Product Context — Belief Engines
Mission
Belief Engines extracts beliefs from long-form content (podcasts, books) and makes them queryable through a terminal-first intelligence system.
Current Focus
Bitcoinology — the active proving ground. A terminal-native Bitcoin intelligence system that turns podcast conversations into searchable belief graphs.
What Exists Today
Agent Architecture
| Agent | Role | Status |
|---|
| Oracle | Route, classify, orchestrate response path | Designed, partially implemented |
| Jackal | Retrieval engine (vector + structured evidence) | Active |
| Playbook | Deep synthesis for complex queries | Active |
| Gate | Citation/claim verification guardrail | Active, hardening |
Tech Stack
| Layer | Technology |
|---|
| Transcription | GPU machine (be-flow-dtd) |
| ETL | be-podcast-etl (extraction, embedding, profiling) |
| Vector search | Qdrant |
| App state | Supabase (Postgres) |
| Frontend | Next.js (be-bitcoinology-v1) |
| Deploys | Vercel |
| Observability | Trigger.dev (step-level tracing) |
| PM | GitHub Projects (4 boards) |
Key Repos
| Repo | Purpose |
|---|
be-bitcoinology-v1 | Main product app (Next.js) |
bit-terminal-blueprint | Terminal UI prototype (Vercel preview) |
be-podcast-etl | Extraction/embedding/profile pipeline |
be-flow-dtd | Transcription + diarization (GPU) |
bit-wiki | This wiki (Docusaurus on Vercel) |
Product Principles
- Terminal-first — not web-heavy. Single-pane-of-glass, retro 8-bit visual language.
- Search quality is the product — retrieval precision and synthesis grounding are non-negotiable.
- Citations are law — unsupported claims get rejected. Gate enforces this.
- Local artifacts — queries produce local file outputs, not just screen renders.
- Speed over ceremony — fast iteration, narrow weekly goals, proof-based shipping.
Tier Taxonomy
| Tier | Name |
|---|
| Entry | Shrimp |
| Growth | Dolphin |
| Power | Orca |
| Enterprise | Whale |
Strategic Priorities (Current)
- Search quality + reliability — this IS the product
- Agentic traceability + citations hardening — trust moat
- Terminal-first UX + local file artifacts — distribution + power-user fit
- Operational speed — pipeline throughput, faster iteration