One rule

Four words do all the work. Real-user: measured on people who visited, not on our laptop. LCP: how long until the biggest thing on the screen showed up. 75th percentile: three quarters of visits were at least this fast. Phone: because that is how most of the web is read, and because desktop makes everyone look good.

Under 1.5 seconds and a site is in the Instant Club. At 1.54 seconds it is not, and no row on this site will pretend otherwise.

It is the closest single number to "the page looks ready". Time to First Byte flatters sites that stream nothing useful; a Lighthouse score is an opinion with a colour on it. LCP is when the thing you came for actually appeared.

One metric, so the verdict cannot be argued with. Everything else on a site page — FCP, TTFB, INP, CLS, page weight — is context for the number, not a second opinion about it.

Three quarters of real visits were at least this fast. An average is dragged around by a handful of terrible loads; a median throws away the slower half of your audience entirely. The p75 is the promise you can actually make: most people got this or better.

If your p75 is 1.4 s, plenty of people still waited longer than that. Which is rather the point.

Real-user data comes from the Chrome UX Report, Google's public dataset of what Chrome users actually experienced, at the 75th percentile on phones. When a site has too little traffic for that, we run a PageSpeed Insights lab test and label it as a lab result.

A lab result is one load on one machine on one connection. It is a stand-in for the real thing, and every row, card and page that shows one says so in as many words.

Every tracked site is re-queried at regular intervals. A site that drops under 1.5 s moves into the Instant Club automatically; one that slips is kicked out, with a reason.

Nothing is moved by hand and there is nobody to lobby. For a kick we also run a fresh lab test and diff it against the last one, so the feed can say something more useful than "got slower" — page weight up 2.1×, third-party requests up 3×, TTFB up 800 ms. That reason is our best reading of the lab run, not a diagnosis of your stack.

Every site we have timed that is currently under 1.5 s, fastest first. Membership is per form factor, so the phone and desktop boards are different lists; the club proper is the phone one, because that is the rule. Filter it by category, search it, or sort it by weight to see which fast sites are fast in spite of themselves.

Everything else we have timed, closest to the line first, each row saying exactly what it has left to shave. It is a queue, not a wall of shame — every site in the club started here, and the kicked-out feed at the bottom of that page is where they go back to.

Check your site on the homepage, then press List it in the club. Pick a category from the list, and add a GitHub or X handle if you want your name on it. That is the whole form.

The title, description and icon come from your own site, read once at listing time — we do not write copy about you and there is nothing to log into. Handles are typed into a form by whoever ran the listing and we do not verify them, which every listed page says out loud.

Slow sites can list too. Listing makes you a contender, which is where everyone in the club started.

There are 29 of them, fixed. A category only appears in the filters once it has at least one site in it, so the chip rows on the homepage, the club and the contenders are all different lengths and all shorter than this list.

Every site page carries an SVG badge with copy-paste HTML and Markdown. It reads the live number each time it is loaded, so it turns amber or red on its own the week you slip. A badge that cannot embarrass you is just a sticker.

No signup, no accounts, no cookies, no analytics script, no third-party request on any page. Nothing about you is stored, because nothing about you is collected.

The only thing kept in your browser is which theme you picked, and that never leaves it. The page-size line in the footer is our own score, measured the same way as yours.