Alignment-aware contemplative cybernetic platform

A closed-loop system that fades itself into obsolescence.

Experimental open research platform. Not a medical device. Not intended to diagnose, treat, or induce altered states.

Sati-Loop is an open-source hardware and software platform for contemplative neurotechnology. Unlike existing meditation devices that reward proxy physiological signals (calm EEG, HRV coherence, breathing resonance scores), Sati-Loop uses negative-only feedback, scaffold withdrawal, and transfer to unassisted practice as the only success criterion.

It is built from the ground up to avoid the contemplative alignment problem formalised in the companion paper.

• No positive reinforcement — the device never rewards calm states
• No gamification — no scores, dashboards, streaks, or progress bars
• Minimal cue only — a single neutral bone-conduction cue when distraction is detected
• Scaffold withdrawal — the device systematically fades its own support as skill develops
• Transfer test — the only metric that matters is unassisted practice improvement

[ Sensors: EEG + PPG + Respiration + IMU ]
                    ↓
[ Layer 1: Attentional Scaffold ]
  EEG-only fast path (500ms windows)
  Mind-wandering onset detection
  Single neutral cue on distraction
                    ↓
[ Layer 2: Autonomic Support (optional) ]
  HRV-guided breath pacing
  Fixed non-contingent ambient audio
                    ↓
[ Layer 3: Experimental Modulation (research only) ]
  taVNS — hardware-isolated, manual current setting
  TDM gating for EEG/stimulation isolation

X  Classifier does NOT tune stimulation
X  No positive reinforcement at any layer

1. Order the entry-level BOM (Muse S + ESP32 + sensors)
2. Flash the ESP32 firmware
3. Install dependencies: pip install -r requirements.txt
4. Run the Python signal pipeline: python signal_pipeline/pipeline.py
5. Complete the 2-week passive calibration (Phase A)
6. Begin active training (Phase B) — device cues distraction and fades over time

For research-grade EEG see the research BOM.

See STATUS.md for what is implemented vs. conceptual.

Bose, J. (2026). The Contemplative Alignment Problem: Reward Misspecification in Closed-Loop Meditation Systems.

sati-loop/
  README.md
  STATUS.md
  ROADMAP.md
  CHANGELOG.md
  CONTRIBUTING.md
  LICENSE
  requirements.txt
  requirements-dev.txt
  .gitignore
  paper/               Link to preprint and figures
  hardware/            Design spec, BOMs, schematics, STL files
  firmware/            ESP32 code including TDM gating
  signal_pipeline/     Python: fast_features, slow_context, classifier
  app/                 Mobile app (in progress)
  protocol/            Transfer test, calibration, taVNS safety
  docs/                Philosophy, known limitations, assembly guide
  safety/              Hardware interlocks, firmware safety rules
  research/            Literature, experiments (future)

• Hardware (schematics, PCB, STL): CERN-OHL-S v2
• Software / firmware: MIT
• Documentation: CC BY 4.0

Use of the name Sati-Loop for any system that rewards proxy contemplative states, claims Tier 3 or Tier 4 attainments, or connects stimulation to the classifier reward loop is prohibited.

Bose, J. (2026). Sati-Loop: Design Specification for an Alignment-Aware
Contemplative Cybernetic Platform, v1.1. GitHub.
https://github.com/joyboseroy/sati-loop

Bose, J. (2026). The Contemplative Alignment Problem: Reward Misspecification
in Closed-Loop Meditation Systems.