Symptom. Scripted physical motion doesn't execute, even though the state/audio/event
layers fire correctly. Idle-motion head glances DO work, so it's choreography/sweep-specific.
Evidence.
- US14 dance:
Dance mode: macarena streams audio, but no state_changed → dance,
no _handle_dance, no rainbow ring, no body movement. Resolves via song-playback path only.
- US5 security:
state_changed → security + security_pose scan_start emitted, capture
loop started, photo taken — but the yaw sweep (−500→+500→0) never executed.
Expected. Dance runs choreography + rainbow ring; security performs the yaw sweep.
Log excerpt.
20:04:02 Dance mode: macarena / streaming singing audio macarena.mid # no state_changed->dance
20:07:14 event state_changed -> security ; event security_pose scan_start # but no physical sweep
Note. US14 "didn't move this time" suggests possible intermittency. audio_capture_pending
in the security loop is the known #31 pending, not part of this issue.
Video verification (2026-07-11 phone recording IMG_1295.MOV, starts 19:41:32).
- Dance window: frames at 20:04:04 / 20:04:12 / 20:04:22 show the head and body in an
identical pose across 18 s while the mouth/speech-bubble animates — audio playing,
zero choreography, no rainbow ring visible.
- Security window: frames at 20:07:17 / 20:07:28 / 20:07:38 show the camera-capture view
then a return to the neutral face with no yaw change across 21 s — capture loop ran,
sweep didn't.
- Nuance preserved: for dance there is no
state_changed → dance in any log, so the
firmware dance state may never have been entered (event emission vs state entry is
unresolved without serial capture); for security the state was entered and the pose
event emitted, yet the sweep still didn't run. Two distinct failure layers.
From the filmed UAT session 2026-07-11 (uat-sessions/2026-07-11/: results.csv, container logs, two video recordings). Investigated and drafted with AI assistance (Claude Fable 5 via Claude Code); human-reviewed before filing.
Symptom. Scripted physical motion doesn't execute, even though the state/audio/event
layers fire correctly. Idle-motion head glances DO work, so it's choreography/sweep-specific.
Evidence.
Dance mode: macarenastreams audio, but nostate_changed → dance,no
_handle_dance, no rainbow ring, no body movement. Resolves via song-playback path only.state_changed → security+security_pose scan_startemitted, captureloop started, photo taken — but the yaw sweep (−500→+500→0) never executed.
Expected. Dance runs choreography + rainbow ring; security performs the yaw sweep.
Log excerpt.
Note. US14 "didn't move this time" suggests possible intermittency.
audio_capture_pendingin the security loop is the known #31 pending, not part of this issue.
Video verification (2026-07-11 phone recording
IMG_1295.MOV, starts 19:41:32).identical pose across 18 s while the mouth/speech-bubble animates — audio playing,
zero choreography, no rainbow ring visible.
then a return to the neutral face with no yaw change across 21 s — capture loop ran,
sweep didn't.
state_changed → dancein any log, so thefirmware dance state may never have been entered (event emission vs state entry is
unresolved without serial capture); for security the state was entered and the pose
event emitted, yet the sweep still didn't run. Two distinct failure layers.
From the filmed UAT session 2026-07-11 (
uat-sessions/2026-07-11/: results.csv, container logs, two video recordings). Investigated and drafted with AI assistance (Claude Fable 5 via Claude Code); human-reviewed before filing.