This has been on my mind for a while, but this comment on Lobsters beat me to the punch and really kicked my ass into getting this written out.¹

AI psychosis is real and it's prevalent. I, in large part, blame it for my total disconnection from my previous job, which drove me to leave it last month without anything else lined up. I've never done that before. It felt really freeing.

While there, though, I sat through (as well as refused to sit through) an absolute slog of reviewing +1,000 LOC PRs, all with these super-fucking-long descriptions always ending with this fucking line:

"Generated by Claude Code 🤖"

My response to this began as a mildly annoyed but otherwise blithe "yeah, man, we can tell" with a little laugh to myself. After the 20th PR like this, though, to say I soured would be pretty charitable. Every single email I got notifying me that my review was requested on a PR speared me with an expectant rage. Which was pretty much always justified.

First I raced to check for its presence on every PR I opened. Scroll down (and boy I had to scroll), yup, there it is. I even took some time poking around other repos' PRs—²lo and behold, that fucking robot was everywhere.

This definitely took a toll on my mental health, in general. Burnout and such. There's one distinct aspect of that toll I'm left with that I have really been enjoying calling AI schizophrenia.³

God, it's everywhere

I’ve become hyper-vigilant and paranoid of AI-generated-anything-anywhere. Scrolling Instagram, watching TV, walking around Chicago. I exhibit markedly decreased joy or motivation and extreme difficulty sustaining activity upon seeing any of it—it’s pretty bad.

Upon opening any GitHub repo, the first thing I look for is an AGENTS.md file at the project root with the same gazelle intensity I looked for that fucking robot emoji. If I see that file, I have the insatiable urge to close out of the tab as if I was throwing something that suddenly burned me. Ditto goes for looking for Claude as a contributor, which is actually so, so, so much worse.

Every restaurant I go to, I check the menus for AI slop. A few weeks ago I stared at the wallpaper of a local ice cream shop because I could have sworn the ice cream cone decal was AI-generated. I think my girlfriend was (rightly) worried about me. I probably sounded insane trying to explain to her why I thought it was.

I’ve been replaying Conker’s Bad Fur Day recently, at an appropriate age for the first time in my life. It’s sick. Some parts didn’t age well, like the camera (holy shit, it’s so bad). I decided, for some reason, to go look at the old IGN review of the game. Dunno why, I just like to read other people's opinions about things that I enjoy. Not far into it, there’s this piece:

This review was originally written in 2001. It was last updated in 2018. It contains exactly 0% AI-generated text. But reading The Facts followed by a list threw me for a loop. Same knee-jerk “toss it out” reaction. This felt like something ChatGPT would spit out if you asked it to write an “honest review” of the game. 2001!

No, I didn’t latch onto the delusion that it actually was written in any amount by AI. I've had too much therapy for that. But I did lose the motivation to actually read the review.⁴

One caveat⁵

At that job that I've since left, I used AI a lot. By my own standards, way, way too fucking much. I had Claude shit out a ton of code that totally worked and is still running in production now. I used it to the point of feeling sick about it, and I still do—yet still I know for absolute certain that I was below the average level of output (not 100%, but also not 0%, of which was indeed slop) in sheer volume.

What was I supposed to do? Not use it? What a joke. The amount of PRs by contributor on each team were shared on an engineering all-hands; you'd be stupid to not tokenmax.

Yes, my name is on those contributions. But they're not mine. They're just not. I was responsible for them, yes, but they're mine in name only. And barely even that—I just never wanted to figure out how to tag Claude as a contributor. I wish I could scrub my name from them and replace it with Claude's, because that would be right.

I know it sounds like I've been shit-talking my previous company, but I want to be clear that that's not at all my intention. I respect the people I worked with there, and their company is consistently profitable, which is incredibly impressive. They're experimenting with what works, throwing as much at the wall as they can bear, and that's fine. It's just not for me. Plus, AI psychosis doesn't care who you are, and it afflicts the smart and talented just as much as the...not smart and untalented. (Not saying anyone in particular are the latter. But some people out there are.)

Recovering

I haven't touched Claude Code since leaving that job and I hope to never do so again. I know that's probably unrealistic—I don't have "FU" money, only "I'm so grateful to have had the opportunity" money—but as long as I can hold out, I will. With inspiration from Nik Suresh, I’m taking the plunge into entrepreneurship with my LLC, Tacet Systems. God willing, I’ll sign a contract some time.

I'll write specifically about why I despise “agentic engineering” eventually, probably, maybe. But I’d also like to mention⁶ that I don't despise people who do use AI coding tools. I think about it sort of like being a chill vegan (which I'm not—a vegan, that is—I generally try to be chill): you can partake in eating animal products, but I'll abstain as long as I don't become a pain in the ass. A bit of cheese or an egg here and there is fine, especially if that's what's at the party.

I guess I have to come to peace with the fact that we're in an AI moment. People are using it for way too many fucking things, from small businesses generating their menu images and wall murals to LinkedIn demons generating corpo-inspo-slop. I'm looking for it all over the place as a habit. I'm hyper-vigilant, for better or worse. I just can't shake my paranoia.

But you know, it’s not paranoia if they’re really out to get you.