SLICE_EXECUTION_BRIEF_2026-07-27.md

pipelines/SLICE_EXECUTION_BRIEF_2026-07-27.md

THE SLICE-SCOPE BRIEF

Synthesis of three analyst passes (slice_scope_cost.md, slice_scope_case.md, slice_scope_reconcile.md)

into one director-facing brief, per the standing decision protocol: shapes → decoupling → recommendation +

strongest objection + falsifiable triggers → what is Josh's vs what executes.

Read-only pass. Nothing here is applied. Contradictions between the three analyses are surfaced, not smoothed.

---

0. STATE CORRECTION — READ THIS FIRST. THE QUESTION IS ALREADY ANSWERED.

The task framed Josh's line as *"thinking aloud — NOT yet a ruling."* That framing is stale. Verified

directly in the working tree:

docs/spine/DECISIONS_PENDING_JOSH.md L30-53 — ✅ RULED 2026-07-27 (fifth sitting):

"Option c. We will do 2-13 first then prologue and chapter 1 then will continue after vertical slice.

After ch 13 we will close with healers child joining and they're off on their travel — to be continued."

And L3-28 — ✅ RULED 2026-07-27 (sixth sitting), the Ch-1 ending sequence authored (the amulet morning,

the familiar's keep-away chase, the rift swallowing both — so the amulet crosses to Earth with the fall,

closing the CH_02 no-inventory question).

Why this was missed by two of three analysts, and why that matters. The file was modified at 09:56;

the last commit (7d1a4bb) is 09:15. The rulings are an uncommitted working-tree edit. The cost

analyst read the current file and reported the ruling. The case and reconcile analysts read a pre-ruling

frame and produced *recommendations against a decision Josh had already made*. This is the

verify-the-artifact-not-the-frame lesson firing again, on a one-hour-old edit.

Confirmation method (per the positive-control rule): my first grep -n "L3-26" returned zero. Rather

than trust the zero, I positive-controlled the file (grep -c "RULED" = 70 hits, file present, 220,943

bytes) and re-searched on content tokens — which found the ruling at line 30. A literal-token search that

cannot match had reported an absence. The rule held.

What this does to the brief's register

This is no longer a decision brief between four shapes. It is:

1. the execution baseline for the ruled shape, and

2. a brief on the three things the ruling genuinely left open — one of which is ship-defining.

The ruling itself orders exactly this: *"the slice-scope workflow's analyses re-point to the execution

re-plan… the systematic scope update lands as its own critic-gated wave, never a hand-scatter."*

The contradiction between the analysts, stated plainly and adjudicated

Adjudication: the two superseded recommendations are *not* discarded. Their analysis is the most

valuable material in the pack — the played-vs-told table, the two-gaps decomposition, the Josh-Gate-per-shape

read, and the slice-decoupling finding all survive the ruling intact and are load-bearing below. What is

dead is only their bottom line. And the reason it is dead is the most elegant fact in this brief:

The case pass's single strongest argument against shape (c) was that building the fairy realm first means

*"building the best thing in the game worst"* — before the art grammar exists to crown it. **Josh's own

sequencing refinement is precisely that objection's remedy**: build Ch 2-13 first so the concept-art

pipeline is *"down and enhanced and enriched and optimized"*, then design the fairy realm. He reached

the analyst's strongest objection independently and answered it inside the same ruling.

The collision the task asked to "work honestly" is therefore resolved by Josh, not open: the morning

realm-directive ordering (fairy realm after all nine mythological realms) is superseded in part — the

fairy realm designs and builds once, at slice-end; the mythological realms stay in the post-slice factory

era; the endgame fairy-realm return (Ch 75-77) extends the slice's build, so the sovereign kit and the

realm art bible are built once and consumed twice. That is a genuine efficiency, not a compromise.

---

1. THE SHAPES

Registers changed: (c) is RULED. (a), (b) are counterfactuals — retained because they make the ruled

shape's price legible and because (b) is a live *ship-boundary* option even though it is dead as a *slice*

option. (d)/(e) are not rivals to the ruling — they are the open sub-question the ruling did not close.

1.1 Cost bands — honest, and banded because that is what the evidence supports

Delay is stated in phase-equivalents against ROADMAP_POST_5090_TO_SHIP.md P1/P2/P3, never weeks — that

roadmap's §0 binds: *"Do NOT time-box to weeks; the phases gate on exit criteria, not calendar."*

The delta in one line (ruled shape): 3.75× the nodes, ESTIMATE 2.5-4× the maps, 9 region pages that do

not exist, 8 system debuts with no slice precedent, ~18 files of amulet integration unapplied, and one

node-pair whose terrain path has never been run once.

The three legs with zero measurement behind them (this is where the variance lives, and it is variance,

not expectation):

1. fairy_realm invented terrain. coordinates: N/A (dimensional). The deterministic

derive_region_bbox → COP30 DEM → WorldCover → import → dress → stage chain — the thing that made six

maps affordable and kept terrain *off* the critical path — has no input here. 100% authored geography

through a path the repo has never executed. Single largest unmodelled item in the plan.

2. The sovereign kit. A second, disjoint playable kit (Sovereign Vril Projection, Rift Sealing, Temporal

Sight, Familiar Communion) for a player character who appears once, at the game's highest presentation

bar — because it *is* the ceiling reference the other 75 chapters climb toward. Plus two encounter classes

(SEAL, ENDURE) the shipped HP-lethal/composure-break/withdrawal model cannot express today.

3. The art choice-board pipeline, whose pilots (Atlantis, the Duat) have not run.

Josh's build-order sequencing is the correct variance control on all three, and it is why the ruled shape

is affordable at all: 11 earthly chapters mature the pipeline, the presentation ladder, the W-SPACE grammar

and the QA belt *before* the one node-pair with no fallback consumes them.

1.2 The played-vs-told beat table

The authored emotional spine (third + sixth sittings), and where each beat actually lands.

P = played · T = told/implied · A = absent · H = hindsight recast of already-played material.

Read by column. (a) delivers the world-promise and the survival engine at full strength and zero of

the authored spine. (b) buys two rows and ships the gesture with the meaning removed — *"I gave away a

nice thing I've always worn"* instead of *the protection that cost my mother her life.* **(c) is the only

column that delivers the authored arc as authored.** The case analyst built this table to argue *against*

(c) on first-impression grounds — but the table itself is the strongest possible evidence *for* the ruling,

and read on the emotional axis alone the case pass concedes it outright: *"That is a better product than

anything the other three shapes describe. No honest analysis should pretend otherwise."*

1.3 Build-order verdict per shape

Two shapes the analyses raised and rejected, recorded so they are not re-proposed:

• Prologue-lite / greybox origin at the front. Rejected twice, independently. It establishes the fairy

realm's visual grammar *at a reduced bar*, so the real Prologue must then match a look already shipped —

the worst of both: it spends the first impression and fails to deliver the ceiling. Building it twice

costs more than building it once.

• Interior-only Prologue. The ceiling is expressed *as the realm's open atmosphere* — *"the light does

not fall from anywhere, it is simply in the air"*; *"he looks and sees his kingdom whole, every corner of

the long plain"*; the rift appears *"far out on the plain."* An interior scope cuts every faculty that

makes the ceiling legible. It is not a reduced Prologue; it is a different scene containing a birth.

1.4 Josh-Gate / P3 / ship implications of the ruled shape

• **The Josh Gate's surface changes from "AAAAA-polished Ch 2-5" to "AAAAA-polished Prologue→13" — 3.75×

larger.** The gate ruling itself is untouched; what it gates is not. See §2 — this is the crux.

• ROADMAP_POST_5090_TO_SHIP.md is stale in ~13 places, starting with its title line and its P1 entry

criterion (*"the slice's regions (flores_island, bali, sumatra_java, angkor_khmer)"* → 13 regions).

• PRE_5090_BUILD_PLAN_VOL2.md §A.0 is now false: *"the whole pre-5090 program has ONE target: CH 2-5

as a full playable vertical slice."*

• DEEP_BOOK_PROCESSING_CONTRACT §4.2 pilot scope moves with P1.
• The density scorecard bands must extend. check_slice_scorecard.py + QT-17/18/19 carry Ch 2-5

early-game targets. Ch 6-13 sit further up the curve; the Prologue is a *ceiling tutorial* and Ch 1 is

declared out of corpus scope — both need declared sentinels or the monotonicity tooth fires a false

regression (the same defect class already caught for Ch 77's exempt demand_ordinal).

• T0_Equipment_Registry is header-only, zero rows. The amulet is the first equipment row in the game.
• Fork C (tiered traversal) now applies across 12 earthly regions, not 4 — and Ch 5 Sumatra/Java → Ch 6

Sri Lanka → Ch 7 South India → Ch 8 Swahili Coast is an Indian Ocean crossing, the hardest instance of

the staged/open-dungeon ruling, never built.

---

2. THE DECOUPLING ANALYSIS — the definitional finding, and the crux of this brief

**The plan uses one word, "slice," for two different things. Every shape reads differently once they are

separated — and the ruling fixed only one of them.**

• THE JOSH-GATE SLICE — what Josh plays first, at P3-exit. Purpose: a ceiling judgement + the arrival

of the human care/feel backstop after a program that has run entirely without one.

• THE SHIP SLICE (EA base) — what strangers buy at P4. Purpose: a commercial and narrative entry.

Today they are welded: the gate sits at P3-exit → P4-entry, and *"his verdict gates the P4 Steam

listing."* Nothing requires them to be the same content, and their sizing arguments point in **opposite

directions**: a ceiling is tested by polish-per-chapter, not chapter count (small and deep), while a

purchase is judged on where it starts and stops (coherent).

2.1 MUST they be the same? — No. And they should not be.

What the ruling fixed: the SLICE (Prologue→13) and the BUILD ORDER (2-13, then Pro+1).

What the ruling did NOT fix: the ship boundary and the gate placement. Under Early Access — which

is itself ruled (D-STEAM-FORMAT) — slice ≠ ship-unit by construction. Both remaining questions are open,

and they are separable from each other.

2.2 The finding neither analyst reached alone

The cost analyst had the ruling but was costing. The case and reconcile analysts had the structural insight

but not the ruling. Composed, they produce a fact none of them states:

**Under the ruled build order, the fairy realm — the single highest-variance, never-run-path, grandest-

display item in the plan — is built LAST, which means the Josh Gate as currently placed delivers his

verdict on it AFTER it is finished.**

The Josh Gate exists so the human backstop's judgement lands on the game's quality. On every earthly chapter

it does. On the **one surface where his judgement is worth the most — his own authored material, the mother,

the King, the fall, at the grandest-display bar, on a path with no precedent and no fallback** — the current

placement guarantees his first contact arrives too late to redirect it. He would be reviewing the crown

jewel as a finished artifact rather than informing it.

That is a real defect in the gate's *placement*, not in the gate's *rule*, and it is invisible unless you

hold the ruling and the decoupling finding at the same time.

2.3 The consequence for the ship boundary

There is also a scheduling fact worth naming plainly: **if the EA base = the full 15-node slice, then first

revenue and first player feedback sit behind the fairy realm** — i.e. behind the one item with zero

measurement. And EA's own ruled rationale is *"chapters ship as updates; the EA feedback channel feeds the

care backstop + density curve"* — a loop whose entire value is getting real players onto a small surface

before the large surface is built. Shipping 12 earthly chapters before any feedback arrives spends that

loop on eight unvalidated chapters, four of them elevated-care and one HIGH (the /Gwi San, a living

Indigenous community).

Against which stands the case pass's sharpest commercial finding, and it is not soft: **the precedent class

for "ship a narrative game without its opening, add the opening later" is close to empty.** BG3, Kentucky

Route Zero, Telltale, Life is Strange all shipped strictly in order. Hades — the closest analogue and the

one that trained its audience to expect story to arrive — was never missing its beginning. The risk has a

specific, permanent shape: *"it starts in the middle and never explains why,"* written once at EA launch, in

the aggregate forever. EA reviews do not expire.

---

3. RECOMMENDATION

**The ruling stands, unamended. Three things execute on it, and one of them is a genuine amendment to a

different standing ruling — the Josh Gate's placement, not its rule.**

R1 — SPLIT THE JOSH GATE. Fire it twice. *(the load-bearing recommendation)*

• Gate 1 — at Ch 2-13 P3-exit, before the fairy realm is designed. The *ceiling and feel judgement*, on

11 AAAAA-polished earthly chapters. This fully honors the Josh Gate rule: he plays nothing that is not

AAAAA-polished, and Ch 2-13 at P3-exit is AAAAA-polished. What it buys is decisive — **his care/feel

verdict lands while the fairy realm is still designable**, and the pipeline is validated by the human

backstop *before* it is aimed at the crown jewel.

• Gate 2 — at full-slice P3-exit. The *origin judgement*, on his own authored material (the King, the

mother, the fall) — the content he most deserves to rule on from the chair. This gate keeps its existing

job: gating the P4 Steam listing.

This is the single highest-value change available, it costs nothing (the content is polished either way),

and it converts the fairy realm from *a thing Josh reviews* into *a thing Josh directs.*

R2 — SHIP BOUNDARY: EA base = the full 15-node slice. Do not cut early.

Recommended against the staged-early shapes (d)/(e), for three reasons that only hold under the ruling:

1. The IOU problem disappears entirely. The origin ships in the designed order, first-in-read-order, for

every buyer. The permanent review sentence never gets written. Under the ruled shape the hindsight-recast

consolation the case pass constructed is simply not needed — and it was always the weaker good: as

its own strongest objection concedes, *"nothing recasts a Prologue the player never played; there is no

memory there to reinterpret, only a gap that gets filled late."*

2. The gift lands provenanced on first play, for everyone. This is the whole point of Josh's question.

3. R1 already recovers most of what staging was for — early human feedback on a small surface — by

putting Josh in the chair at Ch 2-13 instead of at node 15.

Accepted cost, stated honestly: first revenue sits behind the fairy realm, and the EA feedback loop

starts later and larger than its own rationale prefers. R1 is the mitigation, not a cure. Trigger T3 below

is the reversal condition.

R3 — STAGE THE AMULET GIFT AT THE CH-13 DEPARTURE.

The ruling leaves this explicitly open (*"at the join or later"*). The AMULET_PROVENANCE_BRIEF §7 standing

recommendation is *"after the first real scare,"* with Ch-13 as fallback — but the first real scare is

outside the slice, so that recommendation now means *the slice ships without the beat Josh's own question

named as its endpoint.* Recommend the departure, not merely the join:

• Josh's question framed the slice endpoint *as the gift*; the ruling's own words are *"healers child joining

and they're off on their travel."*

• The departure IS the scare-threshold, which answers the brief's craft objection (*"the player gives

away the object before they have spent much time fearing for the person"*) rather than overriding it. The

gift at departure reads: *you are coming with me into danger — wear this.* That is the first fear, and the

gift is its expression.

• It gives the act-break an object as well as a moment, and it closes the coverage gap the brief worried

about (a non-combatant on live fields with no ward).

• Under R2 it lands with full provenance on first play. The guilt engine opens in the Prologue and closes

at the slice's last beat — a complete emotional movement inside one purchase.

R4 — PROMOTE P2.11 (character creation) TO THE FRONT OF THE BUILD ORDER. *(forced, not optional)*

Verified: P2.11 | Character creation — four selections only | BUILD-NOW — *"best-friend gender→healer's-

child companion, familiar species→slot-1 lock, appearance base, name"* — and it is **absent from the ranked

30. Under the ruled build order this becomes forced**, and the reason is a genuine trap:

**Ch 1 — where the four choosings are canonically woven — builds LAST. But Ch 2's opening beat needs a

chosen familiar, and Ch 13's Ntala needs a bound best-friend gender. So the entire 11-chapter earthly

build, and Gate 1, run on selections whose canonical home does not exist yet.**

Without P2.11 landed early, the QA loop plays 11 chapters opening on a bond the build cannot create — the

player grieves a creature they never chose — and Ch 13 ships a companion bound to a selection never offered.

The dividend: neither implementation is wasted. The selection surface is needed at the front; the woven-

as-childhood form is canon and lands with Ch 1 at the back; the selection surface then becomes the

returning-player skip path — the standard shape for a game that adds a prologue. The canon rule *"None

of it is a menu. All of it is a life"* is honored in the version that is canon, at the time that version

exists.

Also add now, while it is cheap: the prepend-safe entry architecture (chapter-select / new-game entry

so a front-prepended Prologue is reachable). SliceChapterId is a constant, NextChapterFor is a chain, and

saves are chapter-keyed — so this is a data-and-content operation today and an expensive one after players

have saves.

3.1 THE STRONGEST OBJECTION

*R1 amends the Josh Gate — Josh's own standing ruling — on a subagent's reasoning, and it does so by

splitting the one thing the gate was built to be: a single, clean, first contact with a finished game. The

gate's real content is not "Josh renders a verdict"; it is "Josh finally gets to play his game, whole." Gate

1 spends that. He will have played 11 of the 15 nodes before the origin exists, so his one irreplaceable

first-playthrough-in-years is now a partial build missing its beginning — which is precisely the experience

R2 argues no buyer should have. The recommendation refuses the IOU for strangers and hands it to Josh.*

*And R2 compounds it: it puts first revenue, first feedback, and the entire care backstop behind the single

item in the plan with zero measurement behind it. If the fairy realm's invented-terrain path goes badly —

and nothing bounds it — the ruled shape ships nothing at all, for a long time, and the EA feedback loop that

was supposed to de-risk eight elevated-care chapters never runs before they are built.*

Where the objection lands, honestly. It lands hard on R1's *framing* and not on its *substance*. The

answer is that Gate 1 must be positioned as a working director's pass, not as The Playthrough — and Gate

2 preserved as the whole-game first contact, in designed order, origin included. If that framing cannot be

held, drop R1 and keep the single gate; the rest of the recommendation stands without it. The objection's

second half is not answerable by argument — it is a real, unbounded risk, and it is what trigger T1 exists

for.

3.2 WHAT WOULD CHANGE THE ANSWER — falsifiable triggers

Each is a measurable event, not a judgement call. Any one firing re-opens the ship boundary.

3.3 The five places the ruled shape will hurt — ranked by risk, with mitigations that are all ungated

1. fairy_realm invented terrain — no DEM, no bbox, no precedent. → **Prototype it during the Ch 6-13

window.** Ungated, no model spend. This is T1's instrument.

2. The companion runtime debuting in the slice's terminal chapter — 12 chapters of soak for everything

else, one for ally AI. → Build and soak against a stand-in early.

3. Nine region pages that do not exist — the measured long pole of the slice was canon-side page

authoring, not terrain. EST 12k-42k lines, serial, critic-gated per section. → **100% ungated. Start now,

in parallel, on the three-subscription capacity Josh named.**

4. Map count, superlinear in spatial spread — four of eight regions are sumatra-class or worse

(central_africa_congo spans 1,331 km; south_africa 545 × 778 km + a SPREAD leg). → **Run the AOI

clustering per region early**; cheap, deterministic. T2's instrument.

5. The soak window — wall-clock, does not fan out, on a 3-4× larger matrix. → The QT-16 Gauntlet fleet is

exactly what the 5090's parallelism is for. This is the one item where the hardware buys throughput

rather than capability.

The calibration that cuts against panic: the pipeline generalization is a one-time cost already paid

(RG.1/RG.2/RG.3 turned three flores-hardcoded drivers into --slug/--chapter drivers; Ch 6-13 inherits it

free); terrain is deterministic, CPU/network-bound and fully parallel; the QT harnesses are

DataTable-driven, so test *authoring* cost is near-flat and only the soak matrix grows; all 8 spine

entries and all 32 boss rows for Ch 6-13 are already complete — zero authoring debt at the spine layer;

and ~70-80% of the earthly work is 5090-independent and can start before the hardware lands.

---

4. WHAT IS JOSH'S AND WHAT EXECUTES

4.1 Josh's — ship-defining, brief-and-present, never hand-applied

J-B, J-C and J-D are the three decision points. J-E is a confirm, not a decision. J-A is closed.

4.2 Executes without asking — creative-within-vision, apply and show the path

Per the ruling's own instruction (*"the systematic scope update lands as its own critic-gated wave, never a

hand-scatter"*), all of the following is execution, not invention:

1. The scope-update wave — re-point ROADMAP_POST_5090_TO_SHIP.md (title, P1 entry criterion, ~13 sites),

PRE_5090_BUILD_PLAN_VOL2.md §A.0, DEEP_BOOK_PROCESSING_CONTRACT §4.2, the window-plan ranks, and the

Josh-Gate surface definition. One critic-gated wave, full gate suite, DOC_MAP rows.

2. Density-scorecard sentinels for Ch 6-13 (further up the curve) and declared exemptions for the

Prologue (ceiling tutorial) and Ch 1 (out of corpus scope) — before the monotonicity tooth fires a

false regression.

3. The AOI-clustering pass for all 8 regions (T2's instrument). Deterministic, cheap, zero model spend.

4. The fairy_realm invented-terrain prototype (T1's instrument). Ungated.

5. Region-page authoring for Ch 6-13 — 8 pages, ungated, parallel, start immediately.

6. The env-profile rail — 80-104 empty T0_Region_Index cells across the 8 new regions.

7. Seed + civ-mark shelves for the 8 regions.

8. P2.11 + the prepend-safe entry architecture into the ranked queue (R4's mechanical half).

9. The ally-runtime stand-in soak (T5's mitigation).

4.3 What must NOT execute without a ruling

• The 18-file amulet integration — it touches two hard-line nodes (Prologue, Ch 1) with show-don't-tell

ABSOLUTE, a new mother row in T0_Character_Index, the first-ever T0_Equipment_Registry row, and the

highest-care scene row in the chapter. Full critic + reveal-discipline verification, and it consumes J-D

and J-E. The brief states of itself that nothing enters the spine until that reconciliation pass runs.

• Any gift-beat staging ahead of J-D.
• Any ward-causality application ahead of J-E.

---

5. THE ONE-PARAGRAPH ANSWER TO JOSH'S QUESTION

Yes — Ch 13 is the healer's child's join (Ntala, CHAR_0002, HL_0070), your memory of the number is exact —

and you already ruled the rest of it this morning: the slice is Prologue→13, built 2-13 first so the concept-

art pipeline is mature before the fairy realm is designed. That ruling independently reached and answered the

strongest objection any of the three analyses raised against it. What it left open is not the slice but the

ship boundary (does Early Access launch with all 15 nodes, or cut earlier and drop forward?), the **gift's

staging** (recommend: the Ch-13 departure — the departure is itself the first scare, so the gift reads *you

are coming with me into danger, wear this*), and one confirm on the ward-off-the-neck causality. The

recommendation is: ship the full slice, stage the gift at the departure, and **split the Josh Gate so you

play the 11 earthly chapters before the fairy realm is designed** — because that is the one surface where

your judgement is worth more as direction than as a verdict, and the current gate placement guarantees it

arrives too late to be either.

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