Open the CSS of this page and you will not find a stack of font files named Light, Regular, Medium, Semibold, Bold. There is one file for the roman of our display face, and it contains every weight from a hairline 100 to a thunderous 900, every width from a pinched 62 percent to an expansive 125 — and, crucially, every point between those points. Weight 437.5 exists. It has always existed, in the design space the type designer drew; the variable font simply lets you address it.
wght),
width (wdth), optical size (opsz), slant, italics, and any
custom axis a designer registers.
The idea is older than the format. Adobe’s Multiple Master fonts of the early 1990s promised the same continuous space and failed on tooling and appetite. Knuth’s METAFONT went further — a typeface as a program, its strokes defined by equations with dozens of parameters — and found almost no adopters outside academia. The lesson the industry drew was that designers did not want parameters. The lesson of 2016 is that they only wanted good ones.
The designer still decides
A common misreading of variable fonts treats them as infinite choice, a typeface stretched like dough until it tears. In truth the axes are the most opinionated part of the design. A type designer chooses which axes exist, where they begin and end, and — through the placement of intermediate masters — how the letterforms behave along the way. The difference between a tasteful heavy weight and a bloated one is drawn, by a person, into the font. What the format changes is the resolution of the conversation: instead of seven fixed instances, the designer hands you a continuous, curated range.
The most humane of the registered axes is optical size. Small text needs sturdy
strokes, open counters, and generous spacing; display sizes can afford hairlines and
tight fits. Metal type did this automatically — a 6-point punch was drawn differently
from a 72-point one — and phototype destroyed the habit, scaling one drawing to every
size. The opsz axis restores it. The serif you are reading now applies it:
at body size it takes the robust station; in a headline it would move, automatically or
by hand, toward the crisp end of the axis.
Typography becomes a runtime
Once the axes are addressable in code, they can be driven by anything: viewport width, scroll position, time of day, a reader’s pointer. This journal’s covers and headlines are built on exactly that premise, and it is worth stating the ethic we hold them to — the axes are expressive, not decorative. Motion may carry a reader into the text; it must never stand between the reader and the words. Every kinetic setting in this issue settles, within a second or two, into a static state you could print.
There is a quiet power shift embedded here. Choosing a weight used to mean choosing a file; now it means choosing a number, and numbers invite interfaces. Some of those interfaces will be gimmicks — fonts that throb with your cursor for no reason. The interesting ones treat the axes as typography’s newly exposed nervous system: grade adjustments for dark mode, widths that respond to column measure, optical size snapping to rendered point size. The typeface stopped being a shipment and became an instrument.
GRAD axis — adjusting stroke contrast without changing set
width — exists precisely so text can be darkened for light-on-dark settings without
reflowing a single line. It is the least theatrical axis and possibly the most useful.
What has not changed is the unit of judgment. A page still succeeds or fails in the reader’s eye, line by line. Variable fonts give the typographer finer control over exactly the things typography is made of — colour of the text block, rhythm of the line, hierarchy of the page. Used with restraint, that is not a special effect. It is simply the craft, continued by other means.