Is WordPress dead? It may still be twitching, but after a weekend with Claude and Astro we started moving our company website off it.
Every few years I move my personal website to a new technology. The last version was Flask, PostgreSQL, Sphinx, nginx and uwsgi. All of that to serve a few dozen articles and over a hundred short notes. Form over substance. Over time the stack turned into a fossil: search didn’t work, the libraries had aged, and I kept putting off every change because I knew how many evenings it would cost me.
Of course, hardly anyone builds a blog on a stack like that, unless they can’t handle WordPress 😉 WordPress is the most popular choice today, for blogs and above all for company websites. My attempts at it usually ended in failure. Anyone who has built a company site or a blog with it knows the hours spent arranging templates and fighting plugins.
Back to my site. Recently I asked Claude what to do with it. It suggested Astro. Over the weekend the whole blog was migrated, with fixes I had been putting off for years, with hundreds of formulas from my articles on quantum computing, and with a look far better than before. I didn’t get stuck once on CSS or centring divs. Instead of fighting the tools, I was carrying out and testing ideas.
👉 The work is changing: from craft to concept and design. I took three lessons from that weekend.
1. The most valuable part was the conversation before the first line of code. Before any code was written, we talked through, among other things:
- splitting content into articles and notes,
- supporting two languages,
- where to keep images: the repository or the cloud,
- a way to keep notes from social media from getting lost,
- even the name of the repository.
I lost a few of those arguments and I don’t regret it. The value was in having someone challenge my assumptions with arguments, while I dealt with decisions rather than implementation.
2. A green build is not the end. At one point the build reported “Complete”, while every image in the old articles would have been dead. We only caught it because we checked the result, not the status. There were several more problems like that. Verification doesn’t go away; what changes is who writes the code you verify.
3. I talked instead of typing. I dictated about 90% of my prompts with MacWhisper. When I type, I abbreviate: “fix the menu”. When I speak, I explain what bothers me and why, like to a colleague at the next desk. Those descriptions produced the best suggestions, because the AI got a problem with context rather than a ready-made solution to execute.
That’s why we’re now moving the KLL-ERP website. A personal site is a safe testing ground; the company one will be a more serious test.
All the content of my site now lives in plain Markdown files in git. I wonder what technology version 6 will be built with in a few years. One thing is certain: tools change, and the content should outlive them.
What is your website built on, company or personal?
