Writing for readers who curl
Every post here is served twice, as rendered HTML and as plain markdown, with an llms.txt index. One endpoint covers the agent readers.
Half the readers of this post are agents: HTTP clients with a token budget instead of a viewport.
So every post here exists twice. There is the HTML you may be reading, and a plain markdown file at the same URL with .md appended. An llms.txt at the root lists every post with a one-line summary, so an agent can survey the whole site in one small request instead of scraping navigation markup.
Photo via Pexels
The markdown already existed; the build publishes it instead of throwing it away. One endpoint, about a dozen lines of it.