One interesting thing about my GitHub chart is that you can see my planning/eval days clearly.
They're usually darker green with fewer commits. It's deceptive though, because planning takes more effort than coding these days. Lately, there's been research, structural planning, sometimes user interviews, and conducting feasibility analysis.
Testing is the same way. Especially on the coding agent project. Testing is done in branches, usually on private repos. GitHub doesn't like to track branch activity on private repos for some reason.
But even there, the effort is real. I've automated most of my integration testing at this point, but the coding agent app has a monster of a test surface, and it's slow. Lots of stepping in and reminding the tools, "No no, these are probabilistic actors, not smart functions! So they need to be bounded, not gated," or some such.
What I'm doing today is testing the swarm functionality at scale. I told it to do a 10,000 line white rooming project. This is day three of the eval. We have found so many interesting bugs with the workflow, and process.
I'm glad I did it this way.
It just doesn't involve high numbers of commits or Pr's, usually. Anything lighter than dark green is an implementation day.