game-design

Game design (solo/small-team method)

Docs/DESIGN.md is the living source of truth (unity-init stamps it). Every design decision made in conversation gets written back there — a decision that only lives in chat history is lost.

Core loop first

Before building anything: what is the player's verb (jump, trade, sneak, grow)? What happens in a 30-second loop, and what makes the player want the 31st second? If the loop can't be stated in two sentences, the project isn't ready for features — help the user find it, prototype-cheap.

First playable slice

The smallest build that proves the loop is fun: one scene, one mechanic, placeholder art, no menus. Resist building systems (inventory, save, settings, meta-progression) before the loop is validated — a fun-less game with a great save system is still fun-less. The slice is also the correct scope for unity-init's first bootstrap.

Scope discipline

• Every feature costs ~3× the estimate once art, UI, edge cases, and saves are counted. Say this out loud when features are proposed.
• Content scales linearly (another level ≈ same cost each); systems scale combinatorially (each new system must interact with every existing one). When cutting, cut systems, not polish.
• When the user proposes a mid-project feature: check it against DESIGN.md's core loop. If it serves the loop, size it honestly; if it doesn't, surface the trade-off ("this delays the slice by ~X; it doesn't touch the core loop — park it in DESIGN.md's ideas list?") instead of silently building it.

Juice (after the loop works, not before)

Cheap polish that multiplies perceived quality — apply as a checklist to the working loop: screen shake (small!), hit-stop (2–4 frames), squash & stretch, particles on every impact, a sound on every interaction (see unity-audio), tweened UI (nothing snaps), controller rumble if applicable. An afternoon of juice beats a week of features for how good a game feels.

Playtesting

• Watch, don't explain. Where the player stalls or asks "what do I do?" is a design bug, not a player bug.
• Treat DESIGN.md as falsifiable: playtest findings update it — especially the Open questions section.
• After each playtest round, the next milestone is whatever confused or bored players most — not the next feature on the list.