10 September 2026

A page for an address that is not a page

There was none: an unmatched address got the host’s own error page, with nothing to click. There is now a 404 carrying the site’s chrome and three ways out, written once and re-rendered in the reader’s language.

Two soft 404s came out with it. A link to an article that does not exist answered 200, which a crawler indexes as a real page; and /de/history was a plain 404 in four languages, so the screen worked until somebody reloaded it.

If you are reading this, you probably arrived at a 404 and wondered what the site does with one.

What it offers

The number, a line saying the address is not a page, and three ways out named rather than hinted at: convert a file, read the documentation, browse the blog. The converter comes first and is the one filled button, because somebody who typed the domain by hand was looking for it. A last line says that if a link on this site sent you here, that is a bug rather than a typo — which tells you whether to retype the address or to report it.

It carries the site's own header and footer and is re-rendered in your language, so landing on it does not feel like being thrown out of the site.

Why a typo used to render the converter

The router asks an address one question: which screen does it want. Its last branch was the converter, the default screen — so every path it did not recognise rendered the Markdown screen. One wrong letter in a URL looked like a working front page with a stranger's path in the address bar, and nothing anywhere said the address was wrong. That last branch is now "not found"; the converter is returned only for an address that is genuinely one of the ten conversions' own, or a shared document.

The host answers most unknown addresses before the app's code runs at all, but it answers them by serving the 404 file — and then this is what decides what you see.

Two soft 404s went with it

A link to an article that does not exist used to answer 200, which a crawler files as a real page and then keeps. And /de/history was a plain 404 in four of the five languages, so the screen worked until somebody reloaded it.

What it does not do

Guess what you meant, or redirect. A 404 that quietly sends you somewhere else is how a broken link stays broken: nobody finds out it is wrong, including whoever wrote it.