pipelines/PIPE_QUEST_WORLD_2026-08-05.md
Lane: how quests are authored, executed and debugged, and how the world is populated with life.
Status: PROPOSAL-TIER (lane-canon). This doc invents no world canon. It is ENGINE per the reuse
directive — game #2 inherits the quest runtime, the facts surface and the population tooling, and
supplies its own registries. Every claim below is either WEB-VERIFIED (source URL given, verification
state declared) or REPO-CITED (file given) — nothing is asserted from memory.
---
**Both independent researchers in the 2026-08-05 comparator-pipeline pass surfaced the same structural
finding, without coordination: no PIPE_* dossier owned quests or world population.** The eight
dossiers minted on 2026-07-29 are all *asset* lanes — art, 3D, animation, video, audio-music, voice,
website, control-plane. The largest content surface in the game, 79 nodes of quest and world content,
had no pipeline dossier, and therefore the pipe-dossiers-bind law (docs/FACTORY_CONTINUATION.md,
class pipe-dossiers-bind-generation-lanes: *every media/tech lane names its PIPE_* dossier as law;
rulings land as in-place supersessions, never parallel charters*) had nothing to bind this lane to.
That is exactly the failure mode that produced the music-lane correction — a lane running off a charter
instead of a dossier. This file closes it for the quest/world lane.
Honest status, declared up front: this dossier is NOT yet a verified stack. The eight sibling
dossiers each open with a web-verified tool/model roster with licences. **No such research pass has been
run for this lane** — the 2026-08-05 pass studied comparator PRODUCTION METHOD, not our tooling options.
So §1 below is a question list, not a stack. Anyone who fills it should follow the sibling format.
Naming a lane's law home before its stack is researched is the correct order; pretending the stack
exists would not be.
---
| # | Question | Why it is load-bearing |
|---|---|---|
| Q1 | What executes a quest graph at runtime — UE's own systems, a custom signal executor over our registry rows, or a data-driven state machine? | The comparator answer (§2, C1-C4) is a signal travelling a graph with structural scope-exit. We hold the DATA shape already; we hold no executor |
| Q2 | Where does the facts surface live, and what writes it? | state_writes is currently prose per quest row. One surface with expiry is the verified comparator shape |
| Q3 | What is the phase-jump debugging entry point, and what does it cost to build? | Ranked the single highest-leverage adoption in the whole pass, because QA replay time is our binding cost until the Josh Gate |
| Q4 | Does world population run as authored instances, or as a spawn/timetable join evaluated at runtime? | Determines whether ambient_population_band_default stays a scalar or becomes a table |
| Q5 | Bake-time or run-time placement? | Named as a director-level fork in GEODATA_TERRAIN_CHAIN.md §8.2; unresolved on purpose |
| Q6 | What is the unit of world regeneration, and is generation deterministic? | Determinism is a hard prerequisite for the nightly rebuild, not a nice-to-have (§2, F3) |
---
The 2026-08-05 comparator-pipeline research for this lane landed in place on the two closest owners,
because this dossier did not exist when it was written. It is NOT duplicated here — one copy, cited, per
the anti-orphan rule. This dossier is its binding home; those sections are its content.
| Section | Lives in | Owns |
|---|---|---|
| Quest authoring + world population | PIPE_CONTROL_PLANE_2026-07-29.md §8 | CDPR quest graph / facts DB / debugger · CDPR communities + action points · Nintendo production machine · Rockstar population data · Bethesda alias templates |
| World building and measurement | GEODATA_TERRAIN_CHAIN.md §8 | Far Cry 5 deterministic nightly rebuild · Horizon run-time placement · BotW density doctrine and heat maps · the CDPR/Rockstar counter-example |
| Systems at scale | PIPE_CONTROL_PLANE_2026-07-29.md §7 | Itemization derivation · drop-table composition · faction data model · Epic live tooling |
| Full cited pack + verification ledger | docs/research/COMPARATOR_PIPELINES_REFERENCE.md | Every source, every verification state, the routed adoption worklist |
1. A quest is a signal travelling a graph, and scope-exit is STRUCTURAL. Verified verbatim on CDPR's
own tool documentation: *"Once the signal passes this all other nodes inside the phase will die."*
Without that rule, branch cleanup becomes per-quest bookkeeping and 79 nodes of content rot.
2. World state is ONE facts surface with expiry, read by conditions and written by quest blocks —
not bespoke per-quest state.
3. Phase-jump debugging is a throughput multiplier, not a nicety. CDPR's debugger highlights the live
signal and lets a designer continue execution from any node's pin. Every hour the AI QA loop spends
replaying to reach a quest state is an hour that tool would have saved.
4. Population is JOINED TABLES, not placed instances — spawn sets × timetables × category-tagged world
points, joined by naming convention. The living world of the most-praised open-world RPG of its
generation was authored in two spreadsheets. That is the strongest external evidence this project has
that its registry-as-authoring-surface bet ships.
5. Behaviour goes on the SLOT, not on the entity. Independently converged on by CDPR (category-tagged
action points) and Rockstar (a five-verb scenario vocabulary), both verified. Bethesda's alias system is
the third leg and remains flagged — see §3.
The Bethesda alias/Radiant material (PIPE_CONTROL_PLANE §8.4, B1-B4) could not be read directly: UESP
returned HTTP 403 on both ck.uesp.net and en.uesp.net across two independent passes. The **convergence
finding in §2.1 item 5 does not depend on it** and is adopted on its two verified legs. The Bethesda
mechanism — the fill-type list, two-layer keyword matching, ordered dependent fill — is **gated behind
re-verification** and must not be built against search-index extracts. Cited-or-cut applies to adoptions
exactly as it applies to canon.
---
Routed items from the full worklist (docs/research/comparator_pipeline_adoptions.json) whose lane is
quest_world. Verdicts and evidence states are the synthesizer's, post-verification.
| # | Technique | Teacher | Verdict | Route |
|---|---|---|---|---|
| QW-1 | Phase-scoped quest graph with structural scope-exit | CDPR | ADOPT — highest leverage | quest runtime; Q1/Q3 above |
| QW-2 | One facts surface with per-fact expiry behind state_writes | CDPR | ADOPT | docs/CONDITION_EXPRESSION_GRAMMAR.md write side |
| QW-3 | --start-at-phase entry point for QA | CDPR | ADOPT | quest runtime; QA loop |
| QW-4 | Population as spawn-set × timetable × category-tagged points | CDPR + Rockstar | ADOPT | T0_Region_Index.ambient_population_band_default becomes a table |
| QW-5 | Composition percentages that must sum correctly, or a gate fails | Rockstar | ADOPT | harness/ population check |
| QW-6 | A tiny closed ambient-verb vocabulary; variety from the join | Rockstar | ADOPT as constraint | persona / dwelling registries |
| QW-7 | Behaviour on the slot, not the entity | CDPR + Rockstar | ADOPT | persona, faction-standing, companion systems re-read |
| QW-8 | Alias/slot-filled quest templates over keyword-tagged locations | Bethesda | GATED — re-verify first | participant_entity_refs fill contract; see §2.2 |
| QW-9 | Quest count drives world size, not the reverse | CDPR | ADOPT as framing | region sizing; T0_Quest_Seed_Registry (74) + T0_Questline_Index (101) |
| QW-10 | Every delayed consequence must produce a player-legible callback | CDPR | ADOPT as a gate | every world_memory_write needs a downstream surfaced read |
| QW-11 | Bug reports emitted by the game carrying coordinates + phase + facts | Nintendo | ADOPT | QA-loop bug emitter payload |
| QW-12 | Deterministic nightly world rebuild on a sector grid | Far Cry 5 | ADOPT | GEODATA_TERRAIN_CHAIN §8.5; the 5090's idle overnight window |
| QW-13 | Three-scale relief (landmark / view-blocking / tempo) as a computed gate | BotW | ADOPT | T0_Region_Index, computed off DEM-derived layers |
| QW-14 | QA traversal heat maps as the standing world-quality instrument | BotW | ADOPT | the AI QA loop is our player; its traversal is loggable |
| QW-15 | Per-node placement posture (hand-crafted in the slice, rule-driven outside) | CDPR vs FC5 | ADOPT as a field, not a stance | T0_Region_Index |
| QW-16 | Bake-time vs run-time placement | Horizon | OPEN — director fork | stated, deliberately unresolved |
---
This lane owns: quest authoring and execution, the facts surface, world population and ambient life,
placement posture, and the world-quality instruments (heat maps, three-scale relief, minimap diff).
It does NOT own: terrain generation and the geodata chain (GEODATA_TERRAIN_CHAIN.md), region art kits
and the zone factory (PIPE_ART §8), creature assets and rigs (PIPE_3D §7), movesets and combat feel
(PIPE_ANIMATION §7), itemization and drop derivation (PIPE_CONTROL_PLANE §7), or job scheduling
(PIPE_CONTROL_PLANE §§3-4).
Standing note for whoever fills §1: the comparator evidence says the counterweight matters as much as
the automation. Far Cry 5's own published lessons, from the team that automated the most, include that
excessive automation causes its own problems and that manual control sometimes outperforms it. The
AAAAA slice (Prologue → Ch 13) is where manual control should win; the procedural chain earns its keep
across the other 66 nodes.