check out the chatGPT subreddit last weekend for the meltdowns from people who were routed away from gpt-4o to the "safety" version of the model when they tried to discuss controversial topics - they sound like somebody confiscated their best friend.

i think more people are treating an LLM like a friend than you might expect - i was certainly surprised.