What does the search surface look like?
- Metadata
- Every page sets its own title, description and canonical URL, derived from the content file rather than typed twice
- Structured data
- JSON-LD on every page. Person and WebSite on the home page. Each case study and post carries its own graph, with breadcrumbs
- Social previews
- Open Graph images are generated per page at build time from the page's own title and summary
- Generated files
- sitemap.xml, robots.txt and the RSS feed are built from the content at deploy time and never hand-edited. Draft posts are excluded from the index, the sitemap and the feed by one rule in the loader, so a draft cannot leak into any of them
- Answer engines
- An llms.txt at the root summarises the site in plain text for retrieval agents, generated from the same files as the pages. The AI crawlers that would read it are named and allowed explicitly in robots.txt
- Real 404s
- The development-only reference pages return a 404 status from middleware in production, not a 200 with a loading skeleton. A soft 404 is the kind of thing this site exists to say I notice