Theo Browne describes using agents to review changesCode review examines proposed source changes to assess their correctness, clarity and suitability before or around integration., respond to pull request feedback and coordinate work across separate threads. He contrasts manual prompting with loops that carry context and repeat review cyclesAn AI feedback loop occurs when AI outputs or their consequences become inputs that influence later model or agent behavior..
He walks through a multi-stage refactor planned as several pull requests and describes an automated workflowWorkflow automation uses software or AI to complete a connected sequence of routine steps with less manual effort. that created, reviewed and merged them. He cautions against applying the experiment uncritically to production systems.
Browne discusses token usage, subscription costs and the risk that loops can spend time and resources following the wrong path. His examples are personal experiences rather than general performance guarantees.
Watch on YouTube
