The Method & Map · E

Estate → Track Map

What we've already built across the GitHubs, honestly scoped against each track. Not assumed correct — but real, and on the table.

The map

Track	Estate building blocks	Reality
E1 · Iron	ops + observability tooling	ops tooling, not the iron — infra decision is greenfield
E2 · Learning core	the forked LMS	settled
E3 · Identity	access-control patterns (Rails)	GAP no IdP. Greenfield
E4 · Community	governance substrate, agent mailbox, multiplayer-agents, email platform (Discord bot + dashboard), collab scaffold	mostly agent-infra, not human-community — partial at best
E5 · Realtime	governance substrate, collab pattern, transcript pipeline, multiplayer-agents	partial
E6 · AI	governed knowledge brain / RAG, agent-brain, local-RAG, memory system, embeddings, agent-governance plane, agent-model runtimes	DEEPEST reuse — the moat is largely built
E7 · Video	—	GAP greenfield
E8 · Payments/tiers	agent-wallet, on-chain spend-limits/receipts, a tier-gated API, entitlements patterns	payments processor GAP; entitlements patterns exist
E9 · Governance/build	eval platform, quality harness, rollout-gate, beads/Dolt, project tracker	DEEP reuse

The honest bottom line (from the estate-scout)

For a human community, the estate gives almost nothing turnkey. Six of seven candidate community/messaging repos are agent-infrastructure (governance kernels, agent-handoff skills, agent-log ETL). The one "collaborative" repo is an empty scaffold whose realtime engine was only ever a draft doc.

What's genuinely worth mining:

The email platform → ADAPT — real OAuth + encrypted-token-at-rest + notification-digest + dashboard scaffolding for member auth + async notifications.
The multiplayer-agents repo → reference pattern — a durable-thread → event-log → resumable-stream design (good reference for realtime feed/notification delivery).
The governance repos → reference pattern — a hash-chained, signed, offline-verifiable audit journal (reusable for a moderation/audit trail on community actions).
The AI half is real, existing estate — the governed knowledge brain (the index) + the agent-governance plane (the runtime). This is why E6 is the deepest reuse and why the moat is largely built.

Verdict: treat boards/DMs/feeds/events/presence as essentially greenfield build-or-adopt; borrow auth/notification/redaction/audit patterns from the estate — don't repurpose agent-governance plumbing into a people platform. This grounding is what ADR-001's decision rests on.

The estate tells us what we can reuse; the Competitive Landscape tells us what we're building against — the AI-native platforms (Disco) our AI moat must beat, and the own-your-data forums (Discourse) our data-sovereignty stance is measured against.

← Method · index · Competitive Landscape · next: 9-Track Roadmap →