My advanced skill with tmux allowed me to run a ton of agents in parallel on a single repository. But they gradually started to conflict. I want to run 3-4 features in parallel in different code snapshots.

There are two solutions:

- git worktrees
- conductor.build or similar

I went with the second option for now. And this made me think that the world of IDEs/code editors is changing again. Now the priority is managing a bunch of agents and code review.

I also have thoughts that we too quickly accepted that agents > autocomplete. But more on that next time 🍿