Week 11 Seminar: The live case

Generative models and knowledge work, read through weeks 1–10. The framework is the argument; the course does not supply the verdict.

3 May 2027

Open the slides

This week applies the course’s method to a case still in progress, and it stops short of a conclusion on purpose. The claim on this page, and everything below it, presents a case and a counter-case and leaves them there. Nothing on this page is the course’s verdict, because the course does not have one to give — the point of the last ten weeks was to build a method sharp enough to use here, not to arrive already knowing what it will find.

The case, and the counter-case#

The case for a loss in progress. Generative models can now produce competent first drafts of writing, code, and analysis that a person used to have to produce themselves in order to develop the underlying judgement. Weeks 6 and 7 both describe how a capability can erode without anyone deciding to erode it: automate the output well enough, and the practice that used to build the underlying skill stops happening as often. If that pattern holds here, the tacit knowledge built by doing the work yourself is exactly the kind of thing week 2 argues doesn’t survive in the output alone — and the output is increasingly all that gets produced.

The case against. Every automation this course has studied changed what skill was needed without necessarily reducing the total skill in the system — Zuboff’s argument is precisely that smart machines demand new capabilities from the people who work with them, even as they remove old ones. Editing, prompting, and judging model output at scale may be a real capability being built right now, not a gap opening up. On this reading, week 1’s distinction cuts the other way: the record (what the model produces) and the capability (what the person using it can actually do) are just as separable in the optimistic direction as in the pessimistic one.

What we do in the seminar#

Split the room. Argue each side for the first half using only material from weeks 1–10, cited by week number. Then argue against your own side for the second half. No side is graded as the “correct” one, because this page doesn’t have one.

One reading#

Zuboff (1988) — written about an earlier wave of workplace automation, and the closest thing this syllabus has to a text that already argues both sides of this week’s question.

A question left open#

This whole page is the open question. Decide, if you want to, on your own time — but notice which of the last ten weeks’ cases you found yourself reaching for first, and ask what that says about which side you were already inclined to argue.