Agentic Discipline 7

It’s time to rethink things. The models have gone through a sea change. They are much more capable than they were just a few months ago. Back then, a single agent could do small tasks, but larger tasks required the discipline of a harness like swarm-forge. But now, the models are so capable that larger tasks are completely within their grasp. So it’s time to step away from the swarm-forge harness and look at a disciplined approach to using a single agent to get such large tasks done.

In this episode we’ll see three of those larger tasks. One will be a case study that we’ll drill down into. We’ll also see a new wrinkle on the Crap, Mutate, and Dry tools. And we’ll take advantage of a tool that allows us to scrutinize what the agents produce at a level just above the code.