Jett Optical Encryption
JOE Twin Roadmap
AGT-first digital twin plan — MOJO native biometrics → VLM data → sell twin API → Helix/J-SPACE.
JOE Twin Roadmap
Canonical source of truth (edit on disk):
C:\Users\joshu\repos\drafts\JOE-TWIN-ROADMAP.md
Hermes skill: joe-twin-roadmap in astrojoe-hermes/skills/joe-twin-roadmap/
Thesis
Ship high-fidelity native AGT + account biometrics first (MOJO), accumulate personal JOE twin data, then expand to webcam/web, then monetize twin calls (token rates / x402 / OpenRouter / HF), then HelixDB / J-SPACE personal RAG for neuromorphic digital twins.
Do not invert the funnel (webcam-first or generic LLM resale).
Phases (summary)
| Phase | Name | Focus |
|---|---|---|
| 1 | NOW — MOJO | ARKit / MediaPipe·ML Kit AGT, Privy, OPTX proof, permissions bootstrap |
| 2 | NEAR — Data plane | Gaze tensors → HEDGEHOG; optional VLM batches |
| 3 | MID — DOJO | Multi-surface; webcam later; IR OpenCV lab |
| 4 | SELL — JOE API | Twin + context; x402 / token rates / $OPTX·$JTX |
| 5 | SCALE — Helix | J-SPACE personal vector RAG for twin calls |
Sensor ladder
- iOS ARKit TrueDepth (gold)
- Android MediaPipe Face Landmarker (JEO Webcam3D class)
- Android ML Kit fallback
- Browser MediaPipe (Phase 3)
- IR OpenCV pupil / Orlosky (DOJO hardware)
Unit of value (Phase 4+)
identity + AGT context + optional Helix retrieval + model outputNot a rented generic chat model.
Related docs
- MOJO — mobile entry tier
- DOJO — training ground
- Gaze authentication
- Draft maps:
drafts/JEO-MEDIAPIPE-KOTLIN-MAP.md