BCI Silent Vibe Coding

Goal: build a local, multimodal biosignal stack that can turn non-speech body/brain signals into text and computer control.

Current hardware target:

• OpenBCI Ganglion (macOS)

Core principles:

• Maximize input bandwidth from body signals (EEG, EMG/ECG where possible).
• Keep acquisition, decoding, and action layers separate.
• Design for realtime latency and safety from day one.
• Be honest about constraints: free-form text directly from non-invasive EEG is a long-horizon goal.

Repo Layout

• docs/SETUP_MACOS.md: software bring-up and first connection steps.
• docs/ROADMAP.md: phased plan from first stream to silent coding demo.
• docs/ARCHITECTURE.md: system design for multimodal decoding.
• docs/QUESTIONS.md: open decisions to lock before implementation.
• docs/ALPHA_SPRINT_PLAN.md: parallel execution plan using worktrees and teammate agents.
• docs/INPUTS_NEEDED_FROM_YOU.md: decisions and hardware details needed before first capture.
• src/bci: signal acquisition, preprocessing, feature extraction, models, realtime loop.
• src/control: action mapping and safety gates.
• src/ui: live dashboard and operator tooling.
• data: raw/labeled/processed signals and trained models.
• experiments: model research scripts/notebooks.
• tests: unit/integration/realtime smoke tests.

Immediate Next Steps

1. Complete docs/SETUP_MACOS.md bring-up.
2. Verify Ganglion stream quality and write Session 001 raw capture.
3. Build baseline decoder that maps biosignal windows to low-level latent intents.
4. Add text generation layer that converts intents to text with language-model assistance.