I'd argue the article's point is "new communication technology encourages a particular form of psychosis, and LLMs are especially prone to encouraging psychosis because they generate such a believable imitation of speech".

I've been coming to believe LLMs dangerous to mental health in general for a lot of reasons, and I thought this was an interesting discussion of how a basic human instinct - to look for patterns and assume rational thought and meaning behind those patterns - has always gone wrong when applied to technology and is particularly dangerous when applied to LLM-generated content.

(Because there is a reason for every LLM-generated utterance, and that reason is "make the company money". LLMs are capitalist speech acts in their purest form.)

BTW, what's wrong with Substack? Is it just the "Substack hosts fascist blogs so everyone using Substack is fascist by association" thing?