Monospace everywhere, and the one exception

Every label, button, input, table header, badge and numeral on this site sets in JetBrains Mono. It is not a code face used decoratively — it is the default UI face, declared once as --default-font-family and inherited by everything.

Paragraphs like this one do not. This is Archivo, a proportional grotesk, and it is the only place on the site where a second typeface is permitted.

That is a rule the direction I picked explicitly forbids. Industrial specifies monospace for body text as well as display. So this is a deviation: a rule the anchor states and this project deliberately breaks. There is exactly one, and the interesting part is not the exception itself but the machinery that stops it spreading.

Why break it

The trade was readability against fidelity, and readability won on evidence rather than taste.

Monospace paragraphs read measurably slower, and they take roughly 15% more horizontal space at the same nominal size. On a site whose whole purpose is to carry writing, that is a real cost paid on every paragraph. Fidelity to the direction is worth a lot, but it is not worth making the writing worse.

What made this decidable rather than arguable was drawing the line somewhere specific. The project has two words for the two kinds of text:

• Chrome — everything that is not long-form prose. Labels, buttons, inputs, headings, table data, numerals, code. Chrome always sets in the monospace face.
• Prose — long-form running text, and the only context permitted the proportional face.

With those two words in place, the question stops being "does Archivo look better here" and becomes "is this chrome or is it prose". That has an answer.

Where it gets interesting

The line runs through places I did not anticipate.

A standfirst is chrome

The two-sentence summary under a title looks like prose. It reads like prose. It is chrome, and it sets in mono.

The reasoning: it is short enough that the proportional face would be spending the deviation for nothing. The readability argument is about paragraphs accumulating — thirty of them, one after another. Two sentences do not accumulate. And a title in mono followed immediately by a summary in Archivo puts a typeface change two lines apart for no gain.

Headings inside prose are chrome

An h2 in the middle of an article body sets in mono, even though everything around it is Archivo.

This one felt wrong until I wrote out why. A heading is a structural label, not running text — it is the same object as a section title or a table header, both of which are unambiguously chrome. If a heading inside prose switched to Archivo, the boundary would stop being "running text" and start being "anything inside an article", which is exactly the widening the deviation is scoped to prevent.

Code was free

Tailwind's preflight already sets code, kbd, samp and pre to the mono family:

code,
kbd,
samp,
pre {
  font-family: --theme(--default-mono-font-family, …);
}

That is an author-origin rule on the element, so it beats an inherited family. Fenced blocks and inline code set in JetBrains Mono inside a prose container without anything asking them to. It looks like an omission in the element map, which is why there is a comment there saying it is not.

The one thing that did need asking: an ordered list's counter. A numeral is chrome, and without an explicit marker:font-mono the count sets in the proportional face that the list body legitimately inherits.

The scope is the whole point

The face is reachable through exactly one component and one utility. Archivo is not available to buttons, labels, inputs, table headers or badges — not because it would be ugly there, but because an exception with no boundary is not an exception.

There is one more consequence I nearly missed. Asking a font loader for a variable axis does not get you an italic. The proportional face needed style: ["normal", "italic"] requested explicitly, because otherwise emphasis in an article resolves to a synthesised oblique — a sheared roman, which at prose size is visibly wrong in a way that is hard to name if you are not looking for it. The monospace face never needed it, because chrome does not have emphasis in it.

Which is the summary, really. The deviation is not "we use two typefaces". It is a single sentence with a written boundary, a named entry point, and a test that fails if the boundary moves. If a button label ever starts setting in Archivo, the anchor has stopped holding — and that is a bug, not an evolution.