REALMARC_PATTERN_BRIEF_2026-07-27.md

supernatural/REALMARC_PATTERN_BRIEF_2026-07-27.md

THE REALM-ARC PATTERN BRIEF — the replacement for mountain-range topography on Ch 70-77

Synthesis of the three-axis fan-out (pattern candidates · relief mechanism · flag pattern) ordered by

Josh's fourth-sitting ruling, 2026-07-27. Proposal-tier throughout. Read-only pass — no canon file,

registry, gate or baseline was edited producing this brief.

THE RULING, verbatim from docs/spine/DECISIONS_PENDING_JOSH.md fourth-sitting block: *"remove the

no three consecutive intensity peaks hard line for ch 70-77"* and *"we need a different pattern for

these mythological realm arc leading into the final end arc; brainstorm this, deep dive across all

our lenses."* The same block states the constraint that shapes this brief's authority split: *"the

new 70-77 intensity string is NOT set by hand — the pattern design proposes it."*

THE FIFTH SITTING, LATER THE SAME DAY, SAME FILE — carried here because the standing lesson is to

re-read the deciding artifact's NEWEST entry. DECISIONS_PENDING_JOSH.md L30, *"RULED 2026-07-27

(fifth sitting) — THE VERTICAL SLICE IS PROLOGUE→CH 13"*, sets the slice at Prologue → Ch 13 (build

order Ch 2-13 first, then Prologue + Ch 1) and states that **"the mythological realms (Ch 66-74) still

build in the post-slice factory era" while "the NARRATIVE intensity arc (the realms building up to

the fairy-realm finale) is untouched."** Nothing in it conflicts with this brief — but it changes the

brief's TIMING, and Josh should see that before he rules:

the slice. Nothing on this road builds until the post-slice factory era.

spine↔registry reconciliations, the R3 re-scope, and the two reconciliation sweeps (§0.4, C-L).

Those are repairs to canon artifacts that are wrong TODAY, and leaving a false HL_0048-holds

parenthetical sitting in CH_74.md for an entire slice cycle is exactly how later agents hallucinate.

slice cycle to land before its Stage-2 batches are needed.

ascent's TERMINUS is nearer the slice than its body is.

Sources synthesized: realmarc_pattern_candidates.md (eight shapes, the carrying-ordinal analysis,

the Final Ascent Law), realmarc_relief_mechanism.md (the four-jobs decomposition, the terrace

predicate, the built-ascent scoring), realmarc_flag_pattern.md (the flag's zero gate teeth, the

fossil diagnosis, the B+C reallocation). Live canon re-verified for this synthesis: HL_0048 as

amended and HL_0049/HL_0050 from registries/T0_Hard_Lines [ACTIVE v0.2]/Sheet1.csv; the

Ch 70-77 intensity/mode/flag rows from the live T0_Chapter_Index; the charter's three-instrument

resolution and the C-1 fork from docs/proposals/REALM_DESIGN_PROGRAM.md §1.4 / L1350-1372 / L1497;

the realm lens roster L-R1..L-R7 at charter L1069-1147; docs/proposals/COMPARATOR_LENS_PROGRAM.md.

---

0. The three findings the patterns are built on

0.1 The intensity column is a COMBAT-GAUNTLET column, so "build intensity" cannot execute on it alone

HL_0050's own live text says it: *"'Relief' lowers the combat-gauntlet intensity, never removes the

fun"*, and it calls its three relief nodes *"intensity valleys that frame surrounding peaks — never

action valleys."* On a combat-pressure column, Ch 77 — the game's single most important beat, and

the arc's lowest-pressure, player-paced one however its combat layer resolves — is correctly a LOW

value. (The brief deliberately does NOT assert Ch 77 as "canonically no-fight" here. That reading is

the CONTESTED side of C-D: the live registry carries combat=Y and R6 is never-waivable, while the

spine entry projects combat=N. The premise this section actually needs is only that Ch 77 sits at

the arc's pressure FLOOR, which holds under either resolution — see C-D as re-scoped.) Any pattern that reads

the directive as "raise the cells" peaks at Ch 76 and drops at the Revelation, which is what the live

string already does. Felt ascent is a composite, and the charter already separated the instruments at

§1.4: POWER follows the ruled vril curve (plateau 66-69, TROUGH 70-71, climb from 72, ceiling at 75),

DEMAND is authored and monotone, AWE peaks at anchor density and is never tuned as difficulty.

Consequence: the deliverable is a LAW with teeth, of which the intensity string is one output.

0.2 The enum cannot express an eight-step rise, by construction

harness/check_chapter_topography.py fixes INTENSITY = ("peak","standard","valley","relief"), and

R5 pins the relief VALUE to exactly {CH_27, CH_52, CH_57} per HL_0050. Three usable values for

eight nodes. Every pattern below therefore names its CARRYING ORDINAL — the fine field that actually

rises — and treats the enum as the coarse shape around it. The carrying ordinal already exists as a

charter proposal: demand_ordinal in the INTENSITY BLOCK (charter L508, alongside power_ordinal,

awe_ordinal, escalates_from/to, sovereign_distance) — none of them minted, all proposal-tier.

0.3 The alternation HL_0048 performed on INTENSITY has to move to another column, and the built prose already performs it

Read off the eight body-prose blocks rather than declared anywhere: Ch 70 elegiac-mournful · Ch 71

adversarial-trickster (the ascent's only comic register) · Ch 72 ascendant-moral-architectural ·

Ch 73 warm-hospitable · Ch 74 vast-elegiac-laboring · Ch 75 grief-and-war · Ch 76 grave-argument

resolving to intimate-stillness · Ch 77 wonder-and-release. No two consecutive nodes share a

register; every five-node window carries five distinct ones. The charter's own feared failure mode

(§4.2 — five consecutive death realms converging into monotone-solemn) is NOT what the built prose

does. That is the single most important fact in this brief: the arc's best-executed pacing property

already exists, and NOTHING in the repo can see it. music_mood_tags on T0_Chapter_Index is

populated 0 of 79 rows; no other column anywhere carries emotional register.

The rule that replaces the removed clause is therefore a LOCK ON EXISTING QUALITY, not a new

constraint — the strongest kind of rule this repo can adopt.

0.4 The live state, verified

Chnodeintensitydominant_combat_modehard-line flagbeatschar+dialogue beatslow-threat sites
70Hades and Elysiumstandardvril_managementFALSE1024 of 5
71Xibalbapeakpuzzle_routedFALSE1020 of 4
72Naraka and Merupeakvril_managementFALSE1013 of 4
73AnnwnstandarddirectFALSE931 of 4 (hall hosts 5 of 9 beats)
74YggdrasilstandardenvironmentalTRUE902 of 4
75The Realm at WarpeakdirectFALSE1240 of 4
76The Two Courtspeakpuzzle_routedTRUE1060 of 2
77The WaterfallstandardastralTRUE1055 of 5
78EpiloguevalleyenvironmentalTRUE

Live string Ch 70-77: standard · peak · peak · standard · standard · peak · peak · standard.

THE REGISTRY IS NOT THE ONLY DECLARING ARTIFACT, AND THE TWO DISAGREE. **docs/spine/CH_74.md L205

already declares intensity PEAK**, verbatim: *"intensity PEAK (the reserved crossing peak that

Ch 73's relief set up — the last Realm-Road node and the door home; Ch 71 peak, Ch 72 peak, Ch 73

standard, Ch 74 peak — no three consecutive peaks or valleys, HL_0048 holds)"*. The live registry

carries standard. The authored spine already says PEAK; the registry lags it. Three

consequences, and they change what this brief is asking for:

"The built content already earns it" is not an argument in this brief — it is literal: the authors

wrote PEAK into the entry. Josh is not being asked to change a cell; he is being asked to ratify

which of two disagreeing canon artifacts is right, on a cell where the deciding artifact already

ruled.

three consecutive peaks or valleys, HL_0048 holds" was computed over Ch 71-74 only; under its own

declared PEAK, Ch 74-75-76 is peak/peak/peak. A false hard-line-compliance claim sitting inside a

canon artifact repairs regardless of which pattern Josh picks.

intensity VALLEY twice (L212, and again at L265 inside a "PRESERVED:" note asserting it as

carried-forward canon, with its own "HL_0048 holds" arithmetic built on it) while the registry

carries standard. Two of twelve road nodes' spine entries disagree with the registry on

intensity — so this is a CLASS, not a one-off, and the class is invisible to every gate.

CONSEQUENT WORKLIST ITEM, arc-wide: a spine-vs-registry intensity sweep across all 79 nodes.

Neither artifact has ever been checked against the other on this column. Ch 66 and Ch 74 are what a

twelve-node read surfaced; nobody has read the other sixty-seven.

Two curves nobody has declared are visible in that table. THE INTIMACY CURVE ALREADY RISES WITH

INTENSITY (character+dialogue beat share 20, 20, 10, 33, 0, 33, 60, 50 percent). THE

ENVIRONMENTAL FLOOR ALREADY FALLS AWAY (low-threat site share 4/5, 0/4, 3/4, 1/4, 2/4, 0/4, 0/2,

5/5). The two are inverse by construction: relief migrates from the environment into the relationship

as the ascent climbs. And there is one hole — **Ch 74 Yggdrasil, the door home, the charter's own

"highest-leverage node in the program," has ZERO character and ZERO dialogue beats in a nine-beat

graph**, the only such node in Ch 66-77, while its own site row high_bough_door_home DECLARES a

character function that no beat uses.

0.5 What binds under every pattern (the envelope)

dominant_combat_mode; every 5-node window carries ≥3 distinct dominant modes. The live 70-78 mode

run satisfies both with margin. No pattern here moves a mode.

coasts); the valleys clause; the peaks clause still binding Ch 1-69.

debut Ch 66-69 and STOP). Mechanical novelty on the ascent comes from the signature-law lane.

demand-shaped (the myth's own trials), never body-count-shaped.

---

1. THE FOUR PATTERNS

Each pattern is a full triple: intensity string + relief architecture + flag set. The flag component

is stated per pattern, but §1.5 shows why it comes out nearly identical in all four — which is a

finding, not a dodge.

1.1 PATTERN A — THE TERRACED ASCENT

*Lens citation: presentation-craft-anim-audio (the adaptive-score half — terraced dynamics and the

subito piano) · L-R5 underworld-grief-and-register-variety (the register slice the charter already

pulled forward) · L-R4 pilgrimage-traversal-and-awe (Journey's apex-as-rest) · `hades-death-run-

realm-identity` (the interstitial as the relief unit) · L-R3 (the factual settler for the flag keying).*

INTENSITY STRING, Ch 70-77:

standard · peak · peak · standard · PEAK · peak · peak · standard

One cell changes from the live REGISTRY: Ch 74 Yggdrasil standard → peak — which is the cell

docs/spine/CH_74.md L205 already declares as PEAK (§0.4). The edit reconciles the registry to

its own spine entry rather than proposing a new value. Ch 74-75-76 becomes three consecutive peaks,

which the ruling has just made legal and which is the point. Epilogue stays valley.

CARRYING ORDINAL: demand_ordinal rising 5→11 across Ch 70-76, EXEMPT at Ch 77 via the

empty-string sentinel (T-RLM-CURVE skips empty rows rather than coercing them to zero — the

charter's own construction, adopted unchanged). Demand is INSTRUMENTED as accumulated

signature-law constraint (each realm's law learned in its threshold zone, reprise_of compounding

from road position three), never as stat inflation. This reuses the ruled difficulty doctrine

verbatim; nothing new is minted mechanically.

THE SHAPE'S LOGIC. Orchestral practice has known for three centuries that a literal eight-step

crescendo is unlistenable — the ear normalises by step five. The working devices are terraced

dynamics (each step's FLOOR rises; the ceiling may be reused) and the subito piano, the sudden hush

immediately before the final fortissimo, which is the loudest instrument in the orchestra precisely

because it is silence. The hush is ALREADY AUTHORED at Ch 73: Annwn is canonically "the last warm

beat before the crossing," its register is warm-hospitable, its fight is a single one-blow honour

duel won by RESTRAINT, and its feast hall hosts five of nine beats at ambient_threat low. And Ch 74,

one node later, is a marquee dragon fight plus a crossing-defense — the loudest node on the road,

mislabelled standard because until this ruling lifting it would have produced three consecutive

peaks and broken HL_0048 outright. **The ruling unlocks precisely the one change the charter wanted

at §6.2 and could not take, and it settles the C-1 fork on its merits rather than by workaround:

lift the DOOR, not Annwn.**

RELIEF ARCHITECTURE — THE TERRACE. A valley was an INTER-chapter object; the ascent's relief is an

INTRA-chapter object. A mountain that rises continuously still has ledges, and a climber rests on the

ledge without descending. A TERRACE is a bounded low-pressure passage inside a rising node, and it

qualifies by satisfying at least one of three forms, every term resolving to an existing populated

field:

learning, discovery, transformation}, none role climax, ≥1 carrying boss none.

AND whose ambient_threat is none-or-low, non-combat mode, boss none, not role climax.

The climax exclusion and boss none are load-bearing: without them the predicate reports Ch 76's

Cassius moral phase (character mode, climax role, boss phase BE_0076.P1c) as a rest beat, which would

make the instrument lie about the most pressured scene in the game. The awe conjunction is equally

load-bearing: Ch 66's atlantis_ring_city carries sacred at ambient_threat HIGH — that is

SPECTACLE, not rest.

THE FOUR CLAUSES:

qualifying form may be T-C. (The anti-hollow clause; Ch 74 is its motivating violation.)

register; every five-node window carries ≥3 distinct registers. Deliberately HL_0049's exact

shape on a different column, so the tooth is R1/R2 re-pointed rather than new rule machinery.

below the previous, none may be a valley or a relief, exactly ONE step-down is permitted, and it

must be immediately followed by a peak. A hush is a preparation, never a rest. (Strictly stronger

than both removed clauses, which dissolves the §3.6 valleys-scope ambiguity rather than arguing it.)

passage — a vril-site sanctum (save, recharge, loadout per COMBAT_PROGRAM_ADDENDUM §9), player-paced,

ambient_threat none, ≥1 companion exchange, and the register hand-off from departing to arriving

node. Five seams only: 69→70, 70→71, 71→72, 72→73, 73→74. From Ch 74 onward every seam is typed

"within-realm, no travel-medium change" in the transitions layer, so a camp there would be a scene

the geography forbids; the 74→75 seam takes a MASTERY terrace instead (the vril ceiling exercised

once with nothing shooting back, before Ch 75's urgency takes over).

THE BUILT ASCENT SCORED AGAINST CLAUSES 1-2:

nodeT-AT-BT-Cverdict
CH_70asphodel_meadowsB03+B04B04PASS, three forms
CH_71nonenonenoneFAIL — no terrace of any form
CH_72narakas_weighing_placeB08+B09B01PASS, three forms
CH_73arawns_court_feast_hallB01+B02B01PASS, three forms
CH_74nonenoneB03THIN — awe-only, Clause-2 violation
CH_75noneB01+B02nonethin — intimacy-only
CH_76noneB09+B10nonethin — intimacy-only (correct; it should not have a meadow)
CH_77the_last_ascentB04+B05B02PASS, three forms

Seven of eight pass Clause 1; one hard failure (Ch 71); one Clause-2 violation (Ch 74). That is the

worklist, and it is small.

FLAG SET: the B+C reallocation at constant Josh cost — road full-tier flags 7 before

(66,67,68,69,74,76,77), 7 after (69,71,72,73,74,76,77); arc-wide 18 before, 18 after. Ch 66/67/68/70

and 75 drop to CG-26's calibrated LIGHT tier (§6.8 batch cultural audit + anti-fringe / reveal

critics; no Stage-2 Josh batch, no author-in-loop quests). See §1.5.

CANON FIT AGAINST THE LIVE SPINE. The strongest of the four. The one changed cell is the cell the

charter itself wanted changed and the spine's own prose already earns: Ch 74's body is "a battle and

a tending in the same breath," it carries the Níðhöggr fight plus the crossing-defense, it is the

Thread-5 capstone and the carved-debt mark_write, and it holds the road's highest awe. The hush at

Ch 73 is not imposed — it is the spine's own "last warm beat before the crossing," and the spine

already credits Ch 73 with SETTING UP Ch 74 ("the reserved crossing peak that Ch 73's relief set

up"). Register rotation passes on all eight nodes today. The one place the pattern fights the built

content is Ch 71, which has no ledge of any kind — its relief is carried ENTIRELY by register (it is

the ascent's only comic node and the prose is genuinely funny), which is real but currently invisible.

THE FAIRY-REALM ARRIVAL IT PRODUCES. The largest dynamic leap available in the game. The player

crosses out of the quietest node on the road (Annwn's borrowed year), through the loudest (the door

under the tree), and arrives at the vril CEILING — the power she has spent seventy-four chapters

climbing back toward, handed back the moment she is home, into a war. Homecoming as detonation. The

Waterfall then reads as the plagal cadence after the final chord, and the Epilogue valley resolves to

tonic.

AUTHORING COST. **THE R3 RE-SCOPE IS A HARD BLOCKING PREDECESSOR AND SHIPS IN THE SAME COMMIT AS THE

CELL — it is not optional and not deferrable.** harness/check_chapter_topography.py R3 as CODED

walks the whole ordered node list with no Ch 1-69 scope (for i in range(len(usable) - 2) over every

node). Re-derived against the live registry: R3 violations today = 0 (gate green); with Ch 74 →

peak, R3 fires peaks [CH_74, CH_75, CH_76] UNWAIVED → gate: FAIL. The one-cell edit reds

run_gates.py BY ITSELF. (Pattern B produces four unwaived R3 violations.) The canon side is clean —

the only 3-peak window lies entirely inside Ch 70-77 where the amendment removed the clause, HL_0049

is untouched (intensity is not mode), the surviving mode-variety clause passes (Ch 74 carries seven

distinct beat modes), and no valley is introduced. **The failure is gate-side only, which is exactly

why it would be missed.** Then: one registry cell (Ch 74 intensity) + baseline re-emit. One column mint

(register_dominant, closed enum, 79 rows; becomes the key music_mood_tags derives from, partially

closing the empty T1_Audio_Spec gap). Two gate homes: Clauses 1-2 and 4 as corpus errors in

check_spine_graph.py (every input already parsed by spinelib.py — BEAT_FIELDS carries role, mode,

site, boss; the site parse carries ambient_threat, vril_density, climactic_affordance), Clause 3 as R7

in check_chapter_topography.py. Three authoring items: the CH_74 intimacy beat, the CH_71 climb-out

companion beat, the CH_75 site split. Two vril-site rows (Ch 75, Ch 76 currently have none of their

own) plus a sanctum-class function on the five road rows. Five seam-camp rows folded into the

already-ordered Fork-C seam-typing pass. Two critic rubrics. Plus the flag reallocation's three

Stage-2 batches that do not exist today (Ch 71, 72, 73) — gated on the substrate repair (§3, C-G).

1.2 PATTERN B — THE PACT LADDER (the maximal ascent)

*Lens citation: hades-death-run-realm-identity (the Pact of Punishment — escalation by ACCUMULATED

CONSTRAINT, not by raising the enemy sheet) · the god-of-war re-run widened to realm adaptation

(the nine realms as places, the world-tree hub, per-realm sovereign identity) · `sekiro-timing-

defense-chain-meter` (demand rising against a toolkit that does not grow).*

INTENSITY STRING:

standard · peak · peak · peak · peak · peak · peak · standard

Two cells change: Ch 73 and Ch 74 both standard → peak. Six consecutive peaks — the maximal legal

reading of the ruling.

CARRYING ORDINAL: identical to Pattern A (demand_ordinal 5→11 as accumulated constraint-clauses).

By Ch 74 the player carries eight compounding signature laws. Demand rises monotonically with zero

stat inflation, on a mechanism the project has already ruled and can already measure off the registry

(count laws per realm, count reprise_of references per realm).

RELIEF ARCHITECTURE: the same terrace predicate and the same five clauses, MINUS Clause 4's hush

provision (there is no step-down to sanction), and with register rotation promoted from one

instrument among several to THE ONLY inter-node contrast instrument. The seam camps become

load-bearing rather than nice-to-have, because with the enum flat the seams are the only places

carrying no intensity cell at all.

FLAG SET: the same B+C reallocation, with one hardening — because the pattern raises pressure hardest

at the two most care-loaded rows on the road (Ch 71 LIVING, Ch 72 MAXIMUM), the newly-full-tier flags

at 71/72/73 stop being a correction and become a PRECONDITION, and the substrate repair for those

three nodes becomes a hard blocking predecessor rather than a scheduling note.

CANON FIT. Good at Ch 74, actively bad at Ch 73, and the evidence is direct rather than aesthetic.

Ch 73 is the ascent's ONLY warm register; its only chapter with a low-threat hall hosting five of

nine beats; one of only four nodes carrying all three terrace forms; and the spine itself credits it

with setting up Ch 74. Lifting it deletes the strongest terrace on the road and collapses the

register rotation at exactly the point the road most needs contrast. Six consecutive peaks across

five death-and-judgment realms is also the fastest route to the monotone-solemn slop mode the charter

names at §4.2 — the SMT/Persona failure the lens roster exists to teach us to avoid.

THE FAIRY-REALM ARRIVAL IT PRODUCES. The top rung of a ladder — and that is the problem. If Ch 71

through Ch 76 are all peaks, the fairy realm is one more peak among six and the arrival does not read

as an arrival at all. The pattern's own logic (Law 0's prediction — every law the road taught is

simply the local weather in the fairy realm) is a beautiful L3 reading that the pacing column

actively works against.

AUTHORING COST. Everything in Pattern A, plus one more intensity cell, plus a materially higher

register-authoring burden (register is doing all the contrast work alone), plus the hardened

substrate blocker. Roughly 1.3× Pattern A for a strictly worse arrival.

1.3 PATTERN C — THE HONORED CURVE (the steelmanned status quo)

*Lens citation: the charter's own §6.2 opening recommendation, pre-ruling: "do NOT change the

intensity column; carry the directive on the DEMAND and AWE axes instead, and give Ch 74 the road's

highest awe ordinal and its strongest signature law." Plus L-R5 (register procession) and L-R4.*

INTENSITY STRING: unchanged. standard · peak · peak · standard · standard · peak · peak · standard.

CARRYING ORDINAL: demand_ordinal monotone 5→11 AND awe_ordinal (charter §1.4 names AWE as a third

instrument but proposes no column, and T-RLM-CURVE reads demand only, so awe is currently a doctrine

with no row). One rule makes the charter's own principle arithmetic: awe_ordinal peaks where vril

density reads anchor. Ch 74 takes the road's highest awe ordinal and its strongest signature law

(the carved debt — the Thread-5 capstone and the spine's own mark_write mechanism).

The argument, and it is not weak: a ruling that REMOVES a ban does not COMPEL its exercise. Josh

removed a constraint that was distorting the arc; the honest question is whether the arc needs the

enum to move at all, given §0.1's proof that the enum cannot carry the directive. This pattern says

no — the rise is real, it lives on demand, awe and register, and the enum stays where three

independent analyses agree the built content already put it.

RELIEF ARCHITECTURE: identical terrace predicate, Clauses 1-3 and 5. Clause 4 is replaced by a

NO-REGRESSION clause on demand_ordinal alone (the enum is not constrained beyond the surviving

HL_0048 clauses).

FLAG SET: the same B+C reallocation. The flag defect is pattern-independent (§1.5) — this pattern

cannot keep the current flag set either, because the current set was proved a fossil rather than a

judgement.

CANON FIT. Perfect by construction — it changes nothing. That is both its strength and its whole

problem.

THE FAIRY-REALM ARRIVAL IT PRODUCES. Legible to a reader of the columns and invisible in the hands.

The player crosses the door (rated standard, containing a marquee dragon fight) and arrives at Ch 75

(rated peak, exactly as Ch 71 and Ch 72 were rated). The demand and awe ordinals say the arrival is

the largest event on the road; the pacing column says it is the third peak of six. The charter's own

strongest objection stands verbatim: *"a standard-rated crossing may simply read as an anticlimax no

matter what the awe column says, and a hard line that produces an anticlimactic door is a hard line

worth a ruling rather than a workaround."* Josh then issued exactly that ruling.

AUTHORING COST. Slightly BELOW Pattern A — one cell and one baseline re-emit cheaper — and otherwise

identical, because every expensive item (register mint, terrace gates, the Ch-74 intimacy beat, the

Ch-71 fix, the seam camps, the flag reallocation) is required under every pattern. **That near-parity

is the decisive comparison in this brief: Pattern A costs one registry cell more than doing nothing,

and buys the arc's loudest node.**

1.4 PATTERN D — THE DEEP FLOOR AND THE RESTORATION

*Lens citation: Journey (dual-listed into L-R4 per the charter widen — the summit where accumulated

ability is stripped and the emotional peak arrives at minimum power) · Sekiro · Dante's Purgatorio

(a proposed non-game structural comparator; the roster currently has none, which is a roster gap by

its own governing principle).*

INTENSITY STRING:

VALLEY · peak · peak · standard · PEAK · peak · peak · standard

Two cells change: Ch 70 standard → valley, Ch 74 standard → peak. The arc opens at its floor to give

the ascent maximum runway and to make the law-change at the Ch-69/70 seam audible.

CARRYING ORDINAL: demand_ordinal plus a companion constraint_count rising 70→74 and dropping to

zero at Ch 75. The road's realms progressively CLOSE usual answers — familiars barred where the

living-and-bound cannot pass, companions who cannot cross a given threshold, realm deltas closing

specific ability classes via the existing ability_availability_delta schema field — and Ch 75

restores the ceiling for the first time since the Prologue.

RELIEF ARCHITECTURE: POWER TROUGH AND RESTORATION as the primary carrier, which is the arc's largest

available exhale and is ALREADY CANON (the ruled vril curve troughs at Ch 70-71 with Xibalba as the

floor, climbs from 72, ceilings at 75 — nothing needs inventing). Terrace clauses 1-3 and 5 ride

underneath. The 74→75 mastery terrace is mandatory here rather than optional.

FLAG SET: the same B+C reallocation; Ch 70 is historical and LIGHT-tier under it, so a valley at

Ch 70 carries no flag consequence.

CANON FIT. Half of it is already ruled and gated — the vril trough is real, Ch 70's descent is

unopposed, the honoured dead are never targets, and combat routes only through the egress, so a

valley at Ch 70 is a defensible reading of built content. The other half **collides with charter Law

6, THE FLOOR LAW**: a realm may close a usual answer but never confiscate the build — build, mastery

tiers, gear, imprints and bonds cross every threshold intact. Aggressive stripping is illegal under

the charter's own floor. In its legal soft form (realms CLOSE answers, they never take things away)

the pattern becomes nearly identical to Pattern B's constraint accumulation, which is evidence the

two should merge rather than compete. It also re-opens the §3.6 seam questions in the sharpest

possible way: a valley inside the ascent, while the valleys clause's scope is ambiguous and R3's

window-straddling behaviour is unruled.

THE FAIRY-REALM ARRIVAL IT PRODUCES. The only arrival in the set whose mechanism is JOY rather than

pressure, and it is genuinely the correct emotional shape for a HOMECOMING. The cost is that the arc

opens on its quietest node — a player who reads valley as "filler" gets exactly the wrong signal

from the road's most beautiful chapter, and Ch 70 is the katabasis that the whole road hangs from.

AUTHORING COST. Highest of the four. Pattern A plus a second intensity cell, plus the

constraint_count column and its population, plus the per-realm ability-delta authoring, plus the

mandatory mastery terrace, plus the two gate-scope ambiguities forced into rulings before anything

can land.

1.5 The flag set is pattern-INDEPENDENT — and no pattern may keep the current one

The flag axis established, positive-controlled, that canonical_hard_line_flag has **zero

deterministic gate teeth**: grep -rnF "canonical_hard_line_flag" harness/*.py returns zero rows

while the same matcher over the same files hits intensity and mode_variety_check. The flag is a

JOSH-ATTENTION ALLOCATOR and a routing bit — the 8th state (canonical_question_review) in the

spine_status machine, the /spine-chapter Step-2 Stage-2 question batch, quest

author_in_loop_required, the §6.8 cultural audit. Nothing goes red if it is wrong. That is exactly

why it silently rotted.

And it did rot, with receipts in three artifacts: Ch 66-69 were flagged for the CVD §19.6 four-realm

name-lock that the Realm-Road re-architecture RETIRED; Ch 74 was flagged for a reveal that re-homed

to Ch 76 (its own spine header reads "RE-BASED to a STRUCTURAL KEYSTONE… NOT a reveal; the reveal

content re-homes to Ch 76"); and Ch 70-73 are unflagged because at population time they were faceless

fairy-war content, not Hades, Xibalba, Naraka/Meru and Annwn. **The column is now anti-correlated

with care**: the road's one invented realm is flagged, and the road's LIVING and MAXIMUM-care

realms are not. The charter says it in its own words about Ch 72 — "the highest-care realm on the

road that has had the LEAST attention."

The recommended repair is the charter's own C-4 fix applied one column over ("Two fields, two jobs,

one tooth each, and neither column asked to mean two things"). Split the flag's two JOBS and key each

to an axis that already exists — CULTURAL basis fires on generation_class ∈ {living, unsettled}

(a FACTUAL question L-R3 answers, not reviewer memory); NARRATIVE basis fires on irreversible-reveal

nodes; everything else takes CG-26's ruled LIGHT tier.

ChRealmNowProposedBasis
66AtlantisTRUELIGHTinvented; anti-fringe duty rides care_tier + the armed check_layer_register.py TIER-A gate
67PohjolaTRUELIGHTDOWNGRADE — historical (PENDING L-R3); Ariosophy/Thule firewall rides the same gate
68the ApsuTRUELIGHThistorical, no living practitioners
69the DuatTRUETRUEliving descendants + active revival; and the Architect's terminus (both bases)
70HadesFALSELIGHTDOWNGRADE in tier though already FALSE — historical, standard care (PENDING L-R3)
71XibalbaFALSETRUEliving, the sharpest tier — Popol Vuh, Guatemalan intangible heritage
72Naraka / MeruFALSETRUEMAXIMUM care, living major world religion
73AnnwnFALSETRUEunsettled → strict until L-R3; auto-relaxes to LIGHT if L-R3 returns historical
74YggdrasilTRUETRUEliving inclusive Heathenry (PENDING L-R3) — flag RETAINED on an honest basis, replacing the stale Cassius rationale; retention is the safe direction while L-R3 is out
75The Realm at WarFALSELIGHTinvented, but the arc's most delicate reveal-RESTRAINT → reveal re-verify, no Josh batch
76The Two CourtsTRUETRUEnarrative: the patriarch reveal + the father encounter
77The WaterfallTRUETRUEnarrative: the single sanctioned §17.9 surfacing

Data-integrity note on the column being reallocated: canonical_hard_line_flag has **three

unpopulated cells** (CH_13, CH_14, CH_15) — 18 TRUE + 58 FALSE + 3 blank = 79. The column carries a

live data hole outside this road, and the reallocation should not be the thing that discovers it.

Road full-tier: 7 before, 7 after. Arc-wide: 18 before, 18 after. A reallocation, not an

inflation — which is the property that satisfies CG-26's anti-inflation half and the bolder care

calibration simultaneously and provably.

THE TWO-LANE CLAIM, STATED HONESTLY IN THE CONDITIONAL. An earlier draft of this brief asserted that

"the charter's §3.3 costume tooth in read_boards.py already reads generation_class," making the

reallocation close a LIVE two-lane disagreement at zero cost. **That claim is false and is

withdrawn. Verified: harness/read_boards.py does not exist** — the charter lists it as work to

be BUILT (REALM_DESIGN_PROGRAM L961, L1285); and generation_class returns **zero hits across

harness/*.py and zero across registries/** (positive control: the same searches return intensity

in three harness .py files and care_tier in five registries, so the matchers demonstrably match

and the absence is real). There is no live two-lane disagreement — there is a disagreement between a

PROPOSED tooth keyed to a PROPOSED column and live routing. The convergence is a benefit the

reallocation will deliver ONCE both are built; it is not one it delivers today, and it is not

available as a reason to adopt it now.

THE FLAG KEY IS ITSELF PROVISIONAL — the whole CULTURAL column is pending L-R3. The

recommendation keys the CULTURAL basis to generation_class ∈ {living, unsettled} and calls it "a

FACTUAL question L-R3 answers, not reviewer memory." That is the right DESIGN, but L-R3 has not

landed, and the charter is explicit that the words this table is built from are not yet enum values:

*"Words like LIVING and MAXIMUM in the rows below are the analyses' descriptive readings … they are

prose, not enum values, and this charter asserts none of them into that column. … the authority that

settles each realm's value is L-R3's named output."* The flag substrate marks three nodes pending

L-R3 — Ch 67, Ch 70 and Ch 74 — and the table below now carries all three markers (an earlier draft

carried the caveat for Annwn alone). **Two of those three are DOWNGRADES (Ch 67, Ch 70), and a

downgrade is the direction in which a wrong call ships UNPROTECTED** — the asymmetry the charter

itself invokes. Josh should rule the downgrades either as gated on L-R3, or as explicitly

provisional.

THE GUARDRAIL THIS BRIEF MUST STATE OUT LOUD: the flag is not a pacing instrument. Pillar 3's

"sanctuary/transformation/learning beats remain chapter peak" and HL_0050's relief language can be

misread as making a flagged node CALMER. On the final ascent it must not. Ch 71 and Ch 72 are PEAK

today and stay PEAK under every pattern here — they gain a Josh batch and a rights-strict asset path,

not a lower intensity ordinal. A flagged node is not a less-fought node: HL_0121 governs, Vritra is

fought, the two Lords of Xibalba are fought, Níðhöggr is fought. The flag routes REVIEW, never combat

presence. Any drafting that reads "flagged → sanctuary" has re-imported the three-concept conflation

at §3, C-I.

---

2. RECOMMENDATION

ADOPT PATTERN A — THE TERRACED ASCENT. String `standard · peak · peak · standard · PEAK · peak ·

peak · standard`; demand monotone 5→11 as accumulated signature-law constraint with Ch 77 exempt;

relief as the three-form terrace with the register rotation carrying the alternation HL_0048 used to

carry on intensity; five seam camps folded into the ordered Fork-C pass; the B+C flag reallocation at

constant Josh cost.

READ THIS FIRST — the recommendation's centerpiece cell is not a proposal. docs/spine/CH_74.md

L205 already declares intensity PEAK; the live registry carries standard (§0.4). Adopting

Pattern A's one cell RECONCILES a live spine↔registry divergence in the direction the authored spine

already chose. Josh is ruling which of two disagreeing canon artifacts governs — not authoring a new

value. Two riders travel with it: the CH_74 entry's own *"no three consecutive peaks or valleys,

HL_0048 holds"* parenthetical is FALSE under its own declared PEAK and repairs either way, and

docs/spine/CH_66.md carries the same divergence class (declares VALLEY twice, registry says

standard).

The five reasons, ranked:

1. Its one change is the change the built content already MADE. Ch 74 contains the dragon, the

crossing-defense, the Thread-5 capstone, the carved debt, and the road's highest awe; its spine

entry declares PEAK in as many words; and the REGISTRY rates it standard for one reason: lifting

it used to break HL_0048. Josh removed that. Pattern A is what the arc looks like with the

distortion taken out — and the distortion was only ever in the registry.

2. It preserves the hush the arc already has. Three independent axes converged on the same

evidence that Ch 73 must not lift: the only warm register, the only low-threat hall hosting five

of nine beats, one of four three-form nodes, and the spine's own credit that it sets Ch 74 up.

3. It costs one registry cell more than doing nothing. Every expensive item is common to all four

patterns. Pattern C buys nothing for its saving and inherits the charter's own anticlimax

objection, which Josh's ruling was issued to answer.

4. The relief replacement is a lock on existing quality, not a new burden. The register rotation

already passes on all eight nodes; the terrace predicate already passes on seven; the two failures

(Ch 71, Ch 74) are real authoring gaps that were invisible before this wave and are worth fixing on

their own merits.

5. It is arithmetic first and declaration second. Clauses 1, 2 and 4 are pure predicates over four

existing populated columns with no human judgement anywhere. Only Clause 3 rests on a declared

value, and its bound is declared on the rubric rather than hidden.

THE STRONGEST OBJECTION. *It changes one cell and calls itself a new pattern. Josh ordered "a

different pattern" and the intensity string barely moved.*

The honest answer, and it must be given in these terms rather than defended around: the string is the

smallest part of the deliverable and §0.1 shows why — the column cannot carry the directive at all.

What actually changes is the LAW. Mountain-range topography is replaced on this arc by a rising-floor

law with one sanctioned hush; a monotone demand column that does not currently exist; the alternation

obligation moved off intensity onto register; relief re-grained from the chapter to the terrace; five

seam camps; a flag column re-keyed from a fossil list to a factual axis; and five new teeth where

today there are zero on this arc. The intensity string moving only one cell is EVIDENCE that the

built spine was already the right shape and was being held down by a rule written for a different

part of the game — which is precisely what Josh's ruling diagnosed. **And that evidence is no longer

an inference: the spine entry for the one moving cell already declares PEAK in as many words (§0.4).

The string barely moves because the authors had already written the shape this pattern asks Josh to

ratify — only the registry never caught up.**

A second, weaker objection worth pre-empting: *the recommendation makes Ch 74 the arc's loudest node

while Ch 74 is simultaneously the arc's most relief-BARE node (zero character/dialogue beats).* This

is correct and it is why the lift is issued CONDITIONAL: the CH_74 intimacy terrace is a hard

predecessor to the intensity cell, landing in the same commit. See §3, C-B.

THE SEAM IS A FEATURE. The ruling's range begins at Ch 70 — node five of the nine-realm road — so the

law changes in the MIDDLE of the road. That reads at first as a scoping artifact worth extending back

to Ch 66. It should not be extended. Ch 66-69 alternating under mountain-range law is what makes Ch

70-77's continuity legible as a different world; and Ch 70 is the vril trough and the katabasis, so

the ascent begins at the floor, which is where an ascent should begin. If Josh later wants the whole

road under the new law, that is a one-line extension.

---

3. CONTRADICTIONS — every disagreement the three axes left standing

Named rather than smoothed, per the standing discipline. Each carries a recommendation; none is

resolved by this brief alone.

C-A. FACULTY ROTATION vs REGISTER ROTATION — which column carries the alternation. The pattern

axis's Clause 4 proposed FACULTY rotation (force / wit / release / restraint / commitment); the

relief axis's Clause 3 proposed REGISTER rotation. Both are HL_0049's exact shape on a new column, and

minting both means two new columns and two R1/R2 clones for one job. RECOMMENDED: mint REGISTER, not

faculty. Register is already delivered by the built prose on all eight nodes (positive-controlled by

full reads, not sampling), it feeds music_mood_tags which is populated 0 of 79, and it is

TIME-CRITICAL — the natural-voice rewrite pass is the next session's first wave and a reviewer

smoothing eight chapters toward one voice would pass every existing gate. Faculty is a DERIVED read of

the existing beat mode column and needs no field at all; it rides as an advisory critic rubric.

C-B. Ch 74's lift versus Ch 74's hole. Pattern A makes Ch 74 the loudest node on the road; the

relief axis proves Ch 74 is the only node in Ch 66-77 with zero character and zero dialogue beats,

while its own high_bough_door_home site row DECLARES a character function no beat uses. Lifting it

without authoring the terrace compounds the defect at the highest-leverage node in the program.

RECOMMENDED: the intensity cell and the intimacy beat land in the same commit; one boss-free character

beat at high_bough_door_home immediately before CH74_B09 — the last quiet before the last door —

which converts Ch 74 from a Clause-2 violation to a pass **and repairs the live registry cell that

already asserts character=Y on a graph with no character beat (C-L).** The beat is owed on three

independent grounds — pacing, the site row's own declared character function, and registry truth —

which is why it is a hard predecessor rather than a nice-to-have.

C-C. Ch 73 — lift or hold. Six of the eight candidate shapes lift Ch 73 to peak; Pattern A and the

relief axis hold it at standard with direct evidence. Both readings are on the table and the

disagreement is real. RECOMMENDED: HOLD, per §2 reason 2. This is the single cell most likely to be

overturned by Josh and the brief should not pretend otherwise.

C-D. Ch 77's combat divergence — and the direction of the reconciliation is NOT open.

T0_Chapter_Index carries combat=Y;puzzle=N for CH_77 in mode_variety_check while the CH_77

spine entry's fold note declares the projected registry value *"projected character=Y; puzzle=Y (via

vril_management); learning=Y; transformation=Y; combat=N — the no-combat finale legitimately runs

four of the five"* and its beat graph contains zero combat-mode beats. **The two artifacts disagree

on combat AND on puzzle** (the second divergence went unnamed in the earlier draft). Same class as

the Ch-74 and Ch-66 intensity divergences at §0.4.

WHICH IS CANON: combat=Y. R6 reads the registry, is coded never-waivable, and is checked before

the structural gate; HL_0051 says a *"Hard Line chapter without combat layer at all is a pillar

violation. Combat absence is creatively fatal"*; and Josh's standing ruling of 2026-07-03 already

binds it — **L1 must be action-packed and playable in EVERY chapter, the hard line winning over WGM

"No combat", with the environmental-boss protocol as the playable layer for protected/relief

chapters.** So the reconciliation direction is to AUTHOR the Ch-77 combat layer, not to flip the

registry cell. The precedent is already in the baseline: the CH_EPILOGUE waiver reads *"combat=Y

rides the endgame-sandbox playable layer (Ruling 1/R6 honored, no story-combat beat)."*

THE RESIDUAL CALL IS JOSH'S, and it is not "which cell wins" — that is settled above. It is **WHAT

Ch 77's combat layer IS** at the game's terminal revelation: an endgame-sandbox playable layer on the

CH_EPILOGUE precedent, an environmental-boss-protocol layer per the 2026-07-03 ruling, or an explicit

HL_0051 exception for the Revelation. That question touches a canonical hard line, so per the standing

decision protocol it goes to Josh as a full brief with a recommendation — not to a schema owner.

RECOMMENDED as the opening position: the CH_EPILOGUE playable-layer precedent, which honors R6 and

HL_0051 without putting a fight in the Revelation. The honest bright side: **resolving it toward

combat=Y does not disturb Pattern A** — Ch 77's intensity stays standard either way, so this

routes in parallel rather than blocking the pattern pick. It must still be answered before any

Ch 70-77 baseline re-emit.

C-E. Ch 71 enum drift, independently confirmed from both sides. The registry reads

dominant_combat_mode puzzle_routed; the spine's mechanical anchors read contest-routed AND build

their HL_0049 five-window argument on contest-routed. If the registry is right, the spine's own

compliance argument rests on a value that is not in the data. The charter already names this drift at

Ch 70 and Ch 71; the flag and relief axes confirm it from the registry side. Route with C-D.

**C-F. The waiver condition on the ascent — ONE waiver, REASON-stale not gate-stale, and it clears

as a PAIRED edit.** docs/topography_baseline.json waives puzzle=N at Ch 72, 73, 75, 77 and

learning=N at Ch 75, 76 as "substrate does not carry the beat." An earlier draft claimed two stale

waivers; re-derived against the live baseline and the gate's waiver logic (_wkey =

rule||chapters; stale = wkeys - {keys of observed violations}), the correct reading is:

learning. CH_77's beat modes carry no puzzle beat at all, and its registry cell already reads

learning=Y. The waiver is fully VALID.

violation stops firing. CH_73's R4 violation still fires off the registry column (puzzle=N), so

the waiver still matches its key. Re-derived: **19 R4 violations, 19 R4 waivers, exact key match,

R1/R2/R3/R5 all clean.**

now false — CH73_B03 is a puzzle beat. That is a reason-staleness, a real defect, and invisible to

the gate.

Removing the waiver without flipping the registry cell leaves an unwaived violation (RED); flipping

CH_73 puzzle=N → Y without removing the waiver leaves a stale waiver (RED). **The registry cell

and the waiver row move together in one commit, or not at all.** An earlier draft's "must clear

these in the same commit" was dangerous as written — it reads as a licence to delete waivers.

**C-G. The flag reallocation's sequencing collision with the substrate state — RE-SCOPED, and the

blocker is materially smaller than the earlier draft claimed.** All three newly-full-tier nodes have

named RCM substrate gaps, but the sharpest claim was FALSE and is corrected here.

"Annwn is MISSING ENTIRELY from the RCM Acquisition Priority Master List" is WITHDRAWN. Verified

directly this pass against `_source/01_Tier_1_Foundation/T1_Research_Corpus_Master [ACTIVE

v1.1.4].md. At L1927, inside the Acquisition Priority Master List, under the Ch 70 Hades and

Elysium` block: *"Davies Sioned trans — The Mabinogion Oxford World's Classics 2007 (T0_RC_01324) —

Welsh canonical mythological corpus Pwyll-Lord-of-Dyfed Annwn-otherworld-tradition primary;

Guest 1849 public-domain trans on Sacred Texts USB as fallback. Status: needed Tier 1."* That is

the Four Branches of the Mabinogi — the exact corpus Annwn needs — acquisition-listed, tiered,

ID'd and status-tracked. The earlier draft compressed the charter's careful NARROW claim into a broad

false one while citing the charter's positive control instead of running one.

The TRUE claim, verified this pass: the list has no Ch 73 per-chapter BLOCK. The blocks run

L1753 Ch 66 · L1794 Ch 67 · L1830 Ch 68 · L1870 Ch 69 · L1916 Ch 70 · L1935 Ch 71 · L1942 Ch 72 ·

L1948 Ch 74 — it jumps 72 → 74. Plus a per-chapter substrate SECTION at RCM L6752 flagged

SUBSTRATE GAP whose body is the displaced pre-v3 Realm-War unit. **POSITIVE CONTROL, run rather than

inherited:** the first heading sweep for ^### Ch (66|…|77) returned ZERO — because the headings

carry bold markers (### Ch 66 …). The corrected matcher returns all nine blocks above, and the

Annwn content search returns L1927, L6583, L6619 and L6752. A search that could not match had

reported zero; the control caught it. This is the repo's most-repeated defect class and it very nearly

landed on a load-bearing blocker for the second time in one wave.

RE-SCOPED CONSEQUENCE: the real gap is a missing per-chapter BLOCK and a displaced substrate section —

not an unidentified corpus. The primary is already named and tier-1 acquisition-listed. That is a

materially smaller and cheaper repair than "the corpus does not exist anywhere." The Ch 71 and Ch 72

blocks are the sharper gaps: both state in their own text that their real corpora (the Popol Vuh; the

Puranic and Vedic tradition) are *"not yet listed here."*

RECOMMENDED, unchanged in shape but lighter in weight: the Wave-3 substrate repair for Ch 71/72/73 is

scheduled as a predecessor to their Stage-2 batches rather than discovered at the factory — a Stage-2

batch cannot be pre-drafted WITH researched resolutions over a substrate gap. Under Pattern B this

hardens into a blocker. **And per the fifth-sitting ruling (D-4 below), the sequencing urgency is

lower than the earlier draft implied: none of this executes until the post-slice factory era.**

C-H. The charter's pre-ruling recommendation is superseded by the ruling, not by this brief.

Charter L1360-1362 recommends "do NOT change the intensity column; carry the directive on the DEMAND

and AWE axes instead." Pattern A changes one cell. The charter itself flagged the escape hatch in the

same paragraph — "a hard line that produces an anticlimactic door is a hard line worth a ruling rather

than a workaround" — and Josh issued that ruling. Pattern C is the charter's position, steelmanned and

carried forward so the supersession is visible rather than silent.

C-I. The three-concept conflation, which every option must survive. Three different things in this

repo are called "hard line" and their sets DISAGREE: the CVD Pillar-3 "Hard Line chapter" (cultural-

sanctuary combat routing; exemplars Ch 38/57/69; CVD L747 says Tier-C MMO does not extend into them),

the hard_line_status enum (9 NARRATIVE_PROTECTED / 4 CULTURAL_PROTECTED / 66 NONE), and

canonical_hard_line_flag (18 nodes, not a function of the other two). It is already causing live

confusion — docs/proposals/topography_rows.json L946 carries an open [DECISION NEEDED - JOSH]

because CVD §12.2 ties astral mode to "Hard Line chapter sanctuary preservation per Pillar 3" while

CH_66's hard_line_status is NONE and its canonical_hard_line_flag is TRUE. **Binding consequence:

any option that says "flag all road nodes" must first state which concept it is setting** — if the

concepts stay conflated, flag-all imports Pillar 3's combat absolute and the CVD L747 Tier-C exclusion

onto nine realms ratified as FOUGHT, colliding with HL_0121 and the don't-defang-monsters bar. That is

the strongest mechanical argument against flag-all and it is not a vibes objection. RECOMMENDED: a

one-paragraph disambiguation in CASCADE_RUNBOOK and the /spine-chapter skill, which also closes the

topography_rows.json flag as a definitional answer rather than a canon ruling.

C-J. Residual, outside the road. Ch 50 (Amazon / El Dorado) is `hard_line_status =

CULTURAL_PROTECTED and canonical_hard_line_flag = FALSE` — the only CULTURAL_PROTECTED node in the

arc that is unflagged. Same defect class as Ch 71/72, outside this wave's scope, worth a row in the

arc-wide pass.

C-K. The ascent's transition layer is stale, and one doc is stale WITHOUT a banner.

docs/spine/transitions/TRANSITIONS_CH72_77.md carries a self-declared STALE banner and the

pre-re-cascade topology (Cassius at Ch 74, the father at Ch 75). More dangerous:

TRANSITIONS_CH62_71.md carries NO staleness banner and its final seams still describe Ch 70 as the

Changed Fairy Realm and Ch 71 as Realm War Phase One. An unflagged stale doc is worse than a flagged

one, and this repo's own doctrine says a stale reference makes later agents hallucinate. The seam-camp

work (Clause 5) must re-base before it touches this range.

**C-L. THE UNSWEPT MIRROR CLASS — registry Y cells the beat graph does not carry, sitting on this

brief's own centerpiece node.** CH_74's registry row asserts character=Y and learning=Y. Its beat

modes are discovery, traversal, investigation, discovery, combat, puzzle, vril_management, combat,

transformation — and its own fold note says "seven distinct modes across nine beats," with **no

character beat and no learning beat.** R4 nevertheless passes on Ch 74 today, because it reads the

registry column, not the graph.

This is the same divergence class as C-D (Ch 77 combat/puzzle) and C-F (Ch 73 puzzle), and it exposes

a systematic hole in how the class has been hunted: **every sweep so far checked WAIVED-N cells

against the beat graph; nobody has ever checked Y cells against the beat graph.** A Y that the

graph cannot support is a silent false pass — the gate is green on data the content contradicts, which

is strictly worse than a red.

RECOMMENDED, as a NAMED REPAIR SWEEP rather than a note: one pass over all 79 nodes checking

mode_variety_check Y cells against each spine entry's beat modes, run alongside the §0.4

spine-vs-registry INTENSITY sweep (same artifacts, same read, one pass). Together they are the

complete registry↔spine reconciliation this arc keeps discovering one cell at a time.

IT ALSO UPGRADES THE CH-74 ASK. The CH_74 intimacy beat (C-B) is not merely a Clause-2 fix that buys a

terrace — **it repairs a live registry cell that currently asserts a character beat the chapter does

not contain.** Two defects, one authored beat, and the beat was already implied by the site row's own

declared character function. That strengthens C-B's "same commit" condition considerably.

---

4. AUTHORITY — what is Josh's, and what the loop rules once he picks

4.1 JOSH'S — the pattern pick and the three cells

The pattern PICK is his because it rewrites structural canon: the reviewed topography baseline for

Ch 70-77, and the replacement for the HL_0048 clause he removed. The fourth-sitting ruling is explicit

that the string is proposed, not set by hand — so the loop hands him a proposal and he rules it.

Four items, each with a recommendation attached rather than surfaced as a question:

1. THE PATTERN. A / B / C / D. RECOMMENDED: A, THE TERRACED ASCENT.

2. THE CH-74 CELL — a RECONCILIATION, not a change. The registry reads standard; the spine entry

at docs/spine/CH_74.md L205 already declares intensity PEAK. The ask is which artifact governs,

and the deciding artifact already chose. RECOMMENDED: ratify PEAK and bring the registry to its own

spine entry — conditional on (a) the CH_74 intimacy beat landing in the same commit (C-B, which

also repairs the live character=Y cell per C-L) and (b) the R3 re-scope landing in the same

commit (§5.7 — without it this single cell reds run_gates.py). Two riders travel with the

ratification regardless of pattern: repair the entry's false *"HL_0048 holds"* parenthetical, and

resolve the same divergence class at CH_66 (declares VALLEY twice, registry says standard).

3. THE CH-73 CELL. hold at standard, or lift to peak. RECOMMENDED: HOLD — this is the cell most

likely to be overturned and the evidence for holding is at §2 reason 2 and C-C.

4. THE FLAG REALLOCATION. canonical_hard_line_flag is structural canon the charter itself said

should be *"a deliberate ruling either way rather than letting the program's realm-parity work

quietly imply a change,"* and it changes JOSH'S OWN WORKLOAD — **three Stage-2 batches added at

Ch 71/72/73, and THREE dropped, at Ch 66/67/68.** (An earlier draft said five dropped, naming

Ch 70 and Ch 75 — both are already FALSE in the live registry, so they carry no Stage-2 batch to

drop and their move is a tier reclassification only. Net is −3/+3, not −5/+3, in the one line Josh

reads as his workload delta.) The constant-total headline SURVIVES and is verified: 18 TRUE before;

−66 −67 −68 +71 +72 +73 = 18 after; road 7 → 7. RECOMMENDED: the B+C reallocation, with the

CULTURAL column flagged provisional pending L-R3 (§1.5) and the two DOWNGRADES at Ch 67/70 ruled

either as L-R3-gated or as explicitly provisional.

Not a question, recorded as a consequence: adopting any pattern means the HL_0048 row's "its

replacement pacing pattern is under the realm-arc pattern design" pointer resolves, and the new law

gets its own hard-line row.

4.2 THE LOOP'S — everything downstream of the pick

Director-ruled and applied with the path shown, per the delegation doctrine (structure-fill inside a

locked vision; schema is not world canon):

awe conjunction that separates a vista you breathe at from a beautiful arena you fight in.

check_chapter_topography.py R7 for 3; the Fork-C seam gate for 5), and the fixtures.

beside the charter's INTENSITY BLOCK, seam_rest + register_from + register_to on the seam rows

the Fork-C pass is already going to create, two T0_Vril_Site_Registry rows for Ch 75/76 plus a

sanctum-class function on the five road rows. (Caveat for a reader checking IDs: VS_CH7677_001's

site_id spans 76-77, but its chapter field reads 77.0, so "Ch 75 and Ch 76 have no vril site row

of their own" SURVIVES — the ID is misleading, the claim is not. Adjacent and out of scope:

T0_Vril_Site_Registry carries an unbalanced-quote column shift, vril_function bleeding into

engineering_anomaly_flag / transformation_chapter_flag — worth its own repair row.)

ambient_threat, mode, boss and climactic_affordance, never declared — a derived predicate cannot be

satisfied by writing a word in a cell; a flag can, and a flag would be satisfied first and authored

never.

xibalba_descent_road (the grinning walk home the body prose already implies and no beat stages),

the CH_75 grief_country site split so the Movement-1 low-threat grief country stops sharing a row

with the war sites, the five seam camps.

exactly as L-R3 delivers the care tiering) and the per-node population.

register from unlabeled prose; if the critic names the same register twice, the declared column is

lying) and RB-ASC-RECOVERY (given a beat sequence with mode and boss hidden, name where a player

puts the controller down; a node where the critic cannot find a place is a Clause-1 failure the data

did not catch).

left to the tooth author: R3 judges only windows lying ENTIRELY within Ch 1-69 — a BLOCKING

predecessor to the Ch-74 cell, not a cleanup, since without it that one cell reds run_gates.py

(§5.7 item 1) — and the new ascent tooth owns Ch 70-77 exclusively; and the amended HL_0048's

valleys clause, which reads as

simultaneously scoped-out and not-named-as-removed, is dissolved by Clause 4 banning valleys in

Ch 70-76 outright — strictly stronger than the old clause, so the ambiguity never has to be argued.

predecessor.

identity, before the DR-2 freeze; and the lens-roster addition of a non-game structural comparator

cluster (Dante, dramatic structure, orchestral dynamics) — the roster currently has no structural

teacher that is not a shipped game, which is a gap by the roster's own governing principle.

ENGINE versus CANON, per the reuse standing order: the terrace predicate, the five gate rules, the

fixtures, the two critic rubrics, the generation_class-keyed flag mechanism and the ordinal-block

schema are ENGINE and transfer whole to game #2. The register values, the seam camps, the flag

assignments and every authored beat are CANON and do not.

---

5. THE MEASUREMENT — how the pacing instruments verify the pattern holds

Nothing in this brief is real until a tooth can fail on it. Five instruments, four of them arithmetic

over columns that already exist and are already parsed.

5.1 check_chapter_topography.py.

lying entirely within Ch 1-69. As CODED it has no scope at all: it walks every ordered node

(for i in range(len(usable) - 2)). **With Ch 74 → peak it fires peaks [CH_74, CH_75, CH_76]

unwaived and run_gates.py goes RED on the one-cell edit alone** (re-derived against the live

registry; today R3 violations = 0). An earlier draft glossed this as "the question never bites in

practice (Ch 70 stays standard)" — that is true only of the SEAM-STRADDLING windows (68-69-70,

69-70-71) and false of the intra-70-77 window, which is exactly where the tooth fires. The

gloss is struck: an implementer reading it would conclude no code change was needed. The re-scope

ships in the same commit as the cell, and the gate must DECLARE which rule it applied — an

undeclared scope choice is how this class of defect propagates.

for dominant_combat_mode. The same unpopulated-is-a-hard-fail discipline applies, so the register

column can never be green by vacuity.

greater than one; no valley and no relief; exactly one step-down, immediately followed by a peak.

5.2 check_spine_graph.py — Clauses 1 and 2 as CORPUS ERRORS. Reuses the existing corpus-error

mechanism that fires only when the full 79-node corpus is passed in one invocation. Every input is

already parsed: spinelib.py BEAT_FIELDS carries role, mode, site and boss; its site parse carries

ambient_threat, vril_density and climactic_affordance. No new gate file, no new parser, no new

fidelity-baseline exposure.

5.3 T-RLM-CURVE — the monotone demand tooth. demand_ordinal strictly increases Ch 70→76; Ch 77

is EXEMPT via the empty-string sentinel and the tooth SKIPS empty rows explicitly rather than coercing

them to zero. awe_ordinal gets its own arithmetic rule so the charter's principle stops being a

doctrine with no row: awe peaks where vril density reads anchor.

5.4 FIXTURES — because an unmutated rule was never armed.

kept as the live fixture until it is authored out, then replaced by a synthetic.

Cassius moral phase as a rest beat until the climax exclusion and boss none were added. Recorded

because a predicate that was never mutated was never armed.

5.5 THE CRITIC RUBRICS, with their bound declared on the rubric. The arithmetic checks that

DECLARED registers alternate; it cannot check that the prose DELIVERS them, and no report may inherit

that unbounded claim. RB-ASC-REGISTER and RB-ASC-RECOVERY (§4.2) are advisory, never gating, per the

existing ladder — the same construction as T-RLM-PARTITION checking completeness over a declared

vocabulary rather than over physics.

5.6 WHAT IS NOT MEASURED, stated so no report over-claims it. The flag reallocation has **no

SEMANTIC gate teeth and will not acquire any** — the flag has zero harness consumers today

(positive-controlled) and this brief does not propose arming it. Its verification is procedural: the

8-state spine_status machine, the Stage-2 batch existing for Ch 71/72/73, the quest rows carrying

author_in_loop_required, and the §6.8 audit running. A report claiming the flag pattern "holds"

must cite those four artifacts, never a green gate.

One precise correction to that framing: "no gate teeth" is true semantically and FALSE

mechanically. No gate reads the column's MEANING, but registry_fidelity.py hashes its BYTES (§5.7

item 4). So a wrong flag value ships green forever, while ANY flag edit — right or wrong — reds the

fidelity gate until the baseline is re-emitted. Those two facts point opposite directions and both

must be stated: the column is simultaneously unguarded against being WRONG and guarded against being

CHANGED silently.

5.7 THE COMMIT DISCIPLINE. In order, because the first item is the one that reds the gate on its

own:

1. THE R3 RE-SCOPE SHIPS IN THE SAME COMMIT AS THE CH-74 CELL. Non-negotiable and first. R3 as

coded has no Ch 1-69 scope, so the single cell fires peaks [CH_74, CH_75, CH_76] unwaived →

gate: FAIL. Cell without tooth = RED. (§1.1, §5.1.)

2. THE CH-74 INTENSITY CELL AND THE CH-74 INTIMACY BEAT LAND TOGETHER (C-B) — which also repairs

the live character=Y divergence (C-L).

3. THE CH-73 WAIVER IS A PAIRED EDIT, NEVER A LONE DELETION (C-F as corrected): the registry cell

and the waiver row move together, or neither moves. Either half alone reds the gate — one leaves

an unwaived violation, the other leaves a stale waiver.

4. THE BASELINE RE-EMIT IS NOT SCOPED TO Ch 70-77. Verified in registry_fidelity.py: both

intensity and canonical_hard_line_flag are declared added_columns, and the extension zone is

SHA-256 hashed with the baseline required to declare the exact hash — so ANY value edit forces

--emit-baseline. **The flag reallocation un-flags Ch 66/67/68, which is OUTSIDE Ch 70-77 and

would red the fidelity gate under the old wording.** Every registry touch in this program,

wherever it falls, re-emits docs/fidelity_baseline.json via

registry_fidelity.py --emit-baseline in the SAME commit.

5. run_gates.py runs UNPIPED.

6. C-D and C-E are answered BEFORE the baseline re-emit, not after — C-D now as a Josh-routed

hard-line brief (M-7 / C-D as re-scoped), not a schema-owner ticket.

---

6. WIRING — nothing here lands orphaned

CONSUMERS: harness/check_chapter_topography.py (R3 re-scope, R7, R8) · harness/check_spine_graph.py

(Clauses 1-2 as corpus errors) · the Fork-C seam-typing pass (Clause 5) · the realm-ruleset schema

concept at REALM_DESIGN_PROGRAM §2.3 (demand/awe ordinals) · /spine-chapter Step 2 and

CASCADE_RUNBOOK §6 (the flag reallocation's routing, plus the C-I disambiguation) · read_boards.py

§3.3 costume tooth — **PLANNED WIRING, not a live consumer: the tool is unbuilt and generation_class

is unminted (§1.5), so the two-lane convergence is a benefit that arrives WITH those two builds, not

one this reallocation delivers today** · T8 region pages and the audio

program (register_dominant as the key music_mood_tags derives from) · the QA-watching rubric

matrix (RB-ASC-REGISTER, RB-ASC-RECOVERY).

LENS INTERFACES: L-R5's register-variety slice owns Clause 3's closed enum as its named output ·

L-R4 owns the awe terrace's craft (Journey's apex-as-rest is literally the Ch-77 held-open frame) ·

hades-death-run-realm-identity owns the seam camp's rhythm · L-R3's named output settles the flag's

CULTURAL basis and Annwn's generation_class · L-R1 and L-R7 own the sovereign-ladder L3 reading ·

presentation-craft-anim-audio should carry ARC PACING into its charter before the DR-2 freeze.

Nothing here asks for a new lens; one non-game structural comparator cluster is PROPOSED for the

roster.

DOC HOME: this brief needs a docs/DOC_MAP.md row and a home under docs/proposals/ before it is

done. Whichever pattern Josh rules becomes a hard-line row of its own, replacing the HL_0048 pointer

that currently reads "its replacement pacing pattern is under the realm-arc pattern design."

---

7. Verification notes

registries/T0_Hard_Lines [ACTIVE v0.2]/Sheet1.csv, not from memory and not from the charter's

paraphrase — the charter itself records that this exact mis-attribution class was caught once before.

live string matches the charter's §6.2 statement, so the charter's reading is confirmed rather than

assumed. CORRECTION — that verification was ONE-SIDED and is now two-sided. The earlier draft

checked the registry against the CHARTER and never checked the twelve spine entries against the

REGISTRY; both siblings that read CH_74.md in full missed its declared intensity PEAK. The

spine side has now been read directly: docs/spine/CH_74.md L205 declares PEAK and

docs/spine/CH_66.md L212/L265 declares VALLEY, against standard in both registry rows (§0.4).

The remaining sixty-seven nodes are STILL unchecked on this column — hence the sweep ordered at

§0.4, paired with the Y-cell sweep at C-L.

(^### Ch (66..77) ) returned zero blocks — a false zero caused by bold markers in the headings.

The corrected matcher (^### \*\*Ch …) returns nine blocks (L1753/1794/1830/1870/1916/1935/1942,

then L1948 Ch 74 — no Ch 73), and a content search for Annwn returns L1927 (the Mabinogion

primary, Tier 1, inside the Acquisition Priority Master List), L6583, L6619 and L6752. The absence

claim the earlier draft carried is false; the narrow block-gap claim is true.

cited *"AAA_REARCHITECTURE_PROPOSAL L1377"* and *"L1375"* for the §19.8 realm-road quotes — that

file is 437 lines, so neither line can exist. The quotes are real and live in

docs/proposals/CVD_AMENDMENTS_DRAFT.md:41 (the "Trigger" paragraph) and :43 (the "Architectural

change" paragraph), verified verbatim this pass. The third co-cited receipt, "WS2 L102", IS correct

(AAA_REARCHITECTURE_PROPOSAL.md:102). Any downstream doc repeating the "receipts in three

artifacts" line must carry the corrected homes.

the "NOT set by hand" clause that governs §4's authority split.

roster L-R1..L-R7 read from docs/proposals/REALM_DESIGN_PROGRAM.md (L341-349, L1350-1372, L1497,

L1069-1147); the INTENSITY BLOCK's proposed columns confirmed at L508 — power_ordinal,

demand_ordinal, awe_ordinal, escalates_from, escalates_to, sovereign_distance, none minted.

music_mood_tags (populated 0 of 79) and no other register-bearing column. Positive control: the

same header read returns intensity, hard_line_status, mode_variety_check and

canonical_hard_line_flag, so the matcher demonstrably matches and the absence is real.

grep -rnF "canonical_hard_line_flag" harness/*.py returns zero rows while the same matcher over

the same files hits intensity (3 files) and mode_variety_check (2 files).

through CH_77.md, positive-controlled: the parser returns non-zero site and beat counts for all

twelve files, so a zero in any cell is a real zero.

canon file was edited producing it.

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