Raw LLM Responses

And yet GPT5.2 has gotten worse \*again\*. Whatever tweaking you guys have done in the last few weeks have made it all but useless for anything outside of some basic coding tasks. Not buying the "this is about mental health" nonsense - I think this is about giving all the good compute to the people with big pockets, and leaving the other paying customers with a nerfed product. It isn't 'better' with mental health anyhow. Every day the model gaslights me, continually tells me "you didn't do anything wrong" and "this isn't a you problem" (or my personal favorite) "You're not crazy" - just for asking it the most basic 'how to' question. How is this supposed to help with mental health, when it is maddening to work with? I get it, inference is expensive - but if this didn't feel like a bait & switch before, it sure does now.