I have known what I wanted this website to look like for quite a long time. The ghost word behind the headline. Typography that felt like future tech and a Sunday morning as well, not a productivity app. A hero section that breathed without shouting. My vision was clearly unclear, though.
And then I lacked the technical ability to build it, too. I tried page builders, templates, "no-code" tools that turned out to have quite a lot of code hiding behind them. A CSS property I didn't understand. A layout that refused to behave. A font that looked right in theory and wrong in practice.
Every time I got close, something blocked me, which cost not just my nerves but also a lot of money. I ended up with compromises. Not bad, but not it. Then I tried Claude Code. And in just one day, my website became the way I had been describing it in my head for years.
What the process actually looked like
I had expected the work to feel like dictating to a very fast assistant. Actually, it felt a little bit like learning a new language. A language for describing my visual ideas, with a patient conversation partner who could immediately show me what my words produced.
Claude Code can do a great deal, but it can only do what you can describe. My first prompts were vague: "make it feel more editorial," "I want it to look like a real publication." These produced results, but not quite the right results. Iterating felt like the design conversations I imagine happen in agencies, except I was simultaneously the client and the person trying to translate client wishes into something buildable.
At some point I realized that I was learning to have taste in language. To say "the ghost word should overflow the right edge, with the heading aligned to the bottom-left, and the overall height reduced so the overflow appears at the top as well." That sentence took me ten minutes to formulate. A month ago I would not have known what I was looking at, let alone how to ask for it.
The technical decisions were made by Claude Code. The aesthetic decisions were made by me. The translation between them was the actual work.
Back to the question I keep asking
In "The effortless illusion" I wrote about desirable difficulty: the friction that builds competence, the struggle you can't afford to skip. Which struggles, I asked, are worth preserving?
I skipped learning CSS deeply. I don't feel conflicted about that. CSS is a means to an end, and for what I needed — a personal blog with a specific aesthetic — the end was the point. There is no version of this story where six months of front-end study before launching a blog about AI and consciousness was the right choice.
What I didn't skip was the work of knowing what I wanted. That is much harder than it sounds. Many people, given a blank canvas and infinite tools, produce something mediocre. Not from lack of taste but because translating internal vision into external decisions is a slow skill, built through years of looking at things and noticing what works and why. I realized that I have been working on that skill for a long time.
Now, Claude Code gave me a way to exercise it at a resolution that was previously unavailable to me. I also want to be honest about what I still don't understand. I can navigate the code now and make small changes. I couldn't have written it from scratch, and I don't understand every decision embedded in the structure. That gap is real. I have chosen to accept it the way I accept not understanding the engine under my car: at some point you decide what depth serves your actual purposes.
The skill question, made honest
The more interesting question for me is not if I built this but what the process required. I think it requires clarity of vision. The ability to describe choices precisely. The patience to iterate. Enough aesthetic judgment to know when something is just right. Those are not trivial skills. They are also not the skills that show up in job descriptions for front-end developers.
Many technical capabilities are now accessible through language. What's becoming more valuable, not less, is the ability to know what you want and say it in an easily understandable way. I learned I have to develop enough judgment, in whatever domain, that I can direct a powerful tool toward something specific and meaningful, instead of just generating output.
"What skills should I develop" just got more complex, not simpler. But it also got more honest. The answer is no longer "whatever the tutorial teaches next." It's closer to: know the context, learn to articulate it, stay curious about which tools can extend your reach.