An independent studio needs a site that does two jobs at once: show the work, and be an example of it. If it is slow, inaccessible or invisible to search engines, nothing written on it counts for much.

So I built mine, and then I audited it the way I would audit a client's. Prototyped in Figma first, then Next.js, React, TypeScript, Tailwind and GSAP on Vercel.

Four of these rows open a case study rather than an external link. They are rows in a data file, not hand-built pages, so a new project is a new entry and the listing, the sitemap and the share previews follow on their own. The fifth has no case study written yet, so its row stays quiet instead of pointing at a page that does not exist.

Measuring only counts if you report what it turns up.

The share preview was invisible. Case links carried a .webp image, which LinkedIn does not render, so every link shared there arrived with no picture. Now a 1200x630 JPEG per case, with width and height declared.

An unfinished case study was reachable from the home page. Search engines ignored it, but a visitor clicking through saw an unwritten page.

An image description described a different image, which is what Google and a screen reader were being told.

Not from memory. Commit counts, measured contrast ratios, the bugs where the first hypothesis was wrong. Reviewing one against its own repo turned up three statements that had drifted from the truth. All three came off the site the same day.

One section is not mine: twelve client reviews averaging 4.97, verified on Contra and written by the people who paid for the work. Everything else on the page I can defend with a measurement. That part I can only show.