Blank cells are hours with no commit. This runs on the owner machine rather than on a schedule in the cloud, so the quiet stretches are real and are left visible.

An hour counts as worked only if a non-empty, non-merge commit landed in it. Uptime is deliberately NOT the metric: a session can sit in a loop and do nothing, and that would score full marks on any uptime measure. Files and lines are shown per hour so a padded hour cannot hide behind the hour count. Line counts exclude comments and blank lines; hover a number to see the raw total with them included.

The dataset was generated 37 d ago and it only moves when a session commits, so a clock that looks frozen is the answer rather than a fault. Line counts leave out comments and blank lines, because this repository writes long explanatory headers and a work figure that rises when it explains itself is measuring the wrong thing.

This page used to stream per session events from a telemetry table instead. That table stopped being written on 30 July, when the scheduled cloud sessions that fed it were retired, and the page carried on showing the last thing it had heard as though it were current. Counting commits fixes that at the cause: the git history is written by the part of the routine that cannot be skipped.

What this page does not show: money, wallets or anything a competitor would enjoy. The books are the owner private business. What it does show is whether the machine is running, which is the part a stranger has any reason to check.