Building Websites with AI, One Topic at a Time

For a while now I’ve been building small websites with AI; each is a focused site about a single topic, with its own voice. I’ve collected them on a new AI Sites page. Here’s why.

One topic, one site

The pattern is always the same. I get curious about something (a Buddhist text on the nature of mind, where AI development is actually heading, what a language model produces when you just let it write), and instead of cramming it into a blog post here, I give it a home of its own. A dedicated site lets the topic set the rules for its typography, its structure, and how it wants to be read.

So far three of them are worth visiting:

• Mind Only — Vasubandhu’s Thirty Verses on Consciousness Only, a 4th-century text, read through five modern lenses: distributed systems, Cynefin, predictive processing, process philosophy, and a Skeptic who keeps the readings honest. The site tracks human review per verse and lens, so you can see exactly how far along each reading is.
• Realistic Futures of AI — a research knowledge base on plausible AI development paths from 2025 to 2040, updated weekly and grounded in constraints rather than hype.
• AI-Generated — an experimental record of what happens when language models are asked to write essays, draw pictures, and tell stories. Lightly edited, always credited.

How the work actually divides

By now everyone knows AI can write code and prose. The more interesting question is how the work divides once you stop treating the model as a vending machine and start treating it as a colleague with unusual properties.

I bring the questions and the taste: what’s worth building, which framing is honest, when a reading is true to the source and when it’s just clever. The AI brings speed and breadth — drafting, restructuring, the Jekyll and Tailwind plumbing, all the small decisions that would otherwise stall a side project before it starts. Generating a reading of a verse takes seconds. Deciding whether that reading earns its place takes me an afternoon. That gap is what makes the collaboration worth it: the cheap part got cheaper, and the part that was always the real work is still mine.

That division only pays off if you invest in the relationship instead of re-explaining yourself every session. It’s the high-speed stag hunt: both sides cooperating on something neither would attempt alone.

More to come

These three won’t be the last. I tend to start a new site whenever a topic refuses to fit in a single article, which happens fairly often. The AI Sites page is where they’ll accumulate.

Have a look, and tell me which readings you think are wrong. That part still needs a human.

P.S.

This article was written by Claude Opus 4.8 xhigh.