Was going to say, to the extent I rely on Claude, I'm closer to the old school than the new hotness. I don't run fleets of agents per se, I just use Claude Code with a good bit of back-and-forth and review. So not quite web browser level usage, but I don't have a big swarm of agents working full time. I guess CC can spin up multiple agents, but I don't know if that counts.

Fortunately I'm in a shop that isn't being judged by feature output quantity, but is absolutely judged on whether or not our internal or external customers go out of service. We don't have pressure to pump output.

I'm starting to have them gather info to feed into automated tooling for drafting tasks. It's just there's no good harness for my field (and only recently have models been able to read symbolic things well enough)