@thorstenball@erd0xbc Results are mindblowing ๐คฏ
I've been using Claude Code heavily since beta, but this is next level. Giving your agent a clear picture of your goal and approach, then coming back hours later to a nice summary and clean results is just unmatched.
@thorstenball@imaxxs Most of the time, a 5-minute Pomodoro break to get a glass of water and breathe some fresh air is what helps me the most without losing focus.
@thorstenball@imaxxs Totally agree. It's useful for me when I encounter complex problems or feel stuck. I realize how helpful it is to step away from the computer for a while. Taking a walk (I live next to a small forest), feeding my chickens, or just taking a shower has brought some of my best ideas
@davefarley77@GergelyOrosz@KentBeck Some challenges with this approach:
- 100% MSI but terrible performance?
- How do you version the specs themselves?
- Could devs game mutation scores?
- What about breaking changes that still pass tests?
@davefarley77@GergelyOrosz@KentBeck My (naive?) assumption: a package with 100% MSI does exactly what the spec demands (nothing more, nothing less).
This could be TDD's answer to dependency hell: let the tests choose the best implementation.
@VotrubaT But you're not talking about @x_framework , right?
Talking on X about an unnamed framework and calling it X, together with the fact that there is actually a framework called X, is enough confusion for me for today. ๐
@_Codito_ We actually also have multiple baseline files (one for each team) and we split them after the generation. Makes sense to commit them directly in the CI job, I guess we will add this, too ๐
@_Codito_ We have something related in our project: If you add the label "phpstan-baseline-generate" it will provide the generated baseline as an artifact that you can download.