Hacker News Reader

Comment by addycb | Hacker News Reader

Ai writing

I immediately suspected AI writing. Then I quickly checked a couple of their older posts, and sure enough: completely different tone, language, grammar.

There may come a day when we can no longer reliably tell the difference, but for now, I'd just prefer not to see this kind of writing popping up on HN.

It's so bad that whenever I click on blog post I don't start by reading the content, first I skim to see if it's even worth my time. This one is not.

> The part I keep coming back to

Immediately caught my eye. Reading in...

> Bitmap fonts make computers feel like computers again. Not abstract “digital products”. Not generic interfaces. Not frictionless panels pretending to be neutral. Actual computers.

Aaand there it is. Tab closed.

It's gotten to the point where I barely even skim first. I sort of unfocus my eyes and can sometimes see the shape of LLM writing. Sort of like when I'm birding and I switch from eagle vision to owl vision, and I can ID a bird just by catching the way the light reflects off its wingflash in the corner of my vision.

>And, just as I can now recognize Christian music in three notes or less, I can spot ChatGPT output without necessarily even reading any one contiguous string of it. I can just tell by the shape or something.¹

1. https://v-n-n-v.github.io/chatgpt-voice.html (bit outdated now, from 2025-03)

Yeah, I got 2 minutes in before a "x isn't just y, it's Z"

It just feels disingenuous. Put a disclaimer at the top, so at least I know. But there's not, it's the author's name.

I can prompt chatgpt to write me this. I want to hear from people who know the tech/history.

Someone here a couple of months ago made a tool[0] to detect AI writing on websites. This one gets categorized as "pure slop" https://tropes.fyi/vetter/d7cebcde

[0] https://news.ycombinator.com/item?id=47291513

Too bad that tool sucks. I ran my own blog through it and it gave me a middling score, even though I’ve never touched it with an AI tool of any kind, even Grammarly.

I rather see the prompt they used to make that article, than the article itself. It overstayed it's welcome to say nothing.

This website has a tool just for this. Here's how it went

https://tropes.fyi/aidr/fc7ff499

From that website: "Every bullet point or list item starts with a bolded phrase or sentence. Extremely common in Claude and ChatGPT markdown output. Almost nobody formats lists this way when writing by hand. It's a telltale sign of AI-generated documentation and blog posts AND README files (especially with emojis)."

That’s bullshit. It’s very common.

I used to do it. Have stopped because AI made it seem uncouth.

Attacking the source of the message instead of the merits.

Ad-HomineLLM

AI writing is worse on the merits: it is lower quality and has concerning externalities associated with its production.

It's also reasonably effective proxy to determine whether somebody is actually passionate about the topic they're writing about. If you've got a very strong interest in a specific niche you're typically able to pour pages and pages of ink down talking about it. If you can't be bothered to take the necessary time to distill your thoughts, it signals to me that your thoughts on the topic aren't as worthwhile as someone who's deeply invested in it.

Of course this proxy isn't perfect, I understand many people use AI to make their writing more comprehensible when English isn't their first language.

Often the commentor wants to take a shortcut and just say it's AI written and hand wave it away.

A comment should argue the merit of the work not attack the source or medium.

The medium is the message. I feel like this should be more obvious in 2026 than at any other time in history.

If the author does not care to take the time to craft their message, why should I care to take the time to read it?

If you used AI, and I cannot tell, I don't care that you used AI. But when it's clear and present almost immediately, I feel as though the author does not respect their audience (of which I am a member).

As every composition teacher would say: "use your own words."

The attention economics are bad more than anything else. LLM articles ask us to put more time into reading it than the LLM put into writing it. Actually committing time to production is the minimum bar which suggests something is worth our time in a world where so much is already vying for our attention.

If a truly amazing thing was released tomorrow that had massive utility you wouldn't care how long it took to create and would just use it.

I get the attention economy is messed up right now, but using it as a justification for being curmudgeonly or abandoning principles is lame.

Yeah but we're not talking about utility, we're talking about content and in this particular case content which basically just boils down to someone's slightly quirky taste in something.

I think utility is a big component of that. I think there's a reason we're discussing these types of things instead of just cat pictures and memes.

We need an update to the HN posting guidelines that addresses this. We already have:

    Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage. They're too common to be interesting.

I'm not welcome to complain that the website is a tiny vertical strip down my screen with 6 inches of whitespace on each side, so we should also not welcome the boring, common "The article is written by AI" criticism, which is going to apply to 99% of articles by the end of 2026. It's already too common to be interesting criticism.

You are wrong in this regard. If you complain about "6 inches of whitespace on each side", you should read this on a proper device (i.e. a desktop computer) or just inject some CSS to fix the site (this is HN, after all).

But knowing that the article is AI generated is useful, because that tells me to not to read it.

an article written by an AI about fonts, when the AI fundamentally does not look at rendered text, is inherently without merit

Those are good arguments to be made regardless of who wrote them. I'm all for actual arguments against the work instead of hand waving it away as AI and being too lazy to say what is wrong with the work.

Can the same argument be made about a writer that is blind? If a blind person submitted work about fonts would we be equally as dismissive?

Who would want to read about the thoughts of an AI?

All it knows about your thoughts are from what text you already fed it with, and it will end up adding things you don't intend or agree with. Even just telling it to fix grammar it can subtly do this.

I think a lot of people are really interested in what an LLM or something that can pass the Turing test "thinks" or generates as output.

Obviously it's for entertainment, but there are many channels where content creators post questions that have asked LLMs [1]

I also think many people are prompting image generators to see what they produce. I can remember a time when many images that involved asking ChatGPT to make someone aggressive would make the people in the image black, whereas using neutral terms would have them generate as white.

I also remember asking early GPT 4 LLMs to explain something to me like a 5 year old from X location and basically seeing the network produce varied responses as it was clear it had an idea of some 5 year olds from one city being inherently smarter than others. Then you can change it to say a 5 year old girl or a 5 year old boy and it would dumb things down a lot more for the boy.

[1]: https://youtube.com/@fatherphi

It reads like a dog wrote it. Whether the writer is terrible or LLM being terrible, get this shit out of here

A dog writing sounds like an interesting story.

Reply on news.ycombinator.com