--- name: parked-2-9-0-update description: "SUPERSEDED as a feature slate — the item list now lives in QUEUE.md. Retained for the durable rulings the queue does not carry: the ONE-MAP consolidation, the per-game map integration requirement, the ENCOUNTER flow decision, the inventory-foundation discipline, and what the owner DROPPED and why." metadata: node_type: project originSessionId: parked-2-9-0-update --- *Correction 2026-07-20: this file used to hold the full curated gameplay-update slate. That list is now carried authoritatively and re-sequenced in `QUEUE.md` — a second copy here could only drift and mislead. The slate has been removed; what follows is the reasoning behind it, which the queue does not record.* The gameplay round was scoped as immersive/gameplay work on the existing games — **all free / bring-your-own-key, game-agnostic, deterministic-native wherever possible.** Owner's locked versioning convention: **a new game means a new major version**, which is what fixes where this round sits relative to the next game. ## ⭐ CONSOLIDATION — ONE MAP (owner 2026-06-30, from the roadmap audit) The map had been designed **three separate times** (a phosphor-cartography SVG minimap · a CSS button-grid "Recon Grid" · a geographic SVG map). **DECISION: build the geographic per-game SVG map as the single target** — most immersive, real per-game geography, the "Pip-Boy map" players actually expect. Use the phosphor-cartography inline-SVG approach as its LOW-RISK foundation and first iteration (coordinate nodes + radar sweep + fog, evolving toward true geography — same SVG lineage). **⛔ DROPPED: the Recon Grid.** It was the awkward middle — still abstract, the worst fit on a narrow phone, the highest regression surface, and it never reached the geographic payoff. The exploration/journal work is a DATA layer on top of the one map, not a second map. **⚠️ REQUIREMENT — adding a new game's map must be STRAIGHTFORWARD, but it does NOT have to be a zero-code drop-in** (owner 2026-07-01, clarifying his own earlier framing). The distinction he drew is the important part: **it is fine and EXPECTED that authoring a new game's actual map — the artwork, per-location coordinates, region shapes — takes a long time. That is inherent labour, not a problem.** What must NOT be hard is the **integration/plumbing**: architect the map with a clean, well-defined per-game hook following the existing per-game definition pattern, so slotting a new game's map in doesn't require a painful refactor or fighting the system. Some per-game config or code is acceptable; a heavy rework to add a game is not. **Goal: when the next game arrives, map integration is a clean understood path and the only big cost is drawing the map.** Pragmatically, not dogmatically. ## ⭐ WASTELAND UPLINK — merge, don't parallel Radio broadcasts and encounter rolls merge into the ONE ambient engine ([[parked-wasteland-uplink-engine]]) alongside the other ambient transmissions, sharing the bulletin bank and seeded-roll infrastructure. Day/night is IN (the owner reversed himself back to yes, with rad-storm). **Hard invariant: the ambient engine touches no stats and writes no saves.** ## ⭐ UX DECISION — ENCOUNTER is one entry point, but the pieces stay à la carte (owner 2026-07-01) Threat assessment + VATS + combat log + loot are ONE combat loop, not four separate panels. Surface it as a **single ENCOUNTER command** that runs the guided sequence — assess threat → VATS turns → auto-write to the combat log → loot — each stage feeding the next (the assessed enemy's HP/DT pre-fills VATS; the VATS result auto-logs; defeat rolls into loot). **KEEP every sub-step independently accessible** for the standalone cases: loot a container with no fight, assess without engaging, manually log a narrative kill. ENCOUNTER is the all-in-one loop for an actual fight; the pieces remain available for edge cases. ## ⭐ THE INVENTORY PANEL IS THE FOUNDATION — and its KEY DISCIPLINE The inventory overhaul (sort/search toolbar + per-row inspect drawer surfacing the item DB + an in-panel loadout header + per-row equip) is the **foundation that crafting and the loadout manager sit on. Do it FIRST, then layer them.** **KEY DISCIPLINE: keep the stored inventory state UNCHANGED — derive everything from the item DB at render time**, so saves and cloud round-trip untouched and there is no migration risk. Related discipline for long lists (inventory, registry, changelog): virtualize as part of that same overhaul. ## ⚠️ RECURRING OWNER PAIN POINT — native features need native input paths Owner: **"not being able to do a native feature without the ai keeps happening, needs fixed."** Every native feature must have a **native input/edit path** — no AI required for data entry. The concrete case that surfaced it: a manual reputation change must be able to record *why* it changed, natively, not via the AI. This is codified as the player-control principle: **the AI narrates and assists ON DEMAND and is never authoritative over player state — the player always drives.** RNG/random modes (encounter rolls, seeded combat, loot) are the one sanctioned exception, and they are always opt-in. See [[player-control-principle]], [[parked-ai-to-native-feature-audit]]. ## ⭐ THE HACKING MINIGAME AS A DIEGETIC GATE — and the real deliverable behind it Owner added the iconic RobCo hack (seeded word puzzle, likeness scoring, attempts/lockout, offline, no AI). **The PAYOFF is the point: a successful hack UNLOCKS a DEV/DEBUG PANEL that must actually be BUILT** — a comprehensive in-terminal debug/test console with everything useful for testing the site: force-fire every ambient feature (this IS the "the dev build must be able to test these features" affordance), inspect and edit state, dump save/version/checksum, toggle every effect, simulate thresholds, force boot variants, view telemetry. **The hack is the diegetic gate; the debug panel behind it is a real deliverable.** Design it so it cannot corrupt saves — additive and confirm-gated wherever it writes. Related constraint carried over: **ambient features must never confuse in-GAME with in-TERMINAL**, and the emergent CRT "condition" effect **MUST be toggleable off** and testable on the dev build. ## MERGE, DON'T BUILD PARALLEL - The Overseer's incident log **merges with the existing session-stats/log telemetry** — owner: *"there are already things tracking this stuff."* - The aid/consumables manager **merges with the existing partial implementation** — don't build a second one. - Session recaps ("Previously on…") are a new **VIEW over the one canonical terminal record**, not a new store. ## ⛔ WHAT THE OWNER DROPPED, AND WHY (record only — do not build) - **Holotape Archive & Audio Logs → DROPPED.** Owner: *"way too many and it's a feature not many would use."* - **Survival / Hardcore Tracker → deferred as a POTENTIAL STANDALONE future update.** Owner judged it big enough to be its own version if ever pursued. On file, not actively planned. - **Removed from the slate** (owner: *"any number I didn't list, remove"*): achievements/challenges tracker, NPC/character codex, standalone encounter/loot roll generator. - **Bench alternates** if anything is cut: a native derived-stats readout, a Pip-Boy snapshot share image, a per-game CRT boot-theme customizer. - The streaming two-phase AI narrator is an **AI-contract rework**, not gameplay work — it belongs with the AI/context work, not this round. ## ★ AI-AS-GAME-MASTER — flagged as needing its own design pass How a solo session actually FLOWS with the AI as your Overseer/GM — scene framing, how AI and native systems talk — is currently scattered across several features and deserves its own first-class design pass. Governed throughout by [[player-control-principle]]. Related: [[parked-deferred-work-program]], [[parked-fallout4-support]], [[parked-wasteland-uplink-engine]], [[parked-ai-feature-eval]], [[player-control-principle]], [[parked-ai-to-native-feature-audit]].