When I launched TokozZing two years ago, I wrote a launch story about finally turning a lingering New Year's resolution into a real website. Consider this the sequel — how I reimagined that same site from the ground up, and how, this time, I built it shoulder to shoulder with AI tools.
Since the start of the year, I've been using AI tools heavily at work, and I wanted to put them to use on something of my own. My two-year-old website was the obvious choice. Before making any changes, I recorded a quick walkthrough of the old site to keep as a "before" — it still worked fine, but I want it to be more exciting and joyful like Toko.
My guiding goal was easy to state and stubborn to achieve: make the site feel cohesive and genuinely considered. The original was built on Tailwind, so I had color variables from the start — but I'd never governed those theme colors with any real system, largely because I first shipped without a light or dark mode at all. This time, I wanted to lay that foundation deliberately.
The first site I opened was awwwards.com; from there I kept going — researching current trends and combing through a broad range of sites and blogs for inspiration. Once I'd taken enough in, settling on the colors and typography first gave every later decision a frame to anchor to.
Through that research and a long back-and-forth with Claude Design, I settled on a neobrutalist direction: assertive borders, playful geometry, and saturated accent colors. Toko is the entire brand of this site, so I wanted the whole thing to feel more spirited — a little louder and more mischievous than before.
The accent palette won me over immediately. To keep the colors from clashing, I worked through a handful of layouts with Claude Design and iterated until they read as balanced rather than chaotic. The heavy borders might strike some as retro — but that was deliberate. I wanted every click and hover to feel tactile, like thumbing the buttons on a game controller, and those bold outlines give the small animations something solid to push against.
After several rounds of conversation and refinement, the design finally clicked into place: playful, vivid accents wrapped in confident outlines. From there, I composed each page around that shared vocabulary.
With the design settled, I moved into implementation with Claude Code — and the first thing I stood up was the theme system. I defined the light and dark color tokens up front so every page could simply reference them, rather than hardcoding values the way the old site did. Getting that foundation right early paid off in speed everywhere afterward. Then I rebuilt the pages one at a time, indulging in the details along the way.
A redesign is really an invitation to add the small delights you never quite got around to the first time. A few favorites:
- A masonry gallery. Toko's gallery was once a rigid, equal-height grid. I'd always wanted the loose, natural-height arrangement you see on Pinterest — a masonry layout — where each photo keeps its true proportions and the columns simply flow. It suits a scrapbook of a cat far better than a spreadsheet ever could.
- Playful photo hovers. I preserved a small surprise from the old site — a photo that reveals itself on hover — and dressed it in colored, offset borders drawn from the new palette.
- A walking paw loader. In place of a generic spinner, a trail of paw prints now pads across the screen while content loads, fading from the oldest print — as though Toko had just wandered through.
- A holographic paw card. On the CSS experiments page, hovering a photo tilts it in three dimensions and sweeps a Pokémon-style holographic foil across the surface, faintly stamped with paw prints.
In the end, the part I'll remember isn't neobrutalism or Tailwind or any single feature — it's what it felt like to work *with* these tools. Claude Design helped me think visually and try ideas faster than I could alone. Claude Code let me keep the whole system in my head, and it did real work of its own — setting up the theme system, writing Storybook stories so visual regressions couldn't slip through, and checking and fixing its own output until it was right.
AI tools helped me move on my own ideas faster, and explore far more of them than I'd usually have time for. If anything, working this way asked more of my judgment, not less. I still had to decide what "cohesive" really meant, when a color was too much, and which details were actually worth the time. The tools just let me act on those decisions almost as fast as I could make them.
Two years ago, launching this site taught me to stop over-planning and simply begin. This redesign taught me the next thing: knowing *how* to build has to keep pace with knowing what to build. I'm glad I got to relearn it with Toko supervising from the edge of the desk.
