Strategic reorient locked 2026-04-27. Re-locked 2026-04-30 evening per
VISION_REVISION_2026-04-30.md: terminology aligned to Ensemble (overworld) / Studio (editor) / Canvas (a creator’s place) / Showcase (the ONE public canvas). Tier limits set to 1+1 free / 1+9 pro. Sprite-walk + character creator pulled into v1.0; multi-tenant overworld stays v1.5.
What this folder is
Eight planning documents that together make Ensemble a 10/10 product across every category. Each addresses one spine; together they’re the contract for the next year.
2026-04-28 evening update — three load-bearing additions are now woven through these docs: canvas-as-primitive (every published unit of work is a Canvas row, distinct from the editable Room workspace; free tier = 1 Room with 3 published Canvas slots, Pro = unlimited), procedural empty-state fill (a brand-new user’s first Canvas is auto-seeded from signup answers + GitHub so the world feels lived-in starting at user #1; spec drafted in Chunk 4, real impl in Chunk 5), and the collective-showcase landing (/index.html v1.0 is a static grid of published canvases, not a manifesto page). The binding contract for the data model is ../CHUNK_2_API_CONTRACT.md.
Read them in this order:
| # | Document | What it answers | Length |
|---|---|---|---|
| 0 | VISION_REVISION_2026-04-30.md |
The 2026-04-30 lock. Terminology (Ensemble / Studio / Canvas / Showcase), tier limits (1+1 free / 1+9 pro), sprite-walk into v1.0, Path-to-10/10 plan. Read this first. | ~3,500 words |
| 1 | MANIFESTO.md |
Why Ensemble exists. The thesis, the audience, the genre, the wedge, the bar, the promise. | ~3,000 words |
| 2 | PRODUCT_SPEC.md |
What ships and in what order. Two-surface architecture, Studio in detail, sprite-walk, component flags, tier limits, component catalog, canvas-as-primitive, procedural fill, collective-showcase landing, six phases A–F. | ~5,000 words |
| 2a | SPRITE_WORLD.md |
The v1.0 sprite-walk spec. Sprite primitive, character creator, walk physics, per-component flags, interaction-prompt UX. Sister doc to V1_5_WALKABLE_WORLD. | ~3,000 words |
| 3 | ARCHITECTURE.md |
How technically. Sandboxing, data schemas (Canvas + walk_flags), Claude API integration, multi-tenancy plan, security. | ~4,500 words |
| 4 | ART_DIRECTION.md |
How visually + sonically. Palette, typography, pixel-art rules, animation, sound, mood-board. | ~4,000 words |
| 5 | GTM_PLAN.md |
People + business. Launch sequence, business model, content, support, safety, success metrics. | ~3,500 words |
| 6 | INTERACTIVE_ASSETS.md |
The breakthrough mechanics. Click-to-launch + composable rooms + live presence + generative rooms + creator-bots. The interactions that make the Studio a platform. | ~5,000 words |
| 7 | PATH_TO_LAUNCH.md |
The launch contract — recalibrated 2026-04-27. Every v1 category scored to A+ / 10/10; ~440h of focused build to raise meaningful-success probability from 25–35% → 50–65%. | ~6,000 words |
| 8 | STAGED_PLATFORM.md |
The release cadence — v1 → v3 staged across nine releases over ~24–30 months, with a single launch-worthy hook in every release. UGC + generative-fallback principle. | ~5,500 words |
| 9 | CREATOR_ECONOMY.md |
The creator-economy flywheel. Ensemble as patron of the indie pixel-art economy, Mana Seed as the de-facto format spec, the v1.1 marketplace MVP scope. Pairs with ASSET_PACK_IMPORT_PLAN + R14_PLAN. |
~2,000 words |
Total: ~38,000 words of locked decisions. Approximately what you’d find in the early-day documents of a serious product company. Each document references back to the manifesto; if any conflict surfaces, the manifesto wins. The data-model commitment in CHUNK_2_API_CONTRACT.md is binding for Chunk 2 onward.
The TL;DR
Ensemble is the place to run, see, and share the work you make with AI — and to walk through it as a sprite you made. Four words carry the product:
- Ensemble — the overworld. The PLACE itself. The public surface a visitor lands in. v1.0 = a static grid of published Showcases. v1.5 = navigable walkable city of rooftops + plazas + paths per
V1_5_WALKABLE_WORLD.md. The first impression is community, not pitch. - Studio — the editor app where creators author canvases. One per creator. Pixel sprites in vector environments, Claude as the design partner in edit mode, procedurally-filled first Canvas so day-zero is never empty.
- Canvas — a creator’s place inside their Studio. Free tier = 1 Showcase + 1 extra (2 canvases); Pro = 1 Showcase + 9 extras (10 canvases). Visitors walk a sprite through it (v1.0); press X to interact with anything wired up. Phase A → Phase C.
- Showcase — the ONE canvas a creator chooses to publish to Ensemble. The doorway visitors walk through from the overworld.
Plus the Dashboard — the operator surface. Power-user widgets, control deck, status, kill switch. Already shipped (Phases 1–2.5); demoted from front door to discoverable surface.
Plus surfaces folded into Ensemble (Map → folded per Maestro decision) or shipped later:
- Showcase / canvas detail — fly-into-canvas view from the grid (Chunk 5).
- Feed — cross-Ensemble timeline (Phase E, multi-tenant).
Audience: solo creative makers using AI as a partner. Not just developers — also musicians, writers, designers, course creators, researchers. United by running multiple AI-augmented projects in parallel and caring about craft.
Genre: Habbo Hotel × crypto-vibe-coding power user × Apple-quality bar. Nobody owns this sweet spot today.
The roadmap
The headline number after the 2026-04-27 recalibration + 2026-04-28 canvas-as-primitive expansion: ~440 hours of focused build for v1.0. See PATH_TO_LAUNCH.md §3 for the category-by-category breakdown and STAGED_PLATFORM.md §3 for the release cadence.
| Phase | Scope | Hours | Status |
|---|---|---|---|
| v1.0 (Phases A–C, Chunks 1–5) | Studio + canvas-as-primitive + procedural-fill stub + sprite-walk + character creator + per-component flags + camera + interaction-prompt UX + tier gates + collective-showcase landing + public flip | ~440 + ~24-30 sprite-walk pull-in | IN PROGRESS |
| v1.5 (Phase D, the cliff) | Multi-tenancy + auth + multi-tenant Ensemble overworld + rooftops + town layout + composable canvases ACROSS Studios + multi-visitor live presence | ~30–50 + recalibration | After v1.0 |
| v2 (Phase E) | Feed + cross-Ensemble discovery + theme/sprite community | TBD | After v1.5 |
| v3 (Phase F) | Creator marketplace (sprites/themes/worlds) | TBD | After v2 |
Each phase ships visible value on Kevin’s Ensemble alone, so we don’t need to wait for the cliff to prove the experience. The Phase A→F mapping from the original spec is preserved internally in PRODUCT_SPEC.md §7; v1/v1.5/v2/v3 is the public-facing cadence per STAGED_PLATFORM.md.
The bar
Every category at A. Every page designed. Every animation earned. Every empty state designed. Every error state designed. No template-feeling components.
If we can’t keep the bar, we don’t ship the feature. Pause and fix, never compromise.
What’s not in this folder
- The existing A-tier plan (
docs/A_TIER_PLAN.md) — covers the dashboard work that’s mostly done. - The session handoff (
docs/SESSION_HANDOFF.md) — operational continuity between Maestro chats. - Per-project charters — live at each repo’s
.claude/PROJECT_CHARTER.md. - Memory files — live at
~/.claude/projects/-Users-kevin/memory/(orchestrator behavior tuning).
This folder is for the new product strategy. The grand plan.
Maintenance
When new strategic decisions land that don’t fit any of the five docs, we add a sixth doc rather than overstuffing one. Possible future docs:
LAUNCH_RETRO.md— written after Wave 4 (the Show HN moment).YEAR_2_PLAN.md— written at month 12 with year-1 learnings.COMPETITIVE_LANDSCAPE.md— written if a real competitor shows up.
The manifesto stays stable. The other docs get versioned in git history; major changes get a note in the doc’s footer.
Compiled 2026-04-27 by Maestro from the strategic reorient conversation. Reviewed 2026-04-27 by Kevin (assumed; locks on first commit).