KILN

On the Casting of Letters

Foundry note №3 · translated from the 1764 manual

Metal remembers the mould long after the mould is gone. A punch cut in steel, driven into copper, filled with an alloy of lead and tin and a whisper of antimony — and what emerges is a letter that will print ten thousand times before its shoulders round off. The foundryman's art was never really about metal. It was about anticipating wear.

Consider the counter — the enclosed white of an a, a g, an e. Too tight and the ink floods it at speed, and the page turns to mud by the third impression. Too open and the letter loses its weight, goes thin and papery beside its neighbours. The correct answer is not a measurement; it is a judgement about how much abuse the page will take.

“Cut for the worst press you will ever meet, not the best.”

Screen type inherited this whole inheritance and then quietly discarded it. The enemy is no longer ink spread but the pixel grid — a letter must survive being resolved onto a lattice that does not care about its intentions. Hinting was our antimony. Then the lattice got fine enough that the problem mostly dissolved.

What did not dissolve is the canvas problem. For thirty years the drawing surface of the web has been able to render a rectangle, a bezier, a bitmap — and a single run of unstyled text, badly. No line breaking worth the name. No bidirectional text. No ligature it did not have to be begged for. Anyone who wanted a chart legend, a game menu, a caption in a shader was told to lay it out by hand, glyph by glyph, and to forget about the screen reader entirely.

Which is the joke of this specimen. The paragraph you are reading is not drawn. It is laid out, by the same engine that lays out the rest of the page, with the same line breaker and the same shaper — and then handed to the GPU as a texture, so that a piece of imaginary glass can bend it. The refraction is a performance. The text is real.