We made a free font called Land Run Caption. Somebody at Bain News Service wrote the captions on their glass negatives in pen, inside the picture, and the Library of Congress still holds the plates. We pulled the master scans, matched the same hand across 13 negatives, and traced 45 letters and marks into something you can install. That page tells you where every mark came from.
This one tells you what did not make it.
That is the more useful half. Anybody can publish the parts that worked. What decides whether a thing like this is honest is the pile beside the desk.
The plate that should have worked
There is a Rock Island railroad negative from 1893. Right decade, right railroad, right subject, the actual machinery of a land run.
The caption on it is typewritten.
So it went in the pile. Not because it was ugly. Because a font that took a typed caption on the grounds that the date was right would be a different object than the one we said we were making. The whole claim is one pen. A typewriter is not a pen, and 1893 is not a good enough reason to pretend otherwise.
That plate is still the one that stings.
Two more hands, and two ruined ampersands
A dust storm at Perry. A cyclone at Ponca. Both Oklahoma, both the right era, both captioned by somebody else entirely. You can see it in about four seconds once you have stared at the real hand long enough. Different pressure, different slant, different way of finishing a stroke. Handwriting is a fingerprint and it does not negotiate.
Then two plates we wanted for one mark. We needed an ampersand, and ampersands are rare in captions because nobody writes an ampersand for fun. It shows up only when a caption names two people at once.
The first candidate sat too close to the edge of the plate and the frame ate half of it. The second sat on emulsion too far gone to read. Both went in the pile.
The ampersand we shipped came from a caption reading "Mr. & Mrs. Borden Harriman." Third try.
Four letters we traced, then threw back
Rejecting a plate is easy. Throwing out work you already finished is the expensive part.
The B. Version one shipped with a B lifted out of the word "BE." In the ink, the B and the E had run together into one shape, and at small sizes nobody noticed. Blown up on a headline it was obviously wrong. We pulled it and went back for a clean B out of the word "BEING" on the same plate.
The U, the V and the W. All three came off a plate where somebody had gone over the caption in white paint, the way archives used to touch up a negative so it would print legibly. Painted strokes are fat. They blob at the joins and they do not match a pen. The V topped out well short of every capital beside it and the W was the widest mark in the whole font, which is exactly the kind of thing that makes a face feel homemade in the bad way.
We went back to the thin pen on a different plate and pulled a U out of "STRUCK," a V out of "SEVERAL," and a W out of "WHO."
Measured after the retrace, off the live font: the V now sits at 81 percent of capital height, in family with the X at 78 and the A at 72, up from 63. The W is no longer the widest mark in the font.
The plate that lost three letters did not get thrown out. It kept the G and both parentheses. It got demoted, not rejected, which is the more common outcome and the less dramatic one.
The one we assembled
The pen never dotted its exclamation point.
We have the plate. The archive's own transcription reads "to-day!" so the mark is there, and under magnification it is a bare stem with nothing beneath it.
