Anthropic's Claude Code engineer Sid Bidasaria: "Stop babysitting your agents. If you're watching Claude work, you've already lost the leverage." In 37 minutes, he shows how to set up agents that run without supervision, and the config below takes it to your phone. Watch the talk, then grab the... full setup below👇show more

The real unlock is offloading the loop entirely so the agent runs while you're away from the desk.

@rileywestreel Yeah loop is the new thing

stop babysitting your agents is easy to say when your agents actually do what you asked. the rest of us are still on iteration 47

@0x_rody honestly, letting go is the hardest part. i always feel like i need to check in on them, but i know it's just slowing things down. gonna give this a shot, thanks!

Sounds like a game-changer! Letting those agents run solo is the way to go. Excited to see what they can accomplish!

agents are only scary until they save you 3 hours

that's the shift right there. supervision is the training wheel you gotta take off eventually. once claude starts running unsupervised, suddenly you're not coding anymore, you're architecting. way different game.

Start task on commute. Come back to a finished PR. That’s not a power user workflow anymore — that’s the baseline expectation Anthropic is setting.

Exactly. If I still have to sit there and watch it work, I already lost the leverage.

Failed to get global typing stats: Public typing stats not found Stack: get typing stats

I will use it later

will read the article after video thank you so much

the leverage isn't 'stop watching', it's the agent leaving a receipt i can read after. i walked away from mine once and got burned by stale env it never flagged. now it tells me what it touched and skipped, or i don't trust the walk-away 👀

Agree: the leverage is letting the loop keep running while you’re away. The missing layer is orchestration: choose workspace, agent, provider profile, then continue from phone/desktop. That’s what I’m building in VibeAround:

Autonomy scales output, but without verification loops you just shift failure downstream.

I still watch the first run like a hawk. After that, logs, tests, and good rollback beat staring at the terminal.