Design for Churn

In the AI era, the rapid pace of code generation is fundamentally changing how developers should approach software architecture. Traditionally, code churn—the frequent modification of code—has been viewed as a negative metric associated with technical debt and poor quality. However, the author argues that as AI makes writing code cheaper and faster, we should shift our perspective from viewing codebases as static buildings to dynamic gardens. By categorizing code into 'annuals' (disposable, fast-changing features) and 'perennials' (stable, core infrastructure), developers can embrace churn as a design input rather than a symptom of failure. The article suggests that by establishing clear contracts, isolating volatile code, and measuring churn by layer, teams can build more resilient systems. Ultimately, the goal is to design structures that allow for frequent, confident updates, treating the codebase as a living environment that requires intentional cultivation rather than rigid preservation.