Boris Cherny from Anthropic said something that really stuck with me:
“I find myself using Loop more and more, I already have dozens of loops running. One is babysitting my PRs, fixing CI, doing auto-rebasing. Another one pulls feedback from Twitter and clusters it every thirty minutes.”
Most people are still writing one prompt at a time and manually checking the output
In just 53 seconds, Boris breaks down how /loop actually works and they’ve just launched Routines (a server-side version that keeps running even when your laptop is closed)
Loop Routines = Claude agents that actually work for you on a schedule, without you constantly watching over them
This also fits perfectly with building a personal Agentic OS, which I talk about later in my article
Save it, read it, and start working with AI in a completely different way