Live Dashboard
Continuous neural-state firewall — fatigue, agency, and a privacy airlock that fail-closes private intention classes so they don’t leave this app until you unlock. Research / simulation only.
Higher → more demand
Engagement proxy
Accumulates slowly
Control confidence
Distribution shift
vs baseline (toy)
Intention stream
Policy status
- No actions this tick.
Privacy airlock
Fail-closed containment for private intention classes. Blocks gated stream output and untrusted sinks so “private thoughts” don’t leave this app until you unlock. Computer-side simulation only — not implant encryption, not Neuralink security.
Baseline drift adaptation
Armed · waiting for stabilitySlow non-stationarity tracking: when anomaly stays low, the auth/anomaly baseline drifts toward the current stream so long sessions don’t false-lock. Freezes during high anomaly. Research heuristic only.
- Baseline
- Warming…
- Stable ticks
- 0/25
- Drift meter
- 0
- Anomaly gate
- 5 < 35
Gate: closed · anomaly OK · α_drift=0.008
Continuous auth & privacy unlock
Toy neural-biometric match vs stream baseline + simulated mental passphrase. Not real identity verification.
- Baseline
- Warming up…
- Match
- 90%
- Session
- Open
- Airlock
- sealed
Prefer the Privacy airlock panel for seal / unlock ceremony. This panel is continuous auth + emergency lock.
Session log
Exports blocked in SEALED. Sample rows include only committed authorship (0 ready). Drafts never leave.
- systemNeuraGuard ready — research simulation only. Not a medical device.
- privacyPrivacy airlock SEALED by default — private intention classes fail-closed.
How estimation works
Transparent heuristics on decoded intention streams — not clinical models, not implant firmware.
- Stream ingest — synthetic / CSV / mock WebSocket
- Feature extractor — sliding-window speed, variance, entropy, confidence
- State estimator — EMAs → load, focus, fatigue, agency
- Policy engine — thresholds → throttle / break / privacy block
- Action dispatcher — gated sample + UI status