Agentic Coding and Merge Conflicts

So when I first started to use Claude code and GitHub copilot I attempted to assign lots of tasks to them both to create prs. There result was chaotic and messy. I lost track of what changes I had asked for on what ticket. On the code reviews I forgot what changes I had asked for and why because the response from the coding agents was verbose. And then off course I got into merge conflicts hell. Each merged branch then had to be merged into the other feature branches. So I spent more time working on merge conflicts than worrying about what the coding agents had done.

That was a while ago (more than a year). Sou the other day I decided to try it again with a smaller number of tickets. Archon using Clyde code and GitHub copilot. Ih have to say the code was better but the cognitive load was still massive and the merge conflicts on what particular Branch was complicated.

So what would I do if these were Debvs. I would get them to talk to each other if they were working in similar areas. Which is a bit difficult to do with multiple competing agents. I would assign tasks so that they would not conflict on the same area. I would also make sure that the feature request or Biggs (where possible) were small. So that is the approach I and going to try with coding agents.

But not yet. I have a Mac mini to setup with deep agent and some local models.