The office that ate itself#
Sixty minutes on what a copy is, before we spend twelve weeks arguing about what happened to one.
Chester Carlson’s problem#
Carlson demonstrated electrophotography in 1938. More than twenty companies declined to license it over the next decade.
The recorded objection was not that it failed. It was that carbon paper already existed, and nobody was asking for anything else.
From David Owen, Copies in Seconds (2004), chapter 1.
Copy room test card 0S1I-B85l#
Four lines, built from characters that look like one another:
0/O · 1/l/I · 5/S · 8/B · 2/Z · 6/G · 9/g
A legibility card concentrates the failure so it can be counted.
One measurement, position by position#
card The clerk in Room 108 filed 1,051 slips before noon.
copy The clerk in Room 1O8 filed 1,05l slips before noon.Two positions differ: a 0 read as O, a 1 read as l.
Every transcription is the same length as the card — no alignment judgement, just a count.
The rule matters more than the number#
A rule you can hand to someone else, who then gets your number, is a measurement.
A number you produced by looking carefully is an impression.
Wednesday’s counting rule is the first thing this course asks you to trust someone else to reproduce.
Four copies, one control#
| Copy | Channel | Operations |
|---|---|---|
| Digital control | digital | 8 |
| Photocopy | photocopy | 1 |
| Photocopy | photocopy | 4 |
| Photocopy | photocopy | 8 |
The control runs the same eight operations as the longest photocopy chain. Not once — eight, checksummed at every step.
What the control buys#
Eight photocopy operations are worse than one. Eight digital operations are not worse at all.
A digital copy made once, set beside a photocopy chain of eight, would prove nothing — the obvious objection is that it was never tested.
Matching the operation count is what makes this a control instead of a gesture.
The finding, stated precisely#
This copying process damages what it copies.
Copying as such does not.
That distinction is the whole reason week 1 opens with a control rather than a dramatic chain of ruined pages.
Two readers, one copy, one disagreement#
After a single photocopy operation, reader R1 records an error at “Room 108”. Reader R2 records none.
Both followed the same transcription procedure.
Three claims that are not the same claim#
- The counting algorithm is deterministic — same transcription, same card, same number, every time.
- Your own repeatability is a separate check: count one transcription twice.
- Readers still differ, because resolving a damaged mark is an interpretation, not a lookup.
Legibility is a relation, not a property#
Reader R1 scores below R2 after four operations and above R2 after eight.
The readers cannot even be ranked consistently.
An honest table reports that spread. Averaging it away would have destroyed the one thing worth knowing.
Before Wednesday#
Download the week 1 pack: card.txt and
counting-rule.txt. Ten minutes.
Workshop: The office that ate itself — you apply the rule to all four copies and both readers, and check the control yourself.
Text walkthrough#
For a reader who cannot use the deck view: Chester Carlson’s photocopier took a decade to be adopted because the problem it solved wasn’t one anyone had asked about. This course’s copy room test card 0S1I-B85l is built from confusable character pairs so a legibility failure can be counted exactly — compare a transcription to the card position by position, since both are the same length. A digital control, copied byte for byte eight times and checksummed at each step, comes back identical, while a photocopy chain of the same length degrades — showing the damage belongs to this photocopier, not to copying in general. Separately, two readers transcribing the same single photocopy already disagree at one position, which is not a flaw in the counting rule (deterministic, by construction) but evidence that legibility is a relation between a copy and a reader, not a fixed property of the copy.