Posted by usefulposter 1 day ago
If you discuss an idea with AI, then close the window and write a post about how you came up with the idea, got stuck, decided to ping an AI for unstuck-ness, describe how the AI’s response got you unstuck, and then continue writing about your idea, that’s not going to be necessarily treated as AI-assisted writing — but people are going to be extremely suspicious of you, because the perception is that 99.9% of people who use chatbots go on to submit AI-assisted writing. That’s probably more like 90% in reality but it’s something to be aware of as you talk about your experiences.
If you use AI in your process and don’t disclose it when writing about your idea and process, that’s generally viewed as lying-by-omission and if egregious enough you could end up downvoted, flagged, and/or banned (see also the recent video game awards / AI usage affair). Better to disclose it with due care than to hide it.
Except it’s bullshitting the whole time. While you think this is what you wanted to convey.
Not sure where I’m going with this, but my point is if I pasted this comment into ChatGPT it would make up an argument I never made to support my case that didn’t exist in the first place. Exploring things is useful but just be aware it’s designed to pull bs out of it’s ass and is distinctly not interested in exploring truth or having a real conversation
So if your layer of cleanup is AI assisted, then it's in violation.
Part of the problem I was getting at is that the requirement of "Don't post AI edited ..." is stricter than necessary to ensure the outcome that "HN is for conversation between humans" because an AI edited post is still a human post.
Anyway, I suspect a lot of people are going to ignore that guideline and will feel free to use their "layer of cleanup" whether it's a basic spellchecker or an LLM, or whatever else they choose, and most people aren't going to be able to tell anyway. The guideline is unnecessarily strict in my opinion, but it doesn't matter in the end.
But I think you and I are on the same page: we both know this isn't a rule that's there to be hard-and-fast enforced because that's completely infeasible. The definition of "AI" is a moving target, as is "generated."
It's a rule that's there to have a rule so when the real problem is "Hey, your content is too low-quality but you dump volumes of it and it's clearly following a procedural template" the mods can call that "AI" and justify limiting or banning the account on prior-stated rules. Which is fine, but I'm glad to call it what it is.
(One unfortunate oversight: we haven't added "posts sounding like they are AI-generaed" to the "Please don't complain about" set. So expect that to become a common refrain now since the incentives to make the complaint against disliked comments are obvious... At least until that becomes annoying enough to justify a rule).
I was thinking, this argument is suspicously cogent!
The real issue isn't just "slop" or bot-spam; it's the cost of entry. HN works because of the "proof of work" behind a good comment. If I’m spending five minutes reading your take on a kernel patch or a startup pivot, I’m doing it because I assume a human actually sat down and thought about it.
When the cost of generating a response drops to zero, the value of the conversation follows it down. If the author didn't care enough to write it, why should I care enough to read it?
The "AI-edited" part of the rule is the trickiest bit, though. We’re reaching a point where the line between a sophisticated spell-checker and a generative "tone polisher" is non-existent. My worry isn't that the mods will ban bots—they've been doing that for years—it's that we'll start seeing "witch hunts" against anyone who writes a bit too formally or whose English is a little too perfect.
Ultimately, I’m glad it’s a rule. I don't come here to see what an LLM thinks; I can get that on my own localhost. I come here for the "graybeards" and the niche experts. If we lose the human friction, we lose the signal.