REALMARC_RELIEF_MECHANISM_2026-07-27.md

supernatural/REALMARC_RELIEF_MECHANISM_2026-07-27.md

THE FINAL ASCENT'S RELIEF MECHANISM — what carries recovery when intensity never drops

Read-only design analysis, 2026-07-27. Axis: THE RELIEF MECHANISM, one lane of the realm-arc

pattern fan-out ordered by Josh's fourth-sitting ruling ("we need a different pattern for these

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

lenses"). Scope Ch 70-77, the span from which HL_0048's no-three-consecutive-peaks clause was

REMOVED. This document authors no canon. Every number below is computed off the live spine entries

and the live registries, and the computation is stated so it can be re-run and refuted.

Grounded on: docs/spine/CH_66.md through CH_77.md (read in full for Ch 70 and by section for the

rest — body prose, site rows, beat graphs, ally rosters, mechanical anchors, seam blocks);

docs/proposals/REALM_DESIGN_PROGRAM.md §1.4, §2.5-§2.7, §4, §6.2; docs/proposals/

COMPARATOR_LENS_PROGRAM.md; docs/spine/DECISIONS_PENDING_JOSH.md (fourth-sitting block);

registries/T0_Hard_Lines (HL_0048 as AMENDED, HL_0049, HL_0050, HL_0051 — quoted from the live

rows, never from memory); registries/T0_Chapter_Index; registries/T0_Vril_Site_Registry;

harness/check_chapter_topography.py (rules R1-R6); harness/spinelib.py (the beat and site parse);

harness/check_spine_graph.py (the corpus-error mechanism); docs/COMBAT_PROGRAM_ADDENDUM.md §9 (the

vril-site sanctum ruling); docs/spine/transitions/.

---

0. The finding, stated once

The mountain-range law's valley was ONE instrument doing FOUR separate jobs, and the reason

removing the peak clause feels dangerous is that all four jobs disappear together. No single

replacement covers all four — verified, not assumed: the road's highest-awe node (Ch 74

Yggdrasil) has ZERO character or dialogue beats in its entire graph, which is the awe mechanism

failing the relationship job in built content.

The replacement is therefore not a mechanism but a GRAIN CHANGE. 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. Relief moves from the intensity

column (chapter grain) down to ambient_threat and climactic_affordance (site grain) and mode

and boss (beat grain) — three columns that ALREADY EXIST, are ALREADY POPULATED across all

twelve realm-arc nodes, and are ALREADY PARSED by harness/spinelib.py. The pattern is a rule over

data the corpus already carries, not a new authoring burden.

And the alternation that HL_0048 performed on the INTENSITY axis moves, intact in shape, to the

REGISTER axis. Intensity may now rise monotonically; REGISTER must alternate. That single

substitution is the answer to Josh's question, and the built prose already satisfies it on all

eight nodes — so the rule is a LOCK ON EXISTING QUALITY rather than a new constraint, which is the

strongest kind of rule this repo can adopt.

Recommended name: THE TERRACE PATTERN. Its measurable core is three qualifying terrace forms, one

corpus-spread clause, and one register-rotation clause. Under it, the built ascent scores 7 of 8

nodes passing, one hard failure (Ch 71 Xibalba), and one thin pass flagged (Ch 74 Yggdrasil,

awe-only). That is the worklist, and it is small.

---

1. What a valley actually did — the four jobs, separated

The replacement cannot be specified until the thing being replaced is decomposed. A valley (and

its HL_0050 subtype relief) did four things at once:

• JOB 1, PHYSIOLOGICAL. Input load drops. Combat density falls, the hands stop, attention

un-clenches. This is the job a player feels in the body and the one that decides whether hour

nine of a climb is exhilarating or numbing.

• JOB 2, EMOTIONAL CONTRAST. Affect resets so the NEXT peak reads as a peak. A peak is a relative

measurement; with no floor, peaks flatten into a plateau. This is exactly the monotone-solemn

slop mode the lens program names at cell 15 and the realm charter names at §4.2 ("five

consecutive death realms converge into the named monotone-solemn slop mode").

• JOB 3, RELATIONAL AIR. The room where relationships get authored. Companion beats need

un-pressured air; a found-family that only ever speaks during objectives is a squad, not a

family. Ch 73's borrowed year is the road's proof that this job needs REAL ROOM, not a line of

banter over a fight.

• JOB 4, STRUCTURAL LEGIBILITY. The player can TELL where they are in the arc. Topography is a

reading instrument as much as a pacing one; a monotone climb with no landmarks is

indistinguishable from a long middle.

Every candidate below is scored against these four. The scoring is the analysis; the mechanism

names are just labels for column sets.

---

2. The measurement — what Ch 70-77 already carries

Method, stated so the zeros are trustworthy. Site rows were parsed with a fixed pattern against the

schema-v3 site line (docs/SPINE_SCHEMA_v3.md line 56); beat rows against the schema-v3 beat header.

The parser was positive-controlled: it returns non-zero site and beat counts for every one of the

twelve files Ch 66-77, so a zero in any cell below is a real zero and not a matcher that cannot

match.

2.1 The per-node shape

The live intensity string for Ch 70-77 is standard, peak, peak, standard, standard, peak, peak,

standard — confirmed by direct read of T0_Chapter_Index, matching the charter's §6.2 statement.

2.2 The two curves nobody declared, and one hole

Two things are visible in that table that no doc in the repo states, and they are the analysis's

foundation.

• THE INTIMACY CURVE ALREADY RISES WITH INTENSITY. Character-plus-dialogue beat share across the

ascent: 20 percent, 20, 10, 33, ZERO, 33, 60, 50. Across the earlier road (Ch 66-68) it is 9, 10

and 33 percent, and T0_Chapter_Index carries character=N in mode_variety_check for Ch 66, 67

and 68 — the character mode switches ON at Ch 69 and never switches off again. The authors, one

chapter at a time, already discovered that as the ground gets harder the recovery has to come

from the people. Nothing named it, nothing measures it, and nothing protects it.

• THE ENVIRONMENTAL FLOOR ALREADY FALLS AWAY. Low-threat site share: 4 of 5, 0 of 4, 3 of 4, 1 of

4, 2 of 4, 0 of 4, 0 of 2, 5 of 5. From Ch 75 the world offers no safe ground at all, which is

correct — a war has no quiet meadow. So the two curves are INVERSE by construction, and that

inversion IS the pattern: relief migrates from the environment into the relationship as the

ascent climbs. The terrace pattern below is the formalization of a migration the corpus already

performed.

• THE HOLE. Ch 74 Yggdrasil has ZERO character and ZERO dialogue beats in a nine-beat graph. It is

the only node in Ch 66-77 with zero, and it is the DOOR HOME — the crossing, the last node of the

road, the charter's own "highest-leverage node in the program." The found-family is textually

present at the crossing (CH74_B09 actors read protagonist, CHAR_0002, CHAR_0017; the ally block

says they are "at the last door of the whole road with her") and they do not get one beat of

their own anywhere in the chapter. This is the single sharpest authoring gap the axis found.

---

3. The five candidate mechanisms, each verified against built content

3.1 RELIEF BY SANCTUARY — the strongest structural lane, and the one with real holes

WHAT THE SPINE ALREADY CARRIES. Every Realm-Road node from Ch 70 to Ch 74 has a canonical vril

site row: VS_CH70_001 through VS_CH74_001, all typed dimensional_threshold with the function set

substrate_perception; recharge_passive; unique. The waystation rhythm exists in the registry

already. Ch 77's VS_CH7677_001 is typed waterfall / recharge_ritual — the true sanctuary type, and

proof that the vocabulary for a sanctum already exists. The addendum §9 ruling makes this lane

canonical rather than inferred: loadout quick-change is performed diegetically "at a waterfall or

vril recharge zone (the vril-site becomes the build sanctum ... the well is save, recharge, AND

loadout home)."

WHAT NEEDS AUTHORING. Two gaps, both cheap.

• Ch 75 and Ch 76 have NO vril site row of their own. Both spine entries carry VS_CH70_001

re-scoped ("the fairy realm at the vril ceiling (VS_CH70_001 carried)"). So the arc's two

highest-intensity chapters — the war and the double climax — have no rest anchor row at all,

in a realm the prose describes as running at the vril ceiling. A ceiling-density realm with no

recharge site is a data gap, not a design decision.

• The road's five rows are typed dimensional_threshold — ARRIVAL GATES. recharge_passive sits

in their function list but nothing types any of them as a SANCTUM in the §9 sense. The

difference matters: a threshold you pass through is not a place you sit down.

JOBS COVERED: 1 strongly (it is the only mechanism that is player-PACED — the player decides how

long to rest, which is what makes it physiological rather than authored), 3 well (it is the natural

home for the found-family beat), 4 weakly. Job 2 not at all.

VERDICT: ADOPT as the pattern's structural spine. It is the cheapest of the five (two registry

rows plus a function-list amendment) and the only one that hands recovery to the PLAYER rather than

authoring it at them. It is also the one that generalizes to the engine and transfers to game #2.

3.2 RELIEF BY INTIMACY — already the ascent's real mechanism, unnamed and unprotected

WHAT THE SPINE ALREADY CARRIES. A great deal, verified in the prose rather than inferred from

counts.

• Ch 73's whole middle IS this mechanism: CH73_B02 (take Arawn's charge, actors protagonist plus

Arawn) and CH73_B05 (keep the borrowed year, "night after night ... until the cauldron warms for

her") are character-mode beats at arawns_court_feast_hall, ambient_threat LOW, vril HIGH,

affordance sacred. That site hosts five of the chapter's nine beats. Annwn is a sanctuary with a

duel at the end of it.

• Ch 72's summit pair, CH72_B08 and CH72_B09, sits at churned_sea_summit: ambient_threat LOW, vista

and sacred, no boss. B09 is a character-mode beat whose whole content is "look at the cup of

never-dying a long moment and know ... that it is not for the taking ... set the cup down full."

That is a textbook terrace, immediately after the chapter's climax fight, inside a PEAK chapter.

It is the proof that a peak chapter can carry a full recovery beat without touching its intensity

value.

• Ch 75 opens with the childhood-best-friend reunion in the grief-country (CH75_B03, the Ch-1

Selection-2 payoff, explicitly held in "the Movement-1 grief-country's tender register

(restraint)").

• Ch 76's back half is almost entirely this: six of ten beats are character mode; CH76_B09 and

CH76_B10 are boss-free character and transformation beats at the frozen court, the room the prose

itself calls "the room the family cannot touch."

WHAT NEEDS AUTHORING. Ch 74, comprehensively (see §6). Also Ch 71 (two intimacy beats, neither

adjacent to another rest beat, in the road's most hostile node) and Ch 72 (one, at 10 percent, the

thinnest share on the ascent, in a MAXIMUM-care node).

JOBS COVERED: 3 definitionally, 1 well when the beat is boss-free, 2 partially. NOT job 4, and not

job 2 reliably — an intimacy beat can be the highest-pressure moment in the game (Ch 76's frozen

court is character-mode throughout and is the hardest scene in the arc). This is why intimacy alone

is not the answer and why the terrace predicate below excludes climax-role beats.

VERDICT: ADOPT as the pattern's emotional spine, with the climax exclusion. It is what the corpus

already does; the job is to name it, measure it, and fill the Ch-74 hole before the natural-voice

rewrite wave touches this prose.

3.3 RELIEF BY AWE — real, already measurable, and structurally incapable of the relationship job

WHAT THE SPINE ALREADY CARRIES. climactic_affordance with vista and sacred values is a live,

populated column on every site row, and the ascent is rich in it: Ch 72 has it on 4 of 4 sites, Ch

77 on 5 of 5, Ch 74 on 3 of 4, Ch 76 on 2 of 2. The prose delivers: the tree "so vast it holds

worlds on its branches, whole countries hung in its crown like fruit" (Ch 74 B01); the churned sea

under Mount Mandara (Ch 72 B08); the reflective pool held open for as long as the player stands

there (Ch 77, "the moment held open under her, waiting, letting the standing be the whole of it" —

which is Journey's apex-as-rest, already authored, already player-paced).

THE SHARPENING THE AXIS ADDS. Awe is only RELIEF when it is un-pressured. Ch 66's

atlantis_ring_city carries sacred and climactic_geometry at ambient_threat HIGH — that is

SPECTACLE, not rest, and counting it as relief would make the instrument lie. So the measurable

predicate must be the CONJUNCTION: an awe affordance AND ambient_threat none-or-low AND a

non-combat mode AND no boss phase AND not a climax-role beat. That conjunction is what separates a

vista you breathe at from a beautiful arena you fight in, and it is checkable off existing columns.

The charter's own instruments align and are adopted unchanged: AWE is a third axis peaking at

anchor density and is never tuned as difficulty (§1.4).

JOBS COVERED: 1 (input load genuinely drops — you look, you do not press), 2 strongly, 4 strongly

(awe is the arc's best landmark instrument). NOT job 3, and Ch 74 is the empirical proof: three

awe sites, the road's biggest vista, zero relationship beats.

VERDICT: ADOPT as the pattern's contrast-and-legibility spine, with the conjunction predicate. Do

NOT let it be the sole qualifying form for any node — that rule exists specifically because Ch 74

currently is exactly that case.

3.4 RELIEF BY MASTERY — the weakest candidate for the road, the strongest unclaimed one for the crossing

WHAT THE SPINE ALREADY CARRIES, AND ACTIVELY REFUSES. The road's vril is authored DOWN, on purpose:

Ch 70 and Ch 71 read vril_density low at every single site (the deliberate trough — "the win by

cleverness, not vril-brute"), Ch 72 climbs out, Ch 73-74 run high-to-anchor, Ch 75 is the ceiling.

So there is no power surplus on Ch 70-72 to take a victory lap with; the mechanism is unavailable

by design for three of the eight nodes, and that design is good (it is what makes Xibalba's wit-win

land).

The one place it exists is the crossing into Ch 75, and the spine both stages it and then withdraws

it in one sentence: "the ceiling she has spent seventy-four chapters climbing back toward simply

given back to her the moment she is home ... But there is no time to stand in it." The single

largest power beat in the game gets a subordinate clause of refusal.

WHAT NEEDS AUTHORING, AND WHERE. Not inside Ch 75 — the urgency in that refusal is a genuine

authorial instinct and overriding it would flatten the war's opening. The correct site is the 74-to-

75 SEAM: the crossing itself is the power restoration, and the seam is where the restored ceiling

can be exercised once with nothing shooting back. That preserves Ch 75's urgency and pays the

seventy-four-chapter climb.

JOBS COVERED: 1 (a fight you cannot lose is rest for the hands), 4 (it is the arc's clearest

progress landmark). NOT 2 — a victory lap raises arousal, it does not reset it. NOT 3.

VERDICT: ADOPT NARROWLY, at exactly one place (the 74-to-75 seam), and do not attempt it on Ch

70-72 where the vril trough is load-bearing. Reported honestly: this is the weakest of the five as

a general mechanism and the analysis does not recommend building a system for it.

3.5 RELIEF BY REGISTER — doctrinally the most important, currently unmeasurable, and time-critical

WHAT THE SPINE ALREADY CARRIES. All eight body-prose blocks were read in full for this axis. The

registers, read off the prose rather than off any declaration:

• Ch 70 ELEGIAC-MOURNFUL. "the sorrow of the place is not cruelty but only loss."
• Ch 71 ADVERSARIAL-TRICKSTER, and the only comic register on the ascent. "she is grinning,

because for the first time on this whole descent the country of the dead did not take anything

from her that she did not give it on purpose to get more back."

• Ch 72 ASCENDANT-MORAL-ARCHITECTURAL. The country IS the climb; "a place for each deed and a term

to serve and then a way onward, nothing eternal, nothing for its own cruel sake."

• Ch 73 WARM-HOSPITABLE. "for the first time in longer than she wants to count, it takes her

somewhere warm."

• Ch 74 VAST-ELEGIAC-LABORING, with apocalyptic pressure at the edges. "a battle and a tending in

the same breath."

• Ch 75 GRIEF-AND-WAR.
• Ch 76 GRAVE-ARGUMENT resolving into INTIMATE-STILLNESS.
• Ch 77 WONDER-AND-RELEASE.

No two consecutive nodes share a register, and every five-node window carries five distinct ones.

The charter's feared failure mode at §4.2 — five consecutive death realms converging into

monotone-solemn — is NOT what the built prose does. That is a positive-controlled finding: the

prose was read, not sampled, and the registers are distinct.

WHAT NEEDS AUTHORING. The instrument. music_mood_tags on T0_Chapter_Index is populated 0 of 79

rows — every cell reads [POPULATE→T8_Region_Pages Section 10 AIVA composition briefs]. There is

no other column anywhere carrying emotional register. So the ascent's best-executed pacing property

is invisible to every gate, every critic rubric, and every QA instrument in the repo. It survives

today only because the eight chapters were authored close together by authors who could see each

other's work.

THIS IS TIME-CRITICAL. The natural-voice doctrine's rewrite pass is named in the standing memory as

"the next session's first wave," and it rewrites player-facing prose across the arc. A rewrite wave

that cannot see the register rotation is exactly how a rotation gets flattened — a reviewer

smoothing eight chapters toward one voice would pass every existing gate. Mint the register column

BEFORE the rewrite wave, not after.

JOBS COVERED: 2 definitionally and uniquely — no other mechanism does emotional contrast without

dropping intensity. 4 well. NOT 1 (a register change does not rest the hands). NOT 3.

VERDICT: ADOPT as the pattern's alternation axis, and mint the field this week. This is the single

highest-value, lowest-cost item in the analysis.

3.6 The coverage matrix — why the answer is a chord, not a mechanism

No row covers all four. Any two rows leave a job uncovered. The minimum covering set is

{sanctuary or intimacy} plus {awe or register} plus register — practically, the pattern must

require at least three instruments and must never let one node lean on a single one. That

requirement is Clause 2 below, and Ch 74 (awe-only) is the case that motivates it.

---

4. THE TERRACE PATTERN — the recommended replacement for mountain-range topography on Ch 70-77

4.1 The substitution, in one line

Mountain-range topography alternated INTENSITY between chapters. The final ascent lets intensity

rise monotonically and moves the alternation to REGISTER, while relief itself drops one grain — from

the chapter to the terrace, a low-pressure ledge INSIDE each rising node.

4.2 The three terrace forms

A TERRACE is a bounded, low-pressure passage inside an ascent node. A node qualifies by satisfying

at least one of three forms. Every term below resolves to an existing populated field.

• FORM T-A, THE SITE TERRACE (the environmental ledge). A site with ambient_threat none or low

that hosts two or more beats, at least one of them non-combat. Covers jobs 1 and 3.

• FORM T-B, THE INTIMACY TERRACE (the fire between realms). Two or more contiguous beats whose

modes are all in {character, dialogue, learning, discovery, transformation}, none of them

role climax, and at least one carrying boss none. Covers jobs 1 and 3.

• FORM T-C, THE AWE TERRACE (the overwhelming-beauty rest). At least one beat at a site whose

climactic_affordance includes vista or sacred AND whose ambient_threat is none or low, in a

non-combat mode, with boss none, and not role climax. Covers jobs 1, 2 and 4.

The climax exclusion and the boss none requirement 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 be the instrument lying about the most pressured scene in the game. This was

caught by running the predicate three ways and reading what it returned; it is recorded because a

predicate that was never mutated was never armed.

4.3 The built ascent, scored

Form spread across Ch 70-77: T-A four nodes, T-B six, T-C five. For reference the earlier road

scores CH_66 three forms, CH_67 two, CH_68 T-B only, CH_69 two — so the thinning is not new at Ch

70, it is a road-long trend that the ascent inherits and worsens.

4.4 The four clauses

• CLAUSE 1, THE TERRACE FLOOR. Every node Ch 70-77 satisfies at least one terrace form.

Currently: 7 of 8. The failure is CH_71.

• CLAUSE 2, THE SPREAD. Across Ch 70-77 each of the three forms appears on at least two nodes, and

no node's ONLY qualifying form may be T-C. The second half of that clause is what prevents an

awe-only ascent, and CH_74 is currently its one violation.

• CLAUSE 3, THE REGISTER ROTATION. No two consecutive nodes Ch 70-77 share a declared dominant

register, and every five-node window contains three or more distinct registers. This is HL_0049's

exact shape applied to a different column — deliberately, so it reuses R1 and R2's proven code

path rather than inventing a new rule form. Currently: PASSES on all eight, once the field exists

to carry the declaration.

• CLAUSE 4, THE SEAM CAMP. Each realm-to-realm seam carries a typed rest passage (see §4.6).

Currently: unbuilt, and the seam documents covering this range are stale (see §8).

4.5 What the pattern explicitly does NOT touch

Stated loudly because the most likely implementation error is a well-meant edit that reddens a

green gate.

• The intensity COLUMN IS NOT TOUCHED. A terrace is a sub-chapter object. Nobody writes

relief into a Ch 70-77 intensity cell: HL_0050 fixes the relief set at exactly Ch 27, Ch 52 and

Ch 57, and harness/check_chapter_topography.py rule R5 fails the gate on any other member. The

pattern adds no intensity values and removes none.

• HL_0049 IS UNTOUCHED. Both its clauses govern combat MODE, and the ascent's modes already clear

them. Nothing in this pattern proposes a mode change.

• HL_0048's SURVIVING CLAUSES BIND. The mode-variety and no-coasting clause holds across all 79

nodes per the amended row; the no-three-consecutive-VALLEYS clause is untouched everywhere, and

the peak clause still binds Ch 1-69.

• HL_0051's COMBAT FLOOR BINDS. A terrace is never a combat-absent chapter. R6 is never waivable.
• The REVIEWED TOPOGRAPHY BASELINE for Ch 1-69 is untouched — structural canon under the cardinal

rule, and outside this ruling's scope.

4.6 The seam camp — the structural addition, and the one that costs no hard line

The realm road has nine nodes and therefore eight seams, and the seams are the one place on the

ascent where recovery costs NOTHING against any intensity value, because a seam is not a node and

carries no intensity cell at all.

The ground for this already exists and is ruled. Fork C (Josh, 2026-07-27 second sitting) ruled the

transition segments PLAYABLE, NAVIGABLE areas, "compressed and enriched for fun," with the important

crossings becoming real staged or open dungeons, and ordered a seam-typing pass over all 78 seams.

Addendum §9 rules the vril site the build sanctum — "the well is save, recharge, AND loadout home."

Put those two rulings together over the road's four realm-to-realm crossings and the found-family at

the fire between realms is not a new invention; it is two ruled systems meeting at a site that is

already scheduled to be typed.

Scoping, honestly bounded by what the seams actually are:

• SEAM CAMPS APPLY to the four realm-to-realm crossings inside the ascent: 70-to-71, 71-to-72,

72-to-73, 73-to-74, plus the 69-to-70 entry seam. These are genuine medium-changing crossings.

• SEAM CAMPS DO NOT APPLY from 74 onward. Every seam from 74-to-75 through 77-to-Epilogue is typed

"within-realm, no travel-medium change" in the transitions layer; the finale is one continuous

ascent. Proposing a camp there would be proposing a scene the geography forbids. From Ch 75 the

relief burden falls entirely on T-B and register, which is exactly what the built prose already

does.

• THE ONE EXCEPTION IS THE 74-TO-75 SEAM, which is the crossing home and the vril ceiling's

restoration. That seam takes the MASTERY terrace from §3.4 rather than a camp.

Each seam camp is one typed row, not a set-piece: a vril-site sanctum (save, recharge, loadout, per

§9), player-paced, ambient_threat none, carrying at least one companion exchange and the register

hand-off from the departing node to the arriving one. Five rows total.

---

5. The measurable form — the pattern the QA and story instruments can read

5.1 Fields to mint, and one deliberately NOT minted

• MINT register_dominant on T0_Chapter_Index, a closed enum, populated for all 79 nodes (Ch

66-78 first). It carries Clause 3 and it becomes the KEY that music_mood_tags derives from, so

the audio program inherits a populated register axis instead of an empty stub — which partially

closes the T1_Audio_Spec gap the lens roster names at cell 12 and the realm charter claims at

§4.3 for the presentation lens. One column, one enum, one population pass.

• MINT two T0_Vril_Site_Registry rows for Ch 75 and Ch 76, and add a sanctum-class function to the

five road rows so recharge_passive stops being the only rest signal on a dimensional_threshold

row.

• MINT seam_rest plus register_from and register_to on the seam rows the Fork-C typing pass is

already going to create. Zero new artifacts; three columns on a table that does not exist yet, so

the cost is a schema line in a pass already ordered.

• ADD awe_ordinal beside the realm charter's proposed demand_ordinal in the

T0_Realm_Ruleset_Registry concept. The charter names AWE as a third instrument at §1.4 but

proposes no column for it and T-RLM-CURVE reads demand only, so awe is currently a doctrine with

no row. Same empty-string sentinel convention as demand_ordinal, and one rule: awe_ordinal peaks

where vril density is anchor, making the charter's own principle arithmetic.

• DO NOT MINT a terrace: true flag. The terrace must be DERIVED from 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. This is the

same reasoning that makes T-RLM-EXPRESSION reject a prose sentence.

5.2 The tooth

Two homes, chosen for reuse rather than tidiness.

• CLAUSES 1 and 2 land in harness/check_spine_graph.py as CORPUS ERRORS, reusing 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, and its

site parse carries ambient_threat, vril_density and climactic_affordance. Implementation is a

predicate over structures the gate already builds. No new gate file, no new parser, no new

fidelity-baseline exposure.

• CLAUSE 3 lands in harness/check_chapter_topography.py as R7, structurally identical to R1 and R2

with register_dominant substituted for dominant_combat_mode. The same unpopulated-is-a-hard-

fail discipline applies, so the register column can never be green by vacuity.

• CLAUSE 4 lands as a row-completeness check inside the Fork-C seam gate when that pass mints one.

5.3 Fixtures — must-fire and must-not-fire, because an unmutated rule was never armed

• Clause 1 must-fire: a synthetic node with every site at ambient_threat high, no vista or sacred

affordance, and no two contiguous non-climax rest-mode beats. The real CH_71 is the live

must-fire case and should be kept as one until it is authored out, then replaced by a synthetic.

• Clause 1 must-not-fire: the real CH_73.
• Clause 2 must-fire: a synthetic eight-node ascent whose every node qualifies only on T-C. The

real CH_74 is the live case.

• Clause 3 must-fire: two consecutive nodes carrying the same register_dominant.
• Clause 3 must-not-fire: the real Ch 70-77 string.

5.4 The critic rubric, with its 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. The delivery check is a critic rubric in the

realm charter's RB- family and it takes their proven shape:

• RB-ASC-REGISTER. Given unlabeled prose passages from two consecutive ascent nodes, name each

node's dominant emotional register. If the critic names the same register twice, the declaration

in the column is a lie regardless of what the gate says. Advisory, never gating, per the existing

ladder.

• RB-ASC-RECOVERY. Given a node's beat sequence with the mode and boss columns hidden, name where

in the chapter a player would put down the controller and feel the tension go out of their hands.

A node where the critic cannot find a place is a Clause-1 failure the data did not catch.

---

6. The authoring worklist, ranked by value

1. CH_74 — THE INTIMACY TERRACE AT THE DOOR HOME. Highest value item in the axis. The found-family

is textually at the crossing and gets no beat. Two candidate sites already exist and are already

low-threat: three_wells_and_norns_hall (low, sacred, functions investigation and learning) and

high_bough_door_home (low, vril anchor, threshold and vista, functions transformation and

character — the character function is DECLARED on that site row and no character-mode beat

uses it). Proposal: one boss-free character beat at high_bough_door_home immediately before

CH74_B09, the last quiet before the last door — which also converts CH_74 from awe-only to a

Clause-2 pass and gives the road's most important node the beat its own site row already

promised. This is structure-fill inside a locked vision and the director rules it.

2. CH_71 — THE ONLY HARD FAILURE. Xibalba has no ledge of any kind: zero low-threat sites, zero

qualifying rest windows, one dialogue beat in the Bat House. 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 invisible and which Clause 1 does not accept alone. Two routes, and the recommendation is

both: declare the trickster register in register_dominant so the mechanism is at least legible,

AND give the descent-road's return leg (xibalba_descent_road already hosts CH71_B10, the exit

beat, and is the chapter's only moderate-rather-than-high site) a companion beat on the climb

out — the grinning walk home with the found-family, which the body prose already implies

("she climbs back up toward the road she came in on, and she is still herself, and she is

grinning") and no beat stages.

3. THE REGISTER COLUMN. Mint and populate register_dominant BEFORE the natural-voice rewrite wave

touches this prose. Cheapest item, most fragile property.

4. CH_75 — THE SITE HONESTY FIX. grief_country is typed ambient_threat moderate and hosts the

arrival, the restraint beat and the childhood-friend reunion — all in Movement 1, which the

prose places explicitly "before the war closes over everything." Splitting the site (a

Movement-1 low-threat grief-country with vista and sacred, and the war sites at high) makes the

data match the prose and converts CH_75 from intimacy-only to a three-form node. Grounded

structure-fill; cite the prose line.

5. THE TWO MISSING VRIL SITE ROWS (Ch 75, Ch 76) plus the sanctum function on the five road rows.

6. THE FIVE SEAM CAMPS, folded into the Fork-C seam-typing pass rather than run as their own wave.

7. THE 74-TO-75 MASTERY TERRACE — the ceiling exercised once, in the seam, before Ch 75's urgency

takes over.

---

7. Adversarial pass — the strongest objections, and the honest answers

• "You have replaced a machine-checked hard line with a vibe in a column." Partly true, and the

bound is declared rather than hidden: the tooth checks DECLARED register alternation, not

delivered register, exactly as T-RLM-PARTITION checks completeness over a declared vocabulary

rather than over physics. The delivery check is a critic rubric and it is advisory. What is NOT a

vibe: Clauses 1 and 2 are pure arithmetic over four existing populated columns, with no human

judgement anywhere in the predicate. The pattern is arithmetic first and declaration second, which

is the right way round.

• "Adding rest to a monotone climb will deflate the ascent." The counter-evidence is Ch 73 itself,

which was authored as an inter-chapter relief and which the spine credits with SETTING UP Ch 74

("the reserved crossing peak that Ch 73's relief set up"). The road's own authors discovered that

the crossing lands harder because something warm came before it. The terrace is that discovery at

a smaller grain, and the charter's own resolution agrees: felt intensity is DEMAND relative to

POWER, and a terrace lowers neither.

• "With the peak clause gone, why not simply lift Ch 73 and Ch 74 to peak and be done?" Ch 74, yes —

that is the sibling lane's call and this axis supports it. Ch 73, NO, and the evidence is direct:

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

nine beats, and one of only four nodes carrying all three terrace forms. Lifting it deletes the

strongest terrace on the road and collapses the register rotation at exactly the point the road

most needs contrast. The removed clause frees the DOOR to rise; it does not require the last warm

room to.

• "Ch 76 fails T-A and T-C, so the pattern says the arc's double climax is under-relieved." It

passes Clause 1 on T-B (CH76_B09 and B10, boss-free, at the frozen court, the room the prose

itself calls untouchable) and that is the correct reading — Ch 76's relief IS its intimacy, six of

ten beats, and it should not have a meadow. An earlier draft of the predicate tried to pass Ch 76

by treating the site row's protected: true as a sanctuary marker; that was rejected on

inspection, because protected tracks §17.1 care-protection across the corpus (Ch 72's

maximum-care Hindu sites carry it, Ch 67's vault carries it) and overloading it would make one key

mean two things — the same error the realm charter refuses when it declines to extend

T0_Environment_Grammar_Registry.

• "Five instruments is over-engineering for one pacing question." The matrix at §3.6 is the answer:

the built corpus already demonstrates that awe alone leaves a node hollow (Ch 74) and that

environment alone runs out at Ch 75. Three instruments is the minimum covering set, not a

preference.

• "The seam camp assumes a playable seam that does not exist yet." True, and it is scoped

accordingly: the seam camp is a CLAUSE ON A PASS ALREADY ORDERED (Fork C's seam typing), not a new

workstream, and if that pass slips, Clauses 1 through 3 stand alone and still cover all four jobs

on seven of eight nodes.

---

8. Adjacent findings surfaced by this axis, reported not acted on

• THE ASCENT'S TRANSITION LAYER IS STALE. 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 encounter

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. Any seam work on this range must re-base first.

• CH_77 COMBAT DIVERGENCE. T0_Chapter_Index carries combat=Y for CH_77 in mode_variety_check and

R6 is explicitly never waivable, while the CH_77 spine entry states "Combat is refused, by design

— there is no fight and no fail state at the Revelation" and its beat graph contains zero

combat-mode beats. The gate is green on a registry value the spine contradicts. Relevant to this

axis because Ch 77 is the arc's terminal relief and the combat floor is exactly where relief and

HL_0051 collide. Not this axis's to rule; it belongs to whoever owns the topography-registry

reconciliation.

• CH_71 ENUM DRIFT, CONFIRMED. The registry reads dominant_combat_mode puzzle_routed; the spine's

mechanical anchors read contest-routed and build the HL_0049 five-window argument on

contest-routed. The realm charter already names this drift at Ch 70 and Ch 71; this is

independent confirmation from the registry side.

• R4 THINNING ON THE ASCENT. The live mode_variety_check carries puzzle=N at Ch 72, 73, 75 and 77

and learning=N at Ch 75 and 76 — all currently waived in docs/topography_baseline.json as

"substrate does not carry the beat." Two of those are now stale as waivers: CH_73's beat graph

contains CH73_B03 in puzzle mode, and CH_77 contains CH77_B04 in learning mode. A stale waiver

FAILs the gate by design, so this is worth a look from the registry-reconciliation owner.

• CH_72 INTIMACY THINNESS. One character beat in ten, in the road's MAXIMUM-care node, which the

charter also flags as "the highest-care realm on the road that has had the LEAST attention." The

two observations are probably the same observation.

---

9. Wiring — nothing here lands orphaned

• CONSUMERS. harness/check_spine_graph.py (Clauses 1-2 as corpus errors);

harness/check_chapter_topography.py (Clause 3 as R7); the Fork-C seam-typing pass (Clause 4); the

realm-ruleset schema concept at REALM_DESIGN_PROGRAM §2.3 (awe_ordinal beside demand_ordinal);

the QA-watching rubric matrix (RB-ASC-REGISTER, RB-ASC-RECOVERY); T8 region pages and the audio

program (register_dominant as the key music_mood_tags derives from).

• LENS INTERFACES. L-R4 pilgrimage-traversal-and-awe owns the awe terrace's craft — Journey's

apex-as-rest is literally the Ch-77 held-open frame and the charter already dual-lists Journey

there. The L-R5 REGISTER-VARIETY slice owns Clause 3's enum and should deliver the closed value

set as its named output, exactly as L-R3 delivers the care tiering. Hades's realm-identity lens

owns the seam camp's rhythm. Spiritfarer under L-R5 owns the Ch-75 grief terrace. Nothing here

asks for a new lens.

• DOC HOME. This axis's output belongs in the realm-arc pattern brief the fan-out synthesizes, and

that brief needs a DOC_MAP row before it is done.

• ENGINE VERSUS CANON, per the reuse standing order: the terrace PREDICATE, the two gate rules, the

fixtures and the two critic rubrics are ENGINE and transfer whole to game #2. The register values,

the seam camps and every authored beat are CANON and do not.

---

10. What is Josh's, and what the director rules

DIRECTOR-RULED, applied and shown rather than surfaced (creative-within-vision, structure-fill

inside a locked vision): the terrace predicate and its three forms; the four clauses; the two gate

homes; the field set including the decision NOT to mint a terrace flag; the CH_74 intimacy beat; the

CH_71 climb-out companion beat; the CH_75 site split; the two vril-site rows; the seam-camp scoping

to five crossings; the two critic rubrics.

GENUINELY JOSH'S, and each already inside a brief the sibling lanes are writing rather than a new

question: whether Ch 73 stays standard (this axis recommends YES, with the evidence at §7) and where

the new intensity string lands, which is the intensity lane's brief, not this one. This axis

contributes one constraint to that brief and no question: whatever the string becomes, Ch 73 is the

ascent's anchor terrace and should not be lifted.

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