FORKC_ANALYSIS_DESIGN_PRECEDENT_2026-07-27.md

pipelines/FORKC_ANALYSIS_DESIGN_PRECEDENT_2026-07-27.md

Fork C — Design Precedent + Our Canon Spaces

Axis: how shipped AAA does playable transitions, and which of OUR canon spaces are born for it.

Read-only analysis, 2026-07-27. Every load-bearing claim cites the artifact it was read from.

---

0. The four facts that reframe the question before any design starts

FACT 1 — There is no cover to defend. It was never built.

C:/dev/Humanity/Humanity/Source/Humanity/Private/Quest/HumanityQuestSliceDriver.cpp L356-365,

PresentTravelCard(), is a UE_LOG and nothing else. Its own comment calls itself "the

honest-minimal travel presentation (the ship-crossing scene-card)" and defers "the full letterbox

travel system … a held title card over the crossing" to W-SPACE. Sunk cost = zero. Josh is not

asking us to tear out a shipped cover; he is asking the question at the only moment it is free.

FACT 2 — The cover was never RULED. The ruled design already says "playable scene".

docs/TRANSITION_TREATMENT.md Layer 2 (Josh, 2026-07-08) states the inter-chapter seams are

"held separate from the chapter bodies because **in the 3D build these are their own arrival and

departure sequences**, not chapter-body prose," and that the consumer is "the 3D build (each seam

maps to a discrete arrival or departure sequence asset)." The transition-screen item

(PRE_5090_BUILD_PLAN_VOL2.md L675, U11.26 "loading screen + hitch-masking") is an *engine

default that drifted in under PG-2*, not a design ruling. **Josh's challenge restores the ruled

design; it does not overturn one.** This also explains why gap-register entry 29 flags the

presentation doctrine as "never ruled — named a Josh gap twice" (window rank 19).

FACT 3 — 78 seams is the wrong cost number. The real number is 3 rigs.

78 authored seam entries across docs/spine/transitions/*.md. Classified by their own

- Passage. lines:

Two conveyances (Ship, Vimana) plus one corridor kit cover all 78. Per-seam marginal cost

after that is *content* — a captain persona, three companion lines, a coastline silhouette — not

engineering. That amortization is the entire affordability argument, and it is the same logic the

rest of the build already runs on.

FACT 4 — Canon already reserved the transition space, by name, at the hardest seam.

docs/spine/CH_05.md L97: the Malacca Strait / Kra isthmian portage / Gulf of Thailand zone is

typed as "**a traversal connector for the excursion, traversed at the top of CH05_B07 (hosts no

primary beat); L2**." Canon deliberately authored a beat-free traversal space exactly where the

engine's largest map hop lands (AngkorKhmer.umap = 377 MB, the biggest asset in

Content/Maps/ by 4.7×). The place we most need to hide a stream is the place canon already told

us to build a journey through. That is not a coincidence — it is the walked-geography doctrine

doing its job.

---

1. The precedent set — what it masks, the verb during the mask, the content cost

Ordered by cost-per-seam, cheapest first. The column that matters most is the verb.

The one rule the whole precedent set agrees on

The mask must carry a RELATIONSHIP, or it is worse than an honest load screen. Every success

above pays its transition in companion voice — Mimir's stories, BD-1's chirps, Peter's phone

calls, Black Flag's shanties. ME1's elevator failed on precisely this axis, not on duration.

That rule is the strongest argument for playable transitions in *this* project specifically,

because our seams are already written as relationship scenes. TRANSITIONS_CH02_21.md gives

every seam a Transition sequence line and a Continuity line tracking, per seam: internal age,

the familiar roster riding along, concealment state, the Sealed Letter, and — the load-bearing one

— "the accumulated sense of travel … and the emotional carry-over" (L19). From Ch 13 the

healer's child is "the continuous traveling companion across every seam" (L26). Ch 4→5 already

names the ship captain, I Wayan Dwija (L62); Ch 5 already names Palembang Port Master

Nakhoda Iskandar (CHAR_0037), rowed explicitly as "the passage onward."

**We have already paid the expensive half of the GoW model — the authored relationship content,

per seam, for all 78 — and a cover screen throws every bit of it away.**

---

2. Our canon spaces that are BORN for playable transition

Ranked by how little invention each requires.

1. The ship / Watercraft — the seams ARE sailing legs in canon.

T1_World_Geography_Master §11 L993-997: Act 1-2 (Ch 2-37) is "walking, riding, river and coastal

sailing"; Ch 36-38 is "sea voyage by sailing ship"; Ch 39-54 is "walking, sea, no flight." Every

one of the ~26 water seams is *already a voyage the protagonist takes*. The engine's two live

cross-map seams are both water: IsCrossMapSeam() returns true only for CH_03→CH_04 (Flores→Bali,

"the range's first water leg," TRANSITIONS L55) and CH_04→CH_05 (Bali→Sumatra, "a captained sea

voyage," L63). **The playable transition is not something we would add to canon; the cover is

something we would subtract from it.**

*Guardrail (CARDINAL C1, DOMAINS_VEHICLES_AND_ENRICHMENT_BACKLOG.md L31): the system is NEVER

called "Vessel" — that is the House-of-Velheim antagonist-agent term. Use Watercraft

(system/registry) or Ship (player-facing).*

2. The deck's verbs are already specced. The Domains doc §(a) L112-113 already rules the

familiar-deployment matrix on a moving craft: "**Flying familiars wheel around a stalled/slow

ship … retreat to the Personal Dimension, re-emerge when slow/idle; Ground familiars ride on

the boat and guard/fight around it**." That is a deck-verb design that exists in canon today and

currently has no home to be expressed in. A cover screen orphans it.

3. The vril-site sanctum — the departure/arrival anchor, free.

COMBAT_PROGRAM_ADDENDUM.md §9 L157-161: loadout quick-change happens diegetically "at a

waterfall or vril recharge zone — **the vril-site becomes the build sanctum … the well is save,

recharge, AND loadout home**." This is structurally Elden Ring's site of grace, which is also that

game's travel node. Anchoring departures at the vril-site means the transition **begins in a room

the player was going to stand in anyway**, and puts a save point exactly where a level transition

needs one. Zero new space. (It also cleanly solves save-on-a-moving-platform — see §5 risk 2.)

4. The journal — the diegetic loading surface we already own.

The healer's child's evidence-journal goes live at Ch 13, is "a hunted and encrypted plot object by

Ch 17-18," and is "read as a canon in miniature by Ch 21" (TRANSITIONS L28); it becomes legal

evidence (Ch 50), is publicly released (Ch 56), and becomes shared after Ch 58. **This is the

single most canon-legitimate loading surface this game has**, W10 already ships book screens, and

it turns every unavoidable wait into an L2 completionist surface the player is *reading with

agency*. See §4 — this is the universal floor that means no seam ever gets a bare cover.

5. The Vimana (Ch 55+) — the transition becomes flight, and the vehicle is a required build.

WGM §11 L997-1001 + §7 L764: from Ch 55 the Vimana enables free travel, underwater navigation

(Ch 64), and dimensional-space mode across the Ch 66-69 realm road. The Domains doc rules it a

home base (§1). ~27 of the 78 seams happen *inside a space we must build regardless*. For

those, "playable transition" costs nothing beyond what the Vimana already owes.

6. Canon's own beat-free traversal connectors. CH_05's Kra portage zone (§0 Fact 4) is the

proof-of-concept, and the coverage architecture's SPREAD mechanism generates more of them: every

SPREAD "rides a route the outbound arc already walks" (WGM L67). These are pre-authored

elevator-equivalents with cultural content on them.

7. Ch 70-77 need no transition at all. Nine seams read "Within-realm — no medium changes and

no journey" / "No spatial move — the SAME grief-stilled hall." **A cover screen at the Ch 75→76

seam would be an active canon defect**, cutting a continuous space in half at the emotional climax

of the arc.

8. Traversal abilities already exist. T1_Ability_Tree: Soaring Glide (T4, "sustained-drain

horizontal-glide chapter-spanning traversal"), Breath Hold (T2, for "underwater and

toxic-atmosphere chapter regions"). The traversal spine is already rooted in the ability tree.

---

3. The three coherent shapes, costed against the factory

Shape A — "The Ship Is The Seam" (full playable vessel leg per water seam)

Build the minimum-viable Watercraft now: a steerable craft, ocean, buoyancy, boarding/disembark,

naval encounters. Every water seam is a real voyage.

• Pro: maximum canon fidelity; the walked-geography doctrine fully expressed; retires the DEV

teleport (rank 21) and the transition cover (rank 19) in one build; amortizes across ~26 seams.

• Con — and it is decisive: this is a vehicle *subsystem* (water rendering, buoyancy, camera,

collision, mid-water save/restore, stuck-recovery), and **Josh already ruled the Watercraft

"capture now, build after 79"** (DOMAINS doc L102-103, ruling 2026-07-10). Building it now to

solve the slice's two seams inverts his own sequencing, on a Steam-Deck floor, before the P0

benchmark harness (window rank 1) even exists to measure it.

• Verdict: REJECT as the near-term shape. Correct destination, wrong decade.

Shape B — "The Threshold Ritual" (playable bookends, voyage elided)

No conveyance. Play a short departure (walk to the quay, board, the last look back) and a short

arrival (walking up the beach into the new region). The voyage itself is elided by a diegetic

time-pass, not a screen.

• Pro: cheapest thing that is honestly "playable." Fits the existing Layer-2 artifact almost

verbatim — every seam entry already has a Departure. and an Arrival. line. No vehicle, no

moving-platform physics, no ocean. Reuses region geometry we are building anyway (a quay is a

site, not a rig).

• Con: the elision has to land *somewhere*, and if that somewhere is a card, Josh's challenge

is only half-answered — the load itself is still a cover. Also risks a hard tonal cut on the

long legs (the Bay of Bengal, the Atlantic crossing) where canon says months pass.

• Verdict: KEEP as a tier, not as the whole answer. It is the right treatment for short water

hops (the Sri Lanka strait) and it is the correct *frame* around every other tier.

Shape C — "The Connector Tier" (hybrid: the mask is chosen per seam) — RECOMMENDED

A four-tier seam taxonomy. Each seam is typed once, in the transitions doc it already lives in,

and the type dictates the treatment. The player never sees a non-diegetic cover at any tier.

The critical piece is T2's definition: an ON-RAILS DECK, not a vehicle.

The ship follows a spline. The player has full character control *on the deck* — walk it, talk to

the captain (I Wayan Dwija is already cast), talk to the healer's child, interact with the familiars

per the deployment matrix already specced, gather from a crate, open the journal, watch the coast

go by — and cannot steer. This is GoW's boat and Spider-Man's train, not Black Flag.

It needs: a moving platform with reliable character attachment, a spline, a cheap water/sky

backdrop, and NPC animation slots. It does not need buoyancy, an ocean sim, steering, naval

combat, or the Trade-10 build/repair ladder. Order-of-magnitude below Shape A.

And it is a harvest, not a duplication. When the real Watercraft lands post-79 per Josh's

ruling, the deck rig *becomes its interior*, and the T2 skin becomes the ship you built — which is

exactly the engine-as-reuse-machine posture (memory: engine-agnostic-reuse-goal). Shape C does

not spend Shape A's budget early; it builds the half of Shape A that is a *space* and defers the

half that is a *simulation*.

---

4. The sleeper recommendation: the journal is the loading surface

Even the best transition design has a residual — the 377 MB Angkor stream on a Steam Deck will not

always fit behind a corridor. The residual must never be a splash card. It should be the

healer's child's journal.

• It is canon, live from Ch 13, a plot object, and the arc's emotional through-line.
• W10 already ships the book-screen UI; the marginal engineering is small.
• The player is *reading with agency* while the world loads — which is the honest, literal answer

to "why does it need a cover instead of something playable."

• It converts every unavoidable wait into an L2 completionist surface and a place the seam's

authored "emotional carry-over" (TRANSITIONS L19) can actually be delivered.

• It scales to all 78 seams as a universal floor, so no seam ever degrades to a bare cover —

including the ones where the design fails, the hardware is slow, or a player skips.

Design rule to hand the engine lane: **the journal opens on any transition that exceeds the

corridor budget, and closes when the stream completes.** Skip (per RDR2's model, available on any

seam after its first traversal) goes to the journal — never to a screen.

---

5. Honest cost, risk, and the strongest objection

Engineering spikes required before committing (each small, each answerable without a 5090):

1. Moving-platform character attachment in UE — the classic problem. Two known solutions

(base-relative movement; or move the world past a stationary deck). One spike settles it. *This

is the single technical gate on T2 and it should be spiked before anything else in this fork.*

2. Save on a conveyance. PG-2 persists ws.* at the seam (PrepareCrossMapSeam, L367-405).

Restoring a player onto a moving platform is a real bug farm. **Mitigation, canon-supported:

conveyances are non-save zones; the vril-site sanctum on either side is the save anchor** —

which is what canon already says the well is for. Clean constraint, no invention.

3. Streaming budget. A connector or deck must hold framerate *while streaming*. GoW's boat and

squeeze-corridors are deliberately visually cheap for exactly this reason. **Proposed rule for

the perf register: a T1/T2 space is budgeted at ~30% of a normal play space's render cost,

because the other 70% is the stream.** Angkor (377 MB) is the stress case and must get T3.

4. Pacing (the ME1 lesson). Every T1/T2 traversal is skippable after its first completion, and

the skip lands on the journal. Duration is capped by the stream, never padded.

The strongest objection to Shape C: *"You are building a vehicle to solve a loading problem,

and Josh ruled Watercraft is post-79."* — Answer: the on-rails deck is a *transition space* that

shares the Watercraft's geometry, not the Watercraft *system*. None of §2's excluded scope (build/

repair ladder, navigation technique, naval combat, ocean sim, module registry) is touched. The

deferred ruling is honored; the deferred asset is pre-earned.

Second objection: *"~26 water seams don't all deserve a deck."* — Correct, and Shape C already

concedes it: short hops take T1 bookends. The tier is chosen per seam, in the transitions doc,

with the whole range in view — the same range-scale judgement TRANSITION_TREATMENT.md already

mandates for Layer 2.

What this fork does NOT resolve (belongs to the other axes / the director): the actual

streaming architecture (World Partition is configured in DefaultEngine.ini L80-81 but the region

maps are separate .umaps today, so T0/T1 continuity across the Flores→Bali boundary may require

a topology change, not just a design one); the measured Steam-Deck budget (window rank 1's

benchmark harness does not exist yet, so every perf number above is a *rule*, not a *measurement*);

and the intra-CH_05 map hops, which the slice driver's own comment (L51-53) flags as "a further

seam layer, not this pass."

---

6. Where this lands in the existing plan

This challenge is window-plan rank 19 × rank 21 colliding — the presentation doctrine

(camera/cut/transition covers, a josh-brief that "precedes CN5.2 template bake") and the

between-site traversal design (route legs, paths/mounts/boats, retire the DEV teleport). Gap

register entry 24 already says "13-23 km legs carry zero content"; entry 29 says the transition

doctrine was "named a Josh gap twice."

Shape C answers both ranks with one taxonomy, and it should be ruled *before* CN5.2 bakes

camera blocking into 11 templates — because the tier a seam gets determines whether its templates

need an arrival camera at all.

The artifact that should carry the ruling is docs/TRANSITION_TREATMENT.md — add the seam-tier

taxonomy as Layer 2b, and type all 78 existing seam entries in the seven

docs/spine/transitions/TRANSITIONS_*.md files (a one-column mechanical pass, zero-token

scriptable from the - Passage. lines that already classify cleanly).

---

Provenance

Read: HumanityQuestSliceDriver.cpp (L30-100, L350-410); Content/Maps/ listing;

Config/DefaultEngine.ini; docs/TRANSITION_TREATMENT.md;

docs/spine/transitions/TRANSITIONS_CH02_21.md + the 7 sibling files (78 - Passage. lines

classified); docs/spine/CH_05.md; T1_World_Geography_Master [ACTIVE v2.3] §7/§11;

docs/proposals/DOMAINS_VEHICLES_AND_ENRICHMENT_BACKLOG.md; docs/COMBAT_PROGRAM_ADDENDUM.md §9;

docs/DESIGN_GAP_REGISTER.md entries 24/29; docs/PRE_5090_BUILD_PLAN.md ranks 19/21;

docs/PRE_5090_BUILD_PLAN_VOL2.md PG-2/U11.22/U11.26; T1_Ability_Tree [ACTIVE v1.4].

Precedent claims are general shipped-title design knowledge, stated factually, no quotation.

Seam classification is reproducible: grep -h "^- Passage\." docs/spine/transitions/*.md.

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