Decisions
Decisions behind this site
Short decision records: the choice, the reason and the price I pay for it.
- 01Next.js App Router with Server Components first
Why: Fast first load and crawlable HTML for SEO, with client code only where interaction needs it.
Trade-off: More care with the boundary between server and client components.
- 02CSS Modules and design tokens instead of Tailwind
Why: Scoped styles, tokens as CSS variables, no extra dependency, and one place to restyle the whole site (see the design system page).
Trade-off: More CSS to write by hand. Tailwind is common in jobs, and I can pick it up with AI support.
- 03GSAP and raw WebGL instead of a 3D library
Why: The 3D effects here (water, orb, companion) are small shaders, so the bundle stays light and Lighthouse stays high.
Trade-off: More shader code to maintain by hand.
- 04next-intl with logical CSS properties
Why: Four languages including right-to-left Arabic from one codebase, with hreflang and correct text direction.
Trade-off: Every layout must be checked in both directions.
- 05Playwright end-to-end tests on the production build
Why: The real risks of this site are layout overflow, languages, SEO tags and interaction, and only a real browser sees them.
Trade-off: Slower than unit tests. Unit tests with Vitest belong to business logic, as in ServiceOS.
- 06Numbers stored with a date, not fetched live
Why: Lighthouse scores, test results and the contribution graph are saved at build time. No third-party request from your browser, and no number without a date.
Trade-off: They can be out of date until I measure again.
- 07No cookies and no analytics
Why: Simple and GDPR-friendly for a personal site.
Trade-off: No usage data until the site is online and Search Console is connected.