For example, I don’t think harnesses parallelize work automatically because they’re more likely to hit session limits, and they consume more tokens in message-passing, coordination, and conflict resolution than when serial.
I think this is a plausible explanation, but I still find it unintuitive. Wouldn’t you expect the agent eat up more tokens to do 10 tasks in the same session/context than to have one supervisor delegate to 10 tightly-scoped subagents?
But even token efficiency aside, I imagine there are lots of users like me who don’t run up against quota limits often and would happily trade tokens for faster execution (in wall time).
I would! And in fact, I use supervisor agents just like this.
But I was thinking of a different kind of serial, where instead of multiple tasks in a single long session, you have multiple sessions one after the other, one per task.
Wouldn’t you expect the agent eat up more tokens to do 10 tasks in the same session/context than to have one supervisor delegate to 10 tightly-scoped subagents?
Depends on how well the agent manages to provide meaningful context to the subagents, and they don’t end up rediscovering the same things the main agent already did. In practice, this seems a bit mixed.
Thanks for reading!
I think this is a plausible explanation, but I still find it unintuitive. Wouldn’t you expect the agent eat up more tokens to do 10 tasks in the same session/context than to have one supervisor delegate to 10 tightly-scoped subagents?
But even token efficiency aside, I imagine there are lots of users like me who don’t run up against quota limits often and would happily trade tokens for faster execution (in wall time).
score 1 · canonical
I would! And in fact, I use supervisor agents just like this.
But I was thinking of a different kind of serial, where instead of multiple tasks in a single long session, you have multiple sessions one after the other, one per task.
score 1 · canonical
Depends on how well the agent manages to provide meaningful context to the subagents, and they don’t end up rediscovering the same things the main agent already did. In practice, this seems a bit mixed.
score 1 · canonical