Sick Internet
2026-01-13
As I've stated before, I use LLMs to help edit my writing (forgive the wonky formatting on desktop, still need to fix it). For a long time, I had a ChatGPT Plus subscription. I stuck with ChatGPT because I was a fan of the memory feature, which kept coherence across threads. It was nice not only for working on writing projects, but also for self-discovery, weirdly enough. I know many people dislike the latter notion, especially with recent "AI psychosis" stories. As someone who has experienced "pure" psychosis, I've ironically avoided that trap. Takes one to know one, I guess.
I had been hearing that Claude's models were better for working on prose. I tried out Claude when it first came out back in 2023. While I enjoyed working with it, ChatGPT won because it was still the only platform with the memory feature. Then I found an AI software that was far superior. It allowed not only for use of all mainstream LLMs, but also had their own refined versions that have proven to be superior to the originals.
On MyNinja, I decided to test out the flagship Claude model and was blown away at how much better it was for editing my Cogito entries. There was a bit of a problem, however. With ChatGPT, it would give me a few edits here and there and make sure to catch glaring errors and typos. With the new Claude model, it gave line-by-line copy edits. Every single sentence had to be sanitized. I can also be a fierce copy editor, so I get it. However, it was clear that the model was over-sanitizing my prose. It was attempting to convert it to the classic AI style we can't stop encountering.
Of course, I took some of its edits—there were plenty of spots that needed improvement. It reminded me a lot of Grammarly. Back what I first started publishing online, Grammarly was the hot tool. I wrote a few pieces with it and grew dissatisfied. While writing, it would give corrections mid-sentence, which halted my flow. Moreover, the edits were not always ones that I preferred. Even when I turned that off and had it look at pieces as a whole, I still wasn't happy. I kept rejecting a lot of the edits because I knew that if I let all of them go through, I'd sound like, well, a fucking robot.
I self-publish most of my writing (I'd love to have other people publish it IF I WAS GOOD ENOUGH AT WRITING) and it's nice to have a cheap second set of artificial eyes look at it and catch my mistakes. Despite countless passes and multiple revisions, mistakes still slip through. That's not just me, every writer deals with it. With how often I publish, and since I don't want to put a bunch of money into it, LLM editing is a solid option for my needs. Working with them has taught me the importance of discernment.
LLMs always offer to rewrite my stuff for me when I pass it through for edits. I despise that. It's a waste of time. I'll try it out occasionally and it's always the same sanitized bullshit I see from idiots who trend on Substack and Medium. It's wild how so many of their readers are blissfully unaware of that. I read comments and feel disgusted. Many have discussed the Dead Internet Theory before. There is a lot of bot traffic online, sure. There's also almost nine billion people on the planet and basically all of them use the internet.
The internet isn't dead. It's sick.
People have become helplessly reliant on LLMs to parse information. It makes sense because there's so much of it. But for many, their discernment has atrophied. Most search engines are completely useless now (I made the jump to pay for mine), and no one is able to recall anything because our attention is constantly shifting. I don't want to harp on points made ad nauseam. None of this is original, anyway. Still, despite the exposure, we don't seem to have any respite. It can be hard to talk with friends or acquaintances about it. Often, what I say falls on deaf ears—at least it can feel that way. I feel confident in my space here on the Small Web.
I just wish it wasn't so small.