Don't be scared by the tag for "slop": this is a human story about humans fighting back against slop. If you're on mobile and reading text feels large, try switching to desktop mode; the writer is...
Don't be scared by the tag for "slop": this is a human story about humans fighting back against slop. If you're on mobile and reading text feels large, try switching to desktop mode; the writer is blind and uses a screen reader
Nobody agrees with me, though. I am an outcast in a gaggle of readers and writers that seem to prefer LLM writing to the jagged edges.
"And this is why," Chad practically yells, "why you don't listen to Luddites! Work smarter, not harder," Chad beams. The smile in his voice is audible, a smile I want to deck. "That's the future. Why churn butter when you can buy margarine, right?"
My favourite sentence at the end:
"This is... this is a first draft," Leo says. His voice shakes, just a little. "It has typos. I didn't run it through a spellchecker. I definitely didn't run it through Claude. And... for a while, I thought I couldn't write without it. I thought my voice was broken. But my friends told me broken things still make sound."
I'm tearing up. Why am I tearing up after reading this story? What the fuck? I'm sitting in a pizza parlor, last nodule of crust grinding between my teeth, clamping down a sob cramp. Why do I feel...
I'm tearing up. Why am I tearing up after reading this story? What the fuck?
I'm sitting in a pizza parlor, last nodule of crust grinding between my teeth, clamping down a sob cramp.
Why do I feel seen, I'm not a writer? Is this hope?
I’m not a writer either and I felt the same. Great writing, and nice insight into the life of someone blind, and great message. It properly conveys what I have been telling my friends who use AI...
I’m not a writer either and I felt the same. Great writing, and nice insight into the life of someone blind, and great message. It properly conveys what I have been telling my friends who use AI for writing. Thanks for sharing.
This is a joke, right? The writing is very AI. I've been having Claude write absurd stories for me for fun for a while, and its many writing cliches have become very noticeable and annoying to me....
This is a joke, right? The writing is very AI. I've been having Claude write absurd stories for me for fun for a while, and its many writing cliches have become very noticeable and annoying to me. It's possible I'm just too sensitive to normal writing cliches now, like how a virus can provoke your body into an auto-immune reaction where it mistakes healthy cells for viruses, but at least Pangram seems to agree with me, saying that 74% of the article's first 2000 words (filling Pangram's free account limit) that I tested are likely AI.
I was pretty sure this was AI from the very first paragraph, but then reaching this part a few paragraphs later made me feel sure enough to make this comment:
I have to suppress the urge to violently groan about his word choice. Instead, I turn my head towards Leo. I can't wait to hear this.
Leo reads. It's a piece about his grandmother's kitchen. He describes the smell of collard greens as a heavy, green blanket that wrestled the air into submission. He describes her laugh as a rusted hinge that still worked perfectly.
It is beautiful. It is imperfect. It stops me cold.
When he finishes, there is a silence. Not the reverent silence of a shared emotional impact, but the uncomfortable, shifting silence of a boardroom that has just been shown a graph with a downward trend.
All I could think reading this is that every single sentence feels exactly like Claude's overuse of bad metaphors and slam-poetry mic drop moments. And then the very next part of this story is comparing the piece to a worse rewrite by "Claud", but both of them and every surrounding sentence sound absolutely AI.
That is an evocative way to write, but the author is absolutely opposed to LLMs and is blind (losing his sight over time), leading to perhaps different sensory descriptions than you might expect...
That is an evocative way to write, but the author is absolutely opposed to LLMs and is blind (losing his sight over time), leading to perhaps different sensory descriptions than you might expect but that doesn't mean they're written by a machine.
He has written a number of things about how he doesn't like AI picture descriptions for his screen reader, didn't appreciate a beta reader feeding his words into an LLM, that it helps him navigate the web but using it for writing seems likely to eliminate the human - and particularly disabled humans - from work. And he finds it rather soulless.
He explicitly states he doesn't use AI in any of his work.
If you hate LLM writing, remember who did it first. Humans. LLM writes the way humans write. Not the other way around. If a human sounds like an LLM, that's because the LLM is plagiarizing their style and their writing style.
Humans will always write like LLMs because humans did it first. I don't know what else to say other than your paranoia is helping them.
I do not believe your criticisms are correct. Whether he's mimicking AI tropes on purpose or writing in his usual style (I'm not familiar enough with his work)as he said it's worth noting that those tropes come from human writers.
I was curious whether this was just their style they've always had, or something they only very recently picked up. I figure if I could find some of their work from before AI was at its current...
I was curious whether this was just their style they've always had, or something they only very recently picked up. I figure if I could find some of their work from before AI was at its current level that was detected as AI or had the same writing style, then it would mean I was wrong about it being AI. I have not found a single piece of writing by them before 2025 so far that was detected as more than 0% AI written, or appeared to me as having similar issues or writing style (where every sentence in a row overuses metaphors and slam-poetry mic drop moments). I subscribed to Pangram so I could use their service without limits. I did judge the writing style of each post myself first before putting it into Pangram, and Pangram always matched my judgment. (They do have some recent posts too that appeared to be 0% AI.) I've tested on several excerpts of their books on their site too.
I searched for other discussions about this article and articles from their domain, and I found this HN thread on a different recent post of theirs where some commenters stated they found the AI style annoying. One commenter linked to a Pangram analysis rating it as 100% AI.
I think they're either being extremely sanctimonious about their own AI use, knowing their anti-AI audience is unlikely to be able to recognize it, or maybe they've recently fallen in love with the style and carefully re-engineered their writing sentence by sentence to be indistinguishable from it while complaining about it.
Also they do have a recent Mastodon post bragging that they found they were better at prompting LLMs than most LLM users. I kind of get the impression they're a little too interested in LLMs for someone who avoids them, but to be clear my judgment above isn't depending on this vague feeling off of their posts.
I think it's most likely that he's writing this way intentionally due to the nature of this story, but again I don't know his style, and I don't trust AI "graders" (the link you shared doubts...
I think it's most likely that he's writing this way intentionally due to the nature of this story, but again I don't know his style, and I don't trust AI "graders" (the link you shared doubts Pangram specifically in fact and makes the point that detectors will cause false accusations.)
I saw his post about covering an LLM start up event for the purpose of fucking around with the prompts on stage (maybe unrealistic). He also just said he was better at the prompting than his neighbors who told the LLM "don't hallucinate". Which, like, I also as a hater could have done better than that. There was no indication this came from his frequent use of them or anything. The post on HN doesn't have that many comments complaining about it being AI or complaining about the style or at least they're not in the top 60% of comments... Update: I read the whole thing it's a very small number of people complaining about AI use. Again maybe I'm wrong about him but I'd have had to search for his name and AI style to even get that HN page to come up and see those comments because they are in the last like 5% of the page.
Mostly the critics are just calling it fake and citing that he writes fiction while ignoring the nonfiction tag. Or saying he blamed her too much or whatever. It's all very reasonable criticism.
I won't be shocked if there are some AI style complaints, and idk how HN upvotes work but they didn't float to the top half*
To me he seems like many authors who dislike AI while he's also particularly passionate about tech - posts about what I will call programming stuff (idk GitHub and stuff) and Linux, and, particularly about accessibility in tech. He's a disabled writer with a journalism background, and he posts a lot on techy topics he cares about as a techy guy online all the time. He'd fit in here well IMO.
I just don't see it as likely that he's lying about all this and secretly using the LLMs to bait people. Especially as a queer dude who is agreeing with posts calling Andy Weir an idiot for thinking that writing doesn't have social commentary.
I think it's fair to criticize the work, I just don't think it's likely to be AI generated or bait given the context. But like maybe he's a big old liar, idk him. However, I think this nuanced conversation is very different from "this is a joke right?".
*Oof the ableism in some of this is gross. I'm reading the whole thread in the interest of fairness. But ick.
Well, keep in mind that AI was trained on, basically, average writing. It learned bad writing from us, so when you see bad writing there's still a very strong chance it was just a regular human...
Claude's overuse of bad metaphors and slam-poetry mic drop moments.
Well, keep in mind that AI was trained on, basically, average writing. It learned bad writing from us, so when you see bad writing there's still a very strong chance it was just a regular human bad writer who wrote it.
Don't be scared by the tag for "slop": this is a human story about humans fighting back against slop. If you're on mobile and reading text feels large, try switching to desktop mode; the writer is blind and uses a screen reader
My favourite sentence at the end:
"This is... this is a first draft," Leo says. His voice shakes, just a little. "It has typos. I didn't run it through a spellchecker. I definitely didn't run it through Claude. And... for a while, I thought I couldn't write without it. I thought my voice was broken. But my friends told me broken things still make sound."
I'm tearing up. Why am I tearing up after reading this story? What the fuck?
I'm sitting in a pizza parlor, last nodule of crust grinding between my teeth, clamping down a sob cramp.
Why do I feel seen, I'm not a writer? Is this hope?
Thank you for sharing. Fuck.
I’m not a writer either and I felt the same. Great writing, and nice insight into the life of someone blind, and great message. It properly conveys what I have been telling my friends who use AI for writing. Thanks for sharing.
I cried, no shame to it.
I think it's the power of friendship. And connection between human beings.
This is a joke, right? The writing is very AI. I've been having Claude write absurd stories for me for fun for a while, and its many writing cliches have become very noticeable and annoying to me. It's possible I'm just too sensitive to normal writing cliches now, like how a virus can provoke your body into an auto-immune reaction where it mistakes healthy cells for viruses, but at least Pangram seems to agree with me, saying that 74% of the article's first 2000 words (filling Pangram's free account limit) that I tested are likely AI.
I was pretty sure this was AI from the very first paragraph, but then reaching this part a few paragraphs later made me feel sure enough to make this comment:
All I could think reading this is that every single sentence feels exactly like Claude's overuse of bad metaphors and slam-poetry mic drop moments. And then the very next part of this story is comparing the piece to a worse rewrite by "Claud", but both of them and every surrounding sentence sound absolutely AI.
That is an evocative way to write, but the author is absolutely opposed to LLMs and is blind (losing his sight over time), leading to perhaps different sensory descriptions than you might expect but that doesn't mean they're written by a machine.
He has written a number of things about how he doesn't like AI picture descriptions for his screen reader, didn't appreciate a beta reader feeding his words into an LLM, that it helps him navigate the web but using it for writing seems likely to eliminate the human - and particularly disabled humans - from work. And he finds it rather soulless.
He explicitly states he doesn't use AI in any of his work.
Here are all his posts tagged with AI.
https://sightlessscribbles.com/posts/tags/ai/
And
Robert Kingett: "I read a blog post the other d…" - Cane and Able https://caneandable.social/@WeirdWriter/116636328695326100
I do not believe your criticisms are correct. Whether he's mimicking AI tropes on purpose or writing in his usual style (I'm not familiar enough with his work)as he said it's worth noting that those tropes come from human writers.
I was curious whether this was just their style they've always had, or something they only very recently picked up. I figure if I could find some of their work from before AI was at its current level that was detected as AI or had the same writing style, then it would mean I was wrong about it being AI. I have not found a single piece of writing by them before 2025 so far that was detected as more than 0% AI written, or appeared to me as having similar issues or writing style (where every sentence in a row overuses metaphors and slam-poetry mic drop moments). I subscribed to Pangram so I could use their service without limits. I did judge the writing style of each post myself first before putting it into Pangram, and Pangram always matched my judgment. (They do have some recent posts too that appeared to be 0% AI.) I've tested on several excerpts of their books on their site too.
I searched for other discussions about this article and articles from their domain, and I found this HN thread on a different recent post of theirs where some commenters stated they found the AI style annoying. One commenter linked to a Pangram analysis rating it as 100% AI.
I think they're either being extremely sanctimonious about their own AI use, knowing their anti-AI audience is unlikely to be able to recognize it, or maybe they've recently fallen in love with the style and carefully re-engineered their writing sentence by sentence to be indistinguishable from it while complaining about it.
Also they do have a recent Mastodon post bragging that they found they were better at prompting LLMs than most LLM users. I kind of get the impression they're a little too interested in LLMs for someone who avoids them, but to be clear my judgment above isn't depending on this vague feeling off of their posts.
I think it's most likely that he's writing this way intentionally due to the nature of this story, but again I don't know his style, and I don't trust AI "graders" (the link you shared doubts Pangram specifically in fact and makes the point that detectors will cause false accusations.)
I saw his post about covering an LLM start up event for the purpose of fucking around with the prompts on stage (maybe unrealistic). He also just said he was better at the prompting than his neighbors who told the LLM "don't hallucinate". Which, like, I also as a hater could have done better than that. There was no indication this came from his frequent use of them or anything. The post on HN doesn't have that many comments complaining about it being AI or complaining about the style or at least they're not in the top 60% of comments... Update: I read the whole thing it's a very small number of people complaining about AI use. Again maybe I'm wrong about him but I'd have had to search for his name and AI style to even get that HN page to come up and see those comments because they are in the last like 5% of the page.
Mostly the critics are just calling it fake and citing that he writes fiction while ignoring the nonfiction tag. Or saying he blamed her too much or whatever. It's all very reasonable criticism.
I won't be shocked if there are some AI style complaints, and idk how HN upvotes work but they didn't float to the top half*
To me he seems like many authors who dislike AI while he's also particularly passionate about tech - posts about what I will call programming stuff (idk GitHub and stuff) and Linux, and, particularly about accessibility in tech. He's a disabled writer with a journalism background, and he posts a lot on techy topics he cares about as a techy guy online all the time. He'd fit in here well IMO.
I just don't see it as likely that he's lying about all this and secretly using the LLMs to bait people. Especially as a queer dude who is agreeing with posts calling Andy Weir an idiot for thinking that writing doesn't have social commentary.
I think it's fair to criticize the work, I just don't think it's likely to be AI generated or bait given the context. But like maybe he's a big old liar, idk him. However, I think this nuanced conversation is very different from "this is a joke right?".
*Oof the ableism in some of this is gross. I'm reading the whole thread in the interest of fairness. But ick.
Well, keep in mind that AI was trained on, basically, average writing. It learned bad writing from us, so when you see bad writing there's still a very strong chance it was just a regular human bad writer who wrote it.