the ruling. It is a ruled-doctrine boundary and therefore

Josh's, not the loop's.

Contested-carried to the director (not Josh): entry 110's bounded-persistence question before it

enters the gate; entry 115's scope against entries 47 and 48; the pop-up-rift density-band reading in

the roll-up; entry 117's class split; the animal-crossing sink-ladder class (WEAKER versus

blocked-on-the-economy-brief); and the T3 creature-roster count currency, which the critic routes to

a director-level apply as an internal inconsistency inside one ACTIVE doc — the director should

confirm that reading before applying, because a roster count is numeric-canon adjacent.

Generated by harness/site/structure_site.py — the URL path is the repo path. review root