Ask HN: Are subagents making flagship AI models feel like managers?

Subagents are one of the newest power features in AI coding assistants, and I have mixed feelings about them. The basic idea is good. If a task is messy, you can split it into smaller threads, run them in parallel, and merge the results. For large refactors, multi-file debugging, or research implementation work, this can save a lot of time. But I am starting to notice a strange manager effect. I ask for a top-tier model because I want the best reasoning and engineering judgment.