A Silver Lining of Slop • Stephen Brennan

To state the obvious: when reading text on the Internet in 2026, there’s really no guarantee that it was written by a human. Sometimes I’ll be reading something, and notice a red flag: heavy em-dash use, common LLM vocabulary, common LLM sentence structures. I realize, in horror, that a human did not write this. Suddenly, I’m frustrated that I’ve wasted my time, annoyed I didn’t see it sooner, and angry with whatever human had a hand in it¹.

It mostly happens in the comments on sites like Hacker News or reddit. That’s not great, but it’s expected; I wouldn’t expect to read anything meaningful in the comments, even if they were all human-authored². Finding an obvious AI comment usually just serves as a reminder to stop reading Internet comments!

But I also notice it in blog posts. And that just rubs me the wrong way. To me, the whole point of a blog is to share your thoughts and opinions in text. These are things that only exist in the context of a human being: LLMs do not have thoughts or feelings. Even in the case of blogs that are technical or information dense, what makes them stand out is their voice, tone, structure, even the framing for why they were written. Your humanity, your thoughts, your opinions: those are a huge part of what make a blog post worth reading.

What’s funny about this is that sometimes, I write something and don’t put it on my blog, because I don’t think it’s good enough. Sometimes I just don’t write it to begin with, because who would read it? But really, why take myself so seriously? After all, I guess there are folks out there having Claude one-shot their blog post, so at least that’s not me.

This blog post may not be a great one, but at least I wrote it myself. It reflects me, and it has my voice. So I guess what I’m saying is (a) I will likely write a bit more small things like this, and (b) I’ll never be using AI to generate blog posts, because then… what’s the point at all?

Yes, I know some folks use LLMs to overcome language barriers, or to help them phrase things. This isn’t what I’m talking about here; it’s not too difficult to tell the difference between wholesale AI-generated content, and genuine use of a tool to help in those situations. ↩
There are definitely exceptions to this, which is why I still look. ↩