Building a simulated user that runs your PR and posts the video. Does this solve a real problem?
retained text
Hi fellow developers, I am trying to build something and want to understand if it actually solves a problem. What it essentially does is it watches a Github repository. When a PR opens it pulls and runs your whole stack on a throwaway VM. A simulated user/agent goes through the diff and understands what changes were made and the affected flows. It uses it like a real customer would and then we post back the video to the PR. We also have a replay session console where you can see logs of all the services started and how each service interacted with each other. We ask for dev environment variables on onboarding and we can provide test credentials in agent instructions. This feels like a loom video + AI code review + evidence for reviewer. I am obviously biased. I want to understand this one thing. Given the outflow of the features you ship using agent how do you test each unit of work
document fetched 2026-10-27 00:41 · canonical https://reddit.com/r/ClaudeCode/comments/1xfszzt/building_a_simulated_user_that_runs_your_pr_and