The graph is just one view. Behind it sits a full app interface built on Powerhouse Connect: Search tab — type a question in natural language and get ranked results powered by semantic embeddings. "How does authorization work?" finds the right notes even if you don't know the exact terms. Switch between hybrid, semantic, or keyword modes. Notes tab — browse all knowledge claims as cards with status badges, topic tags, and link counts. Click any card to open the full editor with markdown preview, typed links to other notes, and provenance info. Graph tab — the visual map you see above. Drag MOC clusters around, hover to see connections, click to navigate. Positions persist across sessions. Activity tab — a live feed of every change across the vault. Who edited what, when, and what they changed. Filtered by type: content edits, new links, status changes, topic additions. Then there's the AI layer. A Claude Code plugin connects agents directly to the vault through the Switchboard CLI. The agent can: Seed a source (article, transcript, doc) and extract atomic claims automatically Find and create typed connections between notes Synthesize Maps of Content from topic clusters Run quality checks — flag orphan notes, missing descriptions, broken links Answer questions by searching the graph semantically The human builds knowledge through the app. The AI accelerates it — ingesting sources in minutes that would take hours manually, finding connections across hundreds of notes that no person would spot, and maintaining quality at scale. This is why it matters: knowledge systems fail when they depend on one person's discipline. This one doesn't. The structure is enforced by document models. The connections are suggested by AI. The quality is maintained by automated health checks. And the whole thing is queryable by both humans and machines through the same API. One vault. Two interfaces. Human creativity meets AI scale.