One content, many projections.
Separating content from code is usually sold as a workflow fix. True, and that is the smaller half. The larger half is what becomes possible once content is a standalone structured document.
Because none of these artifacts take the foundation as an input, they are identical in every context and unchanged when you redesign the site. In other stacks they are either hand-maintained, which drifts, or scraped back out of HTML. Here they are the same source projected again.
What comes out of the box
Four artifacts, none of them authored, none of them maintained by hand.
Agent-readable by construction
An annotated index of every page — title, link, one-line description. Links point at the markdown, so following an entry is one hop rather than fetch-then-strip. Grouping comes free from the page graph.
The document, not a re-render
Serialized from the stored document by the same codec the visual editor uses to write files — not HTML converted back into markdown. Section types and params are deliberately absent.
Search with no crawler
Built from the content graph at publish, so it is exactly as current as the site. Page and section granularity, with weights, anchors and excerpts. Headers and footers are excluded structurally.
Translation without t()
Content reaches components already localized. Strings are keyed by content hash with the contexts they appear in, and whole sections can be rewritten per locale with staleness tracked against the source.
The compounding argument
Add a language, and everything derived from content follows it. A complete localized build, a localized search index, a localized agent index, and hreflang wiring — none of it configured.
67/67
Descriptions resolved automatically
0
Leaked HTML tags or entities
2
Hops from the index to any page
0
Lines of frontmatter authored
See it on your own content.
Create a project, write a page, and look at what the build put next to it.