Canonical source for B.Y.T.E. build phase status, dependencies, and document mapping. Audience: Everyone Status: Active
Phases 1–4 test the core thesis. Phase 4 (Memory + Knowledge) is the falsification point. Phases 5–6 are experimental extensions built on that foundation.
| Phase | Component | Status | Core Documents |
|---|---|---|---|
| 1 | Kernel + Execution Graph | Design complete | GRAPH.md |
| 2 | RPU + Orchestration | Design complete | RPU.md, ORCHESTRATION.md |
| 3 | Signal-to-Intent Pipeline | Design complete | — |
| 4 | Memory + Knowledge + Skill Registry | Design complete | MACROS.md (skill execution) |
| 5 | Macros + Code Registry | Research phase | MACROS.md, REGISTRY.md |
| 6 | Edge + Multimodal | Conceptual | EDGE_ARCHITECTURE.md, MULTIMODAL_INTERFACE.md |
Phase 1 (Kernel + Execution Graph)
↓
Phase 2 (RPU + Orchestration)
↓
Phase 3 (Signal-to-Intent Pipeline)
↓
Phase 4 (Memory + Knowledge + Skill Registry) ← FALSIFICATION POINT
↓
Phase 5 (Macros + Code Registry) ← EXPERIMENTAL
↓
Phase 6 (Edge + Multimodal) ← EXPERIMENTAL
Phases 1–4 are sequential and required for the core thesis test. Phases 5–6 depend on Phase 4 but are not required for the thesis.
The irreplaceable core. Manages execution chains as DAG-based computation graphs, handles scheduling, prioritization, interruption, and resource arbitration. Records all activity as immutable events in the world-state graph.
Sections: TECHNICAL_CONCEPT.md Section 4–5
The reasoning coprocessor and semantic orchestration layer. The RPU receives structured requests and returns structured results. Orchestration primitives provide deterministic tool composition.
Sections: TECHNICAL_CONCEPT.md Section 6, 10
Converts raw signals into actionable intent through three layers: perception processing, situation model generation, and intent estimation.
Sections: TECHNICAL_CONCEPT.md Section 7
The summarization pipeline compresses experience into narrative memories and validated knowledge. Skills provide optional seed behaviors. This is the falsification point — if Phase 4 does not demonstrate a measurable reduction in reasoning cost for equivalent outcomes compared to the Phase 2 baseline, the core thesis fails.
Sections: TECHNICAL_CONCEPT.md Section 8–9.3 Evaluation: EVALUATION.md
The optimization layer. Discovers repeated execution patterns and compiles them into deterministic macros. The code registry provides tested, versioned components for software development.
Sections: TECHNICAL_CONCEPT.md Section 9.1–9.2
Portable Personal Edge Node for mobile perception and the multimodal cognitive interface for sensor fusion. The core thesis does not depend on these existing.
Sections: TECHNICAL_CONCEPT.md Section 12–13
The core thesis — "can accumulated structure substitute for repeated inference?" — is tested by comparing Phase 4 against the Phase 2 baseline:
- Baseline (Phase 2): Structured but not compressed. Planning-first workflows with structured contracts but no memory, knowledge, or macros.
- Test (Phase 4): Memory + Knowledge graphs active. Validated facts eliminate re-derivation. Narrative memories replace raw context.
- Criterion: Statistically significant reduction (p < 0.05) in reasoning cost for equivalent task outcomes.
Skills are available from Phase 4 as a harness completeness feature but are not part of the thesis test — they are a constant across both baseline and test, and therefore cancel out.
Related: README.md for project overview. TECHNICAL_CONCEPT.md for section-to-phase mapping. EVALUATION.md for evaluation methodology.