Low-Context Execution Packet

Use branch stage/d3-02-kinetic-aesthetic-mapping.

Start from the capture and input controls documented by D3-01. Do not guess a physics curve from visual memory; measure at least three timed input trials.

Recommended capture layout:

/mnt/kitchen/collab/d3-tui/captures/d3-02/
  idle/
  hold-left-500ms/
  hold-left-1000ms/
  release-decay/
  snow-idle/
  snow-input/

Frame extraction command:

ffmpeg -y -i input.mp4 -vf fps=60 frames/%06d.png

Create a small analysis script only if it saves repeatable work. Suggested location:

scripts/analysis/d2_motion_probe.py

Minimum data table for docs/D2_PHYSICS_MANIFEST.md:

trial_id,input,hold_ms,start_frame,end_frame,observed_degrees,estimated_deg_per_sec,notes

Measure:

• logo angular acceleration while held
• maximum observed angular velocity
• decay/friction after release
• any clamp, overshoot, or easing
• snow layer count
• snow drift angle and speed by layer
• whether logo motion changes snow/audio/text behavior

Allowed methods:

• optical flow or template matching over captured frames
• visible marker tracking on the logo
• emulator memory watches only if D3-01 documented a safe method

Forbidden methods:

• committing decompiled proprietary source
• presenting raw memory addresses as final logic without a clean-room interpretation

The manifest must give pseudocode with named constants and confidence levels. Use this shape:

velocity += input_axis * LOGO_ACCEL * dt;
velocity = clamp(velocity, -LOGO_MAX_SPEED, LOGO_MAX_SPEED);
velocity *= pow(LOGO_FRICTION, dt * 60.0f);
angle += velocity * dt;

If a value is estimated, label it estimate and state the source capture.