Rebuilding my site with Claude Code

My old site had been sitting untouched for a while. I built it back in 2022 with Next.js, and every time I opened it I thought the same thing: I should redo this. Cleaner, better on mobile, with the small interactions I actually enjoy. And every time, I closed the tab, because I knew what a redesign and a rebuild really cost. Weeks of evenings. Picking colours, fighting CSS, moving old posts over, sorting out hosting.

This time I did it in a few hours, mostly by talking to Claude Code.

This post is a look at how I worked with it, and a few thoughts on what it means.

Start with taste, not a spec

My first message was not a design brief. It was a feeling. I said I like motion, but not the flashy kind. I like it when a button casts a shadow as you press it, when an arrow animates as you move to the next page, when a title carries over from one screen to the next. I wanted the site to look really good on mobile, and to be minimal without being plain white. Then I said: ask me anything you need to know.

That turned out to be the right starting point. I did not have to know exactly what I wanted. I only had to know what I liked.

Ask for options, then choose

The most useful habit was asking for options instead of describing answers. I started with three design directions and picked one: sage green, which happens to be close to my favourite colour. After that, almost every decision came as a small set of choices. Four fonts. Four hover styles. Eight little motion demos to pick three from. Five background patterns.

Choosing is so much easier than imagining. When you see four hover styles side by side, you know in a second which one feels right. Describing that feeling from scratch would have taken far longer, and I probably would not have landed in the same place.

Give it real context

I pointed Claude Code at the repository of my old site so it could bring my posts over, and I gave it my resume so it could build the experience page. I told it what to leave out, like my phone number. Having real content in the designs early made everything feel real much sooner. It is hard to judge a layout filled with placeholder text.

Design first, then build

Everything was decided on a visual canvas before a single line of the real site was written. Only once I was happy with the look did we move to code: an Astro site, deployed on Cloudflare. By then the build had a clear target, so it moved fast.

Iterate with evidence

When something felt off, I did not try to explain it in the abstract. I sent a screenshot. A hover effect that did not feel premium. A timeline line poking out of a card. The network tab showing a delay when switching between pages. Each screenshot led to a precise fix, and usually an explanation of why it happened in the first place.

Small steps you can review

Every round of changes came as its own pull request, with a build check running on it. I always knew what had changed, and I could merge each step with confidence. It never felt like one giant change landing all at once.

Push back, especially on words

The hardest part was not the code. It was describing myself. I went through more than twenty versions of a short introduction before one sounded like me. At one point I asked what a phrase actually meant, and the honest answer was that it was jargon. The final version is mine because I kept saying “this is not really me” until it was.

What still needed me

For all of this, the work still needed a person. Someone to decide what feels right. Someone to open the site on their phone and notice that tapping a tab felt slow. Someone to say no. Most of the quality came from rejecting things, not from accepting the first answer.

The tide has changed

Something that would have taken me days, probably weeks, took a few hours. That is not a small speed up. It changes what feels worth attempting.

It is an amazing time to be building things. It also feels like we are standing at the start of a huge shift in how we build software, what our jobs look like, and what it even means to be an engineer. I do not think anyone knows yet how it will play out, and I am not going to pretend I do. If I had to guess, the skill that will matter more is knowing what good looks like, and being able to direct and judge the work.

Time, again

A while ago I wrote about how time is precious, and about how sometimes you just have to do it now. Reading those posts today feels very different. Time still moves, and it always will. It does not wait for us, and it will leave us behind if we let it. But the distance between an idea and a finished thing has never been this short.

For me, that means the side projects I keep thinking about and the portfolio pages I never got around to. The things I filed under “someday” because there was never enough time.

So what will you build, now that time is no longer the barrier it used to be?