RELAX · reward α (8–12 Hz) up · lower frequency, eyes-closed works best
connecting to live EEG…
audio off · bath · auto 5 min · press A
2-channel flow proxy · div +0.00 · curl +0.00
Posterior N5 Pz + N4 ≈P4/O2 · waiting for signal…
CH1 CH2
band (norm)
0.00
Δ 0.00
reward
0.00
rhythm Hz
—
Δ 0.00
PAF Hz
10.0
Δ 0.00
χ exponent
1.50
Δ 0.00
emg
0.00
EMG detected — feedback frozen
C calibrates your session baseline. PAF drives radial speed; χ drives
high-frequency turbulence.
Eyes closed:A toggles audio and S cycles bath, forest,
and rain; W triggers one capped wake pulse. Broadband rhythm controls
pitch across an expanded 40–3200 Hz audible range, cadence, and stereo orbit; its
short-term variation widens the orbit and melody. Alertness and entropy morph
the tuning from gently tense to consonant. The
two-channel div proxy expands/contracts the field and harmony.
N5 at Pz plus N4 in the approximate right P4/O2 region add independently
artifact-gated posterior vigilance, drowsiness, and exploratory on-task
proxies; these are personal trends, not a detector of thoughts.
A closed-loop arousal deficit controls brightness, tempo, contrast, and sparse
novelty cues; EMG rejection freezes misleading changes. RL learn tests
bounded pitch/light combinations for about 20 seconds, credits the recent
two-minute action history, and stores learned scores plus summarized clean
trial outcomes locally—never raw EEG. Click to hold/resume; Shift-click to
reset learning. RL log shows what has worked across sessions. Above
the arousal target, alerting flashes recede while attention and complexity
reveal more varied orbital families, counter-rotation, and slow precession.
Persistent RL log
No completed clean trials yet.
What consistently works
Recent clean trials
One record per ~20-second clean-EEG trial; history is capped at 500 trials.