A working condition in which the pace of an AI-supported task is kept within the rate at which the human can review the outputs, described as the relationship between output throughput and human checking capacity rather than as advice.
STATE/ratio. Operationalized as a pacing index: the ratio of AI outputs reviewed by the human to AI outputs produced over a task window (0-1). Values near 1 indicate throughput stays within review capacity; low values indicate outputs accrue faster than they are checked. Computed from logs, not self-report.
Proposed measurement protocol (not yet empirically validated): Throughput/review counting from interaction logs: outputs-produced and outputs-reviewed per fixed window, expressed as the reviewed-to-produced ratio. No rater judgement is required; reliability rests on log completeness, which is could be reported as the proportion of outputs with a recorded review status.
https://andreasehstandlicenseofclarityloc.github.io/augmanitai-periodic/#topic-safety-alignment