I designed and built this site with Claude Code. But that doesn’t mean Claude designed it. I brought a lot of the old website’s vibe and content along with me to start, and reshaped it over a few months of tweaking.
I let Claude take the wheel on the code for this one. I modified a lot of the CSS rules and methods as we went, using Chrome Inspector to tweak and then feed values and screenshots back to Claude (so many screenshots).
One of the things I miss about is the endless direction changes and copies that track the progression of your thinking and the places you pushed and explored. With AI, everything is a new version on top of the last one. So as I go, I take screenshots. Some to point out a bug, but a lot, like the ones below, just to document variations. I’d love to build a workflow that automates that step with each major change. But for now it’s just dozens of screenshots in a folder on my computer.
Every time I build something with AI, I have it build me little tools to explore options live on the site. Actually I’ve done this for years in front-end development to compare options and simulate states, but it’s much easier to tell Claude the states I need now.
For this site, a font picker (F), a color variable picker (C), a grid overlay (G), and some other one-off toggles to compare options live in various sections. I love this way of working. It’s like building my own Figma controls. I tried using DialKit for this in previous projects, but I found adding my own tools on the fly was faster and more specific to my needs.
Below is a timeline of what I documented of my design process in screenshots so far. I’m sure I’ll continue to tinker on this for years to come, as is my custom. But I’ll stop the timeline here for now, now that the site is live.