My aha moment with AI was when I heard that it exists to please you rather than provide fact or evidence. Originally that helped me understand it is not a search engine. It is a storytelling engine, telling ‘stories’ to please people and now pleasing specific communities - as you say the substack style.
Side note: The structure you outline is a shorter version of that 1970s paper by Davis ‘That’s Interesting!’ One of my favourite papers on non-fiction writing and what seems to be the Gladwell style that AI is now refining / reducing.
And thank you for this. I will continue to seek the strange and fail, humanly so.
Exactly this - Richard Osman made a similar point when he said that AI doesn’t have a point of view, and that really comes across in the art it produces.
And this is precisely why the tend to use it as your therapist is SO dangerous - there are enough narcissists in the world already! AI is busy making clones of us, by hook and crook…
That's a brilliant point about it trying to please. I'd heard that before, but never thought about applying that to actual groups, as in 'the substack voice.' But it makes complete sense now you've said it, and also why some people love it. I guess that's why Will's prompt 'works' - so interesting.
Hadnt thought of it this way, and this fact makes me feel icky. Humans are extremely skilled in rejecting reality and substituting their own. We dont need help from the tech gods to lie to ourselves.
Tons of people already spend their days avoiding life. Things are going to get much ickier the next few years.
Totally! These comments really get me. As do the ones I'm pretty sure they were actually written by a human who doesn't know they write like the LinkedIn auto-suggest
Absolutely. I teach online and have learned that students don't even have to prompt gen AI for a response: you can program a bot to read posts and generate replies on its own. Reddit has whole subs dedicated to finding people to do that (engagement farming) for each other.
So interesting! I can’t exactly describe it, but I’ve noticed a certain rhythm to the AI stuff. Even if the words are different the rhythm is always the same and it’s so unsatisfying. My favourite writers all use rhythm in interesting ways. I think that’s one of the most enjoyable parts of reading! And maybe that can’t be replicated. Thanks for writing this
I asked it to improve the peroration of Churchill’s ‘never surrender’ speech and it came up with:
“We will press on, unwavering, until the very end. We will fight in France, upon the vast seas, and across the open oceans.
With growing confidence and unyielding strength, we will battle in the skies. We will defend our island, no matter the cost.
We will stand firm on the beaches, hold our ground on the landing fields, We will resist in the fields, in the streets, and on the towering hills. We will never surrender.”
Which is… awful, but really illustrates your point that it’s just can’t hear how words sound and resonate.
Unsatisfying...that is totally the word for it! Every reader hopes for connection, and with AI content that's impossible. Even if I *think* something is written by a human, if I can't find connection there, I'm out.
Yes I ha e noticed this too! I often think- do people not know , it sounds like AI what they’re putting out there? I find, there’s no way to hide AI writing.
I was going to write a post on this. Now I don't have to, as you've already covered the subject so eloquently! (I probably will still write it -- there are people using the "Substack style" on other platforms I visit as well). The thing that concerns me the most about this whole thing is that a lot of people don't seem to care that this dreck is either 100% AI or heavily-assisted by AI. Those comments you quoted... Ugh. 😩
Write it Stephanie. We need more stories to counter this trend. Otherwise people will be walking around imagining they’re deep and profound thinkers, taking themselves way too seriously because they read some AI slop.
I have been lately on a rampage of restacking stuff that is obviously AI like this and calling it out, both posts and notes. My brain almost broke when I found an essay---possibly one you referenced here, actually, I need to check---with 12k likes and people praising it as "vulnerable" and even, hilariously, "human." I've been in several conversations on here about it and I always point out various tells you've mentioned here (which are way more indicting than lots of em dashes). Anyway the point is that you have taken many ideas that have been swirling around my brain lately and articulated them beautifully and logically. This should be required reading for AI literacy going forward.
Also if you wrote this with AI and I'm getting double duped I'll throw myself off a bridge (I can tell you didn't)
Please PLEASE do this study! But don't just take it from a perspective of "what's wrong with these people." I think they are fulfilling some need from this "gruel" that is otherwise not being satisfied. What is that need? How can it be met in other ways? Does it even *need* to be met in other ways? Is it just as "wrong" as the person who enjoys playing a narrative video game is "wrong" for not reading a book instead?
I want to know more about what's driving the desire for these "Substack Style" essays, but I also want to know more about whether they are actually hurting anyone to have them exist (so long as they don't *actually* replace the real human essays with real human experiences for those who want them, just as we wouldn't want the artificial "best seller" book to replace the true "best seller" like you mentioned in the post of yours that brought me here!)
There is a kind of SS newsletter that surprises you by how much you enjoy it the further you are into it. Not the kind that stops you in your tracks with the title necessarily, but the kind that you being by skimming an eye over, and then find yourself cantering through, stifling cries of “yes! YES!”. 😂
So, I'm assuming you've heard of SudoWrite, which I have a friend who played with in, I'm going to say late 2021, early 2022, just to see what it was. At that point it could generate prose that was genuinely interesting, novel, unexpected, and followed the track of your initial input (usually a para), and I say this as a line and dev editor with over ten years paid, 25 years writing with publishing as end goal, and fifteen years industry or industry adjacent. Yes, I'm jaded. And I almost ended up in computational neuroscience.
SudoWrite is based on a single use case highly modified licensed GPT API. In 2022 it could hold the thread for about 5k words. I checked in earlier this year to see what's up with the company and the software.
Well, now they have an enterprise level subscription for over $2,000.00USD year aimed at novelists who put out multiple titles a year (it's in the copy) At this point it can infill entire chapters. (Side note, it also works well because you're encouraged to feed it enough training data to mimic you with enough verisimilitude that it's nearly impossible to tell you DIDN'T write the novel.)
If I wasn't but a poor freelance editor, I had an idea to buy the LUXE package, feed it all of Pynchon as training material (which is encouraged.) And see if it could plop out, well, something that sounds like Pynchon (good lord, even the newest rolling out GPT can't. I'm a formal experimentalist and tinkerer, I do weird things for fun.) It would be interesting, if only because when people think of LLM's and AI for writing, GPT is the go-to. Sudowrite is specifically designed from the ground floor for fiction. (Churn and Burn KDP authors are currently using Sudowrite, I'm positive, with the aims of putting out a thousand books in 2025. And I mean, more than one author has made this proclamation. I'm interested to see if any succeed.) Anyway, you know.
It's very interesting, I'll give it that much. I super look forward to that article (even if I have to buy a month to get it because my ass is poor right now, worth it.)
I just realised I don't use my name - is this me? Am I AI? I tell embarrassing stories about myself. Maybe I am advanced AI. I love hyphens. Apparently this is a sign, yet my sentences are short. Who am I???. I need to be more human, less worried about sounding like a dick. Thank you. (My name is Lucy).
THANK YOU FOR THIS. I’ve been thinking about and calling this out for a while now. I referred to it as “sentimental schlock,” but I love “gruel.” It’s funny because I never necessarily assumed it was all AI generated (although some of it very obviously is), but just a rise in that pseudo-deep speak that comes off so empty because it’s 1) so overdone and (as you said) 2) doesn’t sound how people actually speak. It’s also just frankly boring and really funny that no matter how many different variations of the same thing are posted, people still somehow manage to be (or pretend to be) so imbued with it (aka the fawning comments you referenced), which honestly genuinely rivals the actual content for what comes off less real and genuine lol
The first clue something is written by AI is when someone (probably using AI) re-posts it on notes and says "this is the best thing I have read on substack"
My aha moment with AI was when I heard that it exists to please you rather than provide fact or evidence. Originally that helped me understand it is not a search engine. It is a storytelling engine, telling ‘stories’ to please people and now pleasing specific communities - as you say the substack style.
Side note: The structure you outline is a shorter version of that 1970s paper by Davis ‘That’s Interesting!’ One of my favourite papers on non-fiction writing and what seems to be the Gladwell style that AI is now refining / reducing.
And thank you for this. I will continue to seek the strange and fail, humanly so.
Exactly this - Richard Osman made a similar point when he said that AI doesn’t have a point of view, and that really comes across in the art it produces.
And this is precisely why the tend to use it as your therapist is SO dangerous - there are enough narcissists in the world already! AI is busy making clones of us, by hook and crook…
That's a brilliant point about it trying to please. I'd heard that before, but never thought about applying that to actual groups, as in 'the substack voice.' But it makes complete sense now you've said it, and also why some people love it. I guess that's why Will's prompt 'works' - so interesting.
AI exists to shut people's brains off.
Hadnt thought of it this way, and this fact makes me feel icky. Humans are extremely skilled in rejecting reality and substituting their own. We dont need help from the tech gods to lie to ourselves.
Tons of people already spend their days avoiding life. Things are going to get much ickier the next few years.
The start of every prompt for me:
“Don’t be solicitous”
Thank you! This explains it. Now I understand why I keep getting the sensation that I’m in a literary equivalent of The Truman Show
Good morning, and in case I don't see ya, good afternoon, good evening, and good night!
Oh my gosh yes! Well said.
Its odd isn't it. All these people who want "grow" and create a "brand". Using AI to get rich and famous for writing you did not do.
I do wonder how many of the hundreds of viral comments are ALSO AI generated - and therefore the entire apparent viral success is fabricated.
A lot of these comments read like LinkedIn comments, which are incredibly spammy and often just the first thing LinkedIn auto-suggests
Totally! These comments really get me. As do the ones I'm pretty sure they were actually written by a human who doesn't know they write like the LinkedIn auto-suggest
Absolutely. I teach online and have learned that students don't even have to prompt gen AI for a response: you can program a bot to read posts and generate replies on its own. Reddit has whole subs dedicated to finding people to do that (engagement farming) for each other.
So interesting! I can’t exactly describe it, but I’ve noticed a certain rhythm to the AI stuff. Even if the words are different the rhythm is always the same and it’s so unsatisfying. My favourite writers all use rhythm in interesting ways. I think that’s one of the most enjoyable parts of reading! And maybe that can’t be replicated. Thanks for writing this
I asked it to improve the peroration of Churchill’s ‘never surrender’ speech and it came up with:
“We will press on, unwavering, until the very end. We will fight in France, upon the vast seas, and across the open oceans.
With growing confidence and unyielding strength, we will battle in the skies. We will defend our island, no matter the cost.
We will stand firm on the beaches, hold our ground on the landing fields, We will resist in the fields, in the streets, and on the towering hills. We will never surrender.”
Which is… awful, but really illustrates your point that it’s just can’t hear how words sound and resonate.
This is hilarious
I agree about the rhythm….there’s this cadence and the “ability” to make anything have this artificial, yet supposed to be deep, meaning.
Unsatisfying...that is totally the word for it! Every reader hopes for connection, and with AI content that's impossible. Even if I *think* something is written by a human, if I can't find connection there, I'm out.
Yes I ha e noticed this too! I often think- do people not know , it sounds like AI what they’re putting out there? I find, there’s no way to hide AI writing.
it’s so unsatisfying... ditto
I was going to write a post on this. Now I don't have to, as you've already covered the subject so eloquently! (I probably will still write it -- there are people using the "Substack style" on other platforms I visit as well). The thing that concerns me the most about this whole thing is that a lot of people don't seem to care that this dreck is either 100% AI or heavily-assisted by AI. Those comments you quoted... Ugh. 😩
Write it Stephanie. We need more stories to counter this trend. Otherwise people will be walking around imagining they’re deep and profound thinkers, taking themselves way too seriously because they read some AI slop.
I meant to come back to this comment, but this past week was busy, and it slipped my mind. I wrote the post! https://brokenet.substack.com/p/chatgpt-is-all-over-your-favourite
I have been lately on a rampage of restacking stuff that is obviously AI like this and calling it out, both posts and notes. My brain almost broke when I found an essay---possibly one you referenced here, actually, I need to check---with 12k likes and people praising it as "vulnerable" and even, hilariously, "human." I've been in several conversations on here about it and I always point out various tells you've mentioned here (which are way more indicting than lots of em dashes). Anyway the point is that you have taken many ideas that have been swirling around my brain lately and articulated them beautifully and logically. This should be required reading for AI literacy going forward.
Also if you wrote this with AI and I'm getting double duped I'll throw myself off a bridge (I can tell you didn't)
We have to do a study on the types of people who enjoy that word salad. It is dreadful. I couldn’t even read the full paragraphs in your piece.
Have they minimal exposure to art?
Are they too high on openness to experience?
Are they too high on neuroticism?
Too low on need for cognition?
I’m so curious as to who gets the aesthetic chills from gruel.
Please PLEASE do this study! But don't just take it from a perspective of "what's wrong with these people." I think they are fulfilling some need from this "gruel" that is otherwise not being satisfied. What is that need? How can it be met in other ways? Does it even *need* to be met in other ways? Is it just as "wrong" as the person who enjoys playing a narrative video game is "wrong" for not reading a book instead?
I want to know more about what's driving the desire for these "Substack Style" essays, but I also want to know more about whether they are actually hurting anyone to have them exist (so long as they don't *actually* replace the real human essays with real human experiences for those who want them, just as we wouldn't want the artificial "best seller" book to replace the true "best seller" like you mentioned in the post of yours that brought me here!)
There is a kind of SS newsletter that surprises you by how much you enjoy it the further you are into it. Not the kind that stops you in your tracks with the title necessarily, but the kind that you being by skimming an eye over, and then find yourself cantering through, stifling cries of “yes! YES!”. 😂
You're so right. The unexpected humam experience.
So heartfelt and true! Didn't expect to cry on Substack today.
Haha
So, I'm assuming you've heard of SudoWrite, which I have a friend who played with in, I'm going to say late 2021, early 2022, just to see what it was. At that point it could generate prose that was genuinely interesting, novel, unexpected, and followed the track of your initial input (usually a para), and I say this as a line and dev editor with over ten years paid, 25 years writing with publishing as end goal, and fifteen years industry or industry adjacent. Yes, I'm jaded. And I almost ended up in computational neuroscience.
SudoWrite is based on a single use case highly modified licensed GPT API. In 2022 it could hold the thread for about 5k words. I checked in earlier this year to see what's up with the company and the software.
Well, now they have an enterprise level subscription for over $2,000.00USD year aimed at novelists who put out multiple titles a year (it's in the copy) At this point it can infill entire chapters. (Side note, it also works well because you're encouraged to feed it enough training data to mimic you with enough verisimilitude that it's nearly impossible to tell you DIDN'T write the novel.)
If I wasn't but a poor freelance editor, I had an idea to buy the LUXE package, feed it all of Pynchon as training material (which is encouraged.) And see if it could plop out, well, something that sounds like Pynchon (good lord, even the newest rolling out GPT can't. I'm a formal experimentalist and tinkerer, I do weird things for fun.) It would be interesting, if only because when people think of LLM's and AI for writing, GPT is the go-to. Sudowrite is specifically designed from the ground floor for fiction. (Churn and Burn KDP authors are currently using Sudowrite, I'm positive, with the aims of putting out a thousand books in 2025. And I mean, more than one author has made this proclamation. I'm interested to see if any succeed.) Anyway, you know.
Just a thought.
Thanks Emil - I am planning on having a fiddle with Sudo and writing about it
It's very interesting, I'll give it that much. I super look forward to that article (even if I have to buy a month to get it because my ass is poor right now, worth it.)
One of the best essays I’ve read on AI so far.
Now I wonder: if AI generates average content—and readers love it—does that make them average readers?
Are we seeing a retreat from the hard words?
A flight into the comfort of the average?
I don’t know.
But I do know this: AI will never write McCarthy. Because McCarthy was never average. That’s where AI loses. And that’s the books and essays I read.
Thank you for writing.
All praise McCarthy. I aspire to write like him. Although AI might try, it would never have the soul.
Reading this felt like being held in a way I’ve never been held before. Thank you for your kind words.
LOL
I just realised I don't use my name - is this me? Am I AI? I tell embarrassing stories about myself. Maybe I am advanced AI. I love hyphens. Apparently this is a sign, yet my sentences are short. Who am I???. I need to be more human, less worried about sounding like a dick. Thank you. (My name is Lucy).
I love this so much! I don't use my name either, I guess because I'm shy. I jumped onto your posts and subscribed - they're delicious.
It would be hilarious if this was written by AI.
THANK YOU FOR THIS. I’ve been thinking about and calling this out for a while now. I referred to it as “sentimental schlock,” but I love “gruel.” It’s funny because I never necessarily assumed it was all AI generated (although some of it very obviously is), but just a rise in that pseudo-deep speak that comes off so empty because it’s 1) so overdone and (as you said) 2) doesn’t sound how people actually speak. It’s also just frankly boring and really funny that no matter how many different variations of the same thing are posted, people still somehow manage to be (or pretend to be) so imbued with it (aka the fawning comments you referenced), which honestly genuinely rivals the actual content for what comes off less real and genuine lol
The first clue something is written by AI is when someone (probably using AI) re-posts it on notes and says "this is the best thing I have read on substack"