User Agent
I made more of my blog navigable as pure Markdown by adding .md to the end of most pages. Validating via these .md files still sucks for a human, but agents can use it to save tokens. When the semantic web first came around, I remember reor…
Read here: User Agent
I made more of my blog navigable as pure Markdown by adding .md to the end of most pages. Validating via these .md files still sucks for a human, but agents can use it to save tokens.
When the semantic web first came around, I remember reorganizing my home page to include the tags that would make Google render my blog better in search results. It’s exciting to be able to participate in the web, to change something on your own site and have your changes reverberate through the rest of the standards-based web. To my knowledge, Google no longer does this in any meaningful way, but lately the web has been flipped upside down by agents and things are on the table again.
Maybe there’s a future where my blog only renders simple text and html because styles are controlled by the user agent (by which I means browsers, although I expect the whole concept to continue to evolve). Is it a funny coincidence that we call browsers user agents?
Now we also have user-controlled “agents” that browse the web on our behalf. They render how we want them to render, we control, or can control, the styles and presentation of the content we fetch. This is how it was originally designed too, browsers have a default stylesheet and you used to write your own to suit your style.
Somewhere along the way the browser stopped being the user’s agent and became a tool for companies to render content into a window you mostly don’t control. (You only get to control the “chrome”, the borders and buttons around the outside of the page). The fight over ad blockers makes this clear. Google Chrome effectively removing them means you aren’t allowed to render pages the way you want. GDPR dickovers and other user-hostile features like fields you can’t paste into and things you can’t right-click on, are further evidence that browsers are no longer user agents. The small slice of users that know you can edit the page itself in the inspector, also know that it’s more of a last resort than a feature.
Maybe agents, the LLM kind, will change this? Two things make me hopeful: we pay for them so incentives are aligned with the user, and they summarize what they read, so most dickovers, ads, mess, and user hostile features all gets squashed into a summary. The only thing they do now is increase token spend and cost to serve the page).
I hope my little change to render plain text helps.