PARTY_STRUCTURE.md

npcs/PARTY_STRUCTURE.md

FOUNDATION C — The Party-Structure Derivation

Status: RECOMMENDATION for Josh. Read-only derivation from the arc structure + the ratified travel/loyalty

systems + genre evidence. Nothing here is casting or per-companion loyalty (both are Josh's characterization

canon). This answers ONE ratifiable question: **what party SHAPE does this story want, and therefore what

total-cast RANGE should Josh ratify** — with the loyalty models mapped onto that shape.

The two burned failure modes are named up front so this doc does not repeat them:

• Undisclosed filters behind measured numbers. Every count below is tagged MEASURED (re-derived from a

stable doc, with the filter stated) or RECOMMENDED (a design call). They are never blended.

• Seeds/proposals read as canon. The only companion facts that are canon today are: 5 of 162 Character_Index

rows carry loyalty seeds, of which the roster-proven travellers are Ntala (for_life) and Ana (come_and_go);

everything else is a labelled recommendation.

---

0. The measured ground (what the corpus actually proves about who travels)

All figures re-derived from stable docs (NOT the concurrently-mutating set).

• Ntala CHAR_0002 — the one proven constant. for_life / never. 191 actor-beats across 55 of 79 nodes

(MEASURED, filter = the structural actors: field, 891 total beats parsed)

[SRC: docs/spine/CARDINALS_RATE_AND_ROSTER.md # verification record]. She is the best-friend-seeded travelling

companion (the Ch-1 best-friend choice seeds her starting gender/skills/combat-focus; she is the friend's felt

echo; endgame homecoming reunion pays it off — ratified ground, memory best-friend-seeds-main-companion).

• Ana CHAR_0017 — the one proven recurring guest. come_and_go / major_opposition. **44 beats across 22

nodes** (MEASURED) [SRC: docs/spine/CARDINALS_RATE_AND_ROSTER.md # verification record]. The melee blade; her

slot is covered by CHAR_0148 Wenceslao when she is absent

[SRC: docs/spine/CARDINALS_RATE_AND_ROSTER.md # CARDINAL 2]. She is canon's chosen exemplar for the FIRST

integrity-driven leave/draw beat [SRC: docs/proposals/WHOLE_ARC_CONNECTIVE_WEB.md # Lens C, CL-02].

• The light-contact tier is real and mostly NOT companions. Tunde (CHAR_0127), Amboko (CHAR_0126), Oupa

Gert (CHAR_0125), Wenceslao (CHAR_0148) are "carried lightly" across the modern arc; three of the four carry

no loyalty row at all, and Tunde's drawn_conditional row is a schema example WS7 forbade reading as a

roster assignment [SRC: docs/spine/CARDINALS_RATE_AND_ROSTER.md # CARDINAL 2]. Binti Hassan is a named

light-ally candidate (ratified ground, ROSTER-B).

• The five loyalty models are a live enum: `for_life | invested_sticky | come_and_go | drawn_conditional |

not_companion` [SRC: docs/spine/RELATIONSHIP_TIER_BANDS_BRIEF.md # §3]. Only two of the five have a

spine-proven exemplar today (for_life = Ntala, come_and_go = Ana); invested_sticky and drawn_conditional

have zero exemplars, and drawn_conditional orphaning is a flagged risk

[SRC: docs/spine/CARDINALS_RATE_AND_ROSTER.md # CARDINAL 2 strongest objection].

• The childhood familiar is a SEPARATE axis — do not count it in the companion cast. Lost Ch 2, reunited

Ch 58 (TG), the 22nd familiar slot, Ch-75 aid [SRC: docs/proposals/WHOLE_ARC_CONNECTIVE_WEB.md # Lens C]. The

22 familiar slots are a separate axis from THE BOND by ratified rule — never conflate. It matters below only

because it PROVES a non-human bond crosses into the endgame realms.

The corpus therefore proves a shape, not a headcount: exactly one for-life traveller, one recurring guest,

a handful of stay-at-home local contacts, and a bonded non-human that reaches the endgame. The "~30" figure is

an unratified scope estimate; §5 shows what it most likely was a count OF.

---

1. The three structural forces that dictate the shape

1.1 Region-per-chapter across 79 nodes → you cannot honestly own ~26+ permanent travellers

The ancient arc Ch 2-38 is a physically continuous overland-and-coastal journey, one anchor culture per

chapter, ~26 distinct cultures before the modern arc even begins — and the coverage map SPREADs still more en-route

cultures into those legs [SRC: docs/proposals/EXPANDED_COVERAGE_MAP.md # §1, §4]. A companion recruited in Ch 4

Bali who is still aboard in Ch 56 Arctic is a traveller, not a Balinese; **a permanent companion is, by

construction, an outsider to 25 of the 26 cultures.** So a large permanent roster does not deepen cultural

coverage — it dilutes it. Cultural authenticity per region is carried by a DIFFERENT tier (local guides who

belong to that one place and stay), not by the travelling crew. This is the load-bearing derivation: **coverage

scales with the local-contact tier; meaning scales with the small core.**

1.2 The mothership is a crew home → the FTL / One-Piece / ME2 shape, not the BG2 stronghold shape

Travel is a single transforming vril-flagship mothership + collectible sub-craft; companions travel ABOARD,

crewing stations and away-teams (helm, gunner, cook, engineer, menagerie-keeper), with the Vimana docking as a

peer-flagship from Ch 55 [SRC: docs/proposals/systems/vehicles-fleets.md # §2.7, §2.1, mothership follow-up §6].

A single moving home with a functional crew is the vessel-crew party shape (FTL/Sunless Sea/ME2-Normandy/One

Piece Thousand-Sunny — Josh's own stated reference), NOT the BG2/DAO "recruit sixteen, bench eleven at a

keep" shape. There is no keep to bench at; everyone aboard is present. That caps the *sensible* permanent crew

at what a ship's company reads as — a handful with jobs, not an army.

1.3 The Ch-65 lock + endgame realms → the cast MUST funnel to a tiny realm-worthy core

Ch 66-77 is the Realm Road through real otherworlds (Atlantis/Pohljar/Tiamarath/Maatherion-class, then the Realm

War, Cassius, the Father) [SRC: docs/proposals/ROSTER_PROPOSALS.md # §2.2 rows 66-77]. **Who can even GO into

Duat/Yggdrasil?** Canon proves only two crossers today: the protagonist (elven birthright — she FELL from the

fairy realm in the Prologue) and the childhood FAMILIAR (Ch-75 aid, TGP Sub-Phase C)

[SRC: docs/proposals/ROSTER_PROPOSALS.md # Ch 73, Ch 75]. Every region-bound guest and local contact is

structurally shed at the Ch-65 lock — they belong to Earth and its cultures. The cast is a funnel: wide in

Ch 2-65, collapsing to a 2-to-~4-body realm party for 66-77. This is not a constraint to fight; it is the

natural act structure (gather the crew → test the crew → only the truest cross).

---

2. Genre reference points, mapped to THIS design

Synthesis: the story wants a **One-Piece-sized core nakama crew with ME2 loyalty arcs and an FTL/Sunless-Sea

station grammar, funnelled by the Ch-65 lock — explicitly NOT a BG2-sized roster and NOT a Suikoden army.** The

breadth Josh senses in "~30" is real, but it lives in the local-contact tier, not the travelling crew.

---

3. The recommended party shape — a three-tier funnel

Tier 1 — THE CORE CREW (travel aboard; for_life / invested_sticky)

The nakama. Each is recruited in the chapter that is their culture's home during the ancient arc, carries a

culture-grounded trade job (one of the 10 trades → a mothership station, §2.7), gets a loyalty arc (ME2

loyalty-mission shape, run through THE BOND ladder + the integrity-driven departure rule), and — the hard gate —

must have a reason to cross into the endgame realms (birthright-adjacent, soul-bonded, drawn, or plot-fated),

because Tier 1 is who survives the §1.3 funnel.

• Size RECOMMENDATION: 6-8, anchored by Ntala. Reasoning: One Piece lands at ~9 and ME2 at ~12, but both

front-load recruitment; here each core member needs a full homeland recruitment chapter + a trade + a loyalty

arc + a realm-crossing justification, and each must stay legible across a 30M-word arc. Below ~4 the found-family

aboard a AAA flagship reads thin; above ~8 you cannot give each a realm-worthy justification without cheapening

the "only the truest cross" beat, and each added body dilutes the others' beats (the measured Ntala-saturation

problem — one companion already fills 55 nodes). 6-8 gives one core recruit roughly every 4-6 ancient chapters

— a "gather your crew" cadence with room to breathe. **This is the number Josh's open question most needs

ratified.**

• Loyalty-model split: 1-2 for_life (Ntala is one; a second for_life should be reserved, not assumed —

e.g. a soul-bond or the deepest homeland tie) + the rest invested_sticky (sticky but not immune;

complete opposition still forces them out — the 90-cap invariant, ratified)

[SRC: docs/spine/RELATIONSHIP_TIER_BANDS_BRIEF.md # Ruling 3].

Tier 2 — RECURRING REGIONAL GUESTS (come_and_go, some drawn_conditional)

Ana is the exemplar: a real persona with an arc who joins for their region/era, departs by circumstance or

integrity, and may recur (44 beats / 22 nodes proves a guest can carry serious presence without being permanent).

These are where the integrity-driven leave/draw drama actually fires — the safe, covered path so no departure

is a wall (§8.7 no-dead-end).

• Size RECOMMENDATION: 6-10, roughly one meaningful recurring ally per major region-cluster / era-rhyme

(the gladiator→Olympics-class rhymes) — present 1-4 chapters, some returning late.

• Loyalty-model split: mostly come_and_go; drawn_conditional lives here or at the modern/endgame

edge (see §4) and NEEDS at least one authored exemplar or the ratified model orphans

[SRC: docs/spine/CARDINALS_RATE_AND_ROSTER.md # CARDINAL 2].

Tier 3 — LOCAL CONTACTS / GUIDES (not_companion, or circumstance_only)

The fixers, brokers, elders, and guides who belong to one culture and stay there — Tunde/Amboko/Oupa

Gert/Wenceslao/Binti Hassan class. Each is a full 13-field persona with a job and a relationship state (they

open/close opportunities and hold stations *while you're in their waters*), but they are **NOT BOND companions

and do not travel on.** This is the tier that carries per-culture authenticity at coverage scale, and it is the

honest home of the assimilation mechanic: the local guide is who the protagonist hides from OR confides in.

• Size RECOMMENDATION: 15-30, one to two per culture-node across the ancient + modern + SPREAD regions.

This is almost certainly what the unratified "~30" was counting — the contact roster, misread as a

companion roster (the identical mistake CARDINAL 2 documents on Tunde).

Totals. Travelling companion cast (BOND-bearing, Tiers 1+2) ≈ 12-18. Full persona'd ally population

including local contacts ≈ ~30-48. Endgame-realm party (the funnel's mouth) ≈ **2-4 + the childhood

familiar**. Presenting Josh ONE number would hide the filter; the ratifiable answer is these three, plus the

realm-party.

---

4. Loyalty models mapped onto the structure (and the cadence)

Recruitment / rotation cadence:

• Ancient arc Ch 2-38 = "gather the crew." Recruit the 6-8 Tier-1 core in-homeland, ~1 per 4-6 chapters;

interleave a Tier-3 local guide most chapters; Tier-2 guests begin recurring (Ana-class). The mothership Ship

arrives Ch 8 (dhow) — pre-Ch-8 companions travel on foot, so the earliest 1-2 core join before there's a ship

to crew, which is thematically right (they're why you build the crew).

• Modern arc Ch 39-65 = "test the crew." Core is aboard and deepening; loyalty arcs mature and the

departure/WAVERING/draw tests fire (the CL-02 exemplar). Few or no NEW permanent recruits — at most 1-2

drawn_conditional figures pulled in by high integrity. Tier-3 contacts continue per region. **The Ch-65 lock

is the loyalty crucible** — the corrosion path forces departures (even Ntala departs on the EVIL band per the

live spine), proving "invested is sticky but not immune."

• Endgame realms Ch 66-77 = "only the truest cross." No new travellers. The funnel closes to the

realm-worthy core + the childhood familiar (proven crosser) + the homecoming reunion with the grown best-friend/

Ntala payoff (CF-2). The ONLY sanctioned late-join model here is drawn_conditional (a realm-native drawn by

integrity) — never a fresh recruitment.

---

5. What this crew must wire into (so Tier-1 design is not orphaned)

Each Tier-1 core member is one entity threaded through every ratified system — this is the checklist that makes a

companion "add value" per the mandate:

• The 10 trades → each core member's culture-grounded job = a mothership station (helm/gunner/cook/engineer/

inscriber/menagerie-keeper) [SRC: docs/proposals/systems/vehicles-fleets.md # §2.7]; the away-team system gives

them off-ship agency at the real cost of their absence.

• THE BOND (9 tiers) vs THE STANDING (faction) — never conflate. A companion's bond is personal

(loyalty_score); faction reputation is a separate axis [SRC: docs/spine/RELATIONSHIP_TIER_BANDS_BRIEF.md #

Ruling 4]. A core member may love you (Family) while their home faction is Suspected of you.

• Attackability Tier-3 bonded — bonded companions sit at the protected attackability tier; the emergent-

violence sandbox must respect that.

• Real-science immersive-sim combat + vril-altered physics — allies wield both; each core member should have

a distinct combat identity (Ana = the melee blade; her cover by Wenceslao proves the "slot" design).

• The assimilation mechanic — a travelling companion WITNESSES the protagonist hiding what she is inside each

culture. Design what that costs and buys per core member: the witness who keeps the secret deepens the bond; the

witness whose beliefs the deception violates is a departure trigger. This is the richest untapped companion hook

and it is Tier-specific (a Tier-3 local guide is who you hide FROM; a Tier-1 crew member is who hides WITH you).

---

6. Flags for Josh (the calls this derivation cannot make)

1. RATIFY THE CORE SIZE (the open roster question). Recommendation: Tier-1 core = 6-8 (One-Piece-sized),

Tier-2 guests = 6-10, Tier-3 contacts = 15-30. The single number "~30" is best read as the Tier-3

contact count, not the travelling cast — confirm which tier it was scoping.

2. The 2nd for_life slot — reserve or fill? Recommendation: reserve one for a soul-bond/deepest-homeland

tie; do not default a second for_life.

3. drawn_conditional needs an exemplar or a retirement — author one (Tunde is the candidate, per

CARDINAL 2's Option B) or deliberately retire the model. Casting is Josh's.

4. The endgame-realm party roster — canon proves only the protagonist + the childhood familiar cross today.

WHICH human-world core companions can enter the otherworlds (and by what justification — birthright-adjacent,

soul-bond, drawn realm-native) is characterization + world canon = Josh's. The structure only says the party

there is very small (2-4).

5. §17.1 substrate dependency (do not fake depth). Every Tier-3 local guide and Tier-1 homeland recruit is

only as authentic as the culture's corpus substrate. Where the substrate is THIN (the coverage map's SPREAD/

RESCOPE gaps — e.g. Angkor/Khmer, Korea, per-culture Polynesia, Great Zimbabwe, Great Plains, Haudenosaunee,

Japan-if-rescoped) [SRC: docs/proposals/EXPANDED_COVERAGE_MAP.md # §7], a companion or guide native to that

culture is book-processing-blocked — flag the enrichment dependency, never invent cultural depth.

---

Verification record

• Counts tagged MEASURED are re-derived from stable docs: Ntala 191 beats/55 nodes, Ana 44/22, 5-of-162 loyalty

seeds, the 5-model enum, the 9-tier BOND, the Ch-8 ship arrival, ~26 ancient region-nodes. All Tier sizes and

cadences are tagged RECOMMENDED.

• Citations resolve to STABLE docs only: CARDINALS_RATE_AND_ROSTER.md, RELATIONSHIP_TIER_BANDS_BRIEF.md,

vehicles-fleets.md, EXPANDED_COVERAGE_MAP.md, ROSTER_PROPOSALS.md, WHOLE_ARC_CONNECTIVE_WEB.md. Per the

concurrent-mutation notice I read CH_39/56/57, T1_Integrity, T0_Character_Index and RATE_MISMATCH_ANALYSIS for

CONTENT only and cite NO line numbers from them; the loyalty-seed and departure-rule facts above are anchored

to the stable briefs that quote them.

• No repo file written; read-only discipline held at authoring; the artifact was later committed here (0d2a72b).

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