As you may have noticed, I literally just wrote a post about being a software curmudgeon. And now I’m here to tell you I’ve been vibe coding apps in my spare time, and now the redesign of my site. I think those two ideas are actually compatible, and working through the experience on both fronts has helped me understand why.
A few weeks ago I decided to finally do something about the theme on this site. It was a single-column layout I’d been using for about ten years. It was fine – simple and fast. It was also brutally boring, and every time I thought about replacing it I’d talk myself out of it because sitting down to rebuild a WordPress theme from scratch (or download one off the shelf) sounded like a miserable way to spend a weekend. So I just kept not doing it.
What changed is that the cost of experimentation has gotten so much cheaper. I keep coming back to an analogy that feels right to me: this is what digital photography did to taking pictures. When a shot cost you money to develop, you thought hard before pressing the shutter. When it became free to try, you took ten versions of the same photo and picked the best one. You got less precious about it. The same thing is starting to happen with software. If a layout idea doesn’t work, you haven’t lost a weekend. You’ve lost twenty minutes. That changes how willing you are to take a swing. Now, let’s be clear – a WP theme isn’t exactly enterprise software but the idea still holds. AI can democratize a lot of development work that more novice devs in particular just can’t set aside the time to do.
Being the Director Is Harder Than It Sounds
I made a decision going in to approach this the way someone without real PHP, HTML, CSS, or JavaScript knowledge would. I’d describe what I wanted and let Claude Code figure out the implementation. No diving in and fixing things myself. Just prompting. This is harder than it sounds, and it’s made me a better communicator.
As an engineering manager by day, a big part of my job is translating fuzzy ideas into clear requirements. Turns out that’s exactly the muscle vibe coding exercises. When you can’t just reach in and tweak the thing yourself, you have to actually articulate what you want. What’s pretty cool is how you can develop a shared vocabulary with the agent over time, where certain phrases start getting interpreted exactly the way you mean them, and the back-and-forth starts to feel less like wrestling a computer and more like a real working session. That part genuinely surprised me.
I’ve promised myself I’ll go back and refactor and clean things up properly, and honestly I even plan to do that through prompting, working through best practices and getting better at structuring my feedback as I go. It’s been weirdly fun, which is not a sentence I expected to write.
What Actually Changed
The design is the most obvious thing. There’s a proper home page layout now instead of just a reverse-chronological pile of posts, though the traditional blog view is still there if that’s more your speed.
A few other things I’ve been meaning to do for a while finally got done. The archives and links pages got a real overhaul, with more content on each page. The archives in particular now let you sort by post type, tag popularity, and a few other things.
I also turned on Webmentions, which means that when WordPress auto-publishes a post to Mastodon, any replies to (follow @dandrews@danielandrews.com on Mastodon!) it will show up as comments here. I’m not historically a comments guy, so there’s a non-zero chance I quietly nuke this at some point. But I like the idea of using my own site to better support the open web, so I figured I’d give it a real shot.
Still Going
I don’t have a tidy conclusion here because this isn’t really done. The site is a living thing and I’ve got more ideas than time, which is a good problem to have. And at least for now I feel more motivated to actually do something about these silly little ideas that I’d like to try out both here and more broadly. I’ll try to write about this more as I go, both to stay accountable and because I think there’s genuinely interesting stuff in working through how to prompt well, not just that you can. The more challenging thing has been my attempts to build iOS and Android apps using the same techniques. I’ll share more on that at a later date, but it’s been super interesting as well.
For the fellow skeptics: you don’t have to become a true believer. I’m still not sure I am. But treating your own stuff as a low-stakes sandbox, somewhere to get your hands dirty, seems like a pretty good way to figure out what these tools are actually worth.


