5 comments

[ 0.14 ms ] story [ 19.8 ms ] thread
To be honest, I think it's a lot more simple than that. Human writers want to say something because they believe something that matters to them and they want to get it out. AI doesn't believe in anything and it doesn't care about getting anything out. Even when it tries to emulate doing that, it sounds fake.
The uncanny valley isn't a new concept. It applies across all media an LLM might work in, it's not just about visuals.
I appreciate this article for trying to quantify the strange experience that is reading Claude Code's writing.
It's also picked up a lot of habits from the training data, which is heavily weighted toward academia, that isn't written the way normal people write.
> While AI signatures like “delve” and em dashes are widely associated with “slop”, they are not what makes writing sloppy.

It's so refreshing to see that being said, because I've been using both since time immemorial. I was actually really confused when people started to take aim at the word “delve” specifically. It doesn't feel like such an unusual word to me.