The Atlas

A live map of this blog. Every page is a node, coloured by the part it plays; every internal link is a thread between them. It is built from the writing itself, so it cannot drift out of date.

For the curated, hand-written tour of the ideas, see Maps. For the thinking behind a map like this, The Language of Sound and the dev-log I Built a Live Map of My Own Blog.

How to read it

The map opens already settled (its layout is computed when the site is built), framed to your screen. Then it is yours: hover any node to light up everything it links to, click a node to inspect it (its role, its date, what it links out to in gold, and what links back to it in teal), and click Open page to read it. Drag to pan, scroll or pinch to zoom, use the buttons to zoom and refit, and the ♪ toggles a little synthesized sound.

The bright, labelled anchors carry the structure; everything else is the field of ordinary posts, coloured by kind. The five roles, and what earns a page each one:

  • Pillar (6): a cornerstone idea important enough to earn its own clean, dateless URL, like /coherent-complexity/. The promoted few.
  • Cornerstone (18): an anchor the rest of the site leans on. This is a role, not a category, and it is earned by how many other posts link to it. So a heavily referenced book review counts every bit as much as an essay: the best-supported anchor on the whole map is the Antifragile review, simply because more posts point back to it than to anything else.
  • Hub (7): a curated landing page that points outward to a whole cluster, like Films or AI Development. A hub gives; a cornerstone receives.
  • Category (6): the plain archives, like Essays, Free Writing, and Thoughts.
  • Post: everything else, the ordinary writing, coloured by kind. The threads between posts are what pull the map into its clusters.

Questions this map answers

How it works

A build step reads every published page, tags each with the one role it plays, and walks the posts for the links between them. Pages become 293 nodes; links become 1067 edges; the whole thing is written to one file this page reads. No database, no server: the site already contains the graph, and the script just makes it look at itself, every time the site is rebuilt. The full reasoning is in the dev-log.