The Subagent Tax (systima.ai)

Based on feedback from our last post, we measured Claude Code subagent fan-outs vs. Doing the same work sequentially,

We found that they used between 2 and (almost) 6 x the input tokens and were never faster on any task we timed.