I fear in the future people will read even less because of how often people cheap out and use AI to write.
I know the author of this article tells you never to use a single word it suggests, however I still, while reading, am checking for any amount of AI used.
This makes reading less enjoyable, and even a bit stressful.
If you can't spend the time to write it, why should anyone read it?
Everything I post is from my own finger tips, only to get a reply from someone using a bot. I'm tired.
Excessive generative AI output devalues all communication media unfortunately. As a greater and greater portion of text, imagery, audio, and video come to be synthetic, I don't see why anybody would be motivated to do deep reading, or listening, or analysis of said media. It's frightening to think what that might do to our civilization, almost like putting us back into a weird version of the ancient past when nobody was literate, save the priest.
Maybe humans can adapt to care about generated content and treat it with the same emotional weight that human communication has. I don't know if that's good or not.
> LLM can produce texts but we as people will not be able to understand anyway
LLM generated text oftentimes just do not have the point to be understood. LLM is training you to NOT read, because trying to do close reading of LLM generated text is usually a mistake. There is no deeper there. Either it was correct on the surface things or it is a gibberish.
> If you can't spend the time to write it, why should anyone read it?
Because the value of writing are not the words on a piece of paper but the idea they convey.
Politicians speech are not worth listening too because they have a copywriter polish them?
Teacher assistant homeworks are not worth to do, because they are made by a TA and not by the course professor?
Of course there are writing as: "Hey ChatGPT write me a piece around coding with LLMs" and yeah, those are not worth reading.
But most of the writing is usually, these are the ideas, this is how the idea are structured together, now let's review them and then let's get a nice prose out of it.
> If you can't spend the time to write it, why should anyone read it?
That's actually a legitimate question, and one I've been thinking about, as I've been using LLM tools to compile factual information based on various search tools it uses. The workflow gets sources, fact checks them, filters them according to my needs, and gives me clean urls for source information to hand off to an editor for writing a short listing. It's very hard for the editor to do a much better job at this type of writing; the real benefit of the work was in the research, which was handled by the LLM. The rest is basically just putting the facts in the right order, with some guidance on style, preferred word choices, etc. I'm not comfortable with this at all, as giving work to an LLM that a human could do strikes me as wrong. But using LLM tools gives me a big competitive advantage (for now), in the space I'm doing this in.
>I'm not comfortable with this at all, as giving work to an LLM that a human could do strikes me as wrong
Where does this feeling come from? Is it just because the work is complex? We’re all (excepting maybe the amish) comfortable with using at least some automation, be it mechanical or digital, elsewhere in our lives
Not sure. I don't like LLMs inserting itself into human to human communication, as it can shape our thinking in many ways. Delegating this also makes it much harder for a person to become a subject-matter expert, which is a real loss.
> compile factual information based on various search tools it uses. The workflow gets sources, fact checks them, filters them according to my needs, and gives me clean urls for source information to hand off to an editor for writing a short listing.
What’s preventing the editor from using one of your prompts/tooling and replacing you entirely?
There’s zero value in sharing ai-generated output unless your ai has access to something nobody else’s does on which to base things. If all it’s doing is web research you should share your prompt instead. That way people don’t have to deal with ai-written walls of text.
> What’s preventing the editor from using one of your prompts/tooling and replacing you entirely?
Well, they don't have the many other pieces of the business in place that makes this work meaningful. They certainly could try, though. If they succeeded, this aspect wouldn't be what made the difference, though.
People on my FOSS project couldn't see the issue with their AI slopped documentation contributions.
On one hand I'm like, yeah fair enough it needed writing and something is better than nothing.
On the other I know nobody wants to read it now because it's just word vomit rather than how a human would write something they'd like another human to read.
More than the average person, I suppose. For some reason, AI writing isn't as irritating to me as it seems to be for everyone else.
I treat it the same way I treat human writing: by speed reading most of the article so I can find the points the author is trying to make and decide whether I should read the entire thing.
I don't ever watch TV or streaming shows anymore. It's been a huge improvement in my life. It's the same with reading. Reading is mostly a waste of my time, with some notable exceptions. I'm glad I spend way less time doing each of these!
It's sad that reading has become that way for you.
I love reading, but I mostly read non-fiction and I'm a bit of an amateur history buff. Last time I read a fiction book I was left feeling like I was preached to by an alien about how life doesn't really matter (sci-fi authors will often do that). The state of literature is pretty dire.
Just read people you can trust. Why are you giving yourself all this busy work?
Just make your own tree of material to read. You should have a couple dozen writers/journalists you admire, look at who these people read then start reading those people, then you repeat the cycle.
Don't waste your time reading anything new, most new things are bad. Most ideas we struggle with, are ideas we have been discussing since the written record was invented out of necessity.
Anything new worth reading will assuredly make your way to you if you are at all connected.
Even if you don't want style advice from an LLM (I generally don't, not for blogging), for a technical blogger "please check this post for factual accuracy" is a good way to not look like an idiot. Often it complains about hyperbole, which I tend to ignore since there's no point blogging unless you have some firm opinions, but on at least one occasion it pointed me towards some documentation that undermined my entire post - and led to a much better one a week later once I'd processed and applied that new information. Thanks, LLM.
I do this all the time. It's shocking how well it works. A year or two ago suggesting that someone use LLMs for fact checking would get you laughed out of a room, but today the good ones (Fable, GPT-5.6/6) with a search tool enabled are genuinely excellent at this.
They can even retrieve articles you link to and check that you aren't misrepresenting them.
Hyperbole = strong opinions? If you're stating your 'strong opinions' as fact, then you deserve the push back. If you're stating your strong opinions as strong opinions, then fair game.
You say this but fact and opinion are very similar and not at all easy to distinguish.
For instance, if I said the grass is green vs if I said the grass looks nice vs if I said the grass is doing well and looks how it should vs if I said in my opinion the grass is healthier than its ever been.
If I want to say that iOS is useless for developing Bluetooth 5 features then this is true for all practical purposes and doesn’t deserve elaboration in my text, even if Apple’s API technically exposes one or two tiny pieces.
> Present the original and new writing to the model and ask it which is better.
LLMs have absolutely terrible taste when it comes to writing. I don't find their feedback useful at all, beyond trivial spelling/grammar mistakes, which you don't really need an LLM for in the first place.
Proofreading is all you need.
Edit: I do sometimes ask an LLM for a fact-check, though.
Yeah, the best advice is: Don't. Write it yourself; make judgements; learn.
In general use today, AI primarily functions as a stupidity amplifier: it empowers dumb people to do dumb things more pervasively, with less effort, and at unprecedented scale.
We're moving into an era of artificially enhanced stupidity.
As a "proof" of their poor judgement, take a paragraph you like. Ask the LLM to rewrite it to make it better (which I think we both agree will not make it better), and then in a fresh session ask it which it thinks is best. It'll almost always pick its own writing, even when it sucks.
You are specifically recommending asking the model which of two versions is better (the quote in my top-level comment).
We both agree that they are poor "make it better" machines, but I also believe they are bad A/B testers and I'm using the former to demonstrate the latter.
It doesn't matter who writes what, what matters is that LLMs have a preference for LLM-shaped writing. By A/B testing against an LLMs opinion, you are optimizing in the direction of LLM prose even if the LLM never writes any of the prose itself.
LLM style is not literally anticorrelated with quality. There are some things that they tend to do poorly, but you're not going to do those because you're writing the text yourself. If you have it judge your writing, and are careful to avoid the failure modes that the post goes into, it can be helpful by serving as a competent editor that doesn't share your blind spots.
I don't think it's a useful substitute for actual proofreading though, unless you're truly in a time crunch. When I proofread I don't just look for mistakes, I try to put myself into the position of a prospective audience member. LLMs seem to have absolutely no concept of "theory of the reader's mind". If I'm going for extra high effort, I have friends read it, usually asking them to identify anything that was unclear or hard to follow.
I think it's good advice. I find it a bit funny the author said to not use a single word coming out of the LLM, but they probably did just that because I never heard a human being say "load-bearing" before but now it's all over the place, also used here. Or I don't get the irony.
It’s good—better than the much-revered Elements of Style, imho, which is overly prescriptive and frankly unhelpful at turns, even if it’s well written.
But my all-time favorite is Virginia Tufte’s Artful Sentences: Syntax as Style. It illustrates the variety of English sentence structures in an ordered way with a plethora of examples. The approach is not prescriptive; rather, it allows you inductively to develop a feel for the syntactical rhythms of the language.
Also, I love Stanley Fish’s How to Write a Sentence: And How to Read One. It’s a lovely, entertaining little book, filled with Fish’s insights as an eminent literary critic. What I appreciate most about the book is that Fish walks you through a number of examples and explains how and why they work, so you can walk away as both a better writer and a better reader.
Yeah it's very short, but I can't remember if I read all of it, or just flipped through pieces. It has good advice and I think fairly "famous." (edit: this reminded me I still have my copy :-D)
I know someone in a profession that does a lot of writing, and it blew my mind how clearly feedback was communicated by a superior. Made me wish tech people had better written communication skills. ;-;
Can't vouch for that but I can vouch for `Dreyer's English` as both hilarious (if you're the sort of person who relishes finding le mot juste, like me) and excellent as a style guide.
The frontier live voice models (especially GPT Voice/Live or whatever they call it this week) do not have the typical AI-tells. So I wonder about a few things:
Why is this?
Why couldn’t whatever went into post-training these voice models be applied to the text-gen models to get rid of AI smells?
Could we set up a pipeline that leverages the voice models to clean up text that reeks of AI ?
Till today, I was doing the complete opposite of what the article suggests. Definitely going to try this approach out, pretty helpful.
I have a specific ChatGPT chat that rewrites my posts really well every time; I usually just give it a rough draft and let it clean things up. But it was pretty interesting to realize how good these LLMs are at actually finding mistakes rather than just rewriting things. The point about not letting the model influence your voice was also very, very relevant.
I feel like it normalizes writing faux pas; any use of a semicolon could have just been a new sentence.
I’m thinking that people that get years semi-colon and colon filled paragraphs are going to believe this is a normal and good way to write on their own.
I've seen this in action as well. The biggest value by far is seeing the overused phrases and tics in your writing style. The prompt I like is to tell it to highlight and flag passive voice and run ons since those are my major pain points. I can also tell when its getting out of "real" advice since it'll offer the same "trim 10-15%". Also important to tell it to highlight specifics otherwise it'll give vague editorial advice that isn't anchored. Another fun exercise is to submit other authors writing as well. Say a short story from a magazine. Some of the edits are legit, but do it enough times and you'll start to see the same generic writing advice gets applies to just about any input.
another iteresting thing models can do is to highlight and notice themes and motifs in a piece. I find that helpful when thinking about a piece of work. Might be obvious ones, sometimes it can be subtle themes. Good for stirring your own brain.
Simple example from this piece: I scrubbed 7 instances of the word “just” from it. Obviously, Emacs can do that on their own, but you have to know to look for that word, and every other one.
I’ve used llms in two ways while writing fiction. First I’ll have a story idea and use the llm to explore what has come before that may be related; then I’ll go read some stuff. This has worked great for seeing where my ideas fit and flushing them out. Second I’ll describe settings or plot devices and use the llm to brainstorm consequences or further ideas. This fails miserably. The llms suggestions are devoid of creativity. This is all moot though because I’m really not a writer, I’ve just always thought I might like to be one someday.
I mostly use LLMs for coding, so this is an interesting use case. I’m curious how long an article it can handle in practice before the context starts becoming an issue.
Indeed. I have been experimenting with using frontier models for creative writing.
Using adversarial models to evolve ideas. I have a whole novel of material, and the feedback has indeed been useful. But the actual novel? Not even AI models like their own writing once you give the simple prompt of "account for model sycophancy".
The resulting structure and plot beats are good with AI as copy editor. Only the least discerning reader might not care or notice the actual non-human writing. And hey - maybe those are your readers. Probably not.
Can we all appreciate for a moment the sci-fi dystopia which snuck up on us all? I just used to phrase: " non-human writing". 50 years ago, that would send a chill up your spine.
At one point in my graduate college studies, I tutored qualified (DEI) undergrads 'for free' (for them) as part of a work-study program which paid me a bit for that work.
OTOH, an AI could not have qualified as such a student. But even if it did, before we began I would certainly bargain for considerably more than work-study tutoring paid.
This is similar to how I've come to use them but it's focused too much on the mechanics or writing. I've found them more valuable at role-playing as a specific profile of reader and asking it how they would receive the arguments advanced by the writing and what counterarguments they could present. After some rounds of this process, you either close all the gaps in the argument or realize it's not a strong one.
If English is your second language and you intent to write in your mother tongue, my advice is to use English to communicate with the LLM.
This makes Rule 1 straightforward ("You may not use a single word an LLM suggests to you") and helps avoid the horrible feeling that the machine is taking over your personal voice.
It's darkly amusing that while the author advises us not to let LLMs pick our words, their post includes this chunk of AI slop:
"Those rethinks are load-bearing parts of your voice. Readers won’t put their fingers on what’s wrong, but they’ll sense that you’ve become artificially-flavored."
The metaphor at the start is AI-like weirdness too: "without pasteurizing and jacking it with corn syrup." Pasteurising is good, corn syrup is bad - what the heck am I supposed to make of it? It's nonsense.
It may even have been handwritten by the author, but it's slop nonetheless.
You may be telling yourself that you're disciplined and you're just using the LLMs for review. But while you're reading their output, the slop makes its way directly into your brain anyway - and then it drips out into your writing.
I saw "load-bearing" there and instantly assumed Thomas was making a deliberate point that it's OK to use terms that LLMs have poisoned provided you use them effectively and in a way that adds to the piece.
I'm personally furious that "load-bearing" has become an LLM tell, it's a really useful metaphor!
"Those rethinks are load-bearing parts of your voice."
Loved the (possibly not deliberately ironic) use of the LLM'S favourite analogy, load-bearing.
Nice wee article though. I particularly agree with "Don’t take all of the model’s copyediting advice." I've been using LLMs to review teaching material and often it just doesn't get the point of teaching. It's good for prompting reflection on your writing but it's still an idiot savant.
I fear in the future people will read even less because of how often people cheap out and use AI to write. I know the author of this article tells you never to use a single word it suggests, however I still, while reading, am checking for any amount of AI used. This makes reading less enjoyable, and even a bit stressful.
If you can't spend the time to write it, why should anyone read it?
Everything I post is from my own finger tips, only to get a reply from someone using a bot. I'm tired.
Excessive generative AI output devalues all communication media unfortunately. As a greater and greater portion of text, imagery, audio, and video come to be synthetic, I don't see why anybody would be motivated to do deep reading, or listening, or analysis of said media. It's frightening to think what that might do to our civilization, almost like putting us back into a weird version of the ancient past when nobody was literate, save the priest.
Maybe humans can adapt to care about generated content and treat it with the same emotional weight that human communication has. I don't know if that's good or not.
There is a recent PISA research that students slowly decline in understanding the long reads.
E.g. in one country, 21% of high school students do not even achieve “sufficient reading literacy”…
So the question should be - what are we going to do with this? LLM can produce texts but we as people will not be able to understand anyway…
About PISA research for example here: https://www.edweek.org/leadership/reading-ability-craters-on...
> LLM can produce texts but we as people will not be able to understand anyway
LLM generated text oftentimes just do not have the point to be understood. LLM is training you to NOT read, because trying to do close reading of LLM generated text is usually a mistake. There is no deeper there. Either it was correct on the surface things or it is a gibberish.
You can't even watch funny animal compilations anymore!
Probably the worst part
> If you can't spend the time to write it, why should anyone read it?
Because the value of writing are not the words on a piece of paper but the idea they convey.
Politicians speech are not worth listening too because they have a copywriter polish them?
Teacher assistant homeworks are not worth to do, because they are made by a TA and not by the course professor?
Of course there are writing as: "Hey ChatGPT write me a piece around coding with LLMs" and yeah, those are not worth reading.
But most of the writing is usually, these are the ideas, this is how the idea are structured together, now let's review them and then let's get a nice prose out of it.
"Politicians speech are not worth listening too because they have a copywriter polish them?"
It often makes them less informative and a lot less sincere.
> If you can't spend the time to write it, why should anyone read it?
That's actually a legitimate question, and one I've been thinking about, as I've been using LLM tools to compile factual information based on various search tools it uses. The workflow gets sources, fact checks them, filters them according to my needs, and gives me clean urls for source information to hand off to an editor for writing a short listing. It's very hard for the editor to do a much better job at this type of writing; the real benefit of the work was in the research, which was handled by the LLM. The rest is basically just putting the facts in the right order, with some guidance on style, preferred word choices, etc. I'm not comfortable with this at all, as giving work to an LLM that a human could do strikes me as wrong. But using LLM tools gives me a big competitive advantage (for now), in the space I'm doing this in.
>I'm not comfortable with this at all, as giving work to an LLM that a human could do strikes me as wrong
Where does this feeling come from? Is it just because the work is complex? We’re all (excepting maybe the amish) comfortable with using at least some automation, be it mechanical or digital, elsewhere in our lives
Not sure. I don't like LLMs inserting itself into human to human communication, as it can shape our thinking in many ways. Delegating this also makes it much harder for a person to become a subject-matter expert, which is a real loss.
> compile factual information based on various search tools it uses. The workflow gets sources, fact checks them, filters them according to my needs, and gives me clean urls for source information to hand off to an editor for writing a short listing.
What’s preventing the editor from using one of your prompts/tooling and replacing you entirely?
There’s zero value in sharing ai-generated output unless your ai has access to something nobody else’s does on which to base things. If all it’s doing is web research you should share your prompt instead. That way people don’t have to deal with ai-written walls of text.
> What’s preventing the editor from using one of your prompts/tooling and replacing you entirely?
Well, they don't have the many other pieces of the business in place that makes this work meaningful. They certainly could try, though. If they succeeded, this aspect wouldn't be what made the difference, though.
> If you can't spend the time to write it, why should anyone read it?
100%. We're producing more content, faster, than ever before but ... most of it is just click-producing garbage that won't stand the test of time.
People on my FOSS project couldn't see the issue with their AI slopped documentation contributions.
On one hand I'm like, yeah fair enough it needed writing and something is better than nothing.
On the other I know nobody wants to read it now because it's just word vomit rather than how a human would write something they'd like another human to read.
I hate it so much.
> If you can't spend the time to write it, why should anyone read it?
Because the information in the post could be valuable regardless of human or LLM involvement.
But how much slop are you willing to wade through, hoping to find a nugget of insight?
More than the average person, I suppose. For some reason, AI writing isn't as irritating to me as it seems to be for everyone else.
I treat it the same way I treat human writing: by speed reading most of the article so I can find the points the author is trying to make and decide whether I should read the entire thing.
I don't ever watch TV or streaming shows anymore. It's been a huge improvement in my life. It's the same with reading. Reading is mostly a waste of my time, with some notable exceptions. I'm glad I spend way less time doing each of these!
It's sad that reading has become that way for you.
I love reading, but I mostly read non-fiction and I'm a bit of an amateur history buff. Last time I read a fiction book I was left feeling like I was preached to by an alien about how life doesn't really matter (sci-fi authors will often do that). The state of literature is pretty dire.
Agreed completely. (This reply took me 0.4 seconds to generate.)
Just read people you can trust. Why are you giving yourself all this busy work?
Just make your own tree of material to read. You should have a couple dozen writers/journalists you admire, look at who these people read then start reading those people, then you repeat the cycle.
Don't waste your time reading anything new, most new things are bad. Most ideas we struggle with, are ideas we have been discussing since the written record was invented out of necessity.
Anything new worth reading will assuredly make your way to you if you are at all connected.
Even if you don't want style advice from an LLM (I generally don't, not for blogging), for a technical blogger "please check this post for factual accuracy" is a good way to not look like an idiot. Often it complains about hyperbole, which I tend to ignore since there's no point blogging unless you have some firm opinions, but on at least one occasion it pointed me towards some documentation that undermined my entire post - and led to a much better one a week later once I'd processed and applied that new information. Thanks, LLM.
I do this all the time. It's shocking how well it works. A year or two ago suggesting that someone use LLMs for fact checking would get you laughed out of a room, but today the good ones (Fable, GPT-5.6/6) with a search tool enabled are genuinely excellent at this.
They can even retrieve articles you link to and check that you aren't misrepresenting them.
I do something like this with code all the time, “is this idiomatic <foo>?”, I find it very helpful.
Hyperbole = strong opinions? If you're stating your 'strong opinions' as fact, then you deserve the push back. If you're stating your strong opinions as strong opinions, then fair game.
You say this but fact and opinion are very similar and not at all easy to distinguish.
For instance, if I said the grass is green vs if I said the grass looks nice vs if I said the grass is doing well and looks how it should vs if I said in my opinion the grass is healthier than its ever been.
> If you're stating your 'strong opinions' as fact
Is there anybody on this site who isn't?
If I want to say that iOS is useless for developing Bluetooth 5 features then this is true for all practical purposes and doesn’t deserve elaboration in my text, even if Apple’s API technically exposes one or two tiny pieces.
It's extremely refreshing reading good human prose isn't it! It's like when I would read a newly-written Paul Graham essay 20 years ago.
He also uses rule of three and the phrase "load-bearing", just for fun, which elicited a wry smile from me.
Thanks OP for a pleasant start to my day!
> Present the original and new writing to the model and ask it which is better.
LLMs have absolutely terrible taste when it comes to writing. I don't find their feedback useful at all, beyond trivial spelling/grammar mistakes, which you don't really need an LLM for in the first place.
Proofreading is all you need.
Edit: I do sometimes ask an LLM for a fact-check, though.
Yeah, the best advice is: Don't. Write it yourself; make judgements; learn.
In general use today, AI primarily functions as a stupidity amplifier: it empowers dumb people to do dumb things more pervasively, with less effort, and at unprecedented scale.
We're moving into an era of artificially enhanced stupidity.
That’s what this is: it’s mechanical proofing.
As a "proof" of their poor judgement, take a paragraph you like. Ask the LLM to rewrite it to make it better (which I think we both agree will not make it better), and then in a fresh session ask it which it thinks is best. It'll almost always pick its own writing, even when it sucks.
Right, I agree: that’s exactly not what I’m recommending.
You are specifically recommending asking the model which of two versions is better (the quote in my top-level comment).
We both agree that they are poor "make it better" machines, but I also believe they are bad A/B testers and I'm using the former to demonstrate the latter.
You are writing both paragraphs. You’re specifically not asking a model to make a better paragraph. That would be a load-bearing debacle.
It doesn't matter who writes what, what matters is that LLMs have a preference for LLM-shaped writing. By A/B testing against an LLMs opinion, you are optimizing in the direction of LLM prose even if the LLM never writes any of the prose itself.
LLM style is not literally anticorrelated with quality. There are some things that they tend to do poorly, but you're not going to do those because you're writing the text yourself. If you have it judge your writing, and are careful to avoid the failure modes that the post goes into, it can be helpful by serving as a competent editor that doesn't share your blind spots.
I don't think it's a useful substitute for actual proofreading though, unless you're truly in a time crunch. When I proofread I don't just look for mistakes, I try to put myself into the position of a prospective audience member. LLMs seem to have absolutely no concept of "theory of the reader's mind". If I'm going for extra high effort, I have friends read it, usually asking them to identify anything that was unclear or hard to follow.
I think it's good advice. I find it a bit funny the author said to not use a single word coming out of the LLM, but they probably did just that because I never heard a human being say "load-bearing" before but now it's all over the place, also used here. Or I don't get the irony.
The "load bearing" term has become a bit of a meme, given the author is familiar with these AI-isms, we can assume it's a subtle joke.
I agree with the central thesis, but disagree with some details. I wrote a blog response: https://writing.brlewis.com/posts/2026/llm-copyeditor/
This article was so much better than what I though it would be
Does anyone know about "Style: Lessons in Clarity and Grace" ? Seems like an interesting book
It’s good—better than the much-revered Elements of Style, imho, which is overly prescriptive and frankly unhelpful at turns, even if it’s well written.
But my all-time favorite is Virginia Tufte’s Artful Sentences: Syntax as Style. It illustrates the variety of English sentence structures in an ordered way with a plethora of examples. The approach is not prescriptive; rather, it allows you inductively to develop a feel for the syntactical rhythms of the language.
Also, I love Stanley Fish’s How to Write a Sentence: And How to Read One. It’s a lovely, entertaining little book, filled with Fish’s insights as an eminent literary critic. What I appreciate most about the book is that Fish walks you through a number of examples and explains how and why they work, so you can walk away as both a better writer and a better reader.
Yeah it's very short, but I can't remember if I read all of it, or just flipped through pieces. It has good advice and I think fairly "famous." (edit: this reminded me I still have my copy :-D)
I know someone in a profession that does a lot of writing, and it blew my mind how clearly feedback was communicated by a superior. Made me wish tech people had better written communication skills. ;-;
Can't vouch for that but I can vouch for `Dreyer's English` as both hilarious (if you're the sort of person who relishes finding le mot juste, like me) and excellent as a style guide.
For commit messages and other technical writing, I use a "narrative coherence" checklist to help ensure that a change is clear and understandable:
- Establish a single thesis an external reviewer could recover from the diff.
- Unify vocabulary across code, comments, tests, and commit message.
- Use the same vocabulary consistently to refer to the same concepts.
- Keep every hunk that serves the thesis; consider the removal or deferral of the rest.
- Introduce abstractions at the point of need.
- Align tests to narrate the same story as the implementation.
- Reconcile the commit message and the diff.
- Order changes expositorily, not chronologically.
- Explain what the change does and why. Do not explain the details of the development process.
- Prefer to edit subtractively.
- Recompile and run all relevant tests after edits.
- Iterate.
That checklist sounds like at least as much work as writing a commit message.
The frontier live voice models (especially GPT Voice/Live or whatever they call it this week) do not have the typical AI-tells. So I wonder about a few things:
Why is this?
Why couldn’t whatever went into post-training these voice models be applied to the text-gen models to get rid of AI smells?
Could we set up a pipeline that leverages the voice models to clean up text that reeks of AI ?
Till today, I was doing the complete opposite of what the article suggests. Definitely going to try this approach out, pretty helpful.
I have a specific ChatGPT chat that rewrites my posts really well every time; I usually just give it a rough draft and let it clean things up. But it was pretty interesting to realize how good these LLMs are at actually finding mistakes rather than just rewriting things. The point about not letting the model influence your voice was also very, very relevant.
I feel like it normalizes writing faux pas; any use of a semicolon could have just been a new sentence.
I’m thinking that people that get years semi-colon and colon filled paragraphs are going to believe this is a normal and good way to write on their own.
I've seen this in action as well. The biggest value by far is seeing the overused phrases and tics in your writing style. The prompt I like is to tell it to highlight and flag passive voice and run ons since those are my major pain points. I can also tell when its getting out of "real" advice since it'll offer the same "trim 10-15%". Also important to tell it to highlight specifics otherwise it'll give vague editorial advice that isn't anchored. Another fun exercise is to submit other authors writing as well. Say a short story from a magazine. Some of the edits are legit, but do it enough times and you'll start to see the same generic writing advice gets applies to just about any input.
another iteresting thing models can do is to highlight and notice themes and motifs in a piece. I find that helpful when thinking about a piece of work. Might be obvious ones, sometimes it can be subtle themes. Good for stirring your own brain.
> overused phrases and tics in your writing style
In the extreme these can be annoying, but more often I find them kinda charming, as an expression of the author's personality.
Simple example from this piece: I scrubbed 7 instances of the word “just” from it. Obviously, Emacs can do that on their own, but you have to know to look for that word, and every other one.
You are erasing yourself in doing this, not improving your writing.
If you want to improve your writing, read more.
Agreed and depends on the writing. It's a very helpful tool though to be able to flag and highlight that sort of thing
I’ve used llms in two ways while writing fiction. First I’ll have a story idea and use the llm to explore what has come before that may be related; then I’ll go read some stuff. This has worked great for seeing where my ideas fit and flushing them out. Second I’ll describe settings or plot devices and use the llm to brainstorm consequences or further ideas. This fails miserably. The llms suggestions are devoid of creativity. This is all moot though because I’m really not a writer, I’ve just always thought I might like to be one someday.
How to write with an LLM?
Don’t
I mostly use LLMs for coding, so this is an interesting use case. I’m curious how long an article it can handle in practice before the context starts becoming an issue.
Indeed. I have been experimenting with using frontier models for creative writing.
Using adversarial models to evolve ideas. I have a whole novel of material, and the feedback has indeed been useful. But the actual novel? Not even AI models like their own writing once you give the simple prompt of "account for model sycophancy".
The resulting structure and plot beats are good with AI as copy editor. Only the least discerning reader might not care or notice the actual non-human writing. And hey - maybe those are your readers. Probably not.
Can we all appreciate for a moment the sci-fi dystopia which snuck up on us all? I just used to phrase: " non-human writing". 50 years ago, that would send a chill up your spine.
At one point in my graduate college studies, I tutored qualified (DEI) undergrads 'for free' (for them) as part of a work-study program which paid me a bit for that work.
OTOH, an AI could not have qualified as such a student. But even if it did, before we began I would certainly bargain for considerably more than work-study tutoring paid.
> use them like a copyeditor rather than a ghostwriter
This is what you need to know. Seriously, after 25 years in news (writing), you realize that you need “second eyes”, not the first hands.
This is similar to how I've come to use them but it's focused too much on the mechanics or writing. I've found them more valuable at role-playing as a specific profile of reader and asking it how they would receive the arguments advanced by the writing and what counterarguments they could present. After some rounds of this process, you either close all the gaps in the argument or realize it's not a strong one.
If English is your second language and you intent to write in your mother tongue, my advice is to use English to communicate with the LLM.
This makes Rule 1 straightforward ("You may not use a single word an LLM suggests to you") and helps avoid the horrible feeling that the machine is taking over your personal voice.
It's darkly amusing that while the author advises us not to let LLMs pick our words, their post includes this chunk of AI slop:
"Those rethinks are load-bearing parts of your voice. Readers won’t put their fingers on what’s wrong, but they’ll sense that you’ve become artificially-flavored."
The metaphor at the start is AI-like weirdness too: "without pasteurizing and jacking it with corn syrup." Pasteurising is good, corn syrup is bad - what the heck am I supposed to make of it? It's nonsense.
It may even have been handwritten by the author, but it's slop nonetheless.
You may be telling yourself that you're disciplined and you're just using the LLMs for review. But while you're reading their output, the slop makes its way directly into your brain anyway - and then it drips out into your writing.
I saw "load-bearing" there and instantly assumed Thomas was making a deliberate point that it's OK to use terms that LLMs have poisoned provided you use them effectively and in a way that adds to the piece.
I'm personally furious that "load-bearing" has become an LLM tell, it's a really useful metaphor!
It was a joke.
Recent comments from the simonw account suggest it isn't operated directly by a human anymore /sigh/
What?
The placement of the "Those rethinks are load-bearing parts of your voice" sentence in the article makes me think the author intended the word choice
> Pasteurising is good, corn syrup is bad
Pasteurising something is making it sterile. Safer, but bland and not as interesting. I guess that’s the intended metaphor.
"Those rethinks are load-bearing parts of your voice."
Loved the (possibly not deliberately ironic) use of the LLM'S favourite analogy, load-bearing.
Nice wee article though. I particularly agree with "Don’t take all of the model’s copyediting advice." I've been using LLMs to review teaching material and often it just doesn't get the point of teaching. It's good for prompting reflection on your writing but it's still an idiot savant.
There's no way in hell that wasn't a deliberate joke.
> How to Write with an LLM
TLDR: Don’t.
That's what the piece says.
Ironically, "load-bearing" appears in the main text of the article that says never to use a single word the LLM suggests. Womp womp.
I'm somewhere between 100 and 110% sure the author knew what they were doing.
See https://news.ycombinator.com/item?id=49747070#49748296