Ask HN: Feedback Loop with Coding Agents?

I've been using Claude Code ever since it came out, and I've seen good success with it. However, Matt Shumer's recent article [1] included a piece I'm curious about:

When describing his workflow now, he says "Then, and this is the part that would have been unthinkable a year ago, it opens the app itself. It clicks through the buttons. It tests the features. It uses the app the way a person would."

Unless I missed a major development, do these coding agents have the ability to use feedback loops like this? If so, how have you all built this functionality into your workflow?

Any guidance / advice / documentation I can read would be super helpful!

[1]: https://x.com/mattshumer_/status/2021256989876109403