Savannah Harlan

Introduction: The AI Haters In the early months of 2026, generative AI has now improved (at exceeding speed) to a point where many trademark critiques have become dated. The famous 'gotcha' that AI can never make normal-looking art because it always under or over estimates the number of fingers on a human hand is one example. The disproved claim that AI models will never be competent enough to help reduce the amount of work humans have to do because 'they will always mess it up in a way that makes fixing it longer than doing it yourself' is another, if the rise of vibe coding and AI agents tells us anything. Today I want to talk about a particular thing that has become possible since the invention of modern large language models (LLMs). This is about the other side of it: ‘human slop’. The Consumption of Human Slop However badly written, however absolutely trashy and bizarre your fan-fiction is, however manic of a rambling an essay comes off as: from now on, no person will ever write something that no one else ever reads, unless the writer wants it to be the case. Why does this matter? I don't know. I think in part because I am a writer. Although in bigger part I think maybe because I am a bad writer. I write a lot of human slop—unoptimized, meandering garbage that is crafted neither to be fun to read nor engaging enough to finish. One way you know when you enjoy something is when you continue to do it even when you are bad. In that sense, I mostly write for myself. I don't expect anyone to ever read my stories, and none of it would ever change how much I write or love to write. But as a writer there is also something uniquely fun about getting someone else to read what you have wrote. I'm not entirely sure how to describe it, but it is not about ego. It is something weirder and deeper. When you write something that no one else will ever read, it’s rather solipsistic. Solipsism is this philosophical idea that the only thing you can be certain exists is your