How we know

The tests↑ Why · Someone checks every claim before trusting it Each claim names the test that shows it, and whether it passed.

Every test runs in Docker, one command for each kind

The same on a laptop and in CI, with nothing installed but Docker and Node.

End-to-end tests use real browsers and the site as it ships

Several tabs at once, loading the same files that ship, so a test of two people editing is two tabs editing.

Each test is shown to fail when its promise is broken

A test that can't fail proves nothing.

This site↑ Why · Someone checks every claim before trusting it They answer 'does it really do this?' from the site alone.

The site is the spec: every promise lives here, beside the code that keeps it, so the site can't claim more than the project does.

Four pages, named what readers already call them

A reader holds about four things at once. Home is the hook, Why is the reason for the rest, Features is what it promises, How it works is how.

Every promise is in a part of Features

What shows a promise is worked out, never typed

Each promise opens onto the tests that name it, whether each passed and at which commit, and the code that keeps it.

Every promise links up to its story and across to how it's built

A promise with no story has no reason to exist, and one with no design behind it can't be judged. A line joins each part of Features to each story it serves, drawn at both ends, so either page goes one step to the other.

The comparison is fair and dated

A stale or slanted claim costs more trust than none.

Every page reads with scripts off, and loads nothing from anyone else

Its only scripts are its own live demos.

Every page reads on a phone, light or dark

Most first visits are a link on a phone.

Every page is also markdown

For agents, and anyone who wants the source.

#The tests↑ Why · Someone checks every claim before trusting it Each claim names the test that shows it, and whether it passed.

#Every test runs in Docker, one command for each kindproof

#End-to-end tests use real browsers and the site as it shipsproof

#Each test is shown to fail when its promise is brokenproof

#This site↑ Why · Someone checks every claim before trusting it They answer 'does it really do this?' from the site alone.

#Four pages, named what readers already call themproof

#Every promise is in a part of Featuresproof

#What shows a promise is worked out, never typedproof

#Every promise links up to its story and across to how it's builtproof

#The comparison is fair and datedproof

#Every page reads with scripts off, and loads nothing from anyone elseproof

#Every page reads on a phone, light or darkproof

#Every page is also markdownproof