I build in Next.js, and I have to admit something uncomfortable. On the public data, Next.js sites pass Core Web Vitals less often than Squarespace sites do.

That is from HTTP Archive, which measures real sites in the wild rather than a demo someone tuned for a screenshot.

It took me a while to accept why. Squarespace has a ceiling and a floor. You cannot make it much faster than it is, and you cannot easily wreck it either. A coded stack has no floor. Every third party script, every unoptimised hero image, every font nobody remembered to subset is yours to add, and most people add all of them.

So the honest version of the pitch is not that code is faster. It is that code can be faster, and only if someone is actually watching the numbers after launch.

If you are choosing a platform, the useful question is not which one wins a benchmark. It is who owns the performance in month six when nobody is looking.

Has anyone here measured their own builds against the population data rather than against a Lighthouse run on their own laptop?