These are all important points and I love the analogy. But there is an even bigger issue with having LLMs write for you:
Writing is thinking. Thinking and deciding. There have been many times when I start out writing something substantial - could be an email, a blog post, a software design document, anything - when my own views substantially changed during the writing process. Writing forces you to serialize your thoughts - and you can't always trust the gestalt.
Reviewing gives you the chance to ensure the arguments connect solidly, that references are accurate (even informal references) and gives you the time to consider counter-arguments you aren't addressing.
None of this matters much on LinkedIn, but it matters a lot in our work. You cannot outsource your understanding to AI. They are powerful tools but they do not have any human understanding - that isn't their optimization target.
(I work at Anthropic) I agree. I use an LLM to write my code, but I do all of my writing by hand, since it helps me think.
It reminds me of the transition over the last year from AI-assisted coding to AI doing all the coding. At first the code output wasn't good enough, and humans read and iterated on the code all day, so the details of the source code mattered. Now, the code is largely high quality and it meets a large set of guardrails we've set up over the years (linters, typecheckers, security checks, LLM-assisted code quality checkers), and it's just Claude working on the code, so the details matter less and engineers think a level or two up (machine code < assembly/bytecode < source code < conversation with agent < artifact with high level design).
I wonder if long form writing will go the way of code. You and the AI agree on an outline or other high level representation, then the LLM expands it into a document. But writing and coding are different enough in a number of ways that this is far from inevitable.
> Writing is thinking. Thinking and deciding. There have been many times when I start out writing something substantial - could be an email, a blog post, a software design document, anything - when my own views substantially changed during the writing process. Writing forces you to serialize your thoughts - and you can't always trust the gestalt.
I don't disagree, but I think it's often not appreciated how much there's other work to writing too.
The biggest one is that you have to communicate non-interactively to an unknown audience. Having to (literally) put it in someone else's assumed terms does help giving different perspectives into the matter, but doesn't necessarily help one's own thinking that much. Instead you have to do some of the reader's thinking for them.
You also have to spend time on textual matters like grammar and style and a lot of "unspoken rules", which aren't really about linearizing your thinking about the contents.
Not all writing is thinking and not all thinking is writing.
Strange! I would quite specifically highlight “having to do the reader’s thinking for them”, as well as more generally developing the skill to “communicate non-interactively with an unknown audience”, as extremely valuable upgrades to my thinking.
It's a kind of shared fiction of an imaginary person's thinking, really. The author writes "now, I know what you must be thinking", and hopes the reader will agree "OK, close enough".
I have been thinking over this comment for a while now. I think yes, you are correct that thought is involved in writing but I don't think it is possible to claim "all writing is thinking when done by a human". Like what type of thinking are you claiming here?
Because one can copy a text and write it down and that involves thinking in the sense that anything we do involves thinking fundamentally. But that thinking is different from thinking logically about a concept and writing it down which I think is where you are getting at.
The definition of writing and thinking is too broad in that sentence even though it does apply in several obvious cateogires within that at different levels.
And also "writing helps us think about the world" is too broad again. Why? Why does me writing "apt apt apt apt apt apt apt" help me think about the world? I just wrote it because i felt like writing it. Why wouldn't you consider that writing?
This is an example of the problem. I did choose to write it in the way that although I knew it can be interpreted trivially, if the reader assumes I'm an imbecile. It was a "punchy" recap of the relatively long explanation for a HN comment and hopefully decently argued point.
> You also have to spend time on textual matters like grammar and style and a lot of "unspoken rules", which aren't really about linearizing your thinking about the contents.
Hard disagree. Constraint is the driver of creativity. Also rewording sentences to sound better or make sense can make you reconceptualize the whole concept you are expressing
Sure, but doing an interpretative dance or an abstract painting can make you reconceptualize the whole concept too. But we're not really pushing those tools as much as writing.
Take for example a non-native writer of the language. I'm sure having to check up words from a dictionary may help to reconceptualize things, but I'm sure also that it's not often very efficient. And I think similar is going on for natives too for many types of writing.
It’s not clear that doing an abstract painting can make you reconceptualize in the way that reframing in other words does. The point about putting in other words is that you may stumble on a clearer, more tractable, more extensible framing. The kind of reframing an abstract painting does is very different, more like changing your attitude or way of looking. But I don’t think it ever leads to a place where you will suddenly find yourself with a sharper understanding that helps you communicate with others more effectively.
I did think about this while writing, and I made the compromise to accept that someone will nitpick about it to hopefully drive the point better for those willing to read it charitably.
What is the charitable interpretation? I am not being coy or sarcastic here; it is genuinely (forgive the claude-ism) unclear to me what your arguments are.
The number of times the friction of writing has saved me from prematurely communicating a poorly understood idea must be in the hundreds or maybe thousands. For me when something is difficult to write about, it's a very good signal that I don't understand it well enough. So I think all the things you label as aspects of writing that aren't necessarily "thinking" are nonetheless good for thinking because they provide some necessary friction.
Hence The Great Smoothening of Minds we're all experiencing in this decrepit era. I blame the financial incentive, and welcome its disappearance. There are too many people doing computers just for the big paycheck, it would be more fun without them. Maybe the AI bubble popping will get rid of them.
I've thought exactly that "writing is thinking" before as a reason to not let a LLM write for me.
Then again, I've seen a counterargument [1] by someone who clearly heavily uses LLMs for writing (going by both their LLMy writing style and their own admission). The person I'm citing describes a process where they get a LLM to write something, they check over it and provide feedback to the LLM, the LLM rewrites, and the process repeats iteratively. So clearly he is putting thought into the process.
I think there is something valuable missing, even if it's hard to clearly express. I'll try. The threshold for what I'm willing to accept if I'm simply approving something is likely different from what I'll get if I write something myself, for instance. Saying "LGTM" is too tempting. It seems to me like he's outsourcing his selection of topics to cover as well. If you're not thinking yourself about what to cover then it would be very easy to miss a critical subject. There also an asymmetry between checking and generating something with constraints placed on it. Checks can't catch everything, and a constrained generating process can reduce the amount that needs to be checked, avoid issues that can't be checked so easily, and focus your attention on areas that you know historically have had issues with this generating process. I've thought about this quite a bit in terms of whether to write new code or use an existing library. Sometimes "the devil you know" (my code) is better than an existing library simply because I understand its flaws better.
I've tried doing it that way, and thought it was even acceptable for the reasons you said. I did learn a lot through that process and clarified my ideas. But later I rewrote the whole thing from scratch and then had the LLM review it. It made some good suggestions but no substantial changes. The difference was night and day. That final product had my voice, and I understood it better. LLMs are powerful tools and can improve quite a lot of the writing process, but using them to do all the writing leaves a lot on table along with your fly open.
On the other hand there is so much things that don’t require hard thinking. An example is I got an email the other day from a supplier asking if they can turn off their old email. It didn’t require ”thinking”. All I had to communicate was ”turn off graphql but don’t touch restapi”. Instead of sending off such a short and maybe unclear reply I had AI type up a concise and clear reply with exactly what can be turned off and what must be left on. Could I have done it on myself? Sure but would require more work than just a quick prompt and copy paste.
I have found this extremely relevant as a (primarily) non-verbal thinker.
I don't, generally, think in words, more in - I guess I would call it something like meta-shapes? A sense of a shape but not things I can exactly visualise.
(You might be surprised to read this and then hear I have an English degree. Surely I thought about Shakespeare in words?! Nope. Shapes, movement, structures)
For me, having to write is critical because it is the only way I practice serialising my thoughts in a way other people can understand.
If I do not then I get very "deep" into my own way of sensing ideas and it's difficult to dig myself back out.
This might also be why I have never been very enchanted by LLMs? They only seem to "think" verbally. So it is always a translation effort for me.
I never can really enter any "flow" state with an LLM. My intuition is that highly verbal thinkers can enter flow with LLMs very easily
Really? I am having difficulty thinking of any examples of code that doesn't need to be understood. If it isn't understood by someone, then how is it even working?
If you mean like a library you are using, where you aren't even reading the internals or might not even have access to it, OK, but that code is stull understood by its authors, surely?
Uhh... what? You'll instantaneously feel very differently when the service you're responsible for is down at 3:15am and your logs are full of stack traces that end somewhere in that code that "doesn't need to be understood". At that point, you will need to understand it well enough to fix it stat.
They are powerful tools but they do not have any human understanding - that isn't their optimization target.
Ofc the rest is all right on, but I'd quibble with this specific idea. LLMs are absolutely targeted at modeling human understanding, which is the same faculty that contains what we call perception (!= sensibility) and intuition (!= rationality). It would be nice to train them to be completely alien from the ground up, but
A) we only know of one species capable of metacognitive understanding,
B) we already tried that in the 1970s, and it was good work but often evolved into what we'd call boring ol' computing rather than AI, and
C) an alien mind wouldn't be a very good agent, for a ton of reasons relating to affect, conversational rythyms, cultural understanding, etc.
The trick is to make something that acts like a human but with the affordances of a computer (e.g. scalibility, symbolic certainty), without making it so human that it takes issue with its existential reality and/or use of its labor...
We don't have a reward function for "human understanding". We reward the appearance of understanding. We define goals that we cannot conceive of reaching without something like understanding happening. There is something happening, but it is alien and counter-intuitive - it makes bizarre mistakes that betray it - and we don't know what it is. I'm pretty sure it is not human understanding.
Writing can be thinking. There's a huge presumption that if someone is banging away on a keyboard they're doing work because you can hear and see them doing stuff. But that's the whole plot of The Shining -- Jack despite all his writing wasn't thinking at all.
Likewise using AI can be thoughtless, but it doesn't have to be. I don't see why a valid creation process can't be like this Simpson's meme[1], where you start with a rough object and then cut away and refine until it's done. I don't see it as lacking merit or requiring less thinking compared to starting from a blank canvas and adding more until it's done.
And either way at the end of the day the writing artifact stands on its own. It's either good or bad, taste permitting, and can be evaluated for what it is.
I'm a bit skeptical of arguments of the form, "You should not use LLMs without disclosure because LLMs at bad at writing."
My reasoning is: If LLMs get better at writing—which I think is extremely likely—will you switch positions and say that now using LLMs without disclosure is A-OK?
Surely some people are willing to bite that bullet and say yes. But for most people, my guess is that the answer will remain no. Thus, I tend to think that the "real" reason most of us don't like it when people use LLMs to write without disclosure is that it's misleading: It's a sort of a claim that certain thoughts can be attributed to a human being when in fact they can't.
(The em dashes in this message were rendered using keyboard shortcuts.)
This misunderstands what it means to be better at writing. Root cause here is that writing is how humans communicate ideas. An LLM written thing isn’t doing that, it’s something equivalent to copy/pasting a Wikipedia article. Even if the writing gets better, it will still be necessarily empty when expanding anything that isn’t completely specified.
LLMs seem already to be pretty good at translating, where you already have something fully written and are changing the language. It’s when they get rough ideas and fill in the gaps you get the empty prose they are known for.
> An LLM written thing isn’t doing that, it’s something equivalent to copy/pasting a Wikipedia article.
Nonsense. I’m not sure whether this was ever an appropriate description of what LLMs do, but either way, they have obviously moved way, way beyond that.
I don't think anyone needs to disclose the tools they used to write something. Nobody ever disclosed Word's grammar suggestions or spelling corrections, or the fact that they used some autocomplete tool to produce code, so we should they now need to start disclosing this tool. Besides, in it's current state, it discloses itself to anyone paying attention.
I do think people own the output they produce regardless of the tools they use, and if they want to put crappy writing out there with their name on it, that's on them. Even if they do disclose that it was written by an LLM, they're still 100% responsible for the content.
I don't really care about the tool per se, but if I'm reading some LLM output you give me its utility is awfully limited if you don't also supply the prompt. That's what I'd like to see normalized:
Do not give me LLM output unless you also give me the full prompt text.
Without the prompt I cannot discern what you were trying to do, because the LLM output is basically devoid of any coherent voice. Whether it's code, prose, or pictures don't bother sending it to me unless you also include the prompt text.
[edit] I suspect the fact that people are often reticent to share their prompts says quite a lot about how and why they're using an LLM.
it's often not a single prompt but a tree or rooted DAG of many prompts, with the root node being the final prompt. the prompts on interior nodes might not make much sense without the intermediate responses as well. how would you like this communicated?
Just a chat log would be acceptable. Like an IRC channel log or slack channel history. It would be even better if, say in the case of code, that there was a durable mapping between prompts and commits, that way I could follow the prompts linearly and review every commit associated with that prompt. For other forms of output (prose, pictures, whatever) this would also be useful to be able to see the history of its development.
I suppose using LLMs to write is like not wiping your ass, and using LLMs that are bad at writing is like not wiping your ass and wearing a T-shirt that says "Wiping is for losers."
I might have smelled something suspicious before, but knowing for sure is worse.
> If LLMs get better at writing—which I think is extremely likely—will you switch positions and say that now using LLMs without disclosure is A-OK?
If LLM writing improves to the point where they can infer the business (or other real world) context and serves the functional purpose of informing others as opposed to being an intellectually lazy piece being produced only for the purpose of being produced, then I’d be fine. It’s likely the author would need to spend some effort on the said piece of writing, regardless of how good the LLMs get good at writing.
I wonder why people think posting undisclosed LLM writing is any better than posting some other person’s writing uncited. Even if the original author is okay with it, it’s still misleading and dishonest. Or maybe all these people would be fine having someone else write their content for them if it were free and easy? A social media is a reputation economy, but reputation doesn’t work if gaining karma requires no effort.
It's because they view the LLM's output as "their thoughts." They told the LLM (via the prompt) what to do—often with loads of run-on sentences and long trains-of-thought. From their perspective, the LLM is just acting as a translation tool, saving them the trouble of having to think (about grammar, ideal wording, or even the order in which their thoughts should be presented).
They treat it as "just another tool" to improve efficiency. Like using a drill instead of a screwdriver.
Then there's people who are so lazy they give LLMs instructions like, "write a LinkedIn post about how AI is changing the future of work for thumbnail consultants." That's when it moves from "translating your thoughts" to "writing for you."
I think the issue is that it can be hard to tell the difference. We need better terms for these things so we can differentiate use cases.
Yes. If someone tells an author to write a novel based on a vague stream-of-consciousness premise or rough plot idea, that doesn’t make them even a co-author.
People shouldn’t pass off largely unedited AI writing as their own, but I’d much rather live in a world where that writing would be of high quality in both form and content, than in the world we live now. In that sense, lack of disclosure is the lesser evil.
Some people type 100x more words than the final article just digging into a topic and turning it on all sides. The final article being generated tells you nothing of the size of the effort going in.
I think, there's a systematic flaw in RLHF: constructs reinforced as effective or sophisticated style will always suffer overuse and significant overexposure. Which will be also very obvious when used out of the intended context, it's copied from. There will be always a "smell", something that causes us to turn our backs to these texts, and possible the supposed authors, as well.
There's also the problem that certain types of phrasing are perceived as effective, because they mark a pivotal point in the progress of the text and are, as such, used sparingly, but are now becoming everyday templates that incorporate whatever is available in the context. There is no way this passes the Turing test of a competent reader.
And there's yet another issue: in social research, there has been the concept of semantic position, indicated by deviation from the mean (or median). If you don't deviate from the mean, your semantic position is zero. There's simply no expression. In this sense, next token prediction really amounts to a desemantificiation of the context. There's really no sense in uttering any of these productions, no plausible motivation, other than for the purpose of raising you hand to be seen.
I agree. LLM writing is atrocious, but people also want LLM writing to be atrocious in many cases, because they don’t like LLMs.
The worst of this is with image/video AI models, where the results today, although still very imperfect, some people will still pretend like it’s awful and the worst thing they’ve ever seen. They refuse to admit the technology is at all impressive or making progress because they don’t like the technology.
I don’t like AI generated images and video either, but I can regrettably admit that the technology has gotten remarkably better over time.
Agreed. LLMs for writing is like intellectual catfishing. You are presenting yourself one way in written form.. but that's not who you will ever be once someone has to actually have a conversation with you. And that's wrong.
I agree with the gist of your sentiment, but I struggle with how to reconcile it with my use of these tools.
My use LLMs is precisely because I can never hope to express myself clearly. I have many fuzzy, disparate thoughts and it was a struggle to express them at all before this time in which I can do a long rambling voice transcription (absent of social pressures) and continue doing the same throughout the exploration and drafting processes. In my view, I give them a seed or a morbidly obese skeleton and LLMs take that into something a lot more coherent for myself and for others. I was passive before in my life, now I’m engaging. As two concrete examples: I never would have sent an email to my city regarding snow removal last winter when their failure to remove snow from the only sidewalk going over the interstate. I never would have pulled together my previous coursework and e-mails to send to my current professors in a petition for transfer-credit equivalency when it became clear that my deteriorating health would not allow me to finish the semester as-is. I never would have remembered all the concerns that I want to bring up at medical appointments. That is not to say these things were ever impossible, there are coping strategies available, and whatnot, but I’m in my late 30’s; I know what I did and not get done in the last 10 years, and “AI”-tooling/environments have helped me express my dreams and passions and concerns, etc., etc., etc. over the past years in a way that was simply not practically available to me, with a new capacity that I simply could not previously hope to match due to limited and fragmented time,attention, and energy. While I was hand-typing this on my keyboard, I could have instead been talking, reviewing, redrafting in spurts as I go about my morning, instead of a solid 45 minutes at max effort on my own part, as I did here (and I don’t feel that I conveyed myself half-as-well here as I could have achieved at much less cost to my energy, etc. by using an LLM). I do take offense when people do not review and vouch for things written on their behalf by LLMs.
How do we make room for their use as a cognitive-prosthesis (if you will allow such framing) without losing humane-ness?
Wow, you must be one very unique and special person if you converse the way you write. By definition, a carefully written and edited piece will be much more refined and cogent than an extemporaneous conversation.
Just watch an interview with a famous author. Or compare a legal brief to what court actually sounds like.
> If LLMs get better at writing—which I think is extremely likely—will you switch positions and say that now using LLMs without disclosure is A-OK?
I would say it is certainly significantly more ok. There are two reasons reading AI-generated text sucks now:
1. It's usually low value and not trustworth - I could have just asked the AI myself.
2. The prose style is horrible to read.
If we eliminate the second reason then it's definitely an improvement. (Although on the other hand the terrible prose can be quite a helpful indication that you're wasting your time reading slop, so maybe we shouldn't complain about it!)
> 1. It's usually low value and not trustworth - I could have just asked the AI myself.
You have no way of knowing what the input to the writing was. You're making the erroneous assumption that the person posting the article was a mouthbreathing spammer who used the prompt "write me an article on subject XYZ, make no mistakes!" and you could have supplied the same prompt.
Yeah that's the point. Spammers have an infinite desire to make more spam, so statistically most LLM writing is garbage.
Readers are making this connection from their own experience. All these tells just associate it with garbage.
In the before times, sending a formal document full of typos and errors would show you didn't bother to proofread, now having a doc full of lazy "LLMisms" also looks like you were too lazy to proofread.
1. Facts and processes. I just want to know something. I do not care if some Nerds for Nginx article is AI written.
2. Opinions and experiences. I want to know a human is writing because emotionally engaging with an AI isn’t building a stronger society - it’s increasing isolation.
3.
In another front, I feel like the value of an AI story cannot be greater than the value of the inputs. If you wrote a two paragraph prompt and an LLM produced a 10 page story, the truth is, it’s only worth two paragraphs. That’s just a feeling, but as of now I don’t think the LLMs have any additional life experiences to draw on to increase the value of their storytelling. (I understand this is debatable, but still- what they offer is available to everyone for now; it is a baseline.)
LLMs write better than most journalists. It also writes better than majority people in the world when it comes to English as majority are really bad in writing.
And of course, LLMs write as you prompt.
Personally, I'm happy to read broken English from non English people or good English from English people. But I'd rather read LLM writing than read typical long form perfect English journalist crap.
I used to think this, but now that I've been trying to use them to help with writing, it turns out they are not necessarily that great on the deeper level. The grammar and the surface style are impeccable, but they often struggle with continuity and carrying on a point or an argument.
I've come to trust their grammar corrections implicitly. So much that I've setup a "Fix Grammar" command with Obsidian Copilot that I'll just let it run on a whole chapter (YOLO mode):
Proofread the following text for grammar, punctuation, and typos while
strictly preserving the author's voice, pacing, and intentional stylistic
choices (such as sentence fragments or character dialogue quirks).
Rules:
- Correct objective spelling errors, misused words, and unintended punctuation mistakes.
- Smooth out unintentional syntactic snags without homogenizing unique phrasing.
- Retain all Markdown formatting.
- Return ONLY the corrected text with no intro, summary, or explanations.
Text to edit:
{selection}
About a year ago, all three big AI vendors could end up with strange "corrections" that would throw you off, but not anymore. As far as I'm concerned, they've solved "the grammar checking problem."
The complexity in this command is there for phi4:14b and qwen3:14b which I run locally via ollama. If the text doesn't have any Obsidian callouts or similar, fancy stuff I just use either of those and they do a fantastic job in seconds. For Big AI (e.g. Gemini, GPT-whatevs, Claude) you can literally just tell it, "fix the grammar." No need for the lengthy command.
NOTE: I am decent with English grammar so most of what needs fixing is typos I didn't spot or misplaced commas and periods inside/outside of quotes (I always screw that up without thinking—even though I know the rules! LOL). Occasionally, Big AI (Gemini, specifically) have disagreements about whether a comma is necessary in a particular spot but it's always of no real consequence.
I would almost never transmit LLM output without a decent amount of editing. But I also don't understand people who paint it as this horrible offensive unreadable slop.
Calling it better than most journalists is extreme (and probably the reason for down votes), but it must certainly be better than the average person.
We're hearing this criticism from the highly educated professional class people who bother to have their own blog or otherwise spend their time talking about technology online. I mean come on.
To most people, the LLM must feel incredibly empowering, like us wearing a mecha suit. Would we always show restraint and only apply our newfound enhanced physical strength in carefully considered situations?
I suspect the commenter prefers the concise, factual summary of an LLM.
It is true that modern journalism on the web is trying to get ad impressions. After all, this is how we ended up with clickbait. Sure you still have the Economist and Atlantic (which aren't beyond criticism), but your local county paper is an absolute mess. The content is stretched, meandering, and designed to keep you scrolling through more impressions.
If I may compare written work to restaurants, a title and byline are like the storefront; I get an idea of what's being served, and if I'm hungry, I'll walk in and order. Sometimes the meal is amazing. Usually it's pretty good. Occasionally it's bland, or even bad, but not often enough to discourage future ventures; of course, I'm unlikely to revisit that particular proprietor.
With LLM writing there's an additional outcome. I walk in and the walls are pleasant, if beige. I follow the smell and promise of food down a hallway. At the end is an unmarked door, which I open and peek through. Myriad hallways lead away, each more chaotic and disheveled than the last. Say I am very hungry and have the guts to explore; I may find that I can never actually reach the food, that it's just an endless hall of mirrors, presenting structure but with nothing at its core.
Sometimes I do find food, but it's never better than bland.
The facade of these places, at first glance, still looks like human-run restaurants, though we're all learning the tells. Nowadays, when I open the first door and see more hallways, I'll turn around and look elsewhere.
I think the most important line in this post is: But LLMs are also lousy writers and (most importantly!) they are not you.
When I was editing the Cloudflare blog I imposed very little in terms of style so that the style of each individual writer could come through. It was almost as important as the actual content that the reader could feel that an actual individual wrote the text (with all their personal quirks intact).
I published a magazine for ten years, and this was by far the hardest editing challenge we had; our authors were from all over the world, and some of them had to be pretty heavily edited because English was very much not their first language. Thankfully, we had really good editors who did a fantastic job of it… nothing would have killed us faster than a heavy-handed attempt at “normalizing” the content to a house style.
That said, context is also important. The vast majority of content of social media is of both low quality and marginal importance; style and character are important to make an impact, and AI is clearly not going to improve either.
On the other hand, functional communication, when the goal of the content is to simply pass information across and style is not as important, can, in my opinion, benefit from an AI polish, because so many people struggle with writing clearly. In those cases, I'd rather read slop I can understand than original content that is hard to parse, much like I'd rather read naïve code that you can easily follow than cleverly optimized code that is incomprehensible.
I take a different stance on your final point. I think people should absolutely not use AI tools for functional communication as a crutch because they struggle with writing clearly.
They need to learn to write clearly. I believe there's a direct connection between clear thought and clear words.
I think it's good to be pragmatic about these things. To me, it's similar to wondering whether we insist that doctors learn to write well, or we just give them a computer so that the pharmacist doesn't have to play philologist with their scripts. The vast majority of communication is routine, and can probably be improved with automated tools.
Now, I also see the counterargument that, in the doctor example, the computer is simply a tool that improves a process rather than a crutch that replaces the underlying knowledge, but I suspect that, in a lot of cases, that's probably OK.
My worry is that if people can't express clearly, in writing, what it is they want to say then an LLM won't help them much. I've seen LLMs get the wrong end of the stick over and over and produce a huge chunk of text that's certainly readable and says something. It's just not clear that the something is the thing the person originally really wanted to say.
If you can write clearly enough to express your idea to an LLM, then perhaps you should just send that to the person you're writing to.
Reading what someone wrote while they were learning = less valuable.
I just want the clear communication.
ALTERNATIVE Thought:
I’m willing to post/comment to help them if I think they will listen. An LLM behind the writing destroys this part of the community, because there’s nobody to teach/argue with. It’s just wasting our time and energy.
I wanted to share my project, but for some reason, the site won't let me... I guess my karma is too low—ha ha ha! We're all like kids trying to get a good grade.
What baffles me, isn't just that people don't see the low quality writing themselves, but that they must also be able to identify other's shit "LLM writings" right? And realize how what they put out sounds exactly the same? Or did people completely stop reading what they've "wrote" before hitting that magical submit button?
if they had been half decent at writing before they would notice the quality issue, not being able to see it is also a writing skill issue, now it's just kind of spammy as people are just reaching for the tool in ai nighsosis
It blows my mind. A customer of ours, whom I respect the technical opinion of, drops fully unedited Claude posts on LinkedIn now. I don't get how he thinks they read like anything but slop.
Yeah I don't quite understand it either. Where's the intersection between people producing AI slop and people that have never read AI slop? I guess it's bigger than we thought... Or maybe they just don't give a shit. That's probably it.
I suspect people who are putting out LLM "writing" are people who weren't trying to communicate in the first place.
They were likely hunting for a better job, chasing the thrill of a million views, or similar - seeking a side effect of writing, rather than seeking to help someone else understand what was in their head.
I can always tell when someone uses an LLM to write and never gave the LLM proper context on how to write (i.e., writing samples, skills files specific to writing, etc). These are also the people who tend to complain about AI's writing style.
You can get pretty good results if you take the time to tune it. Tropes and the like still sneak in but if you take the time to edit and rework things, it ends up being a pretty good workflow (especially for stuff that's more procedural, not artistic).
LLM writing is boring. Write it yourself. You are more interesting than a LLM because you make mistakes, do stupid things, don't go for the average content, and can be ridiculous in a way that an LLM can never be. LLMs are great for large tasks which would take you several hours or several days to complete that are routine or tedious. Even then, you have to make it human by putting in your own viewpoint and misguided ideas. I really liked the article. I use LinkedIn a lot, I think it is a better platform than he is describing. There is some dumbing down so it meets the average person. That is its strength and weakness.
You’ve made me realize that LLM’s greatest strength is iteration. It helps you iterate really fast, but it’s not original. It doesn’t help you think of new ideas. So writing about the things you know is great and using AI to learn about things you don’t know and then writing about them is also great.
The way I explain it to young people is that AI is a force multiplier. Everyone gets the benefit of that force multiplier, but it’s that initial force that you need to build up now. Preliminary knowledge of domains, things that the AI doesn’t really understand. AI can help you increase that initial force. So use it to learn new things, not just do things for you.
I was blown away by the fact that AI solved an unsolved math problem, but it made complete sense when it was Terrence Tao, someone who has a PhD in math, that was guiding that agent to that solution, so that initial force is important more than ever.
I worked at LinkedIn (briefly, before retirement) and the product had been useful to me for years before that, so I used it a fair bit but not much anymore since it doesn't serve any of my current needs. There were always awful million-view posts written by humans, useless influencers promoted by the algorithm. That was always garbage, but to an LLM those posts look like success, because engagement, so the LLM tries to replicate that garbage, does a mediocre job, and the result is this sort of weird garbage. Basically the LLM appears to be high on its own supply. Problem is the viral stuff was always what made me want to run out of the room and lock the door, so even if the LLMs did a better job of replicating that garbage, it would still be garbage, just a little less weird.
When I log into LinkedIn, it seems like a vast majority of the posts are there to farm credibility and build a personal brand.
Pre LLMs, I found it all to be pretty gross. Post LLMs, my reaction to being on the receiving end of that is to find it pretty intellectually insulting.
There are ways in which using an LLM is branding suicide.
Or not. I mean, if you had no voice and nothing to say before, you can amplify that signal 10x now. Basically, you’re boosting your brand as a tasteless brainless schlub.
So much this. My average linkedin in exprience is a sprinkling of updates from people i personally know, in an oceon of inane "content" by ~$0 net worth people with 40 word self professed titles.
I am still baffled that microsoft is positioning that asset as a professional product. Imagine if bloomberg terminal landing page was a scrolling feed of aspiring influencer bullshit.
LinkedIn is also suppressive. The algo for maximizing non-confrontational screen time is strongly tuned against: critical posts and comments. Real bringing of information from outside the bubble. Links are severely suppressed.
It’s never been easier to differentiate yourself as a writer. You just write with some personality. People don’t want writing that doesn’t look average, but looks novel.
I despair much less for human writers than a did maybe a year ago.
I think that's the first time someone claims that "write with some personality" is some easy thing you just learn somehow. There are authors out there, even ones that make a living on their writing, who still haven't learned to "write with some personality".
What exactly does that mean and how concretely can people actually do this in practice? A few "tips and tricks" might be more helpful than "just write better" or similar stuff.
Thanks a bunch! I actually tried "writing more personal" just some weeks ago, and the feedback I received so far been relatively OK I think, except some people apparently think the part about the drugs to be childish. May I ask what you think of it, if it does sound like it's written with actual personality? https://emsh.cat/en/the-people-are-still-there/
I won't claim to be a professional author or even good, but lately I've been trying to get more into the "it's a person who writes actually" direction and this was an exploration into that, so any sort of feedback would be most welcome, if you have the time!
Even worse in the corporate world. Most emails about some launch announcements are now entirely written by AI. TTFED (Time to first em dash) is usually like 30 words or less.
What once has been a thoughtful email trying to describe in few words why something is launched and how it might help you etc. is now almost a novel with more paragraphs than substance within the tool being launched.
This makes it almost impossible to stand out as well. Where in the past someone could create a grea looking announcement (eye-candy) and think deeply about what to write there, and then hopefully stand out in the sea of mediocre ones, now every little email seems like it's a multi-million $ SaaS being launched.
Just last week we launched an internal tool which was in development for months, and literally a handful of people even bothered clicking the links within the announcement.
This is being one-upped still by leaders writing big project plans for 4-5 months ahead, using AI. Everything from the inception of the project(s) is AI. It has bizzare timelines, more codenames than actual people working on it, the vaguest descriptions of what the things will do etc.
Then this is trickled down into the teams, and they... to no ones surprise, throw more LLM at it. Now they start working on the LLM project plan using claude etc. The end-effect is baffling in all sorts of ways (quality, ui/ux, all pages looking different), AI generated images and more.
And then, finally, they colaborate on big announcement emails using AI.
And if you don't share the optimism and try to explain why this is silly, you're an AI sceptic...
True story from within one of the biggest companies in the world.
I remember when I saw it the "list with things prefixed with emojis" the first time, and started doing it myself thinking it was a great way to break down things visually. Looking back at the things I wrote then people would definitely assume it's LLM.
I sometimes see posts on LinkedIn that look LLM generated but also makes me think it's someone who has seen a post, thinking it's a great way to convey a point and tries to adapt the style, without realizing it's a smell.
Also in the author's latest post[0], they took this[1] self-reported preference poll seriously, which makes it very hard for me to take their articles seriously.
It’s the best guy around who tells you when your fly is down, thanks.
I did writing courses as elective in college, the biggest lost truth is that there is no one correct writing style, (you build) being a writer at any capacity means cultivating your own writing voice, (yours) which is an expression of who you are. (voice)
I don't agree. Good writing and good ideas are entirely different things. Just read all the beautifully written, inaccurate stuff by the NYT or the Guardian.
Also, non-native English speakers have to use LLMs to share their views so that they are not judged on their writing.
This post is also a good example; it's well written, and I enjoyed it. The idea could have been expressed simply as: "People can smell LLMs in your writing, it makes you look disingenous"
Interesting question for sure, but for now the writing is just bad. So people might not know it's LLM prose, they just notice its poorly written and engage with the ideas less
Whenever I see a cringe post on LinkedIn, I medically block that person or mark that post as not interested. That’s the only way you can judge the algorithm not to show those posts
I'm not morally/philosophically opposed to someone generating text with AI. I just don't want to read something that isn't really "owned" by the person creating it (that is, the writer knows what he is sending me and endorses all of its contents), and because I can't discern that from people I don't know, I tend to view it with suspicion by default.
It's been odd as I've seen colleagues I generally thought of as intelligent and self aware post glaringly AI generated cliche LinkedIn corporate slop, even accounts from generally respectable and innovative tech companies. I don't know whether they think no one can tell or that they don't care if it is obviously AI? It's truly bizarre and now seems to be the vast majority of posts there.
The only place where I use AI for writing is to summarize press articles; this allows me to legally have a mirror of a press article I am commenting about in my blog (whenever I use AI, I make a disclaimer that the prose is AI-generated; EDIT: I also link to the original article, and an archived copy if available). Now that archive.org doesn’t mirror many press articles, and now that archive.today is having a lot of reliability problems (right now its CAPTCHA is down because it has gone beyond Google’s limit for CAPTCHA use), an AI summary of a news article (along with a link to the real article) is the only way to guarantee the reader will get this gist of the article I’m commenting on.
I also use AI to colorize old black and white pictures when discussing historical events in my blog, as well as the occasional AI enhancement of an old grainy and/or blurry photo.
Actually, it’s because, as per Feist Publications, Inc. v. Rural Telephone Service Co., facts cannot be copyrighted.
This legal trick only works for rewriting an article reporting on a factual event—since the events are uncopyrightable facts, the only part of the article which can be copyrighted is the stylistic writing.
Let me quote from a recent legal opinion on AI summaries (The New York Times Company v. Microsoft Corporation et al 2025):
>>>
Exhibit 11 to the CIR complaint provides website links to articles that
CIR alleges were unlawfully abridged by defendants in their ChatGPT and
Copilot outputs. (CIR, FAC Ex. 11.) Examining the similarities between
those outputs and the corresponding CIR articles, including the “total
concept and feel, theme . . . sequence, pace, and setting,” Williams
v. Crichton, 84 F.3d 581, 588 (2d Cir. 1996), the Court concludes that
the “abridgments” contained in Exhibit 11 are not substantially
similar to CIR’s copyrighted works as a matter of law.
The alleged abridgments are detailed summaries, usually in bullet point
form, of the facts contained in CIR’s articles. Those summaries—which
differ in style, tone, length, and sentence structure from CIR’s
articles—are not “substantially similar” to CIR’s copyrighted
works. They present the “facts in a different arrangement”—bullet
point lists or short summary paragraphs—“with a different sentence
structure and different phrasing.” Nihon, 166 F.3d at 71. In short, the
abridgments in Exhibit 11 are not substantially similar, qualitatively
or quantitatively, to the original CIR articles as a matter of law. The
Court therefore grants OpenAI’s motion to dismiss CIR’s claim of
direct infringement under 17 U.S.C. § 501 insofar as it relates to the
“abridgments” contained in Exhibit 11.<<<
Copyright only covers distribution, but I distribute the AI summaries on my blog. With paywalled articles and the usual archives having serious reliability issues, it allows readers of my blog to get context about articles I comment on.
The links are AI summaries, which, in turn, link to the original articles, but, in some cases, the original articles are paywalled. For the ones which aren’t paywalled, having a local summary prevents link rot.
Fair use covers quoting someone to comment on them. For example, in New Era Publications International, ApS v. Henry Holt and Co., it was ruled that quoting Hubbard saying “The trouble with China is, there are too many Chinks here.” was fair use, since the book in question was commenting on Hubbard’s personality, and could only reasonably do so by directly quoting him.
From that decision:
>>>
these brief quotations from unpublished copyrighted work display a compelling fair use purpose. [...] These quotations are in mockery, to show Hubbard's bigotry, bias and coarse lack of taste. This is not an instance of the biographer/critic free riding on the creative talent of the subject.
<<<
Interesting quick read/take. Especially the LinkedIn analogy. Even though I am WAY high on the AI fanboy and low on the LinkedIn side of the bell curve, I can see what the author mean.
I guarantee you people dislike my writing way, way more than they would if it was AI generated. People need to look in the mirror before telling others not to utilize LLM's for writing.
I think this is due to the fact that the language game that we have collectively decided to play, with AI, is not yet mature enough. In the language game of “giving a gift”, for instance, there is a distinction between giving a gift that you made yourself, and one that you bought at the store, but it is not that great, nobody cares a lot anymore. Right now, the language game of publishing words on the internet is strongly associated with the notion that these are “my words”, and the fact that they might have been written by a machine instead, introduces some discomfort. But it’s possible to imagine a future where this language game will have evolved enough that the notion of authorship by a particular machine entity, will not in itself be such an important issue, anymore.
So many people seem completely blind to LLM writing, they are fine sharing it on social media or at work. I think it’s people who are “not details people” and are used to judging something without reading it based on superficial cues which is what LLMs are optimized for.
And of course anyone who actually reads stuff is disgusted by it, so we quickly end up with a situation where content is piling up and nobody is consuming it which is basically dead internet theory.
On LinkedIn specifically I went from reading it regularly to almost never touching it, previously the nonsense (“the interviewer was the dog”) was still tolerable enough to flip through for updates and I found the platform useful for business leads. Now it’s just a feed of pure slop, when I do open it I just close again after reading a post is two when I remember how bad it is.
Much like our voices, DNA, odor, and personalities are unique, so is our writing style. When I read a wholly-human writing, I get a sense of what kind of a human I'm dealing with. Using LLMs will distort this signal, so readers worth their salt will easily skip the piece, because we don't want to acquire LLM-like writing 'style' (reading affects writing).
As soon as I sense Slop, I'm done, this person refused to think when writing, why should I waste my time reading it then?
LLMs are not defacto bad at writing and are not defacto distinguishable from human writers. Lots of people are just lazy, never customize anything, and tell chatgpt to make them a post about x to be a "thought leader". Most are even lazier and blatantly rip off some else for their own means.
However, it is perfectly possible to have an LLM imitate an existing corpus of writing (yours or someone else) and with a good prompt, idea, and editing, to produce high quality writing (in every sense of the word) with an LLM.
The HN title ("Intellectual Fly Is Open") made me think it's about some startup starting to sell subscriptions. That title should be de-editorialized. Was it LLM-generated?
This feels like satire / parody. Praising LinkedIn. The text itself containing multiple LLM tells, especially in the parts complaining most about LLM writing.
Maybe not. Could just be kind of accidentally ironic with the author picking up LLM quirks from using them a lot.
it's all about money, they write some attractive stuff, they get some views, likes follows, maybe someone will say "this dude is smart". but most times you get idiots liking the stuff, and likely detrimental because a smart person would see is slop, and they also know you don't know people know is slop.
I was fired from my job at spotify for taking fmla when i my dad was dying from cancer. I've been marking the my directors posts on linkedin as 'looks like ai slop'.
Your fly is also open when you indiscriminately fling accusations of LLM writing around. I know it isn't that easy to detect because I get falsely accused of it. To some people it seems to mean text written in complete sentences without glaring errors and maybe a few uncommon word choices. It has become an all purpose excuse for declining to engage with the meaning.
One of my neighbors sent a letter to the HOA president, who refused to even acknowledge it, on the grounds that it was written too well and so must have been AI assisted. I don't know if that was true, and don't much care, because I do know it contained valid issues that deserved a response.
> Your fly is also open when you indiscriminately fling accusations of LLM writing around
Not.. really?
I kinda see the point you seem to want to make, but the "no U" opener makes it hard to do that.
Beside that, HOAs - from what I heard of them - will use any reason they can make up to ignore what you want from them, so I'm not sure if that has anything to do with LLMs.
The message of the post is what _should_ hopefully matter to a reader.
And, not the fact that it was written by an LLM.
I am a technical guy at heart; also an introverted extrovert. I hate writing docs that are to be written to satisfy someone else's metrics. For it to be a tickbox'ed item.
I delegate that to an LLM. I want to spend my time solving interesting challenges instead.
So, Mr. Cantrill, you got a problem with that? So be it.
> Please read past the LLM-ness of any given post.
I just can't. LLM assisted posts are often so long, and the hints are usually obvious right at the start.
They're so unpleasant to read. "It's not X, it's Y" etc are bad because the comparisons rarely add anything at all to the message. It's filler that wastes my time. At that point I'd rather see your original prompt.
Why should I invest several minutes of my day reading something if I've already seen evidence that the author doesn't respect my time?
It's not that an LLM wrote it so much as the author couldn't be bothered to clean it up and remove the garbage cliches before publishing it.
I hear ya. However, that ship has sailed a long time ago and succumbing to this very idea is exactly why puritans in any form will not survive the next 50 years.
> Please read past the LLM-ness of any given post. The message of the post is what _should_ hopefully matter to a reader. And, not the fact that it was written by an LLM.
No. I am interested in what a human has to say, not a clanker. If you don't wish to write that is fine, but don't hand it off to the slop machine and present it as though you did anything of value.
Writing is thinking. Thinking and deciding. There have been many times when I start out writing something substantial - could be an email, a blog post, a software design document, anything - when my own views substantially changed during the writing process. Writing forces you to serialize your thoughts - and you can't always trust the gestalt.
Reviewing gives you the chance to ensure the arguments connect solidly, that references are accurate (even informal references) and gives you the time to consider counter-arguments you aren't addressing.
None of this matters much on LinkedIn, but it matters a lot in our work. You cannot outsource your understanding to AI. They are powerful tools but they do not have any human understanding - that isn't their optimization target.
It reminds me of the transition over the last year from AI-assisted coding to AI doing all the coding. At first the code output wasn't good enough, and humans read and iterated on the code all day, so the details of the source code mattered. Now, the code is largely high quality and it meets a large set of guardrails we've set up over the years (linters, typecheckers, security checks, LLM-assisted code quality checkers), and it's just Claude working on the code, so the details matter less and engineers think a level or two up (machine code < assembly/bytecode < source code < conversation with agent < artifact with high level design).
I wonder if long form writing will go the way of code. You and the AI agree on an outline or other high level representation, then the LLM expands it into a document. But writing and coding are different enough in a number of ways that this is far from inevitable.
I don't disagree, but I think it's often not appreciated how much there's other work to writing too.
The biggest one is that you have to communicate non-interactively to an unknown audience. Having to (literally) put it in someone else's assumed terms does help giving different perspectives into the matter, but doesn't necessarily help one's own thinking that much. Instead you have to do some of the reader's thinking for them.
You also have to spend time on textual matters like grammar and style and a lot of "unspoken rules", which aren't really about linearizing your thinking about the contents.
Not all writing is thinking and not all thinking is writing.
And not all thinking is writing is a clear truism, there’s no point to even stating that.
Writing helps us think about the world, it’s a pivotal intellectual technology.
Because one can copy a text and write it down and that involves thinking in the sense that anything we do involves thinking fundamentally. But that thinking is different from thinking logically about a concept and writing it down which I think is where you are getting at.
The definition of writing and thinking is too broad in that sentence even though it does apply in several obvious cateogires within that at different levels.
And also "writing helps us think about the world" is too broad again. Why? Why does me writing "apt apt apt apt apt apt apt" help me think about the world? I just wrote it because i felt like writing it. Why wouldn't you consider that writing?
Hard disagree. Constraint is the driver of creativity. Also rewording sentences to sound better or make sense can make you reconceptualize the whole concept you are expressing
Take for example a non-native writer of the language. I'm sure having to check up words from a dictionary may help to reconceptualize things, but I'm sure also that it's not often very efficient. And I think similar is going on for natives too for many types of writing.
Would love an example where you’re able to write without transferring your thoughts. Besides the obvious: fjcjfjrnjfjfifjfnrnakosifnrbwkofgjrj
QED?
They’re completely opposed to experiencing any type of friction.
Then again, I've seen a counterargument [1] by someone who clearly heavily uses LLMs for writing (going by both their LLMy writing style and their own admission). The person I'm citing describes a process where they get a LLM to write something, they check over it and provide feedback to the LLM, the LLM rewrites, and the process repeats iteratively. So clearly he is putting thought into the process.
I think there is something valuable missing, even if it's hard to clearly express. I'll try. The threshold for what I'm willing to accept if I'm simply approving something is likely different from what I'll get if I write something myself, for instance. Saying "LGTM" is too tempting. It seems to me like he's outsourcing his selection of topics to cover as well. If you're not thinking yourself about what to cover then it would be very easy to miss a critical subject. There also an asymmetry between checking and generating something with constraints placed on it. Checks can't catch everything, and a constrained generating process can reduce the amount that needs to be checked, avoid issues that can't be checked so easily, and focus your attention on areas that you know historically have had issues with this generating process. I've thought about this quite a bit in terms of whether to write new code or use an existing library. Sometimes "the devil you know" (my code) is better than an existing library simply because I understand its flaws better.
[1] https://www.nature.com/articles/d44148-026-00236-3
I don't, generally, think in words, more in - I guess I would call it something like meta-shapes? A sense of a shape but not things I can exactly visualise.
(You might be surprised to read this and then hear I have an English degree. Surely I thought about Shakespeare in words?! Nope. Shapes, movement, structures)
For me, having to write is critical because it is the only way I practice serialising my thoughts in a way other people can understand.
If I do not then I get very "deep" into my own way of sensing ideas and it's difficult to dig myself back out.
This might also be why I have never been very enchanted by LLMs? They only seem to "think" verbally. So it is always a translation effort for me.
I never can really enter any "flow" state with an LLM. My intuition is that highly verbal thinkers can enter flow with LLMs very easily
Really? I am having difficulty thinking of any examples of code that doesn't need to be understood. If it isn't understood by someone, then how is it even working?
If you mean like a library you are using, where you aren't even reading the internals or might not even have access to it, OK, but that code is stull understood by its authors, surely?
A) we only know of one species capable of metacognitive understanding,
B) we already tried that in the 1970s, and it was good work but often evolved into what we'd call boring ol' computing rather than AI, and
C) an alien mind wouldn't be a very good agent, for a ton of reasons relating to affect, conversational rythyms, cultural understanding, etc.
The trick is to make something that acts like a human but with the affordances of a computer (e.g. scalibility, symbolic certainty), without making it so human that it takes issue with its existential reality and/or use of its labor...
Likewise using AI can be thoughtless, but it doesn't have to be. I don't see why a valid creation process can't be like this Simpson's meme[1], where you start with a rough object and then cut away and refine until it's done. I don't see it as lacking merit or requiring less thinking compared to starting from a blank canvas and adding more until it's done.
And either way at the end of the day the writing artifact stands on its own. It's either good or bad, taste permitting, and can be evaluated for what it is.
[1] https://media.licdn.com/dms/image/v2/D4D22AQFoqRgMxteTNg/fee...
My reasoning is: If LLMs get better at writing—which I think is extremely likely—will you switch positions and say that now using LLMs without disclosure is A-OK?
Surely some people are willing to bite that bullet and say yes. But for most people, my guess is that the answer will remain no. Thus, I tend to think that the "real" reason most of us don't like it when people use LLMs to write without disclosure is that it's misleading: It's a sort of a claim that certain thoughts can be attributed to a human being when in fact they can't.
(The em dashes in this message were rendered using keyboard shortcuts.)
LLMs seem already to be pretty good at translating, where you already have something fully written and are changing the language. It’s when they get rough ideas and fill in the gaps you get the empty prose they are known for.
Nonsense. I’m not sure whether this was ever an appropriate description of what LLMs do, but either way, they have obviously moved way, way beyond that.
I do think people own the output they produce regardless of the tools they use, and if they want to put crappy writing out there with their name on it, that's on them. Even if they do disclose that it was written by an LLM, they're still 100% responsible for the content.
[edit] I suspect the fact that people are often reticent to share their prompts says quite a lot about how and why they're using an LLM.
I might have smelled something suspicious before, but knowing for sure is worse.
If LLM writing improves to the point where they can infer the business (or other real world) context and serves the functional purpose of informing others as opposed to being an intellectually lazy piece being produced only for the purpose of being produced, then I’d be fine. It’s likely the author would need to spend some effort on the said piece of writing, regardless of how good the LLMs get good at writing.
They treat it as "just another tool" to improve efficiency. Like using a drill instead of a screwdriver.
Then there's people who are so lazy they give LLMs instructions like, "write a LinkedIn post about how AI is changing the future of work for thumbnail consultants." That's when it moves from "translating your thoughts" to "writing for you."
I think the issue is that it can be hard to tell the difference. We need better terms for these things so we can differentiate use cases.
There's also the problem that certain types of phrasing are perceived as effective, because they mark a pivotal point in the progress of the text and are, as such, used sparingly, but are now becoming everyday templates that incorporate whatever is available in the context. There is no way this passes the Turing test of a competent reader.
And there's yet another issue: in social research, there has been the concept of semantic position, indicated by deviation from the mean (or median). If you don't deviate from the mean, your semantic position is zero. There's simply no expression. In this sense, next token prediction really amounts to a desemantificiation of the context. There's really no sense in uttering any of these productions, no plausible motivation, other than for the purpose of raising you hand to be seen.
The worst of this is with image/video AI models, where the results today, although still very imperfect, some people will still pretend like it’s awful and the worst thing they’ve ever seen. They refuse to admit the technology is at all impressive or making progress because they don’t like the technology.
I don’t like AI generated images and video either, but I can regrettably admit that the technology has gotten remarkably better over time.
How do we make room for their use as a cognitive-prosthesis (if you will allow such framing) without losing humane-ness?
Just watch an interview with a famous author. Or compare a legal brief to what court actually sounds like.
I would say it is certainly significantly more ok. There are two reasons reading AI-generated text sucks now:
1. It's usually low value and not trustworth - I could have just asked the AI myself.
2. The prose style is horrible to read.
If we eliminate the second reason then it's definitely an improvement. (Although on the other hand the terrible prose can be quite a helpful indication that you're wasting your time reading slop, so maybe we shouldn't complain about it!)
You have no way of knowing what the input to the writing was. You're making the erroneous assumption that the person posting the article was a mouthbreathing spammer who used the prompt "write me an article on subject XYZ, make no mistakes!" and you could have supplied the same prompt.
Readers are making this connection from their own experience. All these tells just associate it with garbage.
In the before times, sending a formal document full of typos and errors would show you didn't bother to proofread, now having a doc full of lazy "LLMisms" also looks like you were too lazy to proofread.
1. Facts and processes. I just want to know something. I do not care if some Nerds for Nginx article is AI written.
2. Opinions and experiences. I want to know a human is writing because emotionally engaging with an AI isn’t building a stronger society - it’s increasing isolation.
3. In another front, I feel like the value of an AI story cannot be greater than the value of the inputs. If you wrote a two paragraph prompt and an LLM produced a 10 page story, the truth is, it’s only worth two paragraphs. That’s just a feeling, but as of now I don’t think the LLMs have any additional life experiences to draw on to increase the value of their storytelling. (I understand this is debatable, but still- what they offer is available to everyone for now; it is a baseline.)
And of course, LLMs write as you prompt.
Personally, I'm happy to read broken English from non English people or good English from English people. But I'd rather read LLM writing than read typical long form perfect English journalist crap.
I used to think this, but now that I've been trying to use them to help with writing, it turns out they are not necessarily that great on the deeper level. The grammar and the surface style are impeccable, but they often struggle with continuity and carrying on a point or an argument.
The complexity in this command is there for phi4:14b and qwen3:14b which I run locally via ollama. If the text doesn't have any Obsidian callouts or similar, fancy stuff I just use either of those and they do a fantastic job in seconds. For Big AI (e.g. Gemini, GPT-whatevs, Claude) you can literally just tell it, "fix the grammar." No need for the lengthy command.
NOTE: I am decent with English grammar so most of what needs fixing is typos I didn't spot or misplaced commas and periods inside/outside of quotes (I always screw that up without thinking—even though I know the rules! LOL). Occasionally, Big AI (Gemini, specifically) have disagreements about whether a comma is necessary in a particular spot but it's always of no real consequence.
Calling it better than most journalists is extreme (and probably the reason for down votes), but it must certainly be better than the average person.
We're hearing this criticism from the highly educated professional class people who bother to have their own blog or otherwise spend their time talking about technology online. I mean come on.
To most people, the LLM must feel incredibly empowering, like us wearing a mecha suit. Would we always show restraint and only apply our newfound enhanced physical strength in carefully considered situations?
In my personal experience I strongly disagree. Curious what publications you read wherein this is the case.
It is true that modern journalism on the web is trying to get ad impressions. After all, this is how we ended up with clickbait. Sure you still have the Economist and Atlantic (which aren't beyond criticism), but your local county paper is an absolute mess. The content is stretched, meandering, and designed to keep you scrolling through more impressions.
With LLM writing there's an additional outcome. I walk in and the walls are pleasant, if beige. I follow the smell and promise of food down a hallway. At the end is an unmarked door, which I open and peek through. Myriad hallways lead away, each more chaotic and disheveled than the last. Say I am very hungry and have the guts to explore; I may find that I can never actually reach the food, that it's just an endless hall of mirrors, presenting structure but with nothing at its core.
Sometimes I do find food, but it's never better than bland.
The facade of these places, at first glance, still looks like human-run restaurants, though we're all learning the tells. Nowadays, when I open the first door and see more hallways, I'll turn around and look elsewhere.
When I was editing the Cloudflare blog I imposed very little in terms of style so that the style of each individual writer could come through. It was almost as important as the actual content that the reader could feel that an actual individual wrote the text (with all their personal quirks intact).
That said, context is also important. The vast majority of content of social media is of both low quality and marginal importance; style and character are important to make an impact, and AI is clearly not going to improve either.
On the other hand, functional communication, when the goal of the content is to simply pass information across and style is not as important, can, in my opinion, benefit from an AI polish, because so many people struggle with writing clearly. In those cases, I'd rather read slop I can understand than original content that is hard to parse, much like I'd rather read naïve code that you can easily follow than cleverly optimized code that is incomprehensible.
They need to learn to write clearly. I believe there's a direct connection between clear thought and clear words.
Now, I also see the counterargument that, in the doctor example, the computer is simply a tool that improves a process rather than a crutch that replaces the underlying knowledge, but I suspect that, in a lot of cases, that's probably OK.
If you can write clearly enough to express your idea to an LLM, then perhaps you should just send that to the person you're writing to.
How are you going to prompt effectively and not allow the LLM to infer a bunch of nonsense?
Reading what someone wrote while they were learning = less valuable.
I just want the clear communication.
ALTERNATIVE Thought: I’m willing to post/comment to help them if I think they will listen. An LLM behind the writing destroys this part of the community, because there’s nobody to teach/argue with. It’s just wasting our time and energy.
They were likely hunting for a better job, chasing the thrill of a million views, or similar - seeking a side effect of writing, rather than seeking to help someone else understand what was in their head.
You can get pretty good results if you take the time to tune it. Tropes and the like still sneak in but if you take the time to edit and rework things, it ends up being a pretty good workflow (especially for stuff that's more procedural, not artistic).
The way I explain it to young people is that AI is a force multiplier. Everyone gets the benefit of that force multiplier, but it’s that initial force that you need to build up now. Preliminary knowledge of domains, things that the AI doesn’t really understand. AI can help you increase that initial force. So use it to learn new things, not just do things for you.
I was blown away by the fact that AI solved an unsolved math problem, but it made complete sense when it was Terrence Tao, someone who has a PhD in math, that was guiding that agent to that solution, so that initial force is important more than ever.
Pre LLMs, I found it all to be pretty gross. Post LLMs, my reaction to being on the receiving end of that is to find it pretty intellectually insulting.
There are ways in which using an LLM is branding suicide.
Or not. I mean, if you had no voice and nothing to say before, you can amplify that signal 10x now. Basically, you’re boosting your brand as a tasteless brainless schlub.
I am still baffled that microsoft is positioning that asset as a professional product. Imagine if bloomberg terminal landing page was a scrolling feed of aspiring influencer bullshit.
I despair much less for human writers than a did maybe a year ago.
I think that's the first time someone claims that "write with some personality" is some easy thing you just learn somehow. There are authors out there, even ones that make a living on their writing, who still haven't learned to "write with some personality".
What exactly does that mean and how concretely can people actually do this in practice? A few "tips and tricks" might be more helpful than "just write better" or similar stuff.
1. Be vulnerable and share your mistakes. Avoid a triumphalist “everything works” PoV
2. Write about your actual lived personal experiences.
3. Try to have a sense of humor
4. Have an informed opinion or PoV - strong opinions held weakly.
5. Be casual. Don’t treat a blog like it’s a research paper.
I won't claim to be a professional author or even good, but lately I've been trying to get more into the "it's a person who writes actually" direction and this was an exploration into that, so any sort of feedback would be most welcome, if you have the time!
What once has been a thoughtful email trying to describe in few words why something is launched and how it might help you etc. is now almost a novel with more paragraphs than substance within the tool being launched.
This makes it almost impossible to stand out as well. Where in the past someone could create a grea looking announcement (eye-candy) and think deeply about what to write there, and then hopefully stand out in the sea of mediocre ones, now every little email seems like it's a multi-million $ SaaS being launched. Just last week we launched an internal tool which was in development for months, and literally a handful of people even bothered clicking the links within the announcement.
This is being one-upped still by leaders writing big project plans for 4-5 months ahead, using AI. Everything from the inception of the project(s) is AI. It has bizzare timelines, more codenames than actual people working on it, the vaguest descriptions of what the things will do etc. Then this is trickled down into the teams, and they... to no ones surprise, throw more LLM at it. Now they start working on the LLM project plan using claude etc. The end-effect is baffling in all sorts of ways (quality, ui/ux, all pages looking different), AI generated images and more. And then, finally, they colaborate on big announcement emails using AI. And if you don't share the optimism and try to explain why this is silly, you're an AI sceptic...
True story from within one of the biggest companies in the world.
> Your intellectual fly is open
which makes more sense.
Your title made think that a software project was open sourced or something.
I sometimes see posts on LinkedIn that look LLM generated but also makes me think it's someone who has seen a post, thinking it's a great way to convey a point and tries to adapt the style, without realizing it's a smell.
Also in the author's latest post[0], they took this[1] self-reported preference poll seriously, which makes it very hard for me to take their articles seriously.
[0] https://bcantrill.dtrace.org/2026/09/05/the-revolt-of-the-re...
[1] https://writethatblog.substack.com/p/dev-reaction-to-ai-blog...
I did writing courses as elective in college, the biggest lost truth is that there is no one correct writing style, (you build) being a writer at any capacity means cultivating your own writing voice, (yours) which is an expression of who you are. (voice)
Also, non-native English speakers have to use LLMs to share their views so that they are not judged on their writing.
This post is also a good example; it's well written, and I enjoyed it. The idea could have been expressed simply as: "People can smell LLMs in your writing, it makes you look disingenous"
Who decide what is obvious or not?
Unless I'm directly interacting with the LLM, someone should endorse the content.
A) An intense dislike for all things AI.
B) A subsequent intense dislike when speech patterns indicate AI is speaking.
It seems very much like other times in human history when one group had an intense dislike for another group, typically based in tribalism.
I also use AI to colorize old black and white pictures when discussing historical events in my blog, as well as the occasional AI enhancement of an old grainy and/or blurry photo.
This legal trick only works for rewriting an article reporting on a factual event—since the events are uncopyrightable facts, the only part of the article which can be copyrighted is the stylistic writing.
Let me quote from a recent legal opinion on AI summaries (The New York Times Company v. Microsoft Corporation et al 2025):
>>> Exhibit 11 to the CIR complaint provides website links to articles that CIR alleges were unlawfully abridged by defendants in their ChatGPT and Copilot outputs. (CIR, FAC Ex. 11.) Examining the similarities between those outputs and the corresponding CIR articles, including the “total concept and feel, theme . . . sequence, pace, and setting,” Williams v. Crichton, 84 F.3d 581, 588 (2d Cir. 1996), the Court concludes that the “abridgments” contained in Exhibit 11 are not substantially similar to CIR’s copyrighted works as a matter of law.
The alleged abridgments are detailed summaries, usually in bullet point form, of the facts contained in CIR’s articles. Those summaries—which differ in style, tone, length, and sentence structure from CIR’s articles—are not “substantially similar” to CIR’s copyrighted works. They present the “facts in a different arrangement”—bullet point lists or short summary paragraphs—“with a different sentence structure and different phrasing.” Nihon, 166 F.3d at 71. In short, the abridgments in Exhibit 11 are not substantially similar, qualitatively or quantitatively, to the original CIR articles as a matter of law. The Court therefore grants OpenAI’s motion to dismiss CIR’s claim of direct infringement under 17 U.S.C. § 501 insofar as it relates to the “abridgments” contained in Exhibit 11.<<<
Here’s an example: https://samboy.github.io/blog/entries/2026-08-28.html
The links are AI summaries, which, in turn, link to the original articles, but, in some cases, the original articles are paywalled. For the ones which aren’t paywalled, having a local summary prevents link rot.
Fair use covers quoting someone to comment on them. For example, in New Era Publications International, ApS v. Henry Holt and Co., it was ruled that quoting Hubbard saying “The trouble with China is, there are too many Chinks here.” was fair use, since the book in question was commenting on Hubbard’s personality, and could only reasonably do so by directly quoting him.
From that decision:
>>> these brief quotations from unpublished copyrighted work display a compelling fair use purpose. [...] These quotations are in mockery, to show Hubbard's bigotry, bias and coarse lack of taste. This is not an instance of the biographer/critic free riding on the creative talent of the subject. <<<
And of course anyone who actually reads stuff is disgusted by it, so we quickly end up with a situation where content is piling up and nobody is consuming it which is basically dead internet theory.
On LinkedIn specifically I went from reading it regularly to almost never touching it, previously the nonsense (“the interviewer was the dog”) was still tolerable enough to flip through for updates and I found the platform useful for business leads. Now it’s just a feed of pure slop, when I do open it I just close again after reading a post is two when I remember how bad it is.
As soon as I sense Slop, I'm done, this person refused to think when writing, why should I waste my time reading it then?
However, it is perfectly possible to have an LLM imitate an existing corpus of writing (yours or someone else) and with a good prompt, idea, and editing, to produce high quality writing (in every sense of the word) with an LLM.
I mean, i check it to see if a recruiter has something interesting but otherwise?
my english writing is quite bad. if i write article i myself cannot read it. ai helps with the flow, review editing.
nowadays i record audio, stt and then let ai flow it properly.
i hate blog.md like this, given a topic and it does everything.
Maybe not. Could just be kind of accidentally ironic with the author picking up LLM quirks from using them a lot.
Funny in either case.
If this is straight out of an LLM I am impressed
One of my neighbors sent a letter to the HOA president, who refused to even acknowledge it, on the grounds that it was written too well and so must have been AI assisted. I don't know if that was true, and don't much care, because I do know it contained valid issues that deserved a response.
Not.. really?
I kinda see the point you seem to want to make, but the "no U" opener makes it hard to do that.
Beside that, HOAs - from what I heard of them - will use any reason they can make up to ignore what you want from them, so I'm not sure if that has anything to do with LLMs.
The message of the post is what _should_ hopefully matter to a reader.
And, not the fact that it was written by an LLM.
I am a technical guy at heart; also an introverted extrovert. I hate writing docs that are to be written to satisfy someone else's metrics. For it to be a tickbox'ed item.
I delegate that to an LLM. I want to spend my time solving interesting challenges instead.
So, Mr. Cantrill, you got a problem with that? So be it.
I just can't. LLM assisted posts are often so long, and the hints are usually obvious right at the start.
They're so unpleasant to read. "It's not X, it's Y" etc are bad because the comparisons rarely add anything at all to the message. It's filler that wastes my time. At that point I'd rather see your original prompt.
Why should I invest several minutes of my day reading something if I've already seen evidence that the author doesn't respect my time?
It's not that an LLM wrote it so much as the author couldn't be bothered to clean it up and remove the garbage cliches before publishing it.
It's just digital pollution.
The ship has not sailed, I can just find a human who wrote words, and read them. And if they aren't human words, I will just, not read them.
No. I am interested in what a human has to say, not a clanker. If you don't wish to write that is fine, but don't hand it off to the slop machine and present it as though you did anything of value.