@@ -2,16 +2,10 @@ MODEL: GPT-5.3-codex
22REASONING: high
33
44COMMAND:
5- Create BUILD_PR_LEVEL_17_8_SAMPLES_1605_TO_1608_CORE_TRACK .
5+ Create BUILD_PR_LEVEL_17_9_SAMPLE_1605_DRIVING_CONTROL_FIX .
66
77Goal:
8- Complete the remaining core Phase 16 / 3D sample track by implementing samples 1605 through 1608 as one testable package.
9-
10- Context:
11- - 1601 is visible
12- - 1602 and 1604 visibility defects were fixed
13- - targeted Phase 16 visibility sanity passes
14- - targeted smoke for 1601-1604 passes
8+ Fix Sample 1605 - 3D Driving Sandbox so left/right steering no longer gets stuck.
159
1610Constraints:
1711- one PR purpose only
@@ -20,32 +14,20 @@ Constraints:
2014- no zip output from Codex
2115- keep 2D and networking untouched
2216- do not modify start_of_day
23- - preserve prior Phase 16 samples unless a shared safe improvement is required
2417
2518Implement:
26- 1 . Add and wire up:
27- - samples/phase-16/1605
28- - samples/phase-16/1606
29- - samples/phase-16/1607
30- - samples/phase-16/1608
31- 2 . Reuse existing shared Phase 16 helpers where stable.
32- 3 . Ensure each sample has visible 3D output on first load and one clear teaching purpose:
33- - 1605 driving sandbox
34- - 1606 physics playground
35- - 1607 space shooter
36- - 1608 dungeon crawler
37- 4 . Update samples/index.html so these are live entries when implemented.
38- 5 . Extend the existing targeted Phase 16 runtime sanity only if needed and keep it surgical.
19+ 1 . Inspect samples/phase-16/1605/DrivingSandbox3DScene.js.
20+ 2 . Fix the steering/input-state bug causing left/right navigation to stick.
21+ 3 . Preserve current visible 3D output and forward/reverse behavior.
22+ 4 . Add or extend the smallest targeted behavioral sanity check if useful.
3923
40- Validation:
41- - verify 1605 visible on load
42- - verify 1606 visible on load
43- - verify 1607 visible on load
44- - verify 1608 visible on load
45- - verify intended controls function for each sample
46- - run targeted smoke for 1601-1608
47- - confirm no 2D regression in targeted checks
48- - confirm no networking regression in targeted checks
24+ Validate:
25+ - verify left turn works while held
26+ - verify right turn works while held
27+ - verify steering stops when released
28+ - verify opposite turn can be engaged immediately
29+ - verify sample still renders on load
30+ - run targeted smoke for 1605
4931- update:
5032 - docs/dev/reports/change_summary.txt
5133 - docs/dev/reports/validation_checklist.txt
0 commit comments