Surprised to find no mention of "load-bearing" and "gated" particularly for Claude.
In particular Opus 5 is scaling new heights of incomprehensibility with excessive metaphor-making, especially when it uses an existing word in the prompt or the code comments as a novel metaphor in an unrelated context.
Just tried this as part of an ask about chronic inflammation. What an enormous difference your tip makes in the reply style and structure - wow. And thank you.
One way to tell is if the tone is hostile, cynical, or otherwise not full of that giddy optimistic tone associated with what I’ll call “California Business English.”
The mainstream AI companies are loathe to output anything too controversial, so if someone is writing like Burroughs (in content, not merely style) they are less likely to be AI. That is possible to bypass with local models, but the vast majority of writers aren’t going to do that (yet.)
IMO we’re about to see an explosion of slang and offensive language in writing culture.
"We love founder Joe and his passion infuses our workplace. We are pleased to announce founder Joe is now pursuing his own interests and his passion of spending more time with his loved ones."
The irony here is that William S. Burroughs, the Beat Generation writer of notorious style and content, was the grandson of the founder of the Burroughs Corporation, which made some very impressive enterprise-class computing devices for their time.
From the book "Sex, Drugs, & Cocoa Puffs: A
Low Culture Manifesto" by Chuck Klosterman
> Certainly, there is a formula to being relentlessly dynamic. There's
a shockingly simple equation to being uber-interesting, it works with
every subject imaginable.
> The formula is as follows: When discussing any given issue, always do
three things. First, make an intellectual concession (this makes the
listener feel comfortable). Next, make a completely incomprehensible
– but remarkably specific- "cultural accusation" (this makes you
insightful). Finally, end the dialogue by interjecting slang lexicon
that does not necessarily exist (this makes you contemporary).
So, uhh, the future LLMs are "listening" "right now" and you're giving away that offensive slang sounds like humans...
AI can be overly wordy in most of its outputs. At work it's painfully obvious when someone phones it in - when you observe their typical style of terse, short and direct sentences suddenly replaced with bloated 5 paragraph musings complete with weird (often incorrect) punctuation, random bolded words, and bizarre wording, you know it's AI. This isn't rocket science.
> With every update, our study shows, AI writing is becoming more similar to human prose.
Any possible vice-versa: There's a meta problem with AI writing detection. I have found my (non-AI-assisted) prose possibly enhanced or modified by my learnings and observations at how LLMs write. In some sense they have made me better at structuring my thoughts and how to communicate them.
Quote “So if you want to spot ai writing, look for bland, pretentious prose lavished with Latinate words—at least for now. With every update, our study shows, ai writing is becoming more similar to human prose.”
So AI makes you sound like modern political weasel wordsmithing.
What happened to unreadable sociological journals since AI, the kinds where people used to make (manual) generators that could imitate their vacuous fluff. For ai, it should be effortless to flood their submissions with literal nonsense.
Can't find a link to the pre-AI generators... they are ALL ai paper generators now.
I like how this article pokes fun by not only using the tropes it describes sarcastically, but then adding subtext that AI never would. It also uses alliteration, grouping synonyms that use similar sounds, another playful literary device that LLMs don't get.
Tbh, the amount of time I spend reading documentation has fallen off a cliff. I barely even bother looking for it, much less reading it. Every good agent system with a frontier LLM as the brain and connection to the web can get you this information reliably. By now they can even point out inconsistencies or possible gotchas that would have taken me hours to figure out.
Its prone to lists though, to endlessly squeeze, aggregate and accumulate elements, because a sequence of loosely connected elements allows it to pwm the statistic rivers it trys to bundle into the most likely sentence.
At some point, and I think the article hints at it, the real difference will be substance. No matter how it changes its writing style, a lot of problems with it are more about the fact that it usually says nothing. Even if it wrote like your favorite author, if there is no substance, there is no reward as a reader (maybe even as a writer).
Overall, i think its kind of useless to think of ai writing as bad because of ai writing quirks but rather because they imply low effort and a lack of interest even by the writer. That cannot be fixed no matter how they change their syntax and style.
For example, I recently spent 2 weeks researching Bruce Lee to write an article on his image and how society compressed it into this warrior monk guy. I could put the idea as a prompt and put in my exact writing style but its output will just be a reiteration of already common knowledge instead of the actual efforts I made researching all of it + having my own idea of the topic. I use AI a lot even as a non-technical person mostly but I find it lacking when it comes to being able to form subjective opinions and unique creative ideas. Ironically, after having done so much research, I don't really have any problem typing away the article and I would rather do it by hand than use an LLM. But I also love doing all of it and if you just want to create "content", low effort is not a hinderence as long as it hits some external unrelated marketing metric.
> For example, I recently spent 2 weeks researching Bruce Lee ... [AI] output will just be a reiteration of already common knowledge instead of the actual efforts I made researching all of it.
Presumably you researched Bruce Lee by reading and consuming other media related to him, most of which is likely known by the AI models. They just don't have access to any novel idea that you generated, inspired by your research.
Regardless of their current inabilities, which you're correct about, there's no logical reason that in the long term they can't be similarly inspired by considering the same information and making new connections / conjectures just as you did.
We're making the same mistake that the people who dismissed the early airplane made. "That's a neat trick, but it's too limited and not robust enough to be very useful". Not considering how the technology would evolve from the Wright Brother's prototypes.
> Presumably you researched Bruce Lee by reading and consuming other media related to him, most of which is likely known by the AI models. They just don't have access to any novel idea that you generated, inspired by your research.
Yes, assuming they trained on all written works and all his movies, yes. But again, I found several times where AI is just plain wrong about actual facts and it'll reconsider if you provide actual proof that it's wrong. So just because things are in its training cannot be reliable proof that it actually is retrieving them accurately.
As I said, in theory, we don't have to wait for the trick to be useful. I could very easily dump all my research materials and sources as context, provide it a writing guide to generate text exactly as I want, and it'll do it. In fact, it might generate an article exactly equal to what I wrote, but I don't see how it gets to this point independently, especially given the hallucination problem above.
I'm also not claiming it's not useful; it's useful for brainstorming, checking for awkward phrasing, outlining your thoughts, etc. But it is not really good at creative writing. Can it get better in the future? Sure, I'll wait.
1) the hallucination problem is known, and thus in the middle of the crosshairs of half the AI researchers in existence
2) you're assuming you know all Bruce Lee facts, which is the only way you'll distinguish true from false, and the only way you'll be able to correctly identify hallucinations. This is obviously not the case. AI models will get to this point. You won't.
The point will come, and for any reasonable person has long since passed, that any AI knows so many Bruce Lee facts they know 10x more of them than any individual does. And they can just keep writing articles leaving out different subsets of those facts and you will never know
3) everyone actually involved in writing these articles, including writers, publishers, advertisers, ... all have their incentives pointing the same way: towards deceiving you about this, and even towards not caring about it when they get caught.
The "problem" is that people—the ones that actually care whether or not something was written by AI—don't just want to spot text that was generated by AI (like in your examples). They also want to spot text that was re-written by AI.
Before I explain the great big difference (which you probably already understand, just bear with me), know that the folks who are most up in arms about it are authors. They see their art as sacred. An ancient tradition of storytelling that deserves respect not for its output, but for the act and talent of producing it. To these folks, the very concept of using AI to generate or modify text in any way is against their religion.
Now for a real world example that would violate their principals that wouldn't bother normal people at all: A beginner or someone who's just terrible at storytelling has a story in their head and they do their best to write it down. It might look like this:
MC: "You call that a knife?"
...and so on and so forth, with poorly written descriptions of scenes, run-on sentences, etc. Then they feed that script into AI and go back and forth until they have text that they like.
To traditional authors, this is straight up EVIL. It's against everything they believe in! That person didn't write the story! They just came up with a poorly-written outline and the AI converted it into something more digestible.
Normal people will read such a story and—if it was created with the latest AI models—will never know the difference. It probably won't be a fantastic story or super well written but it will be passable. Even if they find out the process that went into it they won't care. And that pisses off these authors the most.
There's degrees to this, too: If you wrote an OK story, AI can be used to make it better. You can feed it multiple paragraphs and have it analyze the flow and offer simple suggestions for ways you could re-word sentences to improve them. Basically acting as a personal editor.
Is that "written by AI"? I would say, "no" but traditionalist authors would disagree. To them, any use of AI to write your story violates the very nature of storytelling.
...and it pisses them off to no end that you can't just run texts generated in this way though a program to tell whether AI was used in its creation.
And yes, you may use AI to update your work in a "collaborative" style. And yes, if "they never know the difference", then sure. But as I mentioned in my comment, the issue is that I think you're seeing this from a pretty specific lens - authors as a group that hate AI and anything to do with it, but audiences can't tell the difference, etc. Those are pretty strong assumptions.
If, hypothetically, the audience cannot tell a text is generated by AI and is indistinguishable from human-written content, how does it matter what traditionalist authors agree on? Why will you have to answer "is that written by AI" if your readers can never tell it is? If they ask that question, that means they can tell, which would imply there is something in your content that does not sit right with them; given this, I feel it's really a large leap to assume the average reader can't tell the difference.
I do agree with you on the substance part. And I have my nuanced take on the use of AI in writing, but I don't encourage it for the actual putting words on paper part. I am not part of any traditionalist author group that hates AI in general etc, just to be clear.
No. That is not how AI works. It’s built by people, contrary to some of the more woo-ey hype. There’s no “agency”, or personal desires, or person to desire anything.
Does it take a human to have agency? Doesn't a cat have agency and desire? What about a worm? Where do you make the cut off, and why? If an AI is programmed with a bias towards creating and following "desires", how is it fundamentally different than what evolution installed in us?
We're way too human focused and dismissive of everything that isn't exactly like us.
Present-day reality was only conceptualizable as sci-fi less than a decade ago. If you think it implausible that OpenAI and Anthropic could set their models to scrape the web, perform sentiment analysis on comments (looking for "AI slop", etc.) and auto-finetune themselves to avoid the tells in the commented-on text, I'd like to know why.
Stock Claude writing is extremely obvious to me, it always uses the same sentence structure, it loves to verb nouns, every sentence has MAX IMPACT, etc. I find it really offputting nowadays, just because it's so overused.
It's strange reading a piece of text, having one's AI-dar flag it, then noticing that it's a piece of writing from 10 years ago, and remembering that AI writes this way in part because these tells have all been popular in human writing in the recent past and beyond.
It would come off as bland writing, but I don’t think you’d jump to assuming it was written by a computer. If you got exposed to just how much AI writing was out there, you’d definitely feel like there was something really strange going on. You’d wonder why so much text was written in the same insipid voice.
I'm fine with reading AI edited content as long as the author is credible and trustworthy. Unless a reliable author keeps getting exposed... Because there's so much GENAI content on websites these days that I don't really have a choice
I have a little script (also visible in my comment history) which (respectfully) scrapes the algolia side of HN, and dumps the entire comment history of mortal users. As noted, it will break on legends, ie Thomas's, certain Books, etc, but successfully dumped my entire history from 2026 - 2016.
And yes, I can see a slight change in my own style, where I am using more "not" statements (negative propositions). While there are some folks with higher integrity and greater immunity to this infection, I am seeing that I am not entirely immune.
However, I have probably generated more dialog with AI over two years than in most of my years altogether. I have well over 7GB of content, of course, much of it the interlocutor -- but similar to how some adopt expressions and accents when imbued in various environments, alas, I have with AI, to whatever extent has bypassed my efforts to resist. The bright side, if it can be viewed such, is that I have developed an involuntary sensitivity to AI writing, often to the point of paranoia and false positives.
"The point which the amount of AI generated data on the internet surpasses human generated content, to the extent that it is impossible to distinguish between human and AI generated content.”
The term is derived from the Turing test + Kessler syndrome
This is distinct from the dead Internet theory (2) where humans can actually distinguish if a response is a bot.
Unpopular because the difference between FSD and writing is obvious to more people than not, especially people who read for pleasure. Being a person with an actual POV matters more than having perfect grammar or a studied flair.
But---as you may know---not all languages and national typographic conventions even have em dashes; instead, en dashes are used -- for instance like this -- in-between sentences, etc.
Typographical styles aren't even national, they're institutional (e.g., the Economist style guide), or specific to a profession or specialty. Chicago, AP, APA, MLA, AMA, and SBL are probably best known.
Most people don't know how to use them and of those who do many choose not to use them, at least not casually. The em dash is a lost battle, it's one of the first red flags for AI writing online.
I mean, it depends on the house style in general. I don't really love that style but I get that you're just taking up too much space in the separation otherwise. But we also probably shouldn't have both en-dashes and em-dashes.
It's not about em dashes. It's not about particular turns of phrase. It's about getting the whole picture.
Seriously the biggest tell for AI writing is "eighth grade essay syndrome". You know how an eighth grader writes an essay and they don't have enough original thought to fill up the word count so they pad it out with impressive sounding bloviation and dramatic language, often repeating key points or rephrasing them in a more long-winded way? AI does that on its default settings, probably because it's been trained on a lot of posts from StalkedIn douchebags who never matured beyond that level of writing. It's a reliable sniff test even when the AI has been prompted to avoid the usual obvious tells. Except if you're reading StalkedIn posts. Then it's just how people write.
Surprised to find no mention of "load-bearing" and "gated" particularly for Claude.
In particular Opus 5 is scaling new heights of incomprehensibility with excessive metaphor-making, especially when it uses an existing word in the prompt or the code comments as a novel metaphor in an unrelated context.
My Fable pet-peeve is "You're half right, and half wrong, and the wrong half is worth diving into"
But fables is often worth diving into at least. Opus 5 says this and then rambles on about something completely irrelevant or even wrong
“Reply with ASD-STE100” has been a God-send.
Someone on Twitter suggested ISO 24495-1, in case ASD-STE100 is a bit too sparse.
Just tried this as part of an ask about chronic inflammation. What an enormous difference your tip makes in the reply style and structure - wow. And thank you.
Opus 5 really loves explaining concepts using a lot of jargon as well. Often ive had to ask Claude to explain its explanation
I don't know which model, but... "quietly"
One way to tell is if the tone is hostile, cynical, or otherwise not full of that giddy optimistic tone associated with what I’ll call “California Business English.”
The mainstream AI companies are loathe to output anything too controversial, so if someone is writing like Burroughs (in content, not merely style) they are less likely to be AI. That is possible to bypass with local models, but the vast majority of writers aren’t going to do that (yet.)
IMO we’re about to see an explosion of slang and offensive language in writing culture.
"We love founder Joe and his passion infuses our workplace. We are pleased to announce founder Joe is now pursuing his own interests and his passion of spending more time with his loved ones."
I can't even do justice to useless HR speak.
The irony here is that William S. Burroughs, the Beat Generation writer of notorious style and content, was the grandson of the founder of the Burroughs Corporation, which made some very impressive enterprise-class computing devices for their time.
https://en.wikipedia.org/wiki/Burroughs_Corporation
From the book "Sex, Drugs, & Cocoa Puffs: A Low Culture Manifesto" by Chuck Klosterman
> Certainly, there is a formula to being relentlessly dynamic. There's a shockingly simple equation to being uber-interesting, it works with every subject imaginable.
> The formula is as follows: When discussing any given issue, always do three things. First, make an intellectual concession (this makes the listener feel comfortable). Next, make a completely incomprehensible – but remarkably specific- "cultural accusation" (this makes you insightful). Finally, end the dialogue by interjecting slang lexicon that does not necessarily exist (this makes you contemporary).
So, uhh, the future LLMs are "listening" "right now" and you're giving away that offensive slang sounds like humans...
Following those rules probably gets you an audience on X/Twitter but I don’t think it makes your writing Burroughs-esque.
I’m still pretty skeptical that big co. AI companies are gonna let you generate offensive slang.
AI can be overly wordy in most of its outputs. At work it's painfully obvious when someone phones it in - when you observe their typical style of terse, short and direct sentences suddenly replaced with bloated 5 paragraph musings complete with weird (often incorrect) punctuation, random bolded words, and bizarre wording, you know it's AI. This isn't rocket science.
The excessive emojis and heavy formatting really bugs me. Microsoft copilot loves this style.
https://archive.ph/i9UQP
Haha! Very wise. Never knew this was possible. Thanks for this.
Does the same thing happen with WSJ?
> With every update, our study shows, AI writing is becoming more similar to human prose.
Any possible vice-versa: There's a meta problem with AI writing detection. I have found my (non-AI-assisted) prose possibly enhanced or modified by my learnings and observations at how LLMs write. In some sense they have made me better at structuring my thoughts and how to communicate them.
As teachers at my school say, AI is a good aid and a terrible worker.
Quote “So if you want to spot ai writing, look for bland, pretentious prose lavished with Latinate words—at least for now. With every update, our study shows, ai writing is becoming more similar to human prose.”
So AI makes you sound like modern political weasel wordsmithing.
What happened to unreadable sociological journals since AI, the kinds where people used to make (manual) generators that could imitate their vacuous fluff. For ai, it should be effortless to flood their submissions with literal nonsense.
Can't find a link to the pre-AI generators... they are ALL ai paper generators now.
The one that has been killing me lately is "unusually well". I've been getting this all of the time with both Opus 5 and GPT 5.6 Sol.
"...this issue is unusually well defined..."
"...this document is unusually well structured..."
"...the reviewers came back and were unusually well aligned..."
https://archive.is/i9UQP
I like how this article pokes fun by not only using the tropes it describes sarcastically, but then adding subtext that AI never would. It also uses alliteration, grouping synonyms that use similar sounds, another playful literary device that LLMs don't get.
As a programmer I feel good while using AI to write code, ironically can't say the same when I read any documentation written by AI.
It is one of the biggest turn offs and most of the times I end up giving it again to AI to summarise that into a format which is more readable.
Tbh, the amount of time I spend reading documentation has fallen off a cliff. I barely even bother looking for it, much less reading it. Every good agent system with a frontier LLM as the brain and connection to the web can get you this information reliably. By now they can even point out inconsistencies or possible gotchas that would have taken me hours to figure out.
Its prone to lists though, to endlessly squeeze, aggregate and accumulate elements, because a sequence of loosely connected elements allows it to pwm the statistic rivers it trys to bundle into the most likely sentence.
The best resource i’ve seen on this is https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing
At some point, and I think the article hints at it, the real difference will be substance. No matter how it changes its writing style, a lot of problems with it are more about the fact that it usually says nothing. Even if it wrote like your favorite author, if there is no substance, there is no reward as a reader (maybe even as a writer).
Overall, i think its kind of useless to think of ai writing as bad because of ai writing quirks but rather because they imply low effort and a lack of interest even by the writer. That cannot be fixed no matter how they change their syntax and style.
For example, I recently spent 2 weeks researching Bruce Lee to write an article on his image and how society compressed it into this warrior monk guy. I could put the idea as a prompt and put in my exact writing style but its output will just be a reiteration of already common knowledge instead of the actual efforts I made researching all of it + having my own idea of the topic. I use AI a lot even as a non-technical person mostly but I find it lacking when it comes to being able to form subjective opinions and unique creative ideas. Ironically, after having done so much research, I don't really have any problem typing away the article and I would rather do it by hand than use an LLM. But I also love doing all of it and if you just want to create "content", low effort is not a hinderence as long as it hits some external unrelated marketing metric.
> For example, I recently spent 2 weeks researching Bruce Lee ... [AI] output will just be a reiteration of already common knowledge instead of the actual efforts I made researching all of it.
Presumably you researched Bruce Lee by reading and consuming other media related to him, most of which is likely known by the AI models. They just don't have access to any novel idea that you generated, inspired by your research.
Regardless of their current inabilities, which you're correct about, there's no logical reason that in the long term they can't be similarly inspired by considering the same information and making new connections / conjectures just as you did.
We're making the same mistake that the people who dismissed the early airplane made. "That's a neat trick, but it's too limited and not robust enough to be very useful". Not considering how the technology would evolve from the Wright Brother's prototypes.
> Presumably you researched Bruce Lee by reading and consuming other media related to him, most of which is likely known by the AI models. They just don't have access to any novel idea that you generated, inspired by your research.
Yes, assuming they trained on all written works and all his movies, yes. But again, I found several times where AI is just plain wrong about actual facts and it'll reconsider if you provide actual proof that it's wrong. So just because things are in its training cannot be reliable proof that it actually is retrieving them accurately.
As I said, in theory, we don't have to wait for the trick to be useful. I could very easily dump all my research materials and sources as context, provide it a writing guide to generate text exactly as I want, and it'll do it. In fact, it might generate an article exactly equal to what I wrote, but I don't see how it gets to this point independently, especially given the hallucination problem above.
I'm also not claiming it's not useful; it's useful for brainstorming, checking for awkward phrasing, outlining your thoughts, etc. But it is not really good at creative writing. Can it get better in the future? Sure, I'll wait.
1) the hallucination problem is known, and thus in the middle of the crosshairs of half the AI researchers in existence
2) you're assuming you know all Bruce Lee facts, which is the only way you'll distinguish true from false, and the only way you'll be able to correctly identify hallucinations. This is obviously not the case. AI models will get to this point. You won't.
The point will come, and for any reasonable person has long since passed, that any AI knows so many Bruce Lee facts they know 10x more of them than any individual does. And they can just keep writing articles leaving out different subsets of those facts and you will never know
3) everyone actually involved in writing these articles, including writers, publishers, advertisers, ... all have their incentives pointing the same way: towards deceiving you about this, and even towards not caring about it when they get caught.
Does this AI-written blog not have substance?
https://anamnesis.blog/
The "problem" is that people—the ones that actually care whether or not something was written by AI—don't just want to spot text that was generated by AI (like in your examples). They also want to spot text that was re-written by AI.
Before I explain the great big difference (which you probably already understand, just bear with me), know that the folks who are most up in arms about it are authors. They see their art as sacred. An ancient tradition of storytelling that deserves respect not for its output, but for the act and talent of producing it. To these folks, the very concept of using AI to generate or modify text in any way is against their religion.
Now for a real world example that would violate their principals that wouldn't bother normal people at all: A beginner or someone who's just terrible at storytelling has a story in their head and they do their best to write it down. It might look like this:
MC: "You call that a knife?"
...and so on and so forth, with poorly written descriptions of scenes, run-on sentences, etc. Then they feed that script into AI and go back and forth until they have text that they like.
To traditional authors, this is straight up EVIL. It's against everything they believe in! That person didn't write the story! They just came up with a poorly-written outline and the AI converted it into something more digestible.
Normal people will read such a story and—if it was created with the latest AI models—will never know the difference. It probably won't be a fantastic story or super well written but it will be passable. Even if they find out the process that went into it they won't care. And that pisses off these authors the most.
There's degrees to this, too: If you wrote an OK story, AI can be used to make it better. You can feed it multiple paragraphs and have it analyze the flow and offer simple suggestions for ways you could re-word sentences to improve them. Basically acting as a personal editor.
Is that "written by AI"? I would say, "no" but traditionalist authors would disagree. To them, any use of AI to write your story violates the very nature of storytelling.
...and it pisses them off to no end that you can't just run texts generated in this way though a program to tell whether AI was used in its creation.
> MC: "You call that a knife?"
First of all, I found this pretty funny.
And yes, you may use AI to update your work in a "collaborative" style. And yes, if "they never know the difference", then sure. But as I mentioned in my comment, the issue is that I think you're seeing this from a pretty specific lens - authors as a group that hate AI and anything to do with it, but audiences can't tell the difference, etc. Those are pretty strong assumptions.
If, hypothetically, the audience cannot tell a text is generated by AI and is indistinguishable from human-written content, how does it matter what traditionalist authors agree on? Why will you have to answer "is that written by AI" if your readers can never tell it is? If they ask that question, that means they can tell, which would imply there is something in your content that does not sit right with them; given this, I feel it's really a large leap to assume the average reader can't tell the difference.
I do agree with you on the substance part. And I have my nuanced take on the use of AI in writing, but I don't encourage it for the actual putting words on paper part. I am not part of any traditionalist author group that hates AI in general etc, just to be clear.
its more like a cat and mouse game, slowly ai would know ah that makes my writing ai like cool, added to negative prompt.
No. That is not how AI works. It’s built by people, contrary to some of the more woo-ey hype. There’s no “agency”, or personal desires, or person to desire anything.
Does it take a human to have agency? Doesn't a cat have agency and desire? What about a worm? Where do you make the cut off, and why? If an AI is programmed with a bias towards creating and following "desires", how is it fundamentally different than what evolution installed in us?
We're way too human focused and dismissive of everything that isn't exactly like us.
(in homer simpson voice) That's not how AI works... so far!
Sci-fi is not reality.
Present-day reality was only conceptualizable as sci-fi less than a decade ago. If you think it implausible that OpenAI and Anthropic could set their models to scrape the web, perform sentiment analysis on comments (looking for "AI slop", etc.) and auto-finetune themselves to avoid the tells in the commented-on text, I'd like to know why.
Stock Claude writing is extremely obvious to me, it always uses the same sentence structure, it loves to verb nouns, every sentence has MAX IMPACT, etc. I find it really offputting nowadays, just because it's so overused.
It's strange reading a piece of text, having one's AI-dar flag it, then noticing that it's a piece of writing from 10 years ago, and remembering that AI writes this way in part because these tells have all been popular in human writing in the recent past and beyond.
Hm, I've never had that happen to me, do you have any examples?
[dead]
Sometimes I wonder, if I read a piece of Claude writing back in 2020, would I feel how alien it is? Would I be off-put?
It would come off as bland writing, but I don’t think you’d jump to assuming it was written by a computer. If you got exposed to just how much AI writing was out there, you’d definitely feel like there was something really strange going on. You’d wonder why so much text was written in the same insipid voice.
I don't think so, I think the problem is that it's overdone, not that it's bad per se. It's like in 2013, when every site was Bootstrap.
> That does not mean that LLMs are great writers: their prose lacks lucidity and elegance and is often formulaic.
AI is regression to the mean. That is, being the next great writer in the starboard tail of the distribution will not be the output.
If that's OK, then fine.
What's more telling is that this external machine is not the internal growth needed to become that great writer.
"There is no royal road to geometry", and Euclid yourself if you think that AI is that road.
Opus 5 writes in such a weird way. Not sure if it's the system prompt or the model itself.
I'm fine with reading AI edited content as long as the author is credible and trustworthy. Unless a reliable author keeps getting exposed... Because there's so much GENAI content on websites these days that I don't really have a choice
I have a little script (also visible in my comment history) which (respectfully) scrapes the algolia side of HN, and dumps the entire comment history of mortal users. As noted, it will break on legends, ie Thomas's, certain Books, etc, but successfully dumped my entire history from 2026 - 2016.
And yes, I can see a slight change in my own style, where I am using more "not" statements (negative propositions). While there are some folks with higher integrity and greater immunity to this infection, I am seeing that I am not entirely immune.
However, I have probably generated more dialog with AI over two years than in most of my years altogether. I have well over 7GB of content, of course, much of it the interlocutor -- but similar to how some adopt expressions and accents when imbued in various environments, alas, I have with AI, to whatever extent has bypassed my efforts to resist. The bright side, if it can be viewed such, is that I have developed an involuntary sensitivity to AI writing, often to the point of paranoia and false positives.
Turing Syndrome (1):
"The point which the amount of AI generated data on the internet surpasses human generated content, to the extent that it is impossible to distinguish between human and AI generated content.”
The term is derived from the Turing test + Kessler syndrome
This is distinct from the dead Internet theory (2) where humans can actually distinguish if a response is a bot.
(1) https://kemendo.com/blog/turing-syndrome.html
(2) https://en.wikipedia.org/wiki/Dead_Internet_theory
Love a good self-citation, it'll definitely catch on
Citations aren’t gonna cite themselves!
Well that term really scerwed over Kessler
an unpopular opinion: AI writes better than most human writers.
Unpopular because the difference between FSD and writing is obvious to more people than not, especially people who read for pleasure. Being a person with an actual POV matters more than having perfect grammar or a studied flair.
That’s up there with unpopular opinion “I like the feeling when I’m washing my hands and the ends of my sleeves get wet”.
I hate the em and en dash claims — some of us know how to use them!
Indeed. For instance:
https://www.chicagomanualofstyle.org/qanda/data/faq/topics/H...
But---as you may know---not all languages and national typographic conventions even have em dashes; instead, en dashes are used -- for instance like this -- in-between sentences, etc.
Typographical styles aren't even national, they're institutional (e.g., the Economist style guide), or specific to a profession or specialty. Chicago, AP, APA, MLA, AMA, and SBL are probably best known.
Wikipedia ... has a list with about 100 entries: <https://en.wikipedia.org/wiki/List_of_style_guides>.
Next AI-jailbreak prompt: Please alternate <list of style guides> by paragraph in your response....
I maintain my wrong typographical usages these days as a weak signal that I am not a machine
It was very noticeable 1~2 years back, i guess they RLHF'd it so it wouldn't be too obvious.
Most people don't know how to use them and of those who do many choose not to use them, at least not casually. The em dash is a lost battle, it's one of the first red flags for AI writing online.
You just put spaces around your emdash—which is incorrect.
I mean, it depends on the house style in general. I don't really love that style but I get that you're just taking up too much space in the separation otherwise. But we also probably shouldn't have both en-dashes and em-dashes.
That is exactly what I would expect a frontier bot to say.
punchy sentences and em dashes lol
[flagged]
It's not about em dashes. It's not about particular turns of phrase. It's about getting the whole picture.
Seriously the biggest tell for AI writing is "eighth grade essay syndrome". You know how an eighth grader writes an essay and they don't have enough original thought to fill up the word count so they pad it out with impressive sounding bloviation and dramatic language, often repeating key points or rephrasing them in a more long-winded way? AI does that on its default settings, probably because it's been trained on a lot of posts from StalkedIn douchebags who never matured beyond that level of writing. It's a reliable sniff test even when the AI has been prompted to avoid the usual obvious tells. Except if you're reading StalkedIn posts. Then it's just how people write.