diff --git a/docs/workshops/generation-cascade/BRIEF.md b/docs/workshops/generation-cascade/BRIEF.md
new file mode 100644
index 000000000..ff1718655
--- /dev/null
+++ b/docs/workshops/generation-cascade/BRIEF.md
@@ -0,0 +1,131 @@
+---
+title: "Workshop Brief"
+description: "Audit the generation pipeline implementation state, map all layers from heightmap to walkable tile, produce D-records and ticket dependency flows"
+type: workshop
+status: active
+workshop: generation-cascade
+agent: ""
+round: 0
+created: 2026-04-24
+---
+
+# Generation Cascade Workshop Brief
+
+**Goal:** Audit the full generation pipeline from planetary heightmaps to walkable tile environments. For each layer: document what exists, what's stub, and what's missing. Produce D-records and a ticket dependency chain that formally blocks character and apartment work behind the complete pipeline.
+
+**Ticket:** #897
+**Priority:** HIGH — load-bearing for Sprint 38 and all downstream Phase 4/5 work
+**Participants:** Gestalt (systems design), Tyre (architecture/feasibility), Miri (worldbuilding/cultural inputs)
+**Source:** Lead directive (2026-04-24): "we keep cycling back to these topics. I want them parked behind the full cascade from now on."
+
+---
+
+## Context
+
+The generator-architecture workshop (2026-02-27) designed the full data model: DistrictSkeleton, BlockSkeleton, ChunkData, two-phase generation (Phase 1 skeleton, Phase 2 chunk fill). The outcomes are at `docs/workshops/generator-architecture/workshop-outcomes.md`.
+
+**The architecture is designed. The implementation is mostly stubs.** The Rust types in `server/src/simulation/generator.rs` compile but no code actually produces filled instances. The team keeps drifting to character creation, apartment generators, and tycoon starting states — all of which sit several pipeline layers below what's actually built.
+
+This workshop is NOT a redesign. It's an implementation audit and cascade formalization.
+
+### What exists (confirmed working)
+
+| Layer | What | Status |
+|-------|------|--------|
+| Galactic | System definitions, star data | Done — `systems.db`, wiki |
+| System | Body definitions, orbital mechanics | Done — `systems.db` |
+| Planetary surface | Heightmaps, terrain simulation | Done — `tooling/planet-gen/planet_simulation.py` |
+| Atlas markers | City placement, roads, rail, rivers, mountains | Done — `tooling/planet-gen/generate_atlas.py`, `markers.json` |
+| City naming | Gemma-driven cultural naming | Done — `tooling/planet-gen/gemma_naming.py` |
+| Economics | Supply chains, corporations, brands, trade flows | Done — `tooling/economy-db/`, `systems.db` |
+
+### What exists as types only (stubs, no generation logic)
+
+| Layer | What | Location |
+|-------|------|----------|
+| District skeleton | `DistrictSkeleton`, `BlockSkeleton`, enums | `server/src/simulation/generator.rs` |
+| Chunk streaming | Load/unload system | `server/src/simulation/chunk_streaming.rs` |
+| Triangle system | `TriangleAssignment`, `TrianglePurpose` | `server/src/simulation/triangle.rs` |
+
+### What's missing entirely
+
+This is what the workshop must map. Suspected gaps include:
+- Regional/continental subdivision (between heightmap and city)
+- City-to-district decomposition (how does a city marker become N districts?)
+- District skeleton generation (the actual Phase 1 code)
+- Block fill / zoning assignment
+- Chunk tile generation (Phase 2)
+- Infrastructure placement within districts (roads, utilities at local scale)
+- Vertical structure generation (multi-floor buildings)
+
+---
+
+## Key Questions to Resolve
+
+### 1. Pipeline Inventory (all participants)
+
+For each layer from planetary heightmap to walkable tile:
+- What is the layer's input and output?
+- What code/data exists today?
+- What's the minimum viable implementation?
+- What decisions from the generator-architecture workshop apply?
+
+### 2. Layer Dependencies (Tyre)
+
+- What is the strict dependency order? Which layers can be parallelized?
+- Where are the data format boundaries (file vs. runtime, Python vs. Rust)?
+- What's the testing strategy per layer? Can each layer be validated independently?
+
+### 3. Cultural and Worldbuilding Inputs (Miri)
+
+- At which layers do cultural inputs (society profiles, naming, architectural style) enter the pipeline?
+- What wiki/content data is needed before each layer can generate?
+- Are there content gaps that block generation even if the code existed?
+
+### 4. System Interactions (Gestalt)
+
+- How does each generation layer interact with the economics layer?
+- Where do social sites, NPC population, and zone palettes enter?
+- What's the minimum viable "viewable world" — the thinnest vertical slice from heightmap to rendered tiles?
+
+### 5. Ticket Dependency Chain (all participants)
+
+- What tickets exist for missing layers? What new tickets are needed?
+- What is the formal dependency chain that blocks character work (#694, #619) and apartment work (#681, #682)?
+- Which existing tickets (#615, #616) should be re-scoped or deferred?
+
+---
+
+## Workshop Format
+
+**3 rounds:**
+
+### Round 1 — Inventory
+Each participant audits the pipeline from their domain perspective. List every layer, its state (done / stub / missing), inputs, outputs, and the key file paths. Write findings to `docs/workshops/generation-cascade/{agent}-round1.md`.
+
+### Round 2 — Proposals
+Based on the combined inventory, propose: the ordered implementation plan, the ticket dependency graph, and the D-records needed. Identify the thinnest vertical slice that produces viewable output. Write proposals to `docs/workshops/generation-cascade/{agent}-round2.md`.
+
+### Round 3 — Convergence
+Lock the D-records, finalize the ticket dependency chain, and produce the formal blockers. Each participant reviews the proposed D-records and flags disagreements. Write final positions to `docs/workshops/generation-cascade/{agent}-round3.md`.
+
+---
+
+## Required Reading
+
+Before Round 1, all participants must read:
+- `docs/workshops/generator-architecture/workshop-outcomes.md` — the designed architecture
+- `server/src/simulation/generator.rs` — the current data model (stubs)
+- `server/src/simulation/chunk_streaming.rs` — chunk load/unload system
+- `tooling/planet-gen/generate_atlas.py` — what the atlas generator produces
+- `decisions/architecture.md` — D-110 (signed z-levels), D-108 (MobileChunk)
+- CLAUDE.md §Development Cascade — the phase definitions
+
+---
+
+## Expected Outputs
+
+1. **D-record(s)** in `decisions/` defining the generation pipeline layers, their order, and their implementation status
+2. **Ticket dependency graph** — new tickets for missing layers, `ticket_deps` entries blocking character/apartment work
+3. **Updated existing tickets** — #615, #616, #619, #681, #682, #694 re-scoped or formally blocked
+4. **Implementation priority order** — which layer to build next (informs #899 and future sprints)