A year ago I would have agreed but lately, when it comes to stuff linked off of HN, it's actually more likely to be clear and readable if it's AI written.

I don't find the LLM written stuff very readable because after one too many "real"s or "The X Dilemma" my brain shuts off. It's not even voluntary, it just does that on its own.

Is it more likely to be clear and reliable if it is AI-written, or are features associated (both directly and by correlation) with clear writing increasingly misperceived as “AI tells” because they are also favored in LLM training?