settled-reach/docs/workshops/world-generation/sw1-round2-notes.md

title, author, round, workshop, date, status

title	author	round	workshop	date	status
SW1-R2 — Consolidated Round 2 Notes	qatux	SW1-R2	world-generation	2026-03-24	complete

World Generation Workshop — Round 2 Consolidated Notes

Purpose: Synthesise the 5 Round 2 agent outputs. Each agent cross-read the R1b outputs against the product owner interview answers and identified consequences, tensions, and new questions.

Source files read:

• gestalt-sw1r2.md ✓
• miri-sw1r2.md ✓
• nigel-sw1r2.md ✓
• ozzie-sw1r2.md ✓
• tyre-sw1r2.md ✓

---

⚠️ STALE D-RECORDS — ACTION REQUIRED

The following D-records conflict with the product owner interview answers and must be superseded:

D-record	Title	Conflict
D-104	Heritage grammar overlays (abstract roots)	Heritage roots (Frost/Stone/Tide/Vine/Dust/Iron/Salt) explicitly deprecated. Real-world cultural corridors are canonical.
D-105	Heritage grammar overlays (cont.)	Same as D-104.
D-101	ZonePalette modifier system	Modifier axis A = HeritageRoot type — needs remapping to cultural_corridor + founding_culture.
D-107	(Likely references heritage roots)	Requires audit.

Resolution confirmed by PO (R1 + R2 follow-up): Clean break. All four records superseded. The 7 abstract roots are retired. New D-records will describe how real-world corridor identities (Commonwealth, Iberian/Latin American, East Asian, Germanic) map to room grammar, spatial tokens, and NPC behavior.

Note from Miri (R2): Supersession should be paired with new D-records covering:

• Room grammar parameterisation per cultural corridor (after wiki cross-check with Paula/Mellanie)
• Attenuation tier specification (cultural_corridor × heritage_persistence × hop_distance → tier)
• Object list authoring per founding_culture and industrial corridor

---

1. Key Product Owner Decisions and Agent Synthesis

1.1 Fixed canonical galaxy, seeded by simulation drift — not by seed parameters

PO decision: 300+1 fixed systems, same wormhole routes every run. Internal state drifts via background simulation, not via seed. Same canonical starting state every run.

Agent synthesis:

• Nigel: R1b Q1 (fixed canonical galaxy) confirmed; mechanism changed completely. Seed-based internal state variation explicitly rejected. Entire replayability framework shifts: variety comes from simulation trajectories, not from starting parameters. Player knowledge of canonical starting states is durable — veterans know what Van Maanen's Star looks like on day 1. This is a feature (learnable world), but the metagaming surface concentrates in early-game geography.
• Gestalt: Generator's job is to produce canonical state reliably, not produce variety. Variety is the simulation's job. "Generation quality" now means correctness and internal consistency, not seed variety. The generator can be fully deterministic and stable.
• Ozzie: Confirmed. The GTTR arrival window (expectation set by authored text) pairs with simulation-driven reality. The gap between GTTR description and current state IS the archaeology.

1.2 Topography: hard constraint, authored not procedural

PO decision: Mountains block corridors, rivers create district boundaries, coastlines produce port districts. Authored world maps for all planets and moons (not procedurally generated heightmaps).

Agent synthesis:

• Nigel: Confirmed as hard constraint; geography is stable across runs. Players learn the terrain. Replayability comes from simulation state at those fixed locations, not from layout variation.
• Gestalt: This is architecturally simpler for the generator (no terrain generation algorithm) but shifts authoring load. Format of authored world maps is unspecified — this is a load-bearing gap (see section 3 below).
• Miri: Authored topography + wiki cultural data = the generator gets high-quality inputs at Layers 1-3. Cultural territory preferred terrain (first wave takes preferred terrain, later waves adapt) is now derivable from authored world maps + cultural-migration-pressure.md together.

1.3 Heritage roots deprecated — real-world corridors canonical

PO decision: 7 abstract roots (Frost/Stone/Tide/Vine/Dust/Iron/Salt) deprecated as generator inputs. Cultural data already in 301 wiki entries and systems.db uses real-world corridor identities.

Agent synthesis (most detailed from Miri):

• Miri: Full translation table provided (abstract root → corridor equivalent). The mechanisms R1b proposed survive intact — only the vocabulary changes. Room grammar now parameterised by:
  • Layer 1 (spatial): cultural_corridor identity
  • Layer 2 (social): cultural_corridor × attenuation_tier
  • Layer 3 (objects): founding_culture × industrial_corridor × biography_trajectory Four bar descriptions authored (Commonwealth, Iberian/LatAm, East Asian, Germanic/Scandinavian) as concrete room grammar targets.
• Gestalt: The mapping from corridor identity to spatial grammar tokens is the missing authoring artifact. No document yet connects cultural-migration-pressure.md to generator room parameters. Miri's bar descriptions are a strong first draft but need validation against the 301 wiki entries.
• Nigel: Settlement biography concept (R1b Q5) is superseded in its seed-variation framing. Survives as canonical authored content: founding biography is wiki content (Miri/Paula territory), not a seeded generator parameter.

1.4 Location profiles wiki-authored, topography authored

PO decision: Generator reads from authored data, doesn't generate profiles. Every planet and moon gets authored profiles. Same approach for world maps.

Agent synthesis:

• Gestalt: Architectural framing — the generator is now a "reader-and-filler":

  Layer	Source	Generator role
  Galaxy (300 systems)	systems.db — built	Read only
  Location profiles	Wiki — authoring pipeline	Read only
  Topography	Authored world maps — format TBD	Read only
  Arrival	Template from location type + GTTR entry	Template stamp
  Content spidering	LOD/sim tier cascade	Trigger + manage
  Settlement placement	Authored topography + wiki economics	Constraint placement
  District skeleton	Existing data model	Full generation
  Rooms	Full simulation primitives	Full generation

  Generator is richly authored at top, generative at bottom. This is the right architecture: cultural and geographic character too important for procedural approximation.

1.5 Rooms as full simulation primitives with LOD

PO decision: A visited bar needs monthly turnover, cost, employee count, current staff and visitors. Depth scales with player engagement. Economic success follows regional prosperity.

Agent synthesis:

• Gestalt: "Any ambient NPC can evolve" + "rooms are full simulation primitives" together collapse the ambient/narrative NPC distinction. Every generated NPC needs a minimum data bundle sufficient to be an interesting person. The distinction between "real" and "filler" NPCs cannot hold if any filler NPC can become narrative. Required minimum bundle: cultural background, economic role, relationship network seed, behavioral tell config, enough data for a triangle to form.
• Ozzie: Confirmed. The NPC-driven generation hook is "the game" — "the space becomes meaningful by proxy." Spaces are memorable because of what happened there, not because the generator made them notable.

1.6 Replayability from simulation drift, not seed variation

PO decision: NOT seed-based variation. Replayability from simulation trajectory differences. Background simulation (economic + social + political) runs on spare CPU cores.

Agent synthesis:

• Nigel: This is a stronger replayability model IF the simulation stays genuinely variable. Primary concern: simulation convergence risk (see section 4). Identifies 6 replayability surfaces under the new model: simulation trajectory, NPC-driven generation, news-driven generation, economic propagation, exploration order, social evolution.
• Gestalt: Generator's job changes: produce canonical state reliably. Variety is simulation's domain. Important architectural consequence: two playthroughs diverge based on (1) when player visits each location, (2) what player does at each location, (3) which NPC relationships they develop, (4) which news events fire.

1.7 NPC-driven and news-driven generation hooks

PO decision: (1) Know someone from a distant planet → their house, family, friends, neighborhood generate at that location. (2) News reports an event in a town → that town must exist when visited. Sparse frequency on news-driven.

Agent synthesis:

• Ozzie: "THAT'S THE GAME. That's the moment." The NPC-driven hook means every generated space carries potential significance from the moment you enter it. You can't know yet whether the person at the corner table will matter — any ambient NPC can evolve. The magic requires invisibility: player should feel they discovered a real place, not that the simulation made them a personalised level. Open question: does the player know this is happening? (Ozzie's instinct: invisible.)
• Nigel: NPC-driven generation creates unique per-run social geographies — run 1 befriends a dock worker, so Tau Ceti's docking district generates in detail; run 2 befriends a merchant, different district generates. "The player's social map shapes the generated world." Risk: NPC locality bias — if certain NPC archetypes are always most useful to befriend, the same parts of the world always generate in detail.
• Gestalt: Both hooks mean generation is socially-triggered and narratively-triggered, not just player-location-driven. This requires coordination between the generation pipeline and narrative/social simulation layers that don't currently exist in the architecture.

1.8 Environmental archaeology from real simulation events only

PO decision: No fabricated history. If there's rubble, the simulation produced the explosion. If there's a memorial, someone died in the simulation.

Agent synthesis:

• Gestalt: Right design position (integrity > atmosphere) but demands the simulation produce enough traces organically. Pre-play simulation run must produce enough events before player's first move to seed the world with readable history. The background sim workshop must answer: how much pre-play simulation time is needed to make a world feel lived-in?
• Nigel: Settles the settlement biography question — no seeded historical scars. Raises the bar for simulation: must produce interesting events that leave observable traces, at sufficient frequency. Social archaeology (web of relationships visible through NPC behaviour) may be more reliable than event archaeology — generated continuously, not episodically.
• Ozzie: Quiet-but-inhabited is fine IF the NPC social fabric is rich. "A bar that's never had an explosion can still feel alive if the relationships inside it have evolved, fought, reconciled, and changed."

1.9 Arrival as diegetic GTTR window

PO decision: Contextual by travel method. An insert context window opens with system stats, planetary screenshot, and the GTTR narrative entry on arrival.

Agent synthesis:

• Ozzie: GTTR warm arrival > cold orbital arrival. Expectation vs. reality is more interesting than ignorance vs. reality. The GTTR is "writing you a cheque. The world has to cash it." The gap between GTTR description and current simulation state is archaeology — legible if it's visible (you can see WHY things changed). Critical playtest question: is GTTR-vs-reality divergence exciting or jarring?
• Miri: The cultural signal at arrival should be the first delivery on the galaxy map's promise. Settlement patterns at orbital view should be culturally legible — Commonwealth worlds cluster around institutional centres, Iberian/LatAm worlds cluster around water and family-pattern sprawl.

1.10 Galaxy map: full information from start

PO decision: Full information from start. Inserts logically contain a 300-system star chart.

Agent synthesis:

• Ozzie: Full information works IF the dots have visual rhythm and named stories. 300 dots that all look the same are homework; 300 dots where your eye can land on a name and feel something are stories waiting to be entered. The wonder shifts from "what is this?" (ignorance) to "what IS this, actually, given everything I've heard?" (expectation). This requires visual design that hasn't happened yet. Galaxy map as emotional object is the player experience question nobody else has asked.
• Gestalt: Confirmed. The GTTR entries are the galaxy's voice. Every system has a name, a voice, a personality before the player visits.

1.11 Spatial cohesion principle

PO decision: Cohesion at small/middle scale (not messy), variety at middle/large scale (not predictable).

Agent synthesis:

• Gestalt: Codified as a design law:
  • Room → block level: cohesive. Single block feels like it belongs to one identifiable zone/era/culture.
  • District → settlement level: varied. Different districts feel genuinely different.
  • Settlement → system level: varied at settlement identity level, each internally coherent. Era transitions produce variety at district/block boundary, not within a single block.
• Ozzie: Cohesion/variety boundary needs a player-legible version: "every neighbourhood feels internally consistent AND every new district I enter feels genuinely different from the last." Miri's Layer 1 (spatial grammar) and Layer 2 (social pattern) deliver this without asset cost.

---

2. Convergences Across Agents

Theme	Agents	Consensus
Generator is reader-and-filler, not full procedural engine	Gestalt, Miri, Nigel	Layers 1-3 are authored inputs; generator produces spatial expression at layers 6-8.
Simulation drift IS the replayability engine	All four	The background simulation design workshop is the single most important design work after the generator pipeline.
NPC-driven generation hook is the highest-value feature	Ozzie, Nigel, Gestalt	"The game," "the moment," "the most exciting sentence in the interview." Strong convergence.
Background simulation workshop is non-negotiable	Gestalt, Nigel	Replayability is undefined without it. NPC evolution is undefined. Archaeology is undefined.
Heritage roots deprecated → corridor mapping is the missing artifact	Gestalt, Miri	Miri has produced a first draft. Needs validation against wiki content.
Economic cascade confirmed as generator spine	Gestalt, Miri, Nigel	District, block, room all carry economic encoding. EconomicModifier on ZonePalette needs formalisation.
Every NPC needs minimum data bundle	Gestalt	Follows from "any ambient NPC can evolve." Ambient/narrative distinction collapses.
GTTR creates expectation contract	Ozzie, Miri	World must deliver on authored text. Playtest needed to calibrate how much divergence is exciting vs. jarring.

---

3. Tensions and Gaps

3.1 CRITICAL RESOLVED — World map format

PO answer (R2 follow-up): Annotated heightmaps. Grayscale heightmap PNG + metadata JSON per planet/moon, with rivers/coastlines/biome zones as overlay layers. Additionally, a planetary shader plugin generates visual screenshots of each planet type (earthlike, jungle, desert, mars-like, ocean, etc.) for embedding in wiki pages and GTTR entries.

This is Gestalt's option A/B hybrid: annotated image (heightmap PNG) + structured metadata (JSON overlay layers). The JSON overlay layers give the generator the geometric data it needs — coastal boundaries, river paths, biome zones — without requiring a custom drawing tool. The visual screenshots feed the GTTR diegetic arrival window.

Status: Resolved. Authoring pipeline format is decided. Engineering can proceed on the ingestion pipeline (heightmap PNG + overlay JSON → terrain constraint data for district placement).

---

3.2 CRITICAL — Corridor-to-room-grammar mapping is the missing authoring artifact

Heritage roots are deprecated. Cultural corridors are canonical. But there is no document mapping corridor identity to spatial grammar tokens (Layer 1), social pattern tokens (Layer 2), or object selection (Layer 3). Without this document, the generator cannot produce culturally differentiated rooms.

Current state: Miri has authored four bar descriptions (Commonwealth, Iberian/LatAm, East Asian, Germanic/Scandinavian) as concrete room grammar parameterisations. This is a strong start, but:

• Not yet validated against the 301 wiki entries
• Industrial corridor overlay modifier unspecified (affects physical character? NPC composition? Both?)
• Attenuation tiers proposed (Living/Structural/Archaeological) but not ratified
• Intra-corridor diversity (Chinese vs Vietnamese bar in east_reach) not yet expressible at Layer 3
• Compound identity blending function unspecified (corridor × exception archetype × industrial overlay)
• Perpetual Founder Stamp room expression unaddressed

Unblocking path (Miri, post-R2): The wiki cross-check (does my derived room grammar match what Paula and Mellanie actually wrote for the 301 system entries?) should run in parallel with any background sim workshop prep. Paula and Mellanie have the best read on whether the corridor parameterisations match the authored content. This cross-check is the prerequisite before any room grammar authoring work begins.

Owner: Miri (primary), Gestalt (generator integration), Araminta (visual expression of Layer 3) Cross-check: Paula + Mellanie (wiki content validation)

Status: Draft authoring done, wiki cross-check and formalisation pending. Can run in parallel with background sim workshop prep.

---

3.3 REFRAMED — Simulation convergence / replayability model

PO answer (R2 follow-up): Three-part reframe:

1. Player agency is the primary replayability lever. Who you create, what you do, where you go across 300+ systems already provides massive variety. The background sim is secondary variety, not the core mechanism.

2. Equilibrium is signal, not failure. A stable economy shows the effect of the player's choices (or non-choices). That's information, not boredom. Convergence is not a design problem to be engineered away — it is a legible state for the player to read and respond to.

3. The interconnected network IS the propagation mechanism. Any event that tweaks core dials on any system — corporate disaster, portal closing, political shift, successful development effort — propagates and rebalances through the trade/political network. Corporate disasters are ONE entropy source, not THE entropy source. Bulk import/export lists per system inform both simulation (trade flows) and content (what's on store shelves). Set up the propagation properly and emergence follows naturally.

Background simulation workshop design brief (PO): Model bulk trade flows between systems with corporate facility dependencies, political pressure, and disaster injection. Central challenge is not "keep it chaotic" but "set up propagation properly."

Nigel's convergence concern, recontextualised: The invariants-vs-variables framing still holds, but the threat model changes. The question is not "will the simulation converge toward boring equilibria" but "will equilibria be legible enough for players to read and act on." Equilibrium is a valid game state IF the player can see what caused it and what would disrupt it.

Miri's cultural texture concern, recontextualised: The exceptional settlement histories (refugee colonies, Perpetual Founder Stamps, Gate Silence survivors) must still be resilient against normalisation — but this is now an authoring concern (their distinctiveness must be robustly authored in wiki content) more than a simulation tuning concern.

Status: Reframed. The background simulation workshop has a concrete design brief. The central engineering question is now trade flow propagation architecture, not divergence preservation.

---

3.4 Simulation event density for environmental archaeology

Real-event-only archaeology (no fabricated history) is the right design position for integrity. But it demands the simulation produce enough physically-observable events at sufficient frequency to give the world its historical texture. If the background sim mostly produces economic numbers (prices, worker counts), there is nothing to leave physical traces of.

Ozzie: "The world will feel like it has financial history but no human history." Nigel: Need a minimum density of world-marking events per district per in-game year as a simulation design requirement. Gestalt: Pre-play simulation strategy (how much background sim runs before player's first move) is the related question.

Status: Unresolved. Simulation design target needed.

---

3.5 Galaxy map visual design — nobody has asked it yet

Full information from start means the galaxy map is the first impression of the universe. 300 named systems exist. But 300 dots that all look the same are homework. The emotional experience of the galaxy map — colour washes for faction territory, economic signal visible at a glance, visited vs unvisited, visual rhythm that lets your eye land on a name and feel something — is entirely a visual design question that nobody has addressed.

Ozzie: "The reward has to come from VISITING a named place, not from REVEALING it. That's a different dopamine loop. It has to be designed intentionally."

Status: Unaddressed. Flagged for Araminta (visual design).

---

3.6 NPC-driven generation: canonical or seeded neighbourhood?

When a player befriends Kael from Tau Ceti and Kael's neighbourhood generates — is it:

• Canonical: Every player who befriends a Tau Ceti resident gets the same neighbourhood, consistent with the authored wiki state. Creates shared reference points.
• Seeded per run: This Kael's specific neighbourhood is procedurally unique. Creates unique per-run experiences.

Nigel's FQ3. Both are valid design positions with very different comparison-test experiences. Status: Unresolved. Needs product owner input.

---

3.7 GTTR expectation vs. simulation delivery

The GTTR entry is authored at a point in time. The simulation has been running. If reality diverges significantly from the GTTR description, the player either finds it exciting (world has moved on, I can read why) or jarring (I was lied to).

Ozzie's test: "Read a GTTR entry, land, play for 30 minutes. Does the GTTR's description feel TRUE?" If the simulation has depopulated a system the GTTR called warmly communal — what does the player feel?

Status: Cannot be resolved in design. Needs playtest data.

---

3.8 Critical path has shifted — background simulation is the new load-bearing wall

Tyre's primary contribution is a revised cost estimate and a named critical-path shift.

The fundamental change: The pipeline moves from procedural generation grammar → world out to authored content pipeline with simulation-driven drift. This simplifies some layers and adds others.

What got cheaper:

Layer	R1b estimate	Revised	Why
Layer 2: Location profiles	3-4 weeks	1-2 weeks	Wiki ingestion replaces generation grammar
Layer 3: Topography	days-to-5 weeks	1-2 weeks	Authored map ingestion replaces terrain gen
Seed-based variety	implicit cost	removed	No seed variation in the model

What's new and expensive:

System	Estimate	Notes
Background pressure simulation	9-14 weeks	Economic + social + political. Needs own workshop.
LOD simulation primitives	3-5 weeks	4-level business sim, component promotion model
NPC/news generation hooks	2-3 weeks	On-demand generation from simulation events
Authored content pipeline	2-3 weeks	Wiki → data, map → constraints, GTTR compilation

Revised total: ~30-50 weeks (up from 20-30, but better risk profile — open-ended generation grammar risk replaced by well-understood data pipeline work).

Critical path (old): Layer 2 grammar → Layer 6 settlement → Layer 7 districts Critical path (new): Background simulation design → Pressure-to-entity bridge → Phase 1/2 reading pressure state

Don't start background sim engineering until the dedicated workshop produces a pressure model spec. This is Tyre's strongest recommendation and aligns with Gestalt and Nigel.

---

3.9 Background simulation architecture: tokio tasks recommended

Tyre assessed three architectural options for background simulation on spare CPU cores:

Option	Mechanism	Recommendation
A. bevy_ecs gated SystemSets	Same World, conditional run_if schedule	Background gets squeezed when active sim spikes
B. Separate bevy App	Parallel World + channel communication	Data sync complexity; entity state serialization across Worlds
C. tokio async tasks	Outside ECS, snapshot/delta channels	Recommended: CPU capping trivial, no ECS needed for aggregate data

Background sim operates on aggregate data (system-level economics, faction influence, population indicators) — not per-entity ECS queries. An ECS is the wrong tool for "calculate GDP drift for 300 star systems." tokio tasks with a deterministic game-minute clock is the right pattern.

Key D-010 risk: Background sim must be gated on game-minute clock, NOT wall-clock time. Variable CPU availability must not produce different simulation states on different hardware. Deterministic clock → deterministic results (slower machines accrue sim debt but produce identical results when they catch up). This is the frame-independent physics pattern.

Two-layer model:

Layer	Scope	Runs where	Updates
Entity simulation (D-026 tiers)	Per-NPC: movement, dialogue, perception	bevy_ecs main tick	Every tick (Active) or game-minute (Background)
Pressure simulation (new)	Per-system/settlement: economic health, faction balance, social tension	tokio background tasks	Every game-minute or game-hour

---

3.10 Technical risks flagged by Tyre

Risk	Severity	Mitigation
Background sim non-determinism from variable CPU timing	HIGH	Gate on game-minute clock, apply results in fixed order
Background sim design underspecified	HIGH	Don't start engineering until workshop produces pressure model spec
Content authoring bottleneck blocks engineering testing	MEDIUM	Placeholder generation from systems.db attributes; mark as "generated, not reviewed"
Cultural framework in flux delays room grammar implementation	MEDIUM	Lock cultural input format early — labels can change, data shape shouldn't
Authored world maps — format not decided	MEDIUM	See section 3.1; decide format in this workshop
LOD transition edge cases (player sells a business)	LOW	Design demotion rules alongside promotion rules; snapshot on demotion

---

4. New Follow-Up Questions for Next Interview Round

Grouped by theme. Sources noted.

Theme A: Simulation Architecture (pre-empts the background sim workshop)

#	Question	Source	Why load-bearing
A1	What is the simulation's divergence-preservation mechanism? What ensures run trajectories stay different rather than converging toward equilibrium?	Nigel FQ1	Entire replayability model rests on this.
A2	Does the canonical starting state have any run-start stochastic variation, or is it exactly identical every run?	Nigel FQ2	Even small variation (5% band on faction standings) produces trajectory divergence much faster than zero variation.
A3	What is the minimum simulation event density needed per district per in-game year for environmental archaeology to work?	Nigel FQ4, Ozzie	Simulation design target needed before the sim workshop.
A4	What is the pre-play simulation strategy — how much background sim runs before the player's first move? (Options: none / fast-forward / pre-authored canonical events / hybrid)	Gestalt Q4	Without pre-play history, first location has no archaeology.
A5	How does the tycoon interact with simulation drift? Can the simulation close a player-owned bar? Can the player resist economic pressure?	Nigel FQ5	Determines whether simulation drift affects player's story or just background.

Theme B: World Map Format (gates layers 3-6)

#	Question	Source	Why load-bearing
B1	What format do authored world maps take, and how does the generator consume them? (Image / structured GeoJSON-style zones / drawing tool with export / wiki table)	Gestalt Q1	Everything at layers 3-6 depends on generator reading topographic data.

Theme C: NPC Generation

#	Question	Source	Why load-bearing
C1	What is the minimum NPC data bundle at Phase 2 generation time? What must be generated immediately vs expanded on demand?	Gestalt Q2	Every ambient NPC potentially becomes narrative. Bundle determines generation cost and save file size.
C2	Does NPC-driven generation produce canonical or seeded neighbourhoods?	Nigel FQ3	Determines whether social exploration creates shared or unique per-run content.
C3	Is the NPC-driven generation hook visible to the player, or invisible?	Ozzie	Invisible = magic; visible = meta-knowledge. Instinct is invisible, but this needs confirmation.

Theme D: Cultural Cascade

#	Question	Source	Why load-bearing
D1	Who validates the corridor-to-room-grammar mapping against the 301 wiki entries?	Miri Q-A	If derivations conflict with authored content, generator produces rooms inconsistent with the wiki.
D2	Does the industrial corridor overlay (MVG, Gate Corp, DSMC) modify room physical character, or only NPC composition?	Miri Q-B	Determines whether industrial overlays require room grammar authoring work.
D3	How does heritage_persistence in systems.db interact with hop-distance attenuation? Are they independent or redundant?	Miri Q-C	Needed to specify the attenuation function.
D4	How is the Perpetual Founder Stamp expressed at room level? Authored field? NPC encounter only?	Miri Q-D	Perpetual-Stamp systems have unique room identity from individual curation — needs a generation path.

Theme E: Visual Design

#	Question	Source	Why load-bearing
E1	What does the galaxy map look like as an emotional object? Does it show faction territories, economic signals, visited/unvisited?	Ozzie	Full-info-from-start means the reward is VISITING, not REVEALING. This needs a designed dopamine loop.

Theme F: Engineering Fidelity and Scope (Tyre)

#	Question	Source	Why load-bearing
F1	What LOD level does the background simulation need at launch? (LOD 1: economic pressure only / LOD 2: + social / LOD 3: full pressure model)	Tyre Q1	LOD 1 ships in weeks. LOD 3 needs a full design workshop before engineering begins. Data architecture should support all levels from day one but simulation rules can layer.
F2	What format do authored world maps take? (Tagged JSON regions / annotated image + overlay / d2 vector maps / Tiled tilemap) Tyre recommends d2 vector topology first.	Tyre Q2 / Gestalt B1	Generation needs topology data, not heightmaps. Format must be decided before ingestion pipeline is built.
F3	Does the background simulation use a deterministic game-minute clock, or wall-clock scheduling? (This directly affects D-010 compliance.)	Tyre Q3	Variable CPU speed cannot produce different world states. Must be confirmed as part of background sim architecture.

---

5. Agent Summary Table

Agent	Primary contribution	Key tension identified	Follow-ups flagged
Ozzie	Player experience of PO decisions. GTTR warm arrival > cold orbital. NPC-driven generation hook = "the game." Galaxy map as emotional object unaddressed.	GTTR expectations vs simulation divergence. Archaeology event rate. Galaxy map visual design.	A3, E1
Gestalt	Generator architecture synthesis. Reader-and-filler framing. Economic cascade as spine. Heritage deprecation architectural consequences. 4 unresolved design questions.	World map format (gates layers 3-6). NPC universality requirement. Canonical state → deterministic generator.	A4, B1, C1, D1
Miri	Heritage root → corridor translation. Four bar descriptions as room grammar targets. Attenuation tier proposal. Industrial corridor as second modifier axis.	Corridor-to-room-grammar mapping not validated. Compound identity blending function unspecified.	D1, D2, D3, D4
Nigel	Replayability model shift acknowledged and mapped. Six replayability surfaces under new model. Simulation convergence as core architectural risk. NPC-driven generation as "explosive" replayability surface.	Simulation convergence (most serious risk). Cold-start metagaming. NPC locality bias.	A1, A2, A3, A5, C2
Tyre	Revised cost estimate. Reader-and-filler framing confirmed. Background sim as new critical path. tokio tasks architecture recommended. Determinism risk on background sim named.	Simulation convergence + D-010 risk. Content authoring bottleneck. World map format unspecified. Background sim scope underspecified.	F1, F2, F3

---

6. Background Simulation Workshop Brief — Mandatory Agenda Items

The following items must be explicitly addressed in the background simulation design workshop brief. They are architectural load-bearers, not enrichment.

Item	Source	Why mandatory
Invariants vs variables — which aspects of the world are designed to be stable reference points, and which are designed to stay chaotic and resist equilibrium?	Nigel (post-R2)	Without separating these, the simulation cannot be tuned. Invariants need authored starting state. Variables need divergence-preserving chaotic subsystems.
Propagation network architecture — how do events that tweak core dials on any system propagate and rebalance through the trade/political network? Bulk import/export lists per system inform both simulation (trade flows) and content (what's on store shelves).	PO (R2 follow-up) / Nigel R2	"Set up propagation properly and emergence follows naturally." The central challenge is propagation architecture, not divergence preservation.
Consistency layer for generation hooks — when the storyteller generates a news event claiming a location exists, how does that become a real place the generator produces? What is the buffer between "claimed to exist" and "generator has produced"?	Gestalt (post-R2)	NPC-driven and news-driven generation are callable from social and narrative layers, not just player movement. The pipeline needs a consistency mechanism to ensure existence claims are resolved before the player arrives.
Pre-play simulation strategy — what state is the world in when the player's first move happens? How much background sim runs before player control?	Gestalt (post-R2) / Gestalt Q4	Load-bearing for environmental archaeology. If the sim hasn't run long enough before the player starts, the world has no traces and no history. The workshop needs a concrete answer on pre-game sim time or pre-authored canonical events.
Exception community resilience — unusual settlement histories (refugee colonies, Perpetual Founder Stamp systems, Gate Silence survivors) must be resilient against simulated normalization, not just seeded. If the sim smooths these out over time, culturally distinctive room grammar loses its grounding.	Miri (post-R2)	Convergence risk has a direct cultural texture consequence. This constraint must be designed in, not left as an emergent property.

---

7. Upstream Dependencies (in priority order)

Before the generator pipeline sub-workshops can proceed:

1. Background simulation design workshop (product owner flagged, confirmed by Gestalt and Nigel) — Replayability architecture, NPC evolution, archaeology all depend on it. Non-optional.

2. World map format decision (Gestalt Q1) — Blocks all generation below Layer 2. Must be decided before Settlement Placement (SW2 or SW3).

3. Corridor-to-room-grammar document (Miri) — Miri has a draft. Needs validation, industrial overlay specification, Araminta coordination for visual layer. Should feed Phase 2 room-fill specification.

4. D-104, D-105 supersession (flagged by Gestalt, Miri, Qatux) — D-records need formal closure. New D-record covering four-corridor framework to replace them.

5. Minimum NPC data bundle specification (Gestalt Q2) — Blocks Phase 2 NPC fill design. Must be specified before the NPC generation architecture is finalised.

---

— Qatux, 2026-03-24