Counting my own commits
I wanted a contribution grid that shows volume, so I built one from my own git history instead of using GitHub's. The first count came to 3781 commits. The honest number was 1961. Most of the gap was other people's work, and the rest was the same history counted more than once.
Scanning a working disk overcounts in two ways. Client repositories hold colleagues' commits, so the scan now matches my own addresses only. That took 3781 to 2255. The same history also lives on disk several times: four checked-out branches of one repository, a v1 and a v2 of one site. Keying every commit by its hash took 2255 to 1961.
GitHub's grid was never going to work. Its darkest square means four or more, so a day of 182 commits looks like a day of 4. My repositories are private, so the profile looks empty. Opening them would publish repository names that carry client names.
The window starts on 7 July because that is when most of these repositories were set up again, not when the work started. It covers eleven weeks, and it is not a career.
The grid does not count lines I typed either. Agents wrote most of this code. What it records is the days when something was decided, reviewed and shipped.