Live Course: Orchestrating Coding Agents

One coding agent stops being enough at some point. The context window fills up and the agent drifts from its own plan, independent tasks run one at a time, one expensive model does both the architecture and the mechanical renames, and a wrong assumption made early has nothing to catch it. Running more agents fixes some of this and brings new problems: merge conflicts, errors that grow across handoffs, and token costs that grow faster than the number of agents.

In this live course on the O'Reilly platform I go through the main orchestration patterns, from the simplest to the most complex: a scratchpad file that carries state into a fresh context window, a second agent that reviews the first one's work with clear stop rules, several agents working in parallel on one repo, subagents, and orchestration written as a script so a repeated flow runs the same way every time.

Every pattern is shown live on a real repository. You leave able to tell which limit you are hitting, pick the lightest pattern that fixes it, and set up parallel agents so each stream of work ends as its own pull request.