Every technological jump forces us to renegotiate which human effort is valuable and which effort was merely expensive.
William Morris was already wrestling with that in the nineteenth century. Industrial production could make more things, faster and cheaper, but Morris was interested in whether greater production actually meant better work or better lives. His point was not simply that handmade things were morally superior. He was asking what work was for.
Marcel Duchamp challenged a different assumption: that the value of creative work comes mainly from physically making the object yourself. His readymades were ordinary manufactured objects that he selected and presented as art. The object already existed, so the interesting part was no longer how it was made, but why he chose it and what he was asking people to see in it.
Technology can reduce the effort required to produce something without reducing the value of the work itself. The important question is what that freed capacity allows people to do instead.
When production gets easier, the human contribution does not necessarily disappear. As I wrote in The Unit of Work Has Changed, it moves to a different layer of the work: choosing what matters, what needs changing, where an intervention can improve the system, what should be automated, and what requires the next decision.
Generative AI complicates this further because the finished artifact can reveal very little about what happened before it appeared. A polished paragraph, image, prototype, or presentation might contain significant human judgment, almost none, or something in between.
And this is where the word slop enters.
We use one word for several different failures, then argue as though we are all complaining about the same thing. We are not.
There is slop as bad craft: work that is generic, clichéd, repetitive, incoherent, or poorly made.
There is slop as shallow thinking: work that looks competent or complete but contains little insight, reasoning, or understanding beneath the surface.
There is slop as dishonesty: passing off generated knowledge, experience, or thinking as your own.
There is slop as excess: producing far more than the situation calls for simply because generation is cheap and easy.
There is slop as pollution: the collective effect of that excess, where useful information becomes harder to find because more of what surrounds it has to be filtered, verified, or ignored.
These describe different kinds of slop. What connects them, and gives the argument its moral weight, is disrespect.
These complaints can overlap, but they are not interchangeable, and none of them require AI.
Low culture is not automatically disrespectful, and bad taste is not a moral failure. An absurd AI image made for your friend’s group chat might be delightful, while a beautifully formatted six-page AI PRD nobody bothered to verify can be slop. The second one might even look better.
The moral dimension appears when the convenience of producing something is bought at the recipient's expense. The producer gets speed; the recipient inherits uncertainty, unnecessary interpretation, or avoidable verification.
Inside organizations, hierarchy can make that imbalance worse. When low-judgment work moves downward through levels of authority, the people receiving it may have little power to reject the premise or send it back. Instead, they are expected to interpret it, repair it, design around it, or turn it into something workable. By the time the consequences appear in the final product, the dysfunction has travelled far enough that the people closest to execution can end up accountable for quality they did not have the authority to define.
Efficiency is valuable, but efficiency alone doesn't guarantee responsible work. Used well, AI can remove tedious work, accelerate execution, and make expertise easier to apply. The difference is whether judgment still sits behind the result.
The better standard is respect.
Respect for the collective work being served, for the person receiving it, for their time and effort, and for the responsibility to decide what is worthy of entering the work at all.
The current argument around “AI slop” is therefore more interesting than a debate over whether computers should be allowed to make things.
A recent Wall Street Journal report on LinkedIn described the platform wrestling with AI-generated content while users increasingly use “AI slop” as a catch-all accusation. The phrase now carries several complaints at once: artificiality, sameness, carelessness, deception, excess, and low quality.
That confusion matters because the presence of AI is not the useful dividing line. The harder question is whether someone exercised enough judgment to understand, shape, verify, and ultimately stand behind the work.
Those are not nostalgic questions about craftsmanship. They are questions about judgment and responsibility.
The falling cost of production changes how much we can make. It does not reduce what we owe the people affected by what we make.
That is the distinction the language of slop keeps reaching for.
The opportunity in making things easier is the capacity it gives us to focus elsewhere. The opportunity is to use that ease deliberately: automate where it makes sense, build interventions that help work move and scale, and create more capacity for the problems that demand deeper judgment.
That is a higher standard than simply producing more. It asks whether what we make serves the purpose of the work, and whether our tools help us serve it better.
AI can accelerate what we make. It cannot take responsibility for why we make it, what we choose to do with it, or who it is meant to serve.
Respect is still manual.