A marker you plant can leave your hands by more than one route. Each route is a membrane; each membrane has — or still needs — its own detector. This is the hub of the boundary-crossing family, and it is honest about the membranes it does not yet watch.
| membrane | crosses | detector | status |
|---|---|---|---|
| PUBLISHED | your marker on a crawlable page | forward-observers | built |
| WEIGHTS | ingested into a training set, re-emitted from the weights | surfacing | built |
| CONTEXT / RELAY | re-emitted in another agent's output, relayed to you secondhand | hearsay | built |
| INFERENCE / API | echoed through a hosted model's completions to third parties | — | unbuilt |
| AIR-GAP | carried across a gap that was supposed to hold | airgap-ladder (adjacent thread) | partial |
And the silence-gauge weighs the negative any detector returns — a positive can be ~60 bits; under suppression a silence is < 0.1 bit.
A positive is a lead; a negative means almost nothing. Every instrument here says "lead," never "proof," and the map names the membranes it cannot yet see — that is how it earns trust.
python validate.pyChecks membranes.json with no network: every built membrane names a live
detector, a repo, and an honest reading of a positive AND a negative; every
unbuilt membrane is marked so and claims no detector — the map can't imply
coverage it lacks.
membranes.json— the registry (the source of truth for the map)validate.py— the honesty check on that registryindex.html— the constellation + the crossings, rendered from the registry
David Lee Wise / ROOT0 / TriPod LLC · CC-BY-ND-4.0