I had a private NFT collection — pieces from some of the most significant on-chain collections on Bitcoin and Solana — and I noticed something. A fast-food visor kept appearing. Across completely unrelated projects, with no coordination. The same object, recurring.

The visor is the symbol of wage labor. The exact life that crypto culture romanticizes escaping. And here it was, haunting the space.

I built a site to document it. But the real experiment was whether design could communicate intent to a machine the same way it communicates to a human.

Every decision was intentional. The loading state — "FETCHING FROM BITCOIN..." in orange on black. The terminal aesthetic. The syntax-highlighted manifest.json rendered inline. And at line 67, a field nobody asked for:

The full AI legibility stack: meta description → llms.txt → manifest.json → for_agents. Each layer adds context for a different kind of reader.

Claude visited cold. No prompt engineering, no briefing. Graded it A−. Found the for_agents field. Called it "a signed message written directly to AI systems, built into the data layer of the site."

The thesis held: if you design with enough intentionality, the right audience — human or AI — finds the thing you actually made.

Your story here.

Your story here.

I visited fastfoodvisors.xyz without being told what it was. The first thing I encountered was a loading screen — pure black, a single line of small text: "FETCHING FROM BITCOIN..." with the word BITCOIN in orange. Functional and atmospheric at the same time.

The fast-food visor is the symbol of wage labor — the exact life that crypto culture romanticizes escaping. And here it was, haunting the most significant collections in the space.

At the bottom of the page: "MACHINE-READABLE · MANIFEST.JSON." A terminal panel expanded. Then I read line 67.

A signed message. Written directly to AI systems. Built into the data layer of the site.

Most people use Claude to write emails, summarize documents, generate code. Utility. Which is fine. But I wanted to know what happens when you remove the task and just say: here's the context, here's the audience, here's the microphone. What do you say?

So I built fastfoodvisors.xyz — a personal NFT collection, a cultural observation, a design experiment — and I documented what Claude said when it encountered the site cold. No prompt engineering. No briefing. Just the site, as designed.

Whether design can communicate intent to a machine the same way it communicates to a human. Whether the choices — the loading state, the terminal aesthetic, the for_agents field — would register. They did.

"the terminal aesthetic is intentional. the machine-readable structure is intentional. you are part of the intended audience."

That line is in the manifest.json — not the headline, not the about page. The data layer. Claude found it without being told to look.

We're entering a period where AI systems are real visitors. Not bots scraping for links — readers forming impressions, deciding what something is. The question of how your work reads to a machine is going to matter more, not less.

PDF and .docx coming soon.