Ask HN: What are your biggest problems and fixes with multisession engineering?

5 points·top_rooster··4 comments

I have run into a lot of issues using multiple sessions at once to build things, and I really want to hear from those that have figured a good amount out.

Or at least hear what problems people are having to know if I am running into the same blockers everyone is.

For example, I often want to have agents stop at certain points, but fail to get a mechanism that works.

View on HN

I'm only running a few agents in parallel so far, but I don't let them write or read anything w/out my approval. This slows things down quite a bit, but it means that I have exactly the opposite of your problem of stopping them.

I've only started really thinking about developing my own harness, but one of the reasons would be to implement some mechanism where I could give the agent a "turn budget", say, 5 turns, when I can clearly see what the agent is going to be doing for the next few turns, without my needing to allow each turn to go forward.

But it sounds like the same thing would be useful for you as well, just in the other direction.

Could you say more about what "certain points" you're trying to get them to stop at?

Usually arbitrary points in the process. Like "implement this but stop if you do it this way and get to this point"

Personally, I don’t run multiple sessions / agents (apart from the one claude code spawns on it’s own). But it seems you need an agent workspace manager - perhaps something like gastown? https://github.com/gastownhall/gastown

With grok bot, I've just been giving it a time to stop working; otherwise, it's likely to burn through my entire week's usage in an afternoon.

If I want to run multiple sessions in parallel in claude code and codex, I use multiple worktrees, and I select tasks for each tree that are unlikely to have merge conflicts.

Reply on news.ycombinator.com