How I Built a Fashion Website in 48 Hours With Cursor AI and Scored 98 on Lighthouse

I had 48 hours. The site was The Laila Majnu—fashion ecommerce. I used Cursor and Next.js. When I ran Lighthouse, the score was 98.

Not "we aimed for 90." Ninety-eight. Real URL. Real run.

Most "how to get a fast site" posts are theory. This one is a single case: one project, one timeframe, one number.

What Actually Happened

Static export. No heavy runtime. Images sized and formatted so LCP stayed low. I didn't chase every metric—I fixed the things that move the needle. Cursor wrote a lot of the code. I directed it: structure, components, what to avoid.

The friction moment: around hour 36 something broke. A dynamic route I'd forgotten about was pulling in logic that didn't belong in a static build. Lighthouse dropped. I had to strip that path and simplify the data flow. Then the score came back.

So it wasn't magic. It was constraints. Static. Fewer moving parts. And one late fix that I only caught because I ran Lighthouse before calling it done.

Why I'm Writing This

People search "cursor ai website 48 hours" and "lighthouse score" and get generic advice. They don't get "I did it, here's the site, here's the number." I'm putting that on the record.

A logo or brand system isn't the only thing that benefits from clear rules. In how I wrote Cursor rules for the same project, the brand voice lived in a skill file before any UI. Speed and score came from that plus a strict static setup.

If you're building a site and care about performance, run Lighthouse early. Fix the one thing that's dragging the score. Repeat. For design and build work I do exactly that—design and implementation in one flow so performance isn't an afterthought. How I run projects is on my process page.

The Laila Majnu isn't in the portfolio yet. When it is, the case study will live under work. Until then: the URL and the 98 are the proof.