docs(plan): Phase L.3.1 — InterpolationManager core implementation plan
10-task incremental plan with explicit subagent dispatch points:
- Tasks 0+1+2 dispatched in parallel (3 concurrent Sonnet subagents):
Task 0 = decomp dive to settle UseTime head-vs-tail blip ambiguity
Task 1 = InterpolationManager class + ~13 unit tests
Task 2 = MotionInterpreter.GetMaxSpeed() + ~3 unit tests
- Tasks 3-6 sequential GameWindow edits (env-var gated, dual-path):
Task 3 = RemoteMotion gains Interp field
Task 4 = OnLivePositionUpdated MoveOrTeleport routing
Task 5 = per-frame remote tick Interp.AdjustOffset add
Task 6 = OnLiveVectorUpdated.Omega application
- Task 7 = USER GATE (visual verification)
- Tasks 8+9 dispatched in parallel after sign-off (2 subagents):
Task 8 = cleanup commit (delete env-var, dead paths, soft-snap residual)
Task 9 = roadmap update (insert Phase L.3 entry)
Each task has TDD-style steps with exact file paths, code blocks,
build/test commands, and commit messages. Plan honors CLAUDE.md
direct-to-main + commit-after-each-step + visual-verify-on-motion.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
f28240ad19