jesse
b3a1044df6
CI / quality (push) Successful in 54s
- New shared Chassis.luau: geometry constants + per-car tuning from stats - CarManager.buildCarModel rebuilt: prismatic+spring suspension per corner, servo-hinge steering on front knuckles, motor-hinge driven wheels, hidden dense ballast for anti-roll; VehicleSeat kept as input device only - New client ChassisController: drives motors/servos from seat ThrottleFloat/SteerFloat per Heartbeat; speed-sensitive steering lock; engine braking on throttle release - Driver gets network ownership of every unanchored part (constraint assemblies are separate; missing one causes stuttery physics) - Docs: MESH_SWAP.md rewritten for shell-on-chassis workflow; ROADMAP updated
3.0 KiB
3.0 KiB
Roadmap — realistic solo-dev plan
Assumes one part-time developer learning as they go. The MVP is intentionally tiny — just driving — so there's a fun, testable thing FAST. Everything else is layered on only after the driving feels good.
Expect months, not weeks to a polished full game. That's normal. This order guarantees you always have something that runs.
✅ M0 — Drive-only MVP (DONE — this repo)
Spawn on the neon expressway, drive a car, switch cars, respawn. RX-7 demo-locked to owner + son. Cars persist between sessions. CI green on a self-hosted runner.
M1 — Make driving feel GOOD (the real work, do this next)
- Adopt a constraint chassis for real suspension/grip instead of the raw
VehicleSeat: springs + prismatic suspension, servo steering (speed-sensitive
lock), motor-driven wheels, engine braking, ballast anti-roll. Shared tuning
in
src/shared/Chassis.luau; client-sideChassisControllerdrives it (driver owns network physics). Needs a Studio playtest to tune feel. - Import the real RX-7 mesh; weld the shell onto the chassis
Body. Seedocs/MESH_SWAP.mdfor the exact steps (needs Studio for asset upload). - Tune acceleration curve, grip, and the chase camera (after first playtest).
- Sound: engine loop, tire screech, boost. Exit test: the RX-7 is genuinely fun to just drive around. Ship nothing else until this is true.
M2 — More cars + polish (still no racing)
- Import a few more real car meshes.
- Better map: more expressway sections, ramps, tunnels, scenery.
- Day/night, weather mood.
M3 — Racing (the first "game" layer)
- Checkpoints + timing (a new RaceManager — the old one was removed in M0).
- Personal best times.
- Simple point-to-point and drag modes.
M4 — Progression & economy (only once racing is fun)
- Money from races, a garage, buying/unlocking cars, upgrades, reputation. (These systems existed as scaffolding and were cut for the MVP; reintroduce them deliberately when they have a purpose.)
M5 — Systems hardening & monetization
- Leaderboards, anti-cheat pass 2, DataStore backup drill.
- Performance pass (StreamingEnabled, part budget, mobile).
- Fair gamepasses/products — never pay-to-win.
M6 — Traffic AI (deferred hardest system)
- Spline-following dummy cars first (cheap). True AI only if perf allows.
Standing engineering rules
- Every change must pass
stylua --check,selene, androjo build— CI on the self-hosted Gitea runner enforces all three on every push. - Keep the client dumb: it requests, the server decides. The RX-7 allowlist and anti-cheat live server-side.
- Add tunables to
GameConfig, not scattered magic numbers. - Add a save key by adding it to
ProfileTemplate(auto-reconciles); never remove a key without a migration (bumpProfileStoreName_vNin dev to wipe). - Test in Studio after every server change (ProfileStore uses a mock store there).