Machine legibility: one discipline, four readers
The web grew a second audience that reads structure instead of pixels and greps instead of browses. Here is the thesis, and the four claims it rests on.
Writing
Notes on machine legibility, agent-readable interfaces, retrieval, and the design systems work behind this site.
Showing all 5 posts
The web grew a second audience that reads structure instead of pixels and greps instead of browses. Here is the thesis, and the four claims it rests on.
An operating agent's entire perception of a page is its accessibility snapshot. Whatever is not named there does not exist to the agent — which makes semantic HTML a ranking factor with teeth.
The acronyms fight over which reader you optimize for. Roughly 70% of the work is shared — and the fourth reader, the one that acts rather than ranks or quotes, has no acronym yet.
Inside agent harnesses, retrieval shifted from frozen embedding indexes to lexical, live, self-correcting loops. The reason is failure behavior — and the SEO consequence is that exact-match keywords are back.
Snapshot density, internal surfaces, the sr-only inversion, affordances as backlinks, and the fact that nobody's rank tracker sees agent traffic. Each with the one move that captures it.