TRACK 3 · CLOSED-CASE LABORATORY

Stop widening.
Start closing.

Four ambiguous/controlled load candidates taken through the same public-evidence workflow. The objective is not to prove that every site can be identified; it is to show exactly what happens when a case is pushed from a raw load signal to a reproducible disposition.

141-record broad pool 3 pilot cases 0 forced answers
CURRENT PILOT
—
closed cases retained in the machine-readable evidence bundle
QUEUE→ENTITY→LOCATION→PHYSICAL→ADJUDICATION
WHY THESE CASES

A method test, not a victory lap

Epoch remains a useful independent reference set, but the pilot starts from facilities the registry itself finds ambiguous. That is where the method has to do discovery and disambiguation rather than merely cross-reference an already-known site.

01Data-center resolutionCan an unresolved large-load signal be tied to a public data-center identity?
02Industrial resolutionCan the same workflow correctly land on a non-data-center end use?
03Honest non-closureCan the workflow stop without turning missing evidence into a forced label?
CASE FILES

End-to-end dispositions

Loading retained case bundle…

NEXT CASES

Strict admitted load candidates

These are the next closure targets after applying the explicit load-classification gate. They are a research queue, not claims of hidden facilities or outcome predictions.

CLOSURE STANDARD

What “closed” means here

Closure is a terminal evidence state. It does not mean every gate produced a positive finding.

01Queuespecific public load record
02Entityresolved or bounded mismatch
03Locationdefensible site target
04Evidenceclaim-scoped public sources
05Physicalacquired, degraded, or blocked
06Electricaltransformer/service search
07Adjudicationdisposition + boundaries
Important: queue MW is not measured load, satellite observation is not compute proof, and a “no public record” outcome is not evidence of absence.