MasteryOS Ecosystem Vision
The complete map: 8 products, one flywheel, agent swarm replacing the build team.
Ecosystem Map
| Product | Layer | Role in Ecosystem | Status |
|---|---|---|---|
| Expert Extraction System | Create (L1) | Automates expert clone creation — voice, methodology, frameworks extracted from raw content | Local — not yet on GitHub |
| MasteryBook | Extract (L2) | RAG pipeline — uploads → insights, summaries, study materials, audio overviews. SurrealDB + LangGraph + Esperanto | Live |
| NowPage / HC Protocol | Distribute (L3) | HC pages, lead magnets, SEO. Every page = Reveal demo in wild. Assessment → MasteryOS funnel | Live |
| MasteryOS / Reveal | Monetize (L4) | Expert platform. AI assistants. Subscriber access ($297-2,497/mo). Athio JVs. ~85% built | 85% complete |
| Labs | Teach (Meta) | 90-day AI Monetization Lab. Teaches the full 4-layer workflow. April 1 launch, 5 spots. | Launching Apr 1 |
| Credential Vault | Infrastructure | Per-user API key store. Fernet + Supabase KMS. BYOK primitive for all products. | Built, needs extraction |
| Forge / Ralph | Factory | Personal AI OS. Agent swarm orchestration. 24/7 autonomous build capability. | Live (Phase 4A done) |
| Voice Agent | Interface | LiveKit WebRTC. JWT-gated. $10/mo add-on. Embeddable into MasteryOS expert profiles. | Backend done, frontend pending |
Business Funnel
Tech Architecture
| Product | Stack | Auth | AI Layer | Hosting |
|---|---|---|---|---|
| MasteryBook | FastAPI + SurrealDB + Next.js | Supabase Auth | Gemini (Esperanto) + LangGraph | Vercel (frontend) + VPS (API) |
| MasteryOS | FastAPI + MySQL (32 tables, gou_ prefix) + Next.js 15 | JWT (custom) | OpenAI Assistants v2 → migrating to OpenRouter BYOK | AWS EC2 |
| NowPage | Next.js API + Supabase | Bearer token | — | Vercel |
| Labs | PHP + SSL | Custom | — | AWS EC2 |
| Credential Vault | Fernet encryption + Supabase KMS | MasteryOS JWT | — | Embedded in MasteryOS → extracting |
| Forge / Ralph | Grammy bot + TypeScript services + bash | Telegram | ClawdRouter (LiteLLM cascade) | Hetzner VPS |
| Voice Agent | LiveKit self-hosted + WebRTC | JWT-gated | — | TBD |
Integration Design
Three options evaluated. Verdict: Content Pipeline Fabric — each app owns its own DB, shared capability services are stateless and namespace-isolated, content pipeline normalizes and distributes.
Rejected: Shared DB
- Tight coupling — one schema change breaks everything
- MySQL vs Supabase vs SurrealDB — impossible to unify
- Auth models differ (JWT vs Supabase Auth)
✅ Recommended: Content Pipeline Fabric
- Each product owns its DB + auth
- Shared services: rag-service, audio-service (stateless, namespace-isolated)
- Content pipeline normalizes and distributes between products
- Forge Content Pipeline (:5014) already built and live
The 4-Layer Knowledge Stack
Every product maps to exactly one layer. The same source document powers all four.
Create — Expert Extraction System + Raw Content
Expert's raw IP: PDFs, recordings, frameworks, transcripts. Expert Extraction System (local) automates creating the expert clone — voice, methodology, style patterns. This is the pre-processing layer before MasteryBook ingestion.
Extract — MasteryBook
RAG pipeline: chunk → embed → vector store → LangGraph chat. Outputs: source_guide JSON (summary, key_points, topics, insights), audio overviews, study materials, research reports. Universal format consumed by all downstream products.
Distribute — NowPage / HC Protocol
Structured MasteryBook output → HC pages. "Publish to NowPage" button: expert clicks once, lead magnet publishes to their domain with assessment at bottom. SEO compounds. Every page = Reveal demo.
Monetize — MasteryOS / Athio
MasteryBook knowledge base auto-syncs to MasteryOS AI assistant. Subscribers pay $297-2,497/mo to interact with expert's AI. Labs graduates auto-migrate here on graduation day. Athio JVs at top of pyramid.
App Overlaps
| Integration | Connection | Build Priority |
|---|---|---|
| MasteryBook → NowPage | source_guide JSON → HC page template → "Publish" button. Expert uploads once, lead magnet auto-publishes. | P1 — build first |
| MasteryBook → MasteryOS | MasteryOS SSO → MasteryBook login. Expert sources → AI assistant knowledge base. Studio outputs → gated resources. | P2 — after Credential Vault extraction |
| Labs → MasteryOS | Graduation day = MasteryOS launch day. Lab work product → first knowledge base. Lab profile → expert profile. | P2 — before April cohort 2 |
| Credential Vault → All | Extract from MasteryOS → standalone API → BYOK for all products. One key management UI. | P2 — Sumit next |
| Expert Extraction → MasteryBook | Expert Extraction System pre-processes → structured input → MasteryBook ingestion. Automates Layer 1→2. | P1 — when system is on GitHub |
Agent Swarm Strategy
The Strategic Bet
This is not "AI assists developers." This is "AI IS the development team." Sumit, Rohit, Mukesh, Ashwini, Lee stop building new features — they fix what agents break, validate what agents produce, and make architecture calls agents can't. Ralph + Claude Code + Forge do the build.
Current Team Roles (Before)
- Sumit → engineering lead, feature builds, backend APIs
- Rohit → GUI/frontend development
- Mukesh → backend API implementation
- Ashwini → voice agent backend
- Lee → DB, ingestion, JSON→Python flows
New Team Roles (After)
- Sumit → architecture decisions, edge case escalation lead
- Rohit → agent PR review, UI edge case fixes
- Mukesh → agent PR review, API edge case fixes
- Ashwini → voice edge cases, LiveKit configuration
- Lee → DB validation, ingestion edge cases
The Staging Server Model
The forge-lab clone is NOT just "safe experimentation." It's a staging server where agents build autonomously. Proven features reconcile back to production on a regular cadence. Forks diverge — that's expected and managed.
Prove it won't blow up
First 4 weeks: agent swarm builds 10 features in forge-lab. Jason + team review. Pass rate tracked. If >80% of PRs are clean → model is validated.
Establish reconciliation cadence
Weekly merge window. Jason reviews diff between forge-lab and production. Cherry-picks proven features. Rejects anything with unknown risk. Team fixes edge cases before production merge.
Expand autonomy as trust builds
Low-risk task types (bug fixes, copy changes, UI polish) → auto-approve after 2 clean reviews. High-risk (auth, DB schema, payments) → always human review. Trust is earned task-type by task-type, not agent-wide.
Full team role transition
Once forge-lab proves the model, team formally transitions from "builders" to "edge case specialists." New hire criteria change: not "can you build features?" but "can you spot what the AI got wrong?"
Agent Governance Model
| Task Type | Agent Autonomy | Human Role |
|---|---|---|
| UI copy, styling, layouts | Full autonomy | Review final output visually |
| Bug fixes (non-auth, non-payment) | Full autonomy | Review PR diff |
| New API endpoints | Build + flag | Review data model + security |
| DB schema changes | Build + flag | Sumit reviews before merge |
| Auth / session logic | Build only, never merge | Human implements after agent drafts |
| Payment / Stripe flows | Build only, never merge | Human implements after agent drafts |
| Production deploys | Never | Jason approves all production deploys |
Asymmetric Insights
Insight 1 — The Publish Button is the Flywheel Switch
A single "Publish to NowPage" button in MasteryBook turns every expert's content upload into a live lead magnet. One click → HC page → assessment → MasteryOS subscriber funnel. Every expert using MasteryBook is simultaneously building their Reveal funnel without knowing they're doing marketing.
Insight 2 — Labs is the JV Pipeline Supply Chain
Athio requires $250K+ revenue. That's a high bar. Labs at $1,500 produces experts who launch on MasteryOS and hit revenue milestones. The path from Labs → $250K revenue → Athio JV candidate is 12-18 months. Labs isn't a training product — it's a structured pipeline that manufactures Athio JV candidates.
Insight 3 — BYOK Turns the Cost Structure Upside Down
Traditional SaaS: company pays for AI compute, charges per-seat, competes on features. MasteryMade BYOK: customer pays their own AI bill, Jason charges for the platform, margin is 94%+. The bigger the customer's AI usage, the more value they get — but Jason's cost doesn't move. Scale doesn't compress margins; it expands them.
Insight 4 (Corrected) — Agent Swarm Replaces the Build Team, Not Supplements It
The goal is not "AI helps the team build faster." The goal is: agent swarm does the build, humans fix edge cases. Sumit's team stops writing new features. Ralph + Claude Code + Forge write the features. The team becomes a QA and edge-case specialist team. The economics: same or better code output at a fraction of the cost, running 24/7. The forge-lab staging server proves this model before committing to it — agents build freely in the clone, Jason reviews weekly, proven features reconcile to production. Trust is built task-type by task-type. When the model is validated, the team role change is formal: new hire criteria shifts from "can you build?" to "can you catch what the AI missed?"
2nd Order Effects
Active Expert Partners
| Expert | Platform | Status | Next Action |
|---|---|---|---|
| Samuel Ngu | Align360 — coaching frameworks | 5 alpha testers, no trademark yet | 3 assessments, onboarding questions, NowPage lead magnet |
| Brian Muka | Freedom Sherpa — 12 coaching frameworks | JV agreed: 60/40 + $5K upfront (no audience) | MasteryOS profile + first AI assistant setup |
| Brad Himel | TBD | On hold pending direction | Coordinate with Will |
Next Actions (Ordered by Asymmetry)
NowPage MCP → Vercel
3 env vars (RESEND, SUPABASE_URL, SUPABASE_SERVICE_KEY). Gateway into Reveal funnel. JASON-DEP — 5 minutes.
Labs April 1 Launch
5 spots at $1,500. $7,500 immediate. Graduates become MasteryOS experts. JV pipeline supply chain starts here.
Expert Extraction System → GitHub
Push local system to GitHub. Document it. Integrate with MasteryBook ingestion. Layer 1→2 fully automated.
masteryos-forge-lab Repo
Create GitHub repo. Ralph clones + configures. Agent swarm staging server live. Start proving the model.
MasteryBook → NowPage Button
source_guide JSON → HC page template. Expert clicks once, lead magnet publishes. Built in forge-lab first.
Credential Vault Extraction
Standalone API. Sumit's next build. BYOK works across all products. SSO path opens.
Published March 2026 · Command Center · Agent Swarm Deep Dive →