“I’m simply blown away by how much my work has sped up after the addition of sub-agents. Man, I really can’t describe how impressed I actually am. By the way, my work had already improved a lot since I started using this fantastic tool, but now with the addition of the sub-agent, it was one of the best things you guys could have done in any update.** It’s insane, man. I write several tests in minutes; it’s surreal. Running them alongside Composer** is just absurd. Granted, we can see that they consume a lot of tokens, but then we just use our private key with a plan code, and it runs beautifully.”
In my opinion, the best method for using Subagents is to choose the smartest and most convenient model for you as Agent, give it Subagents like it and cheaper Subagents, and let Agent independently distribute and verify work according to the task/task list.
Hey, thanks for the feedback! It’s great to hear that subagents sped things up so much, especially for writing tests together with Composer.
@Artemonim has already shared some helpful insights on using subagents. The strategy of using different model tiers (a main Agent plus cheaper Subagents) really helps cut token usage while keeping quality high.
If you have any questions or feedback on specific parts of subagents, you’re welcome to share!