logoalt Hacker News

gorgmahtoday at 2:32 PM2 repliesview on HN

I apologize for making you sad, seeing all the comments made me realize that I should have been way less aggressive with the AI proofreading ; I wanted the sentences to feel a bit more Terry Pratchetty and thought a lot of Claude's suggestions were really better than what I had made. I actually agree to your point too


Replies

mohamedkoubaatoday at 2:36 PM

I write science fiction as a hobby and am in writing critique groups. One of the first rules of critiquing writing is not to suggest how to say things, but only to say what a certain phrase made you feel (confusion, boredom, etc).

LLMs harnesses try to make them useful to suggest things, but this is the most destructive thing you can do to a writer. You can work around it by just feeding Claude a writing critique skill.

gyomutoday at 2:42 PM

> way less aggressive with the AI proofreading

It’s funny how anytime an article gets called out for being AI slop on HN, the author’s reaction is something like that: “oh yeah sorry I used AI but just for proofreading I swear, I should’ve done just a tiny bit less”.

No one seems to get the message that relying on AI at all is what makes writing shit. Good writers have confidence in what they produce. The fact that you’re willing to incorporate any AI suggestion at all means you’ve already lost the battle.

show 1 reply