The Web Tells Machines What Things Are. It Doesn't Tell Them Why.

I had a thought the other day that won't leave me alone. We've spent decades building the web for humans. HTML gives us structure — headings, tables, buttons, forms. A human looks at a page and immediately understands what's going on. Machines don't. An LLM sees a <table> and has no idea if it's pricing tiers, analytics data, a leaderboard, or a comparison tool. It just sees a table. It has to guess why it's there. And that guess becomes the foundation for AI browsing, accessibility systems, personalized interfaces, and agents operating software — basically everything we're trying to build next. The web gives machines structure, but not intent. And I think that's the missing layer. The Semantic Web Tried This. It Failed. This isn't a new observation. Schema.org, RDFa, microformats, the entire Semantic Web movement — they all tried to make web content machine-readable. They all struggled with adoption. And they all struggled for the same reason: The developer does the work. Someone else