Attention returned, with the maths showing.

The private-pilot weekly reckoning now counts recaps the person marked useful, loops they closed, items SA‑NT held from immediate interruption, corrections they made and recorded failures. A failed delivery or voice turn stays in the same report as the value evidence.

Generated is not accepted. A recap counts only after the person marks it useful through the authenticated capture-client flow. That client choice still needs live usability verification.

The time figure is an estimate, not measured time. It assigns five minutes to each recap marked useful and one minute to each item held from immediate attention. It adds no time for a loop the person closed or a correction they taught SA‑NT.

The estimate stops at two hours in any week, even when the underlying count would imply more. The SMS states the arithmetic beside the result so a person can disagree with the assumption without having to distrust the whole report.

The report reads existing first-party receipts: meeting feedback, loops, capture triage, correction rules and failure logs. It adds no account integration and creates no transcript-scoring or background surveillance stream.

This behavior is available in the private pilot. Recap feedback on the capture client and the usefulness of the weekly wording still require repeated live verification; the estimate is not a promise that every customer saved that amount of time.