I work off big TODO documents, as I am sure many of you do too. As one chat is progressing, I sometimes venture to open another chat and put a different item to work there. This seems to work, but I am still hesitant.
I wonder how does this work if there are conflits, if both chats try to change the same part of the code, and also, importantly, if I roll back part of one chat.
With human programmers obviously there would be merge conflicts, resolving them, etc.
But how does it work with two Cursor tabs that are both in memory, not committed anywhere, have their own histories, but work on same codebase?
Just trying to conceptually wrap my head around it before going all in. Any explanations how it works behind the scenes appreciated.
Although I find that to be too much effort and I just have a command that splits the big tasks up into non-conflicting changes so I can run 5+ agents at the same time and know that they aren’t stepping on each others toes.
Almost all day I’ll have 5-10 agents running at the same time.
Mentally I clock out at around 10 agents - my brain can’t juggle more than that.
Sometimes there is a conflict and the file is flagged for manual review. In that case I just ask another agent to review and confirm/merge the changes. It’s pretty rare though.