Rules
The benchmark I build under, and the same bar for any fair comparison.
These are the constraints I hold NothingFaster.com to. Same tool, same profile, no cherry-picking. If you try your own build under the same terms, I'd like to see it. There is no form, just build and share a URL if you want. The Rank board is one lab snapshot run under this contract.
- Public homepageA real URL anyone can open. No login walls, no localhost, no demo that only runs on your machine.
- Same lab rulerClaims use the same tool and profile as Rank: PageSpeed Insights mobile lab, sorted by LCP. Field data (CrUX) and “this visit” RUM are useful stories; they do not replace the lab floor.
- Worldwide floorSpeed has to hold up from more than one region. If one place is fast and another is slow, the slower reading is the one that matters, not the lucky PoP.
- No best-of-N cherry-pickingPublish a dated run. Do not keep the luckiest single lab hit. Prefer a pass others can re-open in PageSpeed Insights with the same URL and profile.
- The HTML is the productThe document you claim is the page people land on, not an empty shell that bounces into a heavier app, and not a zero-byte trick that is not a site.
- Weight stays on the boardTransfer size and request count matter. A blank “gotcha” page is not a useful comparison.
- Disclose how it is servedEnough hosting or edge detail that someone can judge fairness: for example Workers or Pages at the edge versus a single-origin toy.
- Reproducible by strangersAnyone should be able to paste your URL into PageSpeed Insights and land in the same ballpark. If only you can reproduce it, others cannot verify the result.
If these rules feel fair to you too, try them on your own site and share what you get. I'd genuinely like to see more pages built this way.