Or, How I Learned to Stop Worrying and Love the Bot

Séb Krier recently tweeted a great list of Substack articles he’d fund if he were rich. While we wait for him to accumulate patronage-level wealth, I thought I’d take one on free of charge:

Barriers to entry: how photography at scale has changed the world, and why you should cry about slop less

Regrettably, I’m not equipped to take on the history of photography, but I do feel that we should all cry about slop less. I took a crude stab at this last April, and I’d like to do a better job here, taking Séb’s prompt as a starting point.

On the Origins of Slop

The smartphone put a camera in everyone’s pocket, and the result has been a deluge of photographs, with billions of images uploaded daily. Most of them are terrible—poorly lit brunches, off-center selfies, sunsets that looked better in person. The “democratization of photography” has, in one sense, been a disaster for photography.

And yet… I would not trade the smartphone camera for anything. Not because I care particularly about photography (I don’t) but because the same logic that governs the explosion of bad photographs also governs the explosion of good ones. The bell curve of quality still applies; it simply operates over a larger population. The absolute number of masterpieces grows if the ratio stays constant, or even if it dilutes slightly.

This is the point I want to make about “slop” more broadly.

The Anti-Slop Discourse

There is a growing chorus of voices warning us about AI-generated content. The lament goes something like this: generative AI will flood the zone with low-quality writing, art, and music, making it impossible to find the good stuff. We will drown in mediocrity. Culture itself is at risk.

I have written elsewhere about my concerns regarding the coming era of abundance—concerns I share with Ross Douthat and others who worry about cultural extinction. But I part ways with the slop pessimists on one crucial point: they are worried about the wrong problem.

The worry is not that slop will exist. The worry is that we won’t be able to find the good stuff amid all the slop. This concern overlooks one crucial point: we are already drowning.

The streaming era has already demonstrated that the “recommended for you” algorithm is not your friend. Netflix’s homepage is a graveyard of mediocre original content. Spotify’s Discover Weekly has long since devolved into a promotional tool for whatever the labels are pushing this month. I am consistently baffled with how poorly these companies are using the troves of data they possess.

The signal-to-noise ratio was already catastrophic before a single AI-generated novel hit the market. We are already living in the age of algo-slop.

The Curatorial Responsibility

The slop pessimists seem to operate under the hidden assumption that there was once a golden age when finding great art was easy, when trusted gatekeepers reliably separated wheat from chaff. But when was this, exactly? When the New York Times bestseller list was a reliable guide to literary merit? When the Grammys reflected actual musical achievement?

Curation has already become a personal responsibility. The algorithms have already failed us. We are already in a world where finding meaningful art requires active, intentional effort.

What changes in the AI-slop future is not the difficulty of the task. What changes is the reward for completing it.

If the top 1% of novels published each year is currently, say, 50 books, and AI tools enable ten times as many people to finish a novel, then the top 1% becomes 500 books. That’s 450 more great novels per year. The same curatorial effort that currently yields a handful of masterpieces will yield a bounty. This should be cause for celebration, not mourning.

Humanity at Stake: The Creator-Curator Distinction

Another variety of slop pessimists are worried about the human element of art, fearing that we will lose contact with one another by crowding out personal work with AI slop. I think this camp is conflating two distinct activities: creation and curation.

We tend to think of artists as creators only, but they are curators too. Every artist makes a thousand small decisions about what to include and what to discard, what to polish and what to abandon. The finished work is a curatorial achievement as much as a creative one.

Consider my own modest operation here at the Ragged Frontier. I have written dozens of blog posts that I decided not to publish because they were, in my human judgement, bad. If I had published them, I would be the same creative mind but a much worse curatorial one, and many of you would stop reading my work. The quality of the blog depends not just on my ability to generate ideas but also on my ability to recognize which ideas are worth sharing.

The same logic applies to AI-assisted creation. The idea that a serious writer would simply publish whatever an LLM spits out is incoherent–such writers do exist, but their work is already terrible. The bad AI-assisted novel is not a failure of AI; it is a failure of curation. And if you’re a bad curator, your audience will find out soon enough—with or without AI in the mix.

Consider a thought experiment. Suppose Herman Melville had access to an LLM. Not to write Moby Dick for him—the idea that Melville would simply accept whatever ChatGPT generated is absurd—but to help him draft, to unstick him when he was stuck, to generate a few paragraphs of cetology that he could then revise into something properly Melvillean.

Would we have gotten another epic novel from him?

The historical record complicates this. Melville’s late-career silence was not primarily a productivity problem. It was despair, alcoholism, a reading public that had moved on–fine. The romantic in me wants to say that the struggle was the work, that the friction was essential to the art (please do read the excellent John Stone article if you haven’t already).

But I don’t actually believe this. No one argues that Melville would have written worse novels if he’d had a typewriter instead of a quill pen. No one claims that word processors ruined American literature. I would have gladly given Melville MS Word while he was still productive, so that he could have made better use of that time. Tools are tools. They enable effort, but they do not eliminate it.

What Melville would have done with an LLM is what every serious artist does with every tool: use it in service of his vision, discard what doesn’t work, refine what does. I would kill for another epic from him, and I refuse to believe that better tools would have prevented one.

A Reorientation

I don’t want to be glib about the downsides to all of this. There are real concerns about AI—about labor displacement and alienation, about intellectual property, and (perhaps most crucially) about the homogenization of style. These deserve serious attention.

But the specific worry that we will be unable to find great art in a world of abundant slop strikes me as both overstated and defeatist. The curatorial challenge is already here. The people who are doing it well now will continue to do it well. The people who are failing at it now were never going to succeed anyway.

So my call to action is simple: less cynicism, more openness to possibility. The tools are coming whether we like it or not. If you, like me, are worried about navigating the proliferation of slop, fight fire with fire: develop AI-driven curatorial tools by writing for the AIs now to help them better understand what your pre-slop brain thinks is worthwhile. Find creators and curators who you trust for recommendations, rather than relying on the algorithms.

Or better yet, write your own novel, and make it distinctly human. We will be the generation that gets 10x more novels, and I for one look forward to finding the best of the best and trying to read them all.

Coda: In the spirit of internal logical consistency, the first draft of this essay was written by Claude Opus 4.5 after I fed it an outline and discussed the topic with it extensively. I also fed it all of my prior blog posts to simulate my style. Then, I revised it the way I would any other draft I produced by any other means. If this essay is slop, it’s not the bot’s fault, it’s mine.