A Pragmatic Clean Architecture for Content Sites

Clean architecture promises maintainability, but applied dogmatically to a content site it produces folders nobody wants to open. The goal is boundaries that earn their keep, not layers for their own sake.

Three boundaries that matter

1. Contracts — Zod schemas describe your shapes once. Everything validates against them.
2. Data — repositories read content and hand back validated objects. The UI never touches raw files.
3. Business — small, framework-agnostic functions: reading time, search, submission.

Let the UI stay thin

Components receive typed data and render. No parsing, no fetching gymnastics. When the data layer changes, the UI doesn't care.

Resist the urge to abstract early

YAGNI is a feature. Add an interface when the second implementation actually shows up — not before.