Hacker News Reader

Comment by andros | Hacker News Reader

If you've reached that conclusion, I'm truly proud of my writing technique. I'm sorry to say, though, that your instincts are failing you this time. I write my articles by hand over several days, although it's true that I do consult AI to improve my style, expressions, find synonyms, create tables, and correct spelling mistakes. Thank you for your comment!

View on HN

I don’t think it failed. I was reading it and it did not seem written by a LLM (which I was happy for). But a few sentences did have LLM style and they disrupted my reading flow.

This is my heuristics: Usually when writing a story, authors adopt a fluid flow as they know they have your attention. Same as when telling a story. But LLM tooling usually adopt a highly emphatic tone similar to speeches: Short propositions, emotional crescendo, lots of contrast.

The difference in style is like abruptly going from a conversation to having your interlocutors doing a marketing speech in front of a crowd of one. It’s really jarring.

It’s not the whole piece. Just a few places here and there. If you’ve adjusted a few sentences, maybe try leaving them as is after fixing spelling mistakes.

I think you'd get your point across better if you avoid consulting AI to improve your style, may lead to less distracting threads like this.

FWIW no one can tell anymore. Claude stopped doing all the LLM ticks like 6 months ago. There's a whole industry of people trying to detect LLM writing and they're getting stomped, this can be and is extremely well studied in the literature.

The criticism is that you did something wildly ambitious and pulled it off. The blog is just well written.

People clearly can, though, because they did.

I'm inclined to believe the author, given that they were honest about the machine assist on the engineering? Why would they lie about the post and not the artifact?

More broadly, I'm just getting tired of the low effort snark on this. If someone is going to cite a passage, or examples of a theme, do some serious analysis by all means. I'm not trying to tone police high effort comments.

But this drive by "that's LLM output" thing is the new low effort "javascript is just as hard as c++" a.k.a "i don't know how to do that and i have no intention of learning and it feels bad so i'm going to lash out".

Make a case that something is deficient on its merits or don't, but baseless accusations of dishonesty are not a reasonable default. The parent here did real work that no one paid them to do, donated it to the community asking nothing in return, it's meticulous and high value, and they've already been run over by the FSF goons, you really want to pile on with no evidence?

I'm inclined to believe them too, so when he said he used LLMs to tweak the wording, I'll accept that was the case.

Perhaps the opening post was ill-phrased, but I think the second post clarifies it a bit: https://news.ycombinator.com/item?id=48677597 - and I was certainly getting some vague LLM vibes from the writing, so I was curious about this too.

So, whatever it is in the LLM output that is detectable, it's clearly detectable, even with the latest LLMs, and even when it's present possibly only in trace quantities.

> Claude stopped doing all the LLM ticks like 6 months ago.

Oh God, if only. You’re right that it moved on from the classics, but it has new ones, and they’re just as identifiable.

With how straight forward you were about disclosing 100% LLM generated code, I have no reason to doubt you. Besides, the most riveting parts were the quotes from the ensuing discussion even from Stallman himself.

Reply on news.ycombinator.com