Question
Experiment 002 / Signal becomes state
IntentGate
A pinch is not one frame. Cross a strict distance threshold, hold for 120 ms, then release through a wider gate.
Recent transitions
- UNKNOWNAwaiting evidence
Ready. Camera permission has not been requested.
Experiment record / 002
Frames suggest. Time decides.
Hypothesis
Palm-normalized distance, temporal confirmation, and hysteresis can separate deliberate pinch intent from one noisy frame.
Observed
- Clean pinch activates only after sustained evidence.
- Near miss, threshold jitter, and short tap stay negative.
- Slow release remains stable across the wider continuation gate.
- Missing or impossible geometry returns unknown.
Timing contract
- Activate
- ≤ 0.34
- Confirm
- 120 ms
- Dropout grace
- 100 ms
Failure boundary
Fixture evidence validates state transitions, not varied hands, cameras, sleeves, skin tones, low light, or motion blur. Those cases remain open.
Decision / revise
Keep hysteresis and explicit unknown. Test the thresholds physically.
The state machine is technically legible and rejects noisy fixtures, but its constants remain hypotheses until broader camera validation.