Design / Eng Sync — 2026-09-02¶

Roughly 44 minutes. This page captures the design and technical decisions only.

1. Rune grammar — scope locked¶

Reverse is cut. It behaves like an adjective/modifier, roughly doubles grammar scope, and many cases are not clean opposites. Removed from First Playable.
Enhance is the same risk, deferred. The other adjective-type modifier (boosts other effects, adds a quantity dimension). Parked under Requires Update. Get baseline VERB + NOUN solid first, then revisit — ideally add Enhance without a new grammar dimension.
Grammar is now strictly one VERB + one NOUN. No adjectives in the baseline.

2. Rune interaction — confirmed model ("locked, not changing")¶

The scene is full of runes (on walls, ground, tree-shaped, etc.). Slots are gone.
Flow: walk up → press Space → the rune activates and glows. Walk outside its range → it auto-deactivates (prevents carrying a Level-1 rune into a Level-100 area).
Anything shaped like a rune is a rune, and it is still a physical object (can be moved/rotated/acted on). Same for "rooms": any environment object shaped like a room counts as a room.
Sentence = 2 activations max (one verb + one noun). Two identical → cancels, no effect. Order does not matter.
All effects are instantaneous, not persistent — e.g. rise wall lifts briefly then completes and deactivates. Persistent effects (continuous spin until stopped) are a future system, out of scope for First Playable.
Effect range is centered on the NOUN.
The change period until activate line is dropped.

3. New key mechanic — "matchstick" room assembly¶

Not free drawing. Some rooms are built on a room-base made of "matchsticks." The player removes / moves / rotates parts until it forms the target glyph → the room auto-activates.
With the watch held: writing a verb into it, then touching a scene noun, centers the effect range on the player (small, limited radius — an intentional constraint), because the verb is "in your hand."
Noted inconsistency: verb-in-hand + noun-in-world centers the effect circle on you, not the noun. Agreed workflow: activate the verb first → walk next to the target → create the effect.
Action: assign an engineer to build the matchstick system; a designer to design the detailed form.

4. Memory Palace → "mind world" rework¶

Old orb version replaced. New: press a key → thought-bubble above head → jump into your own mind (a dark, non-architectural, regularly arranged memory space).
Purpose: review collected runes. It is a definition space that gives no literal meaning — it captures a screenshot of each rune/route when first encountered and pins the image beside it. No text.
This replaces the notebook and the 2D UI. The runes themselves are the notebook.
Research needed: the transition into it (load-a-new-level approach; nothing as fancy as high-end teleport transitions).

5. Year scope + level design — "Cascade Town" (codename)¶

One-year target (whole game, current goal): player wakes → goes straight to Cascade Town → climbs to the sky layer → overlooks the town, which itself forms a rune → end. Just this chapter plus the opening piece. Everything else is "if time remains."
Cascade Town is a codename only; it is a geometric building cluster. Starts as ruins → player repairs → ascends. Not literal stairs (e.g. stand on a box, box flies up).
Structure ("pineapple"): 5 sheets × 6–7 zones each × several parts. Section 5 is the final endpoint. One zone is the Garden. Each area carries danger.
Art pipeline: concept art first → a small in-engine scene to test final look/feel (no lighting yet).
Art-direction note: the game has drifted too "architectural." Want a ruined-library feel, cliff/mountain-anchored rather than isolated floating islands — no longer "sky islands"; islands should be bigger, hanging off cliffs.
Visual north star: top-down, like Death's Door.

6. Camera¶

Keep the existing zone-based, fixed-position system. The previous demo moved the camera too often. Mostly follow player, with occasional slight focus (e.g. panning to a mural). Minimize movement, keep the feature, optimize only.

7. Narrative & art — murals vs. carving¶

Narrative delivered via murals / 2D pictures (e.g. the Architect forging runes). Reference: the Babel murals from Act 1.
Memories: entering certain areas triggers a 2D illustration slideshow (story beat). ~2 memory spots now. One artist draws all memories for a unified style. 2D keyframes are doable in UE.
Mural technique — leaning DECAL over ZBrush carving: easier, reusable (same art as wall decal and on paper/book). Carving looks better but is "too fancy" for an aged look; the decision hinges on art-team bandwidth.
Decal + rune combo: if a mural depicts a rune/staircase, it should be touchable/activatable. A rune can be an image: cut it out, place a room with the match left empty, bring the image in. Needs research.

8. Level streaming + VCS — most urgent technical problem ⚠️¶

Flagged as important; to be discussed in more depth separately. Blocker: "without it, designers can't work."

Each designer owns one square map tile. UE levels cannot be merged → if everyone edits one map, submit conflicts are guaranteed.
Fix (not VCS-dependent): split each tile into its own map; each designer commits only their own tile → no conflicts. Add connector maps to stitch tiles; assemble in a master map.
Streaming: don't load the whole world. Load adjacent tiles (1–2 neighbors) so a player sees neighboring rooms; hide farther areas with fog/clouds.
No black-screen on the same layer (seamless walking). Black-screen is acceptable between layers / up stairs.
Proposed UE approach: big level + sub-levels (sub-levels unlit; lighting lives in the master world light and copies down) → points at World Partition / Level Streaming.

VCS note (ties to the ongoing GitHub-vs-Perforce thread)¶

Key realization: the merge pain is identical on GitHub and Perforce — UE levels can't merge either way. The real fix is splitting maps, not switching VCS.
Art/design report that GitHub + UE is painful for non-engineers, which is pushing the room back toward Perforce. Needs a decision after more research.

9. Team structure¶

Concern: structure is too fragmented (1 designer + 1 artist per level, plus design+art per rune → many tiny sub-teams).
Preferred model: don't hard-assign. Sub-teams self-organize; when they have a need they come to us, we assign the task and tag the responsible designer under it.
Staffing: recruiting additional modelers oriented toward natural + ruined-architectural assets.

Decisions locked¶

Reverse cut; Enhance deferred to Requires Update; grammar = 1 verb + 1 noun.
Interaction model (proximity + Space activate, range-based deselect, noun-centered range, instant effects, 2-token sentences) — final, not changing.
Mind-world replaces the orb + notebook + 2D UI.
Year scope = Cascade Town chapter only, ending on the town-as-rune overlook.
Visual target = top-down, Death's Door-like; cliff-anchored, not sky-islands.
Map split into per-designer tiles + sub-level streaming is the path.

Open questions¶

Mural: decal vs. carving — pending art bandwidth.
VCS: GitHub vs. Perforce — art/UE friction pushing toward Perforce; needs a call (see §8).
Same-layer seamless streaming vs. acceptable black-screens — exact rule per transition.

Action items¶

Eng → matchstick room-assembly system.
Eng → mind-world entry + level-load transition research.
Eng → World Partition / sub-level streaming spike (adjacent-tile load, world-light copy-down, fog occlusion).
Eng → camera zone optimization (reduce movement, keep follow).
Design → priority ordering of matrix runes for First Playable; per-tile map ownership plan.
Art → all memory 2D illustrations (unified style); concept art for Cascade Town before the in-engine test.
Lead → VCS decision (GitHub vs Perforce); recruit modelers (natural + ruined-architectural).