I create websitesand write the codeand pick each pixeland make it moveand get it foundand take it liveand write the code

I pin down the real problem and the real need first, not the deliverable that was asked for. Then the deep work: market and competitor research. With AI the chosen direction gets tested against practice in hours rather than weeks.

Then I lay out the structure, the flows and the priorities, and find the shortest route to real progress.

Instead of static wireframes I build a working mid-fidelity prototype in code. You get something you can click through, not a flat mockup.

The client walks through it before a single visual decision is made, so the argument is about behaviour instead of taste. On a five-page site that is the honest amount of testing there is. The rest of the validation comes after launch, from analytics.

This is where taste takes over. AI helps me explore directions and produce assets quickly, but the craft is hands-on: hierarchy, rhythm, and a look that belongs to you rather than to a model's defaults.

I iterate until the design feels deliberate, and until every state is decided rather than discovered later: empty, loading, error, overflow.

Then I build it myself: plain HTML, CSS and JavaScript, written with Claude line by line. There is no builder or framework in the way and no build step at the end, so what ships is what was designed, down to the focus ring.

The launch is mine as well: hosting, the domain pointed at it, tags, analytics and consent wired before the first visitor rather than after. The site is handed over measured, with its sources and every access.