Operator

An operator is the account-level subject associated with AI-work telemetry. Not necessarily a legal person, employer, or unique human. One operator can have devices, submissions, and an optional authenticated account link.

Token counts (input, output, cache_creation, cache_read) from AI coding sessions. No prompt content.

Operator-level aggregates of the four token pillars across sessions, windows, and platforms.

The operator is the unit of analysis for AI processing efficiency — not the model, not the session, not the task.

Compare operator-level metrics across time windows to verify stability and distinguish operator effects from model effects.

Per-operator yield, SNR, leverage, velocity, construction, consistency.

If operator-level metrics show no stability across windows, the operator is not a meaningful unit of analysis.

3,304 models tracked across 17 platforms. Operator-level metrics computed from signed submissions.

One operator may have multiple devices. Operator identity is pseudonymous.

SigRank ontology, Search Authority canon SR-OPERATOR-001. Architecture: mos2es.com/architecture.