adr: substrate ladder — Melvor P0, Stardew Valley P1+

ROADMAP.md

@@ -16,7 +16,7 @@ decisions still ahead, and a dated snapshot of where we are.
 
 | Layer | Is | Milestone 0 choice | Reference |
 |---|---|---|---|
-| **L1** substrate | the world itself | **Melvor** (clean-room); Stendhal / AI Town deferred | `decisions/adr-0003-*` |
+| **L1** substrate | the world itself | **Melvor = P0** (clean-room); **Stardew Valley = P1+** (dual-control); others deferred | `adr-0003-*`, `adr-0007` substrate ladder |
 | **L2** harness + bridge | executor + perception↔action seam | **TypeScript**; runtime (OpenClaw / ElizaOS / Hermes) **unchosen** | `harness/README.md`, `harness/bridge/README.md`, `decisions/adr-0006-*` |
 | **L3** cognition | autotelic drives (core IP) | **Python** | `drives/README.md`, `decisions/adr-0004-*`, `adr-0006-*` |
 
@@ -88,7 +88,7 @@ Evaluations still in motion live in `exploration/` and resolve into ADRs.
 
 | Note | Status | About |
 |---|---|---|
-| `exploration/substrate-selection.md` | converging | the substrate ladder beyond Melvor; Stardew Valley as a strong M0 candidate (native dual-control via co-op) |
+| `exploration/substrate-selection.md` | **resolved → adr-0007** | substrate ladder: Melvor P0, Stardew P1+ (dual-control staged to P1) |
 | `exploration/gateway-selection.md` | converging | operator↔agent channel; MVP = AstrBot over Discord, two-way (transport-not-brain; `discord.py` fallback) |
 
 ## Governance / ops track (parallel to the product)

decisions/adr-pending-76f26c6e-substrate-ladder.md

@@ -0,0 +1,64 @@
+---
+id: pending
+uuid: 76f26c6e
+title: Substrate ladder — Melvor is P0, Stardew Valley is P1+
+date: 2026-06-03
+status: accepted
+---
+
+# ADR-XXXX — Substrate ladder — Melvor is P0, Stardew Valley is P1+
+
+> uuid `76f26c6e` (stable alias; authored as `adr-pending-76f26c6e-substrate-ladder`,
+> numbered on merge). Output of a decision acked by the operator (cyber-ayi),
+> 2026-06-03. Resolves [[exploration-substrate-selection]]; extends
+> [[adr-0003-melvor-clean-room-first]].
+
+## Decision
+
+| Priority | Substrate | Role |
+|---|---|---|
+| **P0** | **Melvor** | Clean-room first. Stand up the research apparatus (T3 logbook · T4 drives · T5 loop) and the **agent-side dyad** (autonomous autotelic play + `private` logbook + agent→operator salient push). Single avatar; **no dual-control** — Melvor has no co-presence. |
+| **P1+** | **Stardew Valley** | The next substrate, where **dual-control** lands: native co-op (operator = Player 1, agent = Player 2, **shared farm, separate avatars**) gives the operator-acts-in-world half of the dyad without the shared-avatar concurrency problem. Adds social/relational depth. Integrate via SMAPI + existing MCP. |
+| deferred | AI Town · 2004scape (private) · Dwarf Fortress (stretch) | per [[exploration-substrate-selection]] |
+
+**Consequence for the MVP scope.** The MVP target is full Milestone 0 *including*
+dual-control. Because Melvor structurally cannot do dual-control, that half is
+**staged to P1 (Stardew)**: P0 proves the apparatus + the agent→operator half on
+Melvor; P1 completes the dyad with operator→world dual-control on Stardew.
+
+## Context
+
+| Constraint | Source |
+|---|---|
+| Substrate complexity is a tax that does not buy research value — value is in L3 | `exploration/substrate-selection.md` |
+| Melvor is lowest-complexity + deterministic — best to stand up the apparatus | `adr-0003`; `harness/README.md` |
+| MVP = full Milestone 0 *with* dual-control (operator + agent act in one world) | operator decision (2026-06-03) |
+| Melvor has no co-presence → cannot exercise dual-control | `exploration/substrate-selection.md` |
+| Stardew co-op = separate avatars in one world → native dual-control, no shared-avatar race | `exploration/substrate-selection.md` (sources: StardewValley-MCP, SMAPI) |
+| The `Bridge` contract is substrate-agnostic → P0→P1 swap is an adapter change | `harness/bridge/`, `adr-0006` |
+
+## Consequences
+
+**Positive**
+- P0 minimizes complexity to prove the loop/drives/logbook deterministically.
+- The substrate-agnostic `Bridge` (T1) makes the Melvor→Stardew swap an adapter
+  change, not a rewrite; a `SyntheticBridge` fixture can develop dual-control
+  logic before either real adapter exists.
+- Stardew's co-op delivers dual-control without solving shared-avatar concurrency.
+
+**Negative**
+- Full dual-control Milestone 0 is **not** reached on Melvor alone — it is staged
+  to P1; the "agent-side" and "operator-side" halves land on different substrates.
+- Two adapters to build (Melvor mod; Stardew SMAPI/MCP).
+
+**Rejected alternatives**
+- **Stardew as P0** — front-loads more complexity (real-time-ish, co-op, SMAPI)
+  before the apparatus is proven; against the clean-room principle.
+- **Skip Melvor, go straight to dual-control** — loses the deterministic clean-room
+  that de-risks T3/T4/T5.
+
+## References
+
+- Resolves: `exploration/substrate-selection.md`
+- Extends: [[adr-0003-melvor-clean-room-first]]; relates to [[adr-0004-autotelic-drive-layer]], [[adr-0006-polyglot-ts-harness-python-cognition]]
+- `TASKS.md` (T2 adapter), `harness/bridge/` (substrate-agnostic contract)

exploration/substrate-selection.md

@@ -1,16 +1,16 @@
 ---
 topic: substrate-selection
-status: converging
+status: resolved
 date: 2026-06-03
 related-adrs: [adr-0003]
-resolves-to:
+resolves-to: adr-0007
 ---
 
 # Substrate selection — which world(s) the dyad inhabits
 
-> Status: **converging**. Not binding. `adr-0003` already locked *Melvor as the
-> clean-room first substrate*; this note evaluates the substrate **ladder** beyond
-> it and surfaces a strong new entrant (Stardew Valley). Resolves into a new ADR.
+> Status: **resolved** → [[adr-pending-76f26c6e-substrate-ladder]] (operator,
+> 2026-06-03): **Melvor = P0, Stardew Valley = P1+**; dual-control staged to P1.
+> The analysis below is the reasoning trail; the decision lives in the ADR.
 
 ## Question