A deliberately boring publishing stack

This publication is built around one recurring task: write a file, review it, and publish it.

There is no database or content management service. The content lives alongside the code, with a schema that catches missing metadata at build time.

The architecture#

Markdown and MDX files flow through Astro content collections into static HTML. Shiki highlights code during the build. Pagefind indexes the rendered articles so search can run without a backend.

The browser gets enough JavaScript to remember a theme preference and, on the search page, search the index. Reading an article does not require a client framework.

The useful tradeoff#

Publishing requires a Git workflow. That is a good fit for a technical publication maintained by an engineer; it would be a poor default for a newsroom full of nontechnical contributors.

The static output can be served by Vercel or another static host. Content and presentation remain portable.

What stays out for now#

Comments need moderation and abuse handling. A text box alone does not create a community feature. The local discussion component reserves the boundary for a future implementation without coupling article templates to a provider.

The same reasoning applies to reactions, accounts, and analytics: add them when there is a question they help answer.