> with the aim of their output being acceptable to human interlocutors.

The more I use LLMs I find that their output drifts toward being acceptable to other LLMs. I find myself quite often telling the model: "this is not human readable.". Especially Opus and Fable write in such a way that it is difficult to follow. The often put sentences out of context, the context itself is fragmented. Often start with what you would think is end of a story so when you start reading you are like wtf it is about.

If think the issue might be that LLMs read text all at once whereas humans are somewhat linear. So when LLM puts out text it understands, to human it is close to random gibberish that looks like English.