Z6 Resonance Lab

v0.4 · autopilot
NOTHING HERE IS VERIFIED. σ=126 is a conjectured, unit-free integer. The engine phase-tags samples, validates the sensor on a known physiological effect, then runs a corrected sequential test to a conclusion — built to reach null.
signal · idle

Instrument

Arm assignment

Blind hides the arm until you reveal it in the Report — keeps every reading honest.

Drive outputs on the on-phase

Tone
sine, live
Vibration
navigator.vibrate
Screen tint
flash on the on-phase

Timing

on/off = half period each≈ 12 min 36 s

Autopilot runs to conclusion

Sequences both campaigns in parallel: validates the sensor, blind-assigns arms, runs sessions back-to-back, feeds each into the corrected sequential test, and stops each instrument on its own verdict. Active tasks (tap/RT) still need you present; passive tasks (HR) just need a fingertip on the camera.
Campaign A · 126 vs 100
tap, reaction time, pulse
Campaign B · 32.5 vs 30 Hz
audio→HR, audio→RT
Passive only
skip tasks needing taps — fully hands-off (finger on camera)
Live hardware check. Nothing logged here.

Microphone

dB rel

Motion

m/s²

Outputs self-test

Capabilities

Pre-registration unlocked

Fixed checkpoints at N/arm = 6, 12, 18, 24, 30. Difference boundary z=2.576 (≈5% type-I across looks); equivalence by TOST 90% CI. This is what stops the tool from finding effects in noise.

Automated conclusions

Sessions 0

No sessions yet.

How the autopilot works

1. Validate the sensor. Slow breathing at ~6/min (0.1 Hz) is a known large boost to heart-rate variability (baroreflex resonance, well documented). The RSA session checks the phone PPG reproduces it. If it doesn't, every 126 null is uninformative — the sensor is just blind.

2. Pre-register. Lock instrument + smallest-effect-of-interest before collecting. Checkpoints are fixed at N=6/12/18/24/30 per arm.

3. Sequential test to a conclusion. At each checkpoint: TOST equivalence (90% CI inside ±SESOI → no effect) or difference (z=2.576 CI excludes 0 → effect), else keep collecting. The z=2.576 boundary keeps the false-positive rate near 5% despite repeated looks — a naive test peeking every session fires false effects 30–40% of the time.

Why these boundaries (grounded in simulation)

feasibility per instrument
  • Tap entrainment — declares equivalence ~87% under H0 by ~18/arm. Strong concluder.
  • Reaction time — detects real effects fast; slow to conclude "no effect".
  • Resting HR (cross-arm) — 1.5 bpm SESOI vs ~4.5 bpm day-to-day noise → basically never concludes. Kept for within-session + the positive control only.
  • Greenness — concludes, but needs ~24 sessions/arm ≈ weeks.

PPG accuracy grounding: fingertip contact PPG ≈ 1 bpm RMSE over 60 s vs ECG, with a rare outlier tail — so the app uses robust windowed stats.

Verified vs conjectured

verified Kernel results under decide/native_decide stay in the kernel. conjecture Everything here — that σ=126 is a physical cadence — is staged to be tested, never asserted.