Agents Make Trunk-Based Development Cheap: One Task, One Branch, Landed Same Day

The article discusses how trunk-based development has become easier to adopt with the rise of coding agents. Trunk-based development involves making small changes to the main codebase regularly, with no more than a few branches active at any time. However, many teams struggle with this practice as their branches can last for weeks or even sprints before merging back into the main branch.

The article argues that coding agents make trunk-based development much cheaper to implement. Each agent task is small, like "fix the flaky login test" or "add a refund endpoint with a test." When an agent finishes a task, it stops working, leaving behind a small diff ready for review. This eliminates the need for lengthy merge events and reduces the risk of bugs and regressions.

The author contrasts this with the traditional approach of long-lived feature branches, which often lead to complex merge processes and increased merge conflicts. The key to successful trunk-based development with agents is to keep each task focused on a single branch, review it immediately, and split the task if necessary. By doing so, teams can maintain three or fewer active branches, merge to the trunk at least once a day, and avoid code freezes.