I would at least consider the possibility that they already are, and this is how it is going. It could be a fundamental architecture problem for LLMs. It's not

I would at least consider the possibility that they already are, and this is how it is going. It could be a fundamental architecture problem for LLMs. It's not like "slop" is a new problem and I'm sure they'd love to announce a new model that generates much less "slop". The fact they've never so much as mentioned it suggests to me that it's not something they've been able to fix.