By skyforbes 
… And the rest is Claude.
I am tired. Honestly, genuinely, deeply tired. Almost every article I read has the tell-tale signs of having been written by ChatGPT or Claude. It’s not the m-dashes — it’s the pointless negation. And short incomplete sentences. Like this one. Like the ones your English teacher probably taught you were in poor taste. And the same names of non-existing people as “experts”, consistent enough to be confidently recognizable as individual LLM models’ watermarks. And the endless repetitions of the same fact throughout the article. The itemized listicles in the middle of prose. The completely out-of-place summaries. The dream-like reality that just doesn’t add up. The lack of any real conclusion. “And those five-word rhetorical questions? Are real.” ™ And yes, I can free-hand mock “your” style, no LLM required.
Worse, the platforms do exactly nothing about it. “We can’t report this article at this time.” Because we already know it is LLM slop. And the top comment on it is LLM slop. And the longest reply to the comment by the author is also LLM slop. Because 90% of everything is crap.
Believe me, I would like to publish three to five articles a day, like some other “authors” on this platform — on every platform, really. I would certainly have enough ideas. But I would feel genuinely uncomfortable subjecting my readers to LLM garbage with questionable facts that reads the same, no matter what the subject is, or who “prompted” the question.
If “nobody has the time” to write the text by hand — who has the time to read it? Just give us the prompt. If it is shorter than the article (and I can guarantee you that in 99% of cases it is), that would save us all time and resources. Maybe someone should write a tool to reverse-engineer the prompt from the article. It should be possible. (After all, that’s literally what the Stable Diffusion algorithm does for images: it reverses the process of generating an image caption.) It would save us all so much time.
Recently, on a whim, I let an LLM expand a two-day research conversation with it into a short-story. I thought, surely after 200k of conversation and several days of corrections and suggestions, the story would be readable and unique. In the end, I had to throw it in the bin. Not because it wasn’t readable. Not because it didn’t bring the point across. But because it read the same as all the other stuff out there. It is not my voice it is written in. It is not even the style in which I ordered it. It is recognizably the model which wrote it. Like, at this point, most of the articles on Medium. And other websites. And youtubers’ scripts. And newsletters, and journal articles, and probably your dog.
Worse, after so many conversations with that model, even the text I write myself, by hand, starts to read like an LLM. See the “Not because…” turn of phrase above: I genuinely can’t tell any more if this is because I am subconsciously mimicking every other effing article out there written by LLMs, or if this is how I already used to write before the LLM text slop invaded my brain, and turned me allergic to my own style. I can, for example, tell you for certain that I have been using dashes since my very beginnings as an author a few decades years ago, and Medium expands them into m-dashes. And I used to use itemized lists for clarity whenever I had to explain anything via email to my colleagues or customers. Which I stopped doing right after the item lists became the hallmark of ChatGPT. Should I sue OpenAI for copyright infringement, I wonder?
So, if you have wondered why I have not been writing anything for a long time, here you have your answer. I felt like I am sick, and I refused to spread the infection. But at this point, maybe the answer is that the patient has long succumbed to its disease. The Internet writing is dead, and what you see are just the zombified corpses of what we used to write when we were still untouched by the GPT brain rot. And it’s not going to get better, because 95% of everything on the internet is already AI slop, and we are all simply stochastic parrots anyway, GPT included.
So, apart from those short fleeting moments when I can gather some original thoughts again, and write them down in original words — not the words of Claude Sonet, Gemini, or ChatGPT, but my own — I will have to try to find some other … medium.
— J. Macodiseas; text proudly written without the help of any automation tool, mistakes and all.
