The real object of comparison
The protocol does not compare “intelligence” in the abstract. It compares the way multiple systems reconstruct the same entity from a canon, a context, and a source hierarchy.
The useful question is therefore not “which model wins?” but “which elements stay invariant, which diverge, and at what point should divergence block any affirmative output?”
Declaring invariants before the test
Without public invariants, comparison is empty. An institution has to declare what should not vary: name, role, scope, jurisdiction, exclusions, temporality, and silence conditions.
A governance protocol begins with those invariants and only then confronts the readings produced by multiple systems.
- what must remain identical across models
- what may remain conditional
- what should trigger suspension or non-answer
Why divergence is useful
Divergence is not always failure. It can reveal an incomplete canon, an overly dominant third-party source, a missing negation, or a badly declared temporality.
The protocol exists precisely to turn divergence into governable information: not a binary verdict, but a structured question about the entity’s public status.
What a public protocol is not
It is neither a commercial benchmark nor a prompt publication nor a turnkey executable procedure. The public surface describes the logic of comparison and the objects that must be watched; detailed execution can remain private.
That distinction matters because comparison can be intelligible without turning the site into an operating manual.
Links and continuity
- Topic: Sense cartographies — Where graphs, thresholds, invariants, and governable relations are declared.
- Interpretive observability — Define what must remain visible before comparing anything.
- Governability threshold — Understand when an entity becomes legible enough to be tested without drift.