building my website with ai
how this website took shape in SvelteKit, with Claude, Gemini and relentless human quality assurance
My old portfolio was an embarrassing cliché. Near-black background, terminal green monospace
font, a blinking cursor in the hero, and mock terminal breadcrumbs
(~/showcase/cascade). It looked like it was generated in four seconds by someone
who had never designed anything in their life.
I wanted a site that looked like collage, one of the design styles I love most. I also wanted it to run well on weak phones, so I picked SvelteKit, which updates the page directly instead of going through a virtual DOM. The site should feel smooth even on an old iPhone 6s like the one I have here, which I used a lot for testing.
So I opened an empty folder and spent the next few weeks building and improving it progressively with Claude and Gemini.
escaping the training set
When I asked for designs without any guardrails, I got whatever sits at the center of the training data. Depending on the company, that can mean purple gradients and bento grids, the dominant style across the web for the last few years. With Claude, the default leans toward an exaggerated editorial look: giant serif headlines, 1990s magazine typography, an earthy palette, and faux-artisanal touches.
To get off those defaults, I asked Claude to go and research schools of design and art, looking at artists and designers from different eras and regions. Then it drafted standalone HTML boards with redesign options: Curitiba's transit grid, risograph printing, concrete-art planes and a few others.
The one I decided to keep was the one Claude called "Antropofagia", based on Oswald de Andrade's 1928 manifesto of the same name: loosely, the idea of taking what you learn from elsewhere and making something of your own with it rather than copying it. That felt close to how the work actually happens for me: Swift learned from Apple, then design from a mostly European canon, then psychology from foreign books, and in the end I, a Brazilian, have to make things out of the lot.
In the SvelteKit components, that idea became literal. Instead of putting screenshots inside
fake browser frames, the headline letters swallow the newest project cover directly into the
typography with background-clip: text. The rest of the site stays quiet with its
newsprint and serif type, with the collage surfacing in just two other places: the section
titles and the torn halftone thumbnails.
For the project showcase grid, I also took inspiration from Molleindustria's website, which puts the studio's games in a grid arrangement with taglines that I've always appreciated.
An earlier draft stamped the homepage with Tupi or not Tupi — Andrade, 1928, but we cut it. It was kind of cringe.
AI doesn't have eyes, but it can use a ruler
The homepage's headline was one of the first places Claude led me straight into a wall. A broken line of code is easy to spot. Valid CSS that comes with a convincing explanation and still renders wrong in a real browser is much harder.
Because the hero letters contain covers from different projects, they need a guaranteed
contrast floor so the words remain legible against the newsprint background
(#fffdf9). Claude combined background-color, a multiply blend mode,
and text clipping, assuring me the contrast would never drop below 3:1. In Chromium,
background-color is silently dropped from the blend group when text clipping is
active. Over a dark cover like Cascade's, the letters turned to mud, while browser devtools
reported styles that looked completely fine.
Arguing with Claude about it in chat got me nowhere. It couldn't see the screen, so it kept
explaining why its math should work. What finally worked was an automated check. I had it write floor.js with Playwright and Sharp. It opens a headless
browser, takes a screenshot of the headline, trims the antialiased edges off each letter, and
checks every interior pixel. If any pixel drops below 3:1 contrast against the background, the
script exits with an error.
Once the model could run the script and see real numbers in the terminal, the problem took ten
minutes. Linear gradients inside background-image stay in the blend group,
so layering a gradient under the cover clamped the lightness without touching
background-color.
buttons on paper
The buttons had their own version of the problem. Most websites answer a hover by inflating the
element with scale(1.04), the ones LLMs make most of all, but paper doesn't swell
when a finger gets close. I wanted buttons that felt like scraps of paper sitting on newsprint, with a 3px
hard edge underneath. When you click, the scrap sinks flush with the page.
When we first wired it up, half the buttons didn't budge. Entrance animations were holding
their keyframe transforms in place and overriding the click state. Moving the press offset to
the CSS translate property fixed it, since translate works
independently of keyframe transforms.
To keep that behavior from breaking later, Claude cooked up scripts/press.js. It runs
Playwright, holds the pointer down on every button across six routes in both themes, and
checks that the element actually moves 3px on click.
certified ink
The slowest problem to notice was drift. Each component Claude wrote worked fine on its own, but from one session to the next it lost track of the rest of the site.
On this site, for example, each section uses its own ink: charcoal for home, blue for showcase, green for
blog, and red for contact. Every link, hover state, and filter chip is supposed to inherit
that section's --accent color. A stylesheet with four colors still looks
reasonable at a glance, though, so for weeks the blog index had a blue back-link and a black
filter chip, until I got around to it. Claude never noticed.
To stop that drift, Claude wrote two more scripts, colors.js and look.js.
Instead of checking screenshots by hand or re-reading CSS, the scripts inspect the live DOM in a
headless browser. They print a table of computed values, verify that every element uses the
right section ink, and measure column widths. If a route borrows a color from
another section, the test fails.
fitting words
The site is bilingual, which brought a layout problem. Portuguese words run noticeably longer
than English ones. DEVELOPER has nine letters; DESENVOLVEDOR has
thirteen. If you pick a fixed font size or guess with breakpoints, one language will sometimes overflow the
column or the other will look loose.
Rather than guessing font sizes or breakpoint values, I got Claude to write a small Python script with
fontTools to calculate the exact advance width of each headline word down to the
thousandth of an em. We stored that value as a custom property (--k) and used
container query units (cqi) to let the headline size itself to the column:
.head { --k: 15.56; font-size: calc(88 / var(--k) * 1cqi); }
Both languages lock to the column, preventing overflow.
what it changed
Building the site this way didn't let me skip learning how browsers work. I needed to know more, if anything, because when Claude offered five plausible reasons an animation broke, I still had to work out which layer had actually failed.
It also changed how I prompt. Long paragraphs trying to talk a model into having taste didn't get me far. A script that exits non-zero when a rule breaks did, and after that my prompts got short: the test failed, here's the output, fix it.
And the speed is real. Claude refactored something like thirty components in one go when we
renamed the design tokens, and Gemini tracked down why Mobile Safari needed an empty
touchstart listener in +layout.svelte before it would register a click.
Each of those took minutes. The blog's back-link, meanwhile, stayed blue for weeks.