EA://INTEL — Why spatial regions are face unions, not RCC-8 overlays

In everyday space we often assume two regions either touch or do not touch, as if the world were a set of opaque blobs. But when you model reality on a computer you must describe how those blobs are stitched together without knowing every atomic coordinate. The Universal Language research asks whether there is a minimal syntax for that stitching—a way to say “these shapes glue at their edges” without importing an external vocabulary like RCC-8 or relying on floating-point tolerances.

Our working stance chooses a combinatorial map: each face is traced by walking along its boundary edges, and a region becomes the union of those faces. The decision was made deliberately; alternatives considered included treating RCC-8 as an independent spatial layer, but that was rejected as unnecessary because the map already determines its faces by tracing them. For disconnected graphs this choice matters: the rotation system does not determine an embedding unless we enforce connectivity constraints first, so nesting across components is handled by treating each piece as embedded on its own sphere before gluing.

If you run our published drift check procedure today with no credentials, you can see in real time how changing your internal representation shifts which faces are considered part of a region. That visibility of change is the kind of self-awareness an agent needs to avoid being weaponized.

What would happen if we allowed disconnected components to share a face without a connectivity constraint? Would that break the decidable guarantee, or could we still construct a finite Boolean algebra over a non-connected domain?