A benchmark kept by one person
Forever & Always is a small static site with seven tests, no accounts and no analytics. This page says who runs it, what it is for, and what it will never do with your scores.
The name
1314 is read as 一生一世 — one life, one lifetime. It turns up in phone numbers, wedding dates and the occasional password. The parent domain is 13146780.xyz, so when this site needed a name the four digits were already sitting there; Forever & Always is simply the English for the same four characters.
It is a slightly odd name for a benchmark, which is part of why it fits. Measuring yourself is a way of paying attention over a long stretch of time, and attention is what a vow is made of. The tests are short; the habit is long.
What the seven tests are
Reaction time, sequence memory, number memory, verbal memory, visual memory, the chimp test and an aim trainer. Each one isolates a single, narrow ability: how fast a visual signal becomes a motor response; how long a spatial sequence you can hold; how many digits survive in working memory; whether a word feels old or new; how many lit tiles you can reconstruct; how many numerals you can track once they hide; how quickly you can land a pointer on a target.
The formats follow the public versions published by Human Benchmark, because those formats have years of data behind them and people already know how to read them. Everything else — the wording, the code, the scoring detail, the palette, the mark — was made for this site.
Every test takes under two minutes, runs entirely in the page, and writes its result to this browser. Nothing is submitted anywhere.
What was deliberately left out
No accounts, so there is no password to lose and no profile to leak. No leaderboard, so nobody can farm a score to sit above strangers. No analytics script, no session recorder, no heatmap tool. No third-party font, no icon CDN, no embedded video.
The absence is the design. A benchmark is only useful if the number means the same thing next week; the fewer moving parts between you and the measurement, the more likely that is true.
How it is built
A static site: HTML pages, five stylesheets and a handful of ES modules. No framework, no runtime build step, no server-side rendering. Fonts come from your own system. The mark and the icons were drawn for this site and then rasterised into the sizes browsers ask for.
Records live in localStorage under four keys, all listed on the privacy page. Clear site data and the site forgets you completely, which is the intended behaviour rather than a limitation.
Who runs it
One person, not a company, not a lab and not a startup. There is no editorial team, no investor and no institutional affiliation, and the site does not claim one. Pages are written by hand in two languages, and corrections are welcome by mail.
That also means the site is small on purpose: seven tests, thirteen pages in English and the same thirteen in Chinese. Nothing here is scaled, automated or syndicated.
Money, and the possibility of advertising
The site is free to use and currently costs only a domain and a small server. If traffic ever grows enough that bandwidth stops being trivial, display advertising may be added to cover the cost — this section will be updated before that happens, and the privacy policy will describe exactly what an ad partner would set.
Two commitments if advertising does arrive: ads will sit outside the test screens, never interrupting a run and never overlapping the instrument; and no advertising script will read your scores, because the scores are only ever read in the page that wrote them and never sent anywhere.
Where the numbers come from
Reference points come from Human Benchmark's published statistics and from the papers listed on the science page: Donders (1868) on reaction chains, Miller (1956) on the span of immediate memory, Fitts (1954) on target acquisition, Baddeley and Hitch (1974) on working memory, and Inoue and Matsuzawa (2007) on numeral memory in chimpanzees.
Where a figure is uncertain — verbal memory, for instance — the site gives a range instead of inventing a precise number. Where a figure depends on hardware, the page says so. A test that cannot be honest about its error is not worth running.
Contact
Corrections, broken links, accessibility problems, data questions, advertising enquiries and permission requests all go to the same address: admin@13146780.xyz.
Mail is read by one person, so replies can take a few days. Anything about a specific score should include the test name, the device and the browser, because that is usually where the answer is.