This line from a piece by Brian Phillips, in which he makes the case for “slop” as word-of-the-year, is just perfect:
As if all art ultimately aspires to be Fortnite, an endless slog of reshuffled IP.
Various folk have made the point that slop has in some senses long preceded the generative model phenomenon—Jay Springett’s coining of the term “gunk” for the human-made equivalent comes to mind—but Phillips addresses the distinction with the first of his six defining attributes: slop exists at scale.
Whether by intention or (un)happy accident, its scale reinforces the metanarrative of technological inevitability: how are those of us who loathe this stuff supposed to compete or resist? I made preemptive efforts early on, and have switched off or circumvented “AI” functions in browsers and search engines wherever possible, but that only amplifies the shock I get when I use a machine that hasn’t been firewalled in this way; the now-default offerings of a G**gle search make it hard to recall how genuinely exciting and transformative that company’s original service once was. As Warren Ellis notes, “there’s now probably grounds for the argument that 2025 was the last year of the internet”, and he’s far from the first to say so.
Worst of all, that’s still true even if you have firewalled your interfaces, because so much of the material now being indexed is itself slop, with whatever reliable sources still remain submerged deep beneath a tide of search-optimised generative rubbish.
I’m finding myself very grateful for my years working in libraries, because the basic skills of library science are only going to become more valuable. The promise of the chat interface is a quick answer to a question posed in natural language, but the first thing you learn on the enquiry desk is that, to be blunt, people don’t know what they actually want; they only know what they think they want.
Even assuming a clanker answers their question correctly—an assumption of considerable and utterly undeserved generosity—that question may be malformed, misinformed, or both. With apologies for raising the shade of Rumsfeld, people mostly don’t know what it is that they don’t know. Being the front-end to an archive of knowledge requires this assumption of imperfect knowledge as a starting point; it requires an understanding of the epistemological limits of humans (including oneself!), and the patient diplomatic work of refusing to play sycophant.
The second thing you learn on the enquiry desk is that some enquirers really do not want to be confronted with their own epistemological limits, no matter how politely and helpfully you attempt to do it. Which is to say: I fully understand why the sycophancy-as-a-service of chatbots is so successful, and why they are particularly successful with, let’s say, a certain sort of person.
Such people were perhaps always beyond saving. What’s rather more sad are those people, presumably far more numerous, who simply don’t have access to the better option any more, assuming they ever knew it existed. Per Alan Jacobs, there’s literacy, and then there’s literacy—and the one that means more than simply being able to read and write was perhaps never so widespread as was assumed.
So much for competing: there will still be call for the skills of comprehension and discernment, though it’s looking like it will revert to being (even more of) a thing that’s only of interest to (and accessible by) a certain sort of elite.
What of resistance, then? Jay has the right of it, I think, as he often does: deny the slop your attention, and try to make good things without what Crowley called “lust for result”. Art—a category that I am here using very broadly—can and should be made in hope of its finding an audience. But when audience is the only motivating consideration, well: that’s how we got where we are now, playing King Canute to a rising tide of slop.
(The title of this post is a refrain stolen from the Brian Phillips piece I first linked.)
Leave a Reply