What Do People Actually Discuss with AI?

I’ve been seeing more and more articles recently about people getting LLM psychosis from interacting with AI models regularly. The first major touchstone for this I believe was ChatGPT’s 4-o model. The current wave as of January 2026 is the Claude Opus 4.5 model, specifically when using Claude Code. For this wave, it’s not just people chatting with the AI and getting stuck in some ideological echo chamber – it’s people using it and thinking it’s on a path to taking over the world. To be fair, I do think it’s pretty good at what it does. Having said that… how are people getting ideologically captured by AI? I have no idea. I’m not going to pretend like I’m somehow better than everyone, but… I dunno, it just seems like common sense to take everything the bot says with a grain of salt? Another aspect that makes it difficult to figure out how people are falling into these traps is the inherent privacy of their conversations. Whatever feedback loop they get stuck in stays strictly between them and the AI model. I imagine the way people use AI varies, drastically. While I can’t look into other people’s chatting styles, I can at the very least share my own. So here are some threads where I explore my interests.

Learning about geopolitics:

https://gemini.google.com/share/4c581bf13dde

Playing a weird movie associations game:

https://claude.ai/share/a3b95d84-1a9d-4ec1-bee5-0542eb5a4250

If anyone reads these, please share your own in the comments – I’d love to take a gander.