i’m still talking to gpt-5.5 a lot and interestingly its writing seems to get much worse when it “tries harder”? it casually drops sentences that make me pause and go “whoa” in the middle of yapping, but whenever it tries to do a pithy summarization of what it just said it’s awful, just pure contrastslop

i find this mystifying, wonder if it’s some sort of weird artifact of something in training. like reward hacking is kicking in harder when it goes for the summary? i have just gotten used to tolerating this for now but it would be genuinely nice if they put some effort into cleaning this up for future models, it seems pretty fixable