The argument in brief· 3–5 min read· full book 19 pp
How AI slop transfers its cost to everyone else.
AI slop is a failure of judgment, not a method of generation. The generated sentence is not the offense. The unexamined one is — and the cost of skipping that judgment does not disappear. It lands on whoever reads, reviews, or merges the result.
LinkedIn now has a feedback option called “Seems like AI slop.” Choose it on a post and the platform treats that as a signal the post felt low-value — not a takedown, not a policy finding.
The name invites a shortcut: hear “AI slop,” infer “text an AI wrote.” That is too broad to be useful. An AI can help a person research, outline, translate, edit, or stress-test an argument, and none of that makes the result slop. LinkedIn’s own definition is better than its button’s name:
Not how the content was created, but whether it adds value and reflects a person’s perspective, experience, or expertise.
Imagine three people publishing the same article about AI agents: one pastes a model’s 1,500 words straight into a post; one supplies a thesis and rewrites the draft; one decides which sources matter, checks the claims, and stands behind the conclusion. The visible artifact can look almost identical across all three. The process behind it is not. Authorship is the work of choosing — the question, the evidence, the omissions, the sentence you’re willing to defend — not the count of keystrokes.
The complaint about slop reads as aesthetic — too many bland posts, too much sameness — but the deeper cost is attention. Every plausible-looking, low-value artifact asks a reader to run a small investigation: is this true, is it copied, is there a specific person behind it? Each check is cheap alone. At scale it makes the whole information environment tiring, and readers respond by skimming harder and trusting less — a tax paid by every careful writer sharing the same channel.
The same asymmetry shows up off the feed, inside a codebase: generating a pull request is cheap; understanding it, maintaining it, and recovering its rationale six months later is not. The technical debt isn’t “AI wrote this function.” It is the gap between what a repository contains and what its humans can actually account for.
There is no reliable detector for care. So the standard has to be behavioral, not forensic.
The book proposes five questions in place of a detector: what is the point, what is new here, what was checked, what was removed, and who owns the result. None of them cares how many tokens a machine produced. All five are answerable by a person who did the work of deciding the piece was worth publishing — and unanswerable by one who didn’t.
This is an argued essay, not a claim that the boundary of “AI slop” can be measured precisely. Its factual claims — about LinkedIn’s own guidance, the research literature on slop, and the reporting on AI-generated content farms — are footnoted to the source that supports each one, so a reader can check them directly rather than take the essay’s word for it.
It does not carry a verification ledger the way some of the other titles on this shelf do. Where the research is unsettled, the book says so, rather than rounding a contested claim up to a fact.
Eight chapters: what the word actually names, why detecting AI-written text is the wrong question, what changes when slop reaches a repository instead of a feed, the five-question responsibility test in full, and what to do about it on Monday. Read it → or listen to it →.