Skip to content
C.W.K.
Stream
Lesson 03 of 04 · published

This Is Just How Books Are Made

~11 min · publishing, process, history, craft

Level 0Wet Clay
0 XP0/36 lessons0/12 achievements
0/100 XP to next level100 XP to go0% complete

Nothing Here Is New

Read the previous two lessons back and they can sound like an elaborate ritual invented to manage a machine. They are not. A real author and a real publisher have run exactly this pipeline for as long as publishing has existed, often several readers deep, and every stage of it exists for the reason this workshop rediscovered by measurement.

The author writes. A developmental editor reads for whether the thing holds together — and is deliberately not the author, because the author cannot see the gap between what is on the page and what is in their head. A copy editor reads for language, sentence by sentence, and is given the manuscript rather than the author's intentions. A proofreader reads the final proof without the manuscript it came from, precisely so that they cannot unconsciously repair a defect by remembering what was meant. Beta readers read cold and report where they got confused, not why.

What the Industry Version Tells You That the Software Version Forgets

Three things, and each maps onto a decision this workshop had to make the hard way.

The roles are separated by what they may see, not by seniority. A proofreader is not a junior copy editor. They are a reader with strictly less information, on purpose, because the reduction is the instrument.

Nobody treats it as an insult. An author who objected that a copy editor was implying they could not write would be misunderstanding the craft. The boundary is about position, not competence, and publishing worked that out centuries before anyone had to explain it to an engineering team.

The passes are named, ordered, and priced. They are not a checklist bolted onto the end of a schedule; they are the schedule. A house that skips the copy edit does not save a pass, it ships a different quality of book, and everyone in the building knows which houses do that.

The Only Real Difference

In publishing, the boundary is maintained by the fact that people are separate. Here it has to be built, because the same system can be asked to do all the roles and will say yes to each of them. That is not a reason to invent a new process. It is a reason to implement the old one with a harness where the industry has staffing.

Before designing a process for machine collaborators, check whether a human profession already solved it. If it did, the version that survived a century of commercial pressure is better than the one you will design this week — and its role boundaries will already be drawn along the lines that actually matter, which is who is allowed to know what.

Code

The same pipeline, two industries·text
PUBLISHING                        THIS WORKSHOP
--------------------------------  ------------------------------
author writes the manuscript      authoring stage, full context

developmental editor              plan gate + the structural
  reads for structure             half of cold review
  is NOT the author

copy editor                       the cold repair pass
  gets the manuscript             gets the target language only
  not the author's intentions     not the source, not the brief

proofreader                       the sampling reader
  reads the final proof           gets passages as TEXT
  WITHOUT the manuscript          never learns a path exists

beta readers                      the review round
  report where they stumbled      rejection log, mandatory
  cold, no briefing               fresh instance, cold artifact

# Left column: enforced by employing different people.
# Right column: enforced by a harness, because the same system
# would cheerfully accept all five roles.

External links

Exercise

Take one process you are designing for machine collaborators and find the human profession that has been doing it for at least fifty years. Write down that profession's role list and what each role is deliberately not shown. Then compare it to your design and note every role you have collapsed into one. Each collapse is a decision you made without noticing, and at least one of them is why something keeps slipping through.
Hint
Look for professions where being wrong is expensive and slow to discover — publishing, accounting, aviation, surveying, clinical trials. Their process is old because their errors were costly, which is precisely the condition that produces separations worth copying.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.