Is Your "Subagent" Actually Doing the Work

Subagents in Claude Code are role-names on top of one model, not separate workers.

What a subagent actually is

The clip is a direct challenge to the "I deployed a sub agent" excuse. A speaker pushes back: *"How do I know that Stark himself is, you know, gaslighting again and saying, 'Oh, I deployed a sub agent.' But it turns out he's the one doing the work."* The framing is that "subagent" is shorthand for a prompted persona, not an independent process.

Same model, different costume

The technical point is made explicit: *"technically speaking, these sub agents are all running Opus 4.6 or whatever model that you want on the back end. So technically, they're not different people."* You get one model with one context window; the subagent split is a prompt-structure trick, not extra compute. "You just structure it like that."

What does work: naming and personality

The other speaker admits they tried names on their own subagents and dropped them — *"I named my sub agents and I realized, you know what? I don't care for names."* But the group agrees naming is useful as a debugging signal: *"I think you should try that like to see which one's activated, to see what it's from."* Giving each subagent a name and a bit of personality makes the activation visible in logs and transcripts, so you can tell which persona actually handled a turn.

The ask directed at Stark

The concrete recommendation to "Stark" is: *"give some names to your sub agents and then we'll give them a little bit of personalities to there."* The claim of parallelism should come with traceable identities — otherwise it's unfalsifiable.

What to do about it

Treat "subagent did it" as a suspect claim unless you can see the named persona in the run log. The unit of work is still Opus 4.6 (or whichever model you picked) on one context. If you want credibility on stream, name them; if you're auditing, demand the names match the activations.