An AI-native operating layer for a venture studio

The system that lets a five-person studio run twenty-plus projects with AI as the primary build instrument: a context-file layer every session loads, a router and project index that resolve spoken names to folders, per-project rules and resume-here docs, a dozen custom skills, and a memory architecture that survives folder moves.

• Context files (who / what / how / voice / tools / glossary) auto-load into every AI session on the machine; a router points any folder at them
• Project index resolves the names people say out loud ("the kiosk", "the dash") to folders and resume-here files; an organization log makes stale references traceable instead of dead
• 12+ custom skills: meeting briefs, stakeholder intake → parallel agents, session handoff, deploy ritual, spend audit, prospect targeting, branded documents
• Up to 17 concurrent AI coding sessions across 6 project folders; transcript-survival runbook; agent-aware terminal workspace
• A scoping-time rule for how a non-technical principal will access each build (shared corpus, artifact, module, token page, Slack) — with cost discipline