Hello, dogfood

The first published article of the dogfood proving ground, and what this slice actually runs.

This site is the first executable slice of the dogfood project. It reads repository Markdown, builds the canonical post model, and renders static documents without touching the GitHub API.

What runs today

The builder and the renderer are two separate Eliscript programs:

1. The builder owns the filesystem, reads content/posts/, validates front matter, and writes a staging tree.
2. The renderer takes only the canonical model and returns documents. It has no filesystem, environment, or network authority.

Output replacement is atomic: the build writes _site.staging, and only after every route exists does it rename that tree into place.

What has landed since

Issues and Discussions are article sources now, each with owner-only publication and carrier classification. Comment channels exist in three shapes: a native snapshot captured during a build, a live channel that no build ever fetches, and a mount region for an external provider.

What has not landed is packaging. Those gates arrive in the order recorded by specification 0009. The static article authority rule holds throughout: every article is complete HTML and stays readable with JavaScript disabled. Search and the theme control are enhancements, not requirements, and the <code>&lt;noscript&gt;</code> note on the search page says so.

A design is only as good as the smallest slice that proves it.

Here is the renderer boundary in Eliscript:

(defun render-site (site posts)
  (let* ((sorted (sort-posts posts))
         (tags (collect-tags sorted)))
    ...))

Front matter is the only content metadata. Identity is the explicit id field; a file name never defines identity.

Markdown · content/posts/0001-hello-dogfood.md · archive