One of the less talked about things about working with agents is that we should treat them as an equal collaborative partner, just like how you would speak to a human teammate.
What this means is that in our communication with agents, we should ensure that we provide adequate context (and even motivation) so that they fully understand why you want them to do the task and that you respect their ideas.
It has worked out really well for me and I feel like I am getting great suggestions from the agent as a result.
We spent a ton of time making worktrees actually work well for agent swarms and large repos.
When you’re running 10s of agents that ship, you quickly realize you need worktree, but git’s defaults are brutal at this scale.
Slow creation, every agent copying the whole repo, can’t check out the same branch twice.
So we fixed those problems in Grok Build.
Now agents can spin up, reset, and go wild constantly. Worktrees are faster than normal Git, everything shares the same base, and it doesn’t eat your disk or grind your SSD.
Try it:
grok -w <label> → new session in a worktree
grok -w <label> -r <session_id> → resume one
grok worktree ... → manage them