Matthew Berman explains coding-agent loopsAn agent loop repeatedly observes a task's state, chooses an action, checks the result and decides whether to continue or stop. as workflows initiated by a triggerWorkflow automation uses software or AI to complete a connected sequence of routine steps with less manual effort. and guided by a verifiable goal. He gives examples involving pull requests and scheduled checks.
Matthew Berman discusses why a clear end state matters and why open-ended goals can be harder to specify and verify. He notes the risk of consuming tokens indefinitely and the cost of repeated agent workAI inference economics examines the revenue, cost, utilization, pricing, and margins involved in serving trained models..
Matthew Berman distinguishes his definition of a loop from a simple automation and offers speculative views on future autonomous software development. His predictions are presented as opinions rather than established outcomes.
Watch on YouTube
