Infolayer · a build-state floorplan
A floorplan of ROVA's office/co-founder build. Solid rooms are built; hatched middles are shipped fronts with hollow interiors; dashed outlines are empty lots. The corrected read, per-room.
Plan → CHAMBERLAIN assert_consented_or_raise() → detect_drift() every phase. The gate is a type: it literally cannot dispatch an un-consented plan./forge_cofounder, cofounder:5085. But "plan" is a markdown string — no seal, no drift-check. Held by an operator at deploy, by design. Same door, softer gate.plan.py) and left empty; nothing constructs one. The typed hallway exists; this wing was never threaded through it./flow/api/genome, live.companion_key provisioned per room; context-binding is Phase-2./talk; dedicated UI queued.The building is more finished than the first draft claimed. Both build-wings exist and are live. The real gap isn't a missing spine — it's an asymmetry in the type: the tenant wing is gated by code that cannot dispatch an un-consented plan; the co-founder wing — the one carrying THE_TWO_BUILDS, where the person is a founder being handed a partner — is gated by prose and a human at deploy. The agent_birth slot that would thread it through the typed hallway is cut and empty.
The genuinely empty lots are three: the v3 womb (project-gestation), the governance floor (Gates, WARDEN, the Panel), and the District beyond the Office (Legal, Adoption, Lobby). Everything else is either standing or a door-with-a-hollow-middle.
Honesty note, kept visible: the "door live / middle hollow" pattern isn't this survey's discovery — docs/recon/2026-07-17-promise-vs-build-census.md named it first and evidenced it far harder (13 of 17 inter-agent contracts half-built, with production row-counts). This floorplan is one instance at the typed-contract altitude, drawn so you can hold the shape — not a competing claim.