#ai · #platform · May 2026 · 6 min read

Building a website
with AI.

Most people use AI the way they use autocomplete. You type something, it finishes the sentence, you move on. That’s not what happened here.

This site was built by treating Claude as a technical collaborator — someone to think with, not just a tool to generate output. The distinction matters more than it sounds.

The brief was simple. The execution wasn’t.

I needed a portfolio site. Not a template, not a WordPress install, not an agency build. Something that reflected how I actually think — commercially direct, editorially restrained, built properly.

The design brief took three sessions to get right. Not because the AI kept getting it wrong, but because I kept refining what I actually wanted. That’s the honest version of working with AI: it surfaces your own unclear thinking faster than anything else. You can’t be vague with it. Vague in, vague out.

What Claude Code actually did.

Every file on this site was written by Claude Code. The HTML, the CSS, the scroll behaviour, the responsive breakpoints, the git commits. I didn’t write a line of production code.

What I did write was precise instructions. That’s the skill — not prompting tricks, not magic words. Knowing what you want clearly enough to direct a capable collaborator. It’s the same skill that makes a good creative brief, or a good agency relationship, or a good conversation with a developer who knows more than you do.

What Claude Design did.

The visual direction came from Claude Design. I described the aesthetic I wanted — editorial, not corporate, warm off-white, Fraunces as the display typeface, a single coral accent used sparingly — gave it competitive references, and iterated until the output matched the thing in my head.

That took judgement, not prompting. Knowing when something was wrong and being able to say why. Knowing when to push back and when to accept a suggestion that was better than your instinct.

The parallel to running digital teams.

I’ve spent twelve years directing digital work — platforms, agencies, internal teams. The best outcomes always came from the same place: clear direction, fast iteration, honest feedback, no ego about who had the idea.

Working with AI is the same model compressed. The feedback loop is faster. The output is immediate. The requirement for clarity is unforgiving.

The people who will get the most from AI are the ones who were already good at directing other people’s work. It’s not a replacement for judgement. It’s a multiplier of it.

What’s next.

This site will keep being built in public, in this way. Case studies, articles, the Payload CMS setup when it comes — all of it documented as it happens.

Not as a tutorial. As a record of what the process actually looks like when you use AI properly, on a real project, with real constraints.

The CV is the work. This is how the work gets done.