How to Write with an LLM
sockpuppet.org
sockpuppet.org
It slows things down a bit, but in the best possible way. It has helped immensely to improve the depth of my understanding of the agent-generated code. When agents are doing everything its way too easy to “skim” diffs and not really absorb them.
I always prided myself on my technical writing, and commit messages and PRs were a great place to hone that skill. I found that I missed it and my work is better now I’ve reclaimed that part of my old job back.
"No AI PR Descriptions" is a great rule because it does 3 things:
1. It's a hard binary that's easy to recognize and enforce
2. It establishes personal ownership for the submitter, making it psychologically difficult to submit something you don't understand. If a person has to describe what a change does, how, and why, then they must have looked at it and made a real attempt to understand it, because you can't map a territory you have never seen. This alone prevents "claude-code run amok" scenarios where devs have completely abdicated responsibility that are a real pain to clean up.
3. It informs the reviewer of the change's intent. A human PR tells reviewers what a thing is intended to do, as well as the authors belief about what it does, which they can use as a framework to critique the actual substance. It becomes easier to notice missed edge cases, code behavior that subtly differs from intent, or changes that would block off or complicate an intended future project direction or reduce maintainability.
And as a bonus, at time of writing, AI pr descriptions basically always suck. They are usually full of irrelevant implementation details, misplaced emphasis, weird presuppositions that seem to come out of nowhere and are really annoying to read, etc. That won't be true forever, they are always getting better, but it is true now, so that's a temporary 4th benefit.
We have one dev who uses LLMs to write the code, but still commits by hand. Most of his messages are of the type above, and none of them are useful.
I have this in my claude.md along with guidance on (not) writing comments but it’s still dumps multi-paragraph comments of Claude speak everywhere.
I think a lot of us are talking past eachother, but what I and I think others are complaining about is doing things well, and being minimal. LLMs don’t really do that yet, they can’t look at and simplify a codebase well, even with guidance. They always seem to add rather than subtract. And eventually it becomes an issue. I think they were actually better in this regard with like Opus 4.8, and are now getting even worse (but score better, and are more autonomous).
At home I use Codex, and it’s better as far as language goes at least.
Anyway, it’s like the logic piece necessary to do good work is still missing, and hurt by recent reinforcement fine-tuning. And it leads to massive bloat since more comments = higher scores, but walls of text and the tokens (or cognition if you’re the sad human reading them) aren’t free.
For PR/commit descriptions, I mainly use Claude Sonnet 4.5. It isn’t perfect, but it produces significantly less of this weird gibberish than 5.x models or even Opus 4.x do
I also use an iterative process in which it writes the description, I read it, and then either manually edit it or ask it to make changes
Models will hallucinate. They are also quite good at spotting hallucinations in other models' output (with some more hallucinations thrown in). With a threshold for confirmation, and a few iteration loops, you arrive at a fixed point where every claim is supported.
Tune your model parameters to what is right for the task, not the highest you can afford.
And even then… writing is personal expression. Here people are talking about commit messages. That’s fine but AI doing writing for anyone and I WILL NOT READ IT unless it’s literally basic tech manual.
We read to hear and engage with people’s thoughts. If someone outsources that to AI then they should be shunned.
So long as there are in fact those things, and so long as it didn't sneak something else in there at the same time, it being just on the knife-edge between sense and word-salad is better than "fix".
Buuuuuut far to often it says it fixed a bug I reported, when it only touched one superficial failure mode rather than the root cause.
Yesterday's issue: Why is zoom/pan randomly failing? It told me it was because it was applying a transformation matrix with every input and sometimes JavaScript gave it a non-invertible matrix (why?) which then propagated NaNs everywhere and you can't update a matrix filled with non-numbers.
Why was it doing that in the first place? Seems to be because it's too motivated to perform quick wins and not sufficiently motivated to do good engineering.
Good thing this was just a game editor. Spiky intelligence: superhuman on some dimensions, total noob on others.
You probably had those for the humans in your team for when new hurts join, right? Just direct your LLMs to them and you’ll have a much better experience.
Just like you can’t expect a new junior dev to know what to do without direction, the same applies with LLMs.
Also:
1. It's often a premature extrapolation from diffs and tickets which you could probably generate later, at need, and likely with even better results.
2. If those commit messages can ever influence future work, then it's just carcinogenic cargo-cult cruft. Pruning outputs (future inputs) is important to prevent weird unwanted drift.
But AI commit messages are still bad. Way too verbose and focused on the wrong level: that of code mechanics. That's just wrong. Messages should focus on the level of intent and design, with the primary purpose being to aid human review. They should include a high level overview of the change, decisions, caveats, information not obvious from reading the diff.
So I wrote a skill that captures these principles and allows the agent to even research past related commits and to ask focused questions in order to uncover the intent rather than guessing and writing a bad one.
Now the AI writes better commit messages than it used to, and even writes better than most humans (who can't be bothered to write a good one). Not better than a good manual message, but you can't have everything.
But sometimes even the good writers are tired or didn't think things through. Being able to compare with the AI's version is still useful.
Here is the skill for anyone interested: https://github.com/FooBarWidget/ai-skills-and-principles/blo...
Depends on my "documentation principles" skill: https://github.com/FooBarWidget/ai-skills-and-principles/blo...
Even before coding harnesses become mainstream, a lot of tools offered to automatically generate commit messages based on the diff (I think JetBrains IDEs started to offer it very early) and I always cringed when I've seen it.
The reason why I don't like diff-base commit messages is because they are redundant. If I want an LLM-generated summary of the diff, I can easily generate it myself, there is absolutely no reason to put it in the commit message.
What I would like to see in the commit message is some additional context that is not a part of the diff. I don't need to read what has changed, because I can already can see it in the diff (or get an LLM to summarize it for me). But I often do need to understand why this change was made. What were they trying to achieve? That's the important part that is not contained in the diff itself. And this is the part that diff-base commits rarely contain.
I think that even something simple like a link to a Jira (or other bugtracker ticket) is much more helpful than diff summary. Maybe instead of "fix" or "update" you could write just a couple of words about what are you trying to fix and why does it need updated. Still better, than a diff-summary.
I usually have it lead with a paragraph or two of context on why the change was made (though I often use it as just a tool to turn my rambling on that question into proper English), followed by a couple paragraphs of "abstract" explaining the change (because an LLM-generated summary reviewed by the person who made the commit is better than something which the person reading can get themselves).
This is the reason (and IMO the only reason) code should be commented. The why can often be hundreds of characters explaining technical limitations at the time, business considerations, or even just "we know this will break in a non-trivial number of instances but it's an incremental step forward to our end goal so we're just doing it."
I have the opposite issue - some of my team members now submit mini-essays generated by the LLM. Like 300-500 word commit messages with everything from the essence of the change up to philosophical design trade off discussions.
Like most writing, what is left out is as important as what is included.
You are not a serious person if you recommend that.
Check back in a couple years and I’m sure they’ll be there but today’s frontier models 100% absolutely are not ready to fully own a nontrivial production codebase with no human involvement.
If your work has little liability then you can afford to not care beyond “does it work”.
If you have to worry about quality and ensuring you don’t get sued, you make sure the process works.
If it has to maintainable, you you need the mental model to be present in someone’s head.
I'm a happy little bottleneck, what's wrong with that? How do you even know how to ask it to build stuff for you if you don't understand what foundations are you building on?
Writing commit message (aka. what has changed) gives me a sense of control that I know what's happening and I can confidently build the next thing.
Commit messages are not exactly in this category for me, I view them mostly as a work log to be later inspected by another LLM to gather context. I do usually review them before approving a given plan, so I do care about their structure and content, but find the LLM sufficiently competent at writing them.
If they already knew what function or detail is involved, they would be already be doing a "all commits that touched this line" filter, and my comment about affecting $thing would probably be superfluous.
I also don't need to tell them such details in the commit, because that is better expressed by the actual diff.
So in a sense, I'm trying to provide good keywords, about transactions, errors, logging, button color, whatever.
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.
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.
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
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.
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.
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.
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 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.
It's not clear what you mean by "sufficient reading literacy," but I guess it's one of the PISA proficiency levels. Those are simply equally-spaced intervals on the bell curve. For example, level 1c in reading is 2.4 to 3.1 standard deviations below the reference mean and 21% of participants in Iraqi Kurdistan and Lebanon scored at that level. Relative to the baseline, they are definitely close to the bottom. However, even they must've gotten some questions right, or they would've ended up in the "below level 1c" bucket (like 3% in Iraqi Kurdistan and 5% in Lebanon did). (https://www.oecd.org/en/publications/pisa-2025-results-volum... Table I.B1.2a.11)
If you wanted to know how many people achieve "sufficient reading literacy," you would need a test specifically designed for measuring it (i.e. not PISA), where people with sufficient ability can be expected to get all questions right. For example, the literacy rate is typically defined via the ability to "read and write a simple statement about everyday life," a test which 99% of adults in wealthy countries pass, 92% in Lebanon and 84% in Iraq: https://ourworldindata.org/grapher/cross-country-literacy-ra...
Of course a test that 99% pass just becomes table stakes and boring, whereas PISA with its large differences between individuals is much more exciting, but you shouldn't let that mislead you into thinking that students scoring relatively low on PISA necessarily must be lacking even those table-stakes skills.
As society/technology/the world moves faster and becomes more complicated, more and more of them will be left behind. 1 in 5 Americans are already functionally illiterate, and about 20% of working age men are already out of the labor force.
Minneapolis Fed: https://www.minneapolisfed.org/article/2022/whos-not-working...
Bipartisan Policy Center: https://bipartisanpolicy.org/article/why-are-prime-age-adult...
Do you have any numbers you'd like to provide to counter the claim that a significant portion of the population of working age individuals aren't already out of the labor force?
Not that I disagree much, but for myself I have an actual cause connecting the two parts, and it's not something inherent to AI.
Basically, synthetic content is like those fancy desserts - they look nice and the top layer possibly has an exquisite flavor (probably on purpose), but once you cut deeper, it's some weird fruit or nut or other foodstuff that belongs next to potatoes and meat, not in a cake. So I learned to avoid such desserts and stick to basic stuff I know are good entirely, like IDK a cream pie or cream-filled chocolate stuff or such.
Same is with content. My problem isn't that it's AI, but that it's wrong, full of mistakes human authors don't make. Hallucinations mid-text that make half of it convincingly argue bullshit. Or the uncanny or repetitive factor in art that kicks in after few seconds of looking.
Like with desert, I'm avoiding disappointment and wasted hopes. When synthetic context gets better, I'll be consuming it just like "vanilla" one, for the same reasons.
I think it's also more about active vrs passive consumption these days. People these days consume media mostly passively (watch whatever is on a streaming service, hear a podcast/audiobook while driving, let AI summarize an article instead of reading it), whereas that wasn't really possible previously (go to a theater to watch a movie, actually read stuff with attention, etc.).
So yes, people might be reading less, but I don't think it's some massive fall in literacy as people make it out to be. We're much more busy and don't have time to actually sit down and actively focus on media consumption these days as well (assuming you're employed in a somewhat capitalism-based system).
False, at least from an American or Northern European perspective.
In the 1800s reading was an extremely popular activity. Not something restricted to a few. Printing press had been around for ages and the majority of the population read.
Literacy rates in the 1800s across different regions in Western Europe: https://brewminate.com/the-growth-of-literacy-in-western-eur...
> According to data compiled by Our World in Data and the World Bank, the literacy rate of the world's population from secondary school age onward was only 12 percent in 1820 - around one person in ten. In 1900, it still barely exceeded 20 percent. From the 1950s on, world literacy began to take off, hitting 42 percent in 1960 and 70 percent in 1983.
https://www.weforum.org/stories/education-and-skills/reading...
For instance, in Sweden, to cherrypick a stat, in 1875 only 1% of military recruits were found unable to read [1].
We really must stop thinking that people of yestertimes where so much inferior to people today. Yes, there were differences between men and women, but this was also starting to change.
[1]: https://history.state.gov/historicaldocuments/frus1876/d308
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 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.
Try to think a little more critically here. If you can't trust the opinion of people you supposedly like, already established in their craft, who can you trust?
Have no idea why you think this would increase influencer culture, but thanks for telling us all publicly how you interact with social media: in an sloppy uncritical way.
My advice for you is to pull the plug and read books that are older than 125 years.
> Have no idea why you think this would increase influencer culture, but thanks for telling us all publicly how you interact with social media: in an sloppy uncritical way.
This is a personal attack that is unsubstantiated by what I wrote, read the guidelines and follow them or go elsewhere.
Reading trusted authors is likely to be valuable, but they may devalue over time. We don't know if our chosen authors maximize our value, nor whether exploratory reading does, etc.
https://en.wikipedia.org/wiki/Exploration%E2%80%93exploitati...
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.
The problem here is that the volume of low-mid content is making it harder to find the best content.
Because the information in the post could be valuable regardless of human or LLM involvement.
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.
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.
One person was arguing that they need llm because english isn't their first language and then it helps them write better and be understood. I was incapable of reading their (extremely long) email of arguments; I tried but my mind just went elsewhere as my eyes were moving through the lines.
At the very end there was a disclaimer that llm had been used to "aid" in writing that very email.
Poor guy thinks he's communicating better but all he's doing is making his emails completely unreadable. It's a sad state of affairs.
However most likely they just don't understand what good communication looks like, LLMs are perfectly capable of turning anything into clear concise communication, in fact an easy tell that something is LLM written is that it's just too precise and clear, humans are much more rambling.
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.
It often makes them less informative and a lot less sincere.
I know lots of AI-enabled writers believe this, but I've never actually seen it be true.
what do you think this process is? why don't we just write lists of ideas to each other, in the raw format you presumably started with? is there any value in it?
my opinion is that the process is literally nothing more than thinking deeply about the ideas and trying to phrase them in the clearest possible way such that you lead someone else into understanding them.
LLMs have awful, terrible judgement of what other people know, or are likely to infer from a piece of text.
This isn’t your thought; you are merely regurgitating a sentence we have all heard many times before, practically word for word.
Presumably you think there is value in posting yet another copy of that other person’s thoughts in this discussion, despite having spent no time or effort in thinking it up yourself.
Why is that? Is it because you think there is inherent value in the sentence you transcribed, in spite of the near zero effort you spent creating it? If you think anybody here should read it, then you have undermined your own point. If you don’t think anybody should read it, then you shouldn’t post it.
If the spouses-to-be at a marriage ceremony say "I do", would you leap up from the audience to denounce them as frauds because obviously they heard the vow-formula before a hundred times? Hopefully not, and not only for reasons of decorum.
In contrast, most of the LLM slop I deal with from coworkers involve statements that--while less formulaic--are ones they can't even recall saying and are unwilling to defend when questioned.
If we go with your interpretation instead, then sm-silversight simply misread the comment they replied to and their post doesn't follow. No one doubted OP felt sincere about not wanting to read AI.
I mean, how can you say that he didn't think it himself? I thought this same thing and realised apparently I'm not the first. You have no idea how many of my blog posts ideas have come in HN first before I could publish. People can have the same thoughts. And have em all the time as well.
Psss.. See how @sm-silversight who commented on your post said the same thing I did? That's realtime example right there.
Coming back to the topic at hand. I agree with the OP. If you don't care to write it, why should anyone read it? AI has good usecases. But am not sure writing is one.
Writing is not just about the end result. It's also about the thought process. I learn about the topic more as I write. I correct my incorrect understandings as I write. When am pissed, my article has a different tone. Am rude, too direct. When am content/happy, different one. All of this is sterilised into something robotic where the author won't even remember what he/she wrote? What's the to like? For brownie points?
You are a senior software architect, you "prompt" 10 junior architects to write a software architecture proposal and then pick one which you send the client.
Would you say it's worthless, because the senior architect didn't care to write it?
There are plenty of industries where "taste" is a real job.
Or you think that for some reason prompting a LLM and forwarding it's output after vetting is different than prompting a human and forwarding the output after vetting.
Lemme ask you a counter question, the senior architect writes the proposal and the junior writes it. Which one do you think client likes better?
Moreover, Your focus is on the end-result. Sure, in corporate environment, most people don't care about what they do as long as the end results checks some boxes. That wasn't the discussion here. The discussion is about writing and reading in general. Whether it's worth my time or not if you don't mean (forget mean, don't even remember) what you wrote.
Also, in your own example, the client don't care as well. As long the client's checks all the boxes his/her/their manager said should satisfy. Nobody really gives a damn about the quality.
Also, are you saying you'll be fine paying for Financial Times, WSJ and NYTimes KNOWING all they do is shove the story through an LLM? Neither will there be a soul, nor will there be any material effect on people who reads it.
> Or you think that for some reason prompting a LLM and forwarding it's output after vetting is different than prompting a human and forwarding the output after vetting.
There is difference no? I forward the output to get good feedback and correction from someone BETTER. And there is a process involved in it which showcases the "TASTE" you mentioned from the person whom you are forwarding your output to be vetted and that person comes up to you with feedback. And guess what? You learn from the person's feedback. You don't go back again and again forever to learn the same thing. That's exactly what am trying to point out.
Maybe I just want the news.
> nor will there be any material effect on people who reads it.
This is a ridiculous thing to say - "because the Weather app in my phone is 100% automated with no human "soul" in it, I will not take it into account into how should I dress today"
> the client don't care as well
This applies here too. If you didn't pay for author's output, are not a subscriber to their Substack, why should they bother to hand craft you a quality human written post?
> If you didn't bother to write it, I won't bother to read it
So the argument is that effort should be somehow matched. Ok, let's go with that. It's well know that writing something takes 10x more effort than reading it. So if we use the market argument, if it takes you 10 min to read a blog post, the author should spend 1 min writing it, because writing is 10x more intense, and this is only possible with a LLM. You might say "but there is one author, and 1000 readers". That gets us back to the paying argument.
Because the same sentence has been repeated practically word-for-word ad nauseam all over social media for years. What do you think is plausible - that they repeated something that is damn near omnipresent in the discourse, or that they never saw any of the many, many times it has come up and independently came up with the exact same formulation? They are perfectly able to step in and claim that it was their original thought if they want to.
Both are fully possible.
We're talking about a simple sentence that sounds a lot like something a primary school teacher or parent of young children would say. "If you won't do/don't/haven't done X, why should I Y?" is very normal moral/ethical phrasing.
I suppose I heard someone say it before I spelled it out in that way, but I am pretty confident I would ultimately have used a sufficiently similar expression, because I have been on the receiving end of the problem.
And that is putting aside the communicative compression of aphorisms. People don't use shared aphorisms to avoid ever doing the thinking themselves; they use them to shorthand more complex arguments for a listener who has already heard them. They are shared macros for agreed concepts. (Using them correctly requires having done the thinking they package up at least once).
Do you think that they came up with it independently? If you aren’t willing to say you believe this, you are just asking me to waste my time arguing with a phantom opinion nobody holds.
“Just do it” is an incredibly common, generic phrase, but if a new shoe manufacturer chose it as their slogan, it isn’t plausible that they came up with it independently of Nike.
OK then. BOTH ARE PLAUSIBLE. Clear enough?
It's absolutely not up to me to decide whether they came up with it independently.
But then again, it isn't up to you, Jim.
> If you aren’t willing to say you believe this, you are just asking me to waste my time arguing with a phantom opinion nobody holds.
This is the weirdest thing to put in a comment you could simply not have made. If your time is so valuable why are you here commenting on HN?
Again: aphorisms have value regardless of how many times they are used. People can use them without you having the right to demand proof they thought them up for themselves like some prep school debate bully picking on a younger child.
> This is the weirdest thing to put in a comment you could simply not have made. If your time is so valuable why are you here commenting on HN?
It’s basic intellectual honesty. If nobody believes that they came up with it independently, putting it forward as if it’s a real argument is just a form of trolling. If I don’t believe it, you don’t believe it, and even the person who posted it doesn’t claim it’s original, you’re just baiting me to respond to a position nobody holds.
Do you believe that they actually came up with it independently? If you do actually believe this, then a discussion is not a waste of time. But if you aren’t willing to claim that position, it’s pointless me arguing against it because it’s just you deliberately wasting my time on something you don’t believe.
> aphorisms have value regardless of how many times they are used.
This is a cornerstone of the point I was making. You’re literally supporting the point I was making.
That's more or less the same thing expressed more nihilistically. But nihilism is, like, just a word man (to 'The Dude' it up a bit). Perhaps we are either fiddling on the deck of the Titanic, or just staring at the waves?
What do you think the results will be of the machines consuming only their own slop and Habsburging themselves intellectually, and then all the humans abandoning any attempt to either talk or listen? As a global civilization of humans? I'm not sure we've explored the likely consequences of 'if you can't spend them time to write it why should anyone read it'.
Maybe it's gone too far already.
That is completely different from generating words that are not an expression of your thoughts or at the very best take some concise thoughts from the prompt and stretch them into a lengthy word soup.
Of course there are people who just parrot other people's thoughts without understanding them. That's closer to using an LLM. But you can often see how well someone understands something and how much reasoning of their own they put into it through their writing. And for the writer writing is a very good way to reflect on their own thoughts and to get them into order. Proper writing is thinking.
But this is what people said about the printing press. And the internet. Less effort to spread the memes. Has it been a net negative? Maybe in some respects. Can you stop it? Probably not.
Im sorry for you. It must be hard to detach yourself with a fixation. I also cant help but spend my attention spotting where the magician is hiding the coin instead of trying to enjoy the show
> If you can't spend the time to write it, why should anyone read it?
Most people in these threads repeat this question. As if pre-ai all writing was from deep thinkers full of genuine intention.
As a result, when I see an AI wall of text I just don’t bother reading it and DM the submitter asking what it’s about.
I don't think writing it is the whole process, research could be 10 or 100x longer.
Couple weeks ago I had someone linking to a doc page in my ticket "Made this summary, hope it helps". It was 100% AI generated, the person clearly had no understanding of the problem, it only mentioned UI changes although the thing required changes in many systems. Complete waste of time, I don't think I will humor anyone who makes AI summaries anymore.
They can even retrieve articles you link to and check that you aren't misrepresenting them.
But the 50% I do take into account, improves the text! And, like TFA, I never ask it for concrete text. It only helps me diagnose the issues, I prescribe the medicine!
Is there anybody on this site who isn't?
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.
It’s kind of like designing the high level software architecture yourself and have the LLM write the code for each component.
Not bulletproof, requires some iteration, but miles better than what it would produce on its own.
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.
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.
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.
In other words, you need to read a style manual or two, and you need to read and have opinions about other people's work. You must have taste. Otherwise, you can't really distinguish between good and bad advice. You'll anchor to the suggestions made by the LLM and it will turn your writing into LLMese. I guess that works - it sure gets clicks on HN - but it won't be your voice anymore.
There's still value in getting a second pair of eyes on what you write. But here's another cool trick that gives you a similar benefit: just let it stew. Go for a walk, grab a lunch. Re-read what you wrote. You'll be surprised.
I don't care whether the LLM likes these constructions or not. I don't like them. I usually don't want to keep them in my work. It's my call what to change; I just want issues flagged.
Your prompts would be different. Or, they would if you decided that having an LLM merely point things out about your own writing wasn't itself a form of contamination. You do you!
I wrote this article in about 45 minutes and didn't workshop it with anybody else, so I'm not surprised it's gone over people's heads; it was just something I was talking about on Twitter that I wanted to get down on a page. But it's pretty obvious that a lot of people here have missed the point.
People are beginning to wisen to how to more skillfully deploy generative models.
Sure they can also be used for outsourcing the writing process and producing text that doesn't resemble your own voice, just as any tool can be used for multiple purposes. Tools aren't responsible for how particular humans choose to use them.
Your process is similar in spirit to mine, although mine is a bit more fluid. I don't think it's necessary to disqualify words and phrases based on provenance per se, but rather on whether they're genuinely the correct words for what I intend to say and how I intend to say it. That's basically the test I use; if I know that I've intentionally chosen every detail of the text, that I've maintained consistency with my own manner of writing and speaking, and that I've faithfully encoded my own bona fide thoughts, then as far as I'm concerned it's my writing and I see no need to qualify that based on whatever tools may or may not have been used.
And, as a fan of the "let it stew" technique, it's not a full substitute for an actual other perspective.
A good approach may be to present an LLM with an article without saying its yours, and then have it write a thoughtful polemic. But it's not a very efficient way of doing things.
Step two: close the LLM
Step three: pick up a pen and write
Writing with LLMs is a great way to make yourself stupid: https://lambdaland.org/posts/2026-08-07-ai-writing-stupid/
I'm glad the OP acknowledges that you have to have to do the writing yourself—though they say this more as a means to avoid getting pegged as having used an LLM, whereas I see the problem as being cognitive surrender.
The best advice for writing for other humans is: don't use LLMs.
If you're writing for machines, liking coding, then fine go for it. If you're writing for processes with formal highly structured content like manuals, specifications, form content, procedures, information, that sort of thing, then also ok to use LLMS. But if you're writing for a human mind to ingest and extract meaning from, then LLMs are poison.
This is an unnecessary argument both ways. Whether you use a LLM or write it yourself or read it yourself (and then comment on it), you're putting your signature beneath it. I don't care whether you use your natural brain or an artifical brain (LLM)
I wouldn't. LLMs don't summarize things very well. If you don't care if someone wrote it, and you don't plan to read it, why are you even here?
Not in full. I plan to read the LLM's summary. Most people could benefit from brevity.
author : writes the summary (prompt). LLM expands it to article.
reader : LLM summarizes the article. reader reads it.
both author and reader are free to use, or not use LLM. I don't see the big deal here.
When you review code, you're not taking it all in end to end like an email or memo or blog, you're looking for important symbols, recognizing structures and patterns, jumping between different files at different levels of reference and abstraction, and building a model in your head of the principles that make it work and their consequences for the overall procedure.
Code is "meant to be read" in the same way as a law, blueprint, policy document, or manual is, ie: jumping between relevant sections to get small pieces of information relevant to some task. The act of "reading" never starts at the beginning, it starts with a ctrl-f or flipping through an index or table of contents for a specific relevant symbol that is the only reason you are looking at the document.
Code is invention, not art (and should be covered by patent law, not copyright law, I will die on this hill)
For example the code I would write for somebody who just started to learn programming will not be the same as the code who is an expert even though both will pass the specifications.
Consequently taking into account who (if anyone) will actually review, and thus read, the code changes the code itself.
Manuals, laws, and dictionaries are not designed to be read, but they are designed to be interpreted. In fact interpretability is what they maximize for. And interpretability isn't an absolute value, it is relative to an audience, this is why household equipment often comes with separate owners, installers, and service manuals, and why Webster publishes Elementary, Collegiate, International, and Mass-Market versions of its Dictionary.
Documents that are "meant for human consumption" in the way the top comment intended generally don't optimize for interpretability like code, manuals, and dictionaries do. Novels, letters, emails, internet comments, blogs, articles, nonfiction books, etc often forgo interpretaiblity for practical, expressive, or rhetorical effect. The best moments of these is even when they are confusing on purpose, because imprecise language is a gateway to deeper levels of meaning through fourth-wall breaks, metaphor, analogy, and humor. (all things that LLMs are notoriously bad at and don't seem to be improving on over the years)
Those other types of writing are intended to be read in full, every word, end-to-end, in order to educate, entertain, inform, or influence the reader who, and i think this is the critical part, does not drive the experience.
When I'm reading a news article, and you realize that you're reading the voice of chatgpt, I'm annoyed because it means that i've been strung along by a stupid machine that somebody else handed me off to to keep me occupied.
When I'm reading code, and I notice that it seems AI written for whatever reason, I might then choose to apply more scrutiny, maybe I try to suss out what the prompter intended to happen, so that I can validate whether that actually happened and if it was accomplished in an optimal way. But it's not insulting in the same way, because I'm still in the driver's seat of the experience, deciding on my own which parts to read and which parts to ignore as irrelevant to my investigation, because the only thing I want to get out of it an understanding of a part of the mechanical system that the code creates.
I don't agree. Because those metaphors and other word plays reach deep into the human mind (at least for the purported audience), while most technical writing try to be more explicit.
Here is the introduction for Laravel Socialite
In addition to typical, form based authentication, Laravel also provides a simple, convenient way to authenticate with OAuth providers using Laravel Socialite. Socialite currently supports authentication via Facebook, X, LinkedIn, Google, GitHub, GitLab, Bitbucket, and Slack.
A century has passed since the gods fought and drove themselves to extinction. Now only their bones remain, promising great power to those brave enough to seek them out.
As whispers of war echo across the land of Vigrid, fate follows in the footsteps of three warriors: a huntress on a dangerous quest, a noblewoman pursuing battle fame, and a thrall seeking vengeance among the mercenaries known as the Bloodsworn.
All three will shape the fate of the world as it once more falls under the shadow of the gods.
It’s perfectly possible to consider e.g. working, LLM generated, human approved code acceptable, while considering LLM writing that hasn’t undergone major human editing unacceptable because of its repetitious, formulaic, marketing-oriented style - a criticism that doesn’t even make sense for LLM generated code.
Yeah... no. The people doing this lack the communications training, see the output has the necessary information, and regurgitate it with no effort or care. This needs to stop.
We spent decades format building to make it easy quick and easy to get through something like a runbook. If your commands are bulleted instead of numbered and code blocked, it's wrong. If you didn't crawl through the playbook, it's immoral to hand that to me, you're wasting my time with untested slop.
This is a hill I will die on or absolutely start slaughtering people on. I just refuse to deal with this crap.
A lot of documentation generated in corporations has marginal real value.
> If your commands are bulleted instead of numbered and code blocked, it's wrong.
That’s easy to instruct an agent to do. Put it in a skill.
> If you didn't crawl through the playbook, it's immoral to hand that to me, you're wasting my time with untested slop.
Using an AI doesn’t absolve the user of their responsibilities. This is an easy problem to solve, just make sure teams know what’s expected of them and encourage everyone to push back (professionally) against offenders.
To be completely fair though, I'm against the behaviors in information transfer I've been seeing. I've pass along AI generated runbooks, but they look nothing like the default outputs of these models. It's because I took time to apply all the writing knowledge I like to see in my curation. If people are doing this, I can't even tell it's AI writing. My work is done in minutes instead of deciphering so BS pseudo language they developed in their AI workspace (people really need to turn off those memory features).
---
Edit: Also if I'm the guy receiving a security report and it's AI generated and poorly formatted, I'm failing you short of producing something for a human to parse. Simple as that.
Should you think it is wise to trust the machine that can't differentiate subject matters in a chat styled context, you have fun with that fluster cluck when it blows up.
Like Fable is highly useful, but it's really bad at keeping it's responses straight.
In fact, that "it's not X it is Y" pattern always crops up when it reasoned about the idea of X and I never fed it that. It's literally doing that because it can't predict that I'm a different entity despite it being able to say I am a different entity.
Edit: Clarification by removal of incomplete sentence fragment. Edit2: Clarification on the "proven in whole" thing.
It’s clear that you’ve never worked for a compliance company and probably have never been involved in a compliance project. As such, you don’t have the context needed to participate usefully in this discussion.
I can't write anything for humans using LLMs, because there is so much to change that it ends up like the Ship of Theseus, I would have been faster writing it myself than delegating the task.
it is great at analyzing the argument, finding inconsistencies, helping you think through what parts should be cut, helping you refine examples or fix the occasional “how do i get this phrase to work correctly in this transition?” kinds of stuff. but anything where LLMs are the prima materia… that stuff literally only makes sense _to me_. which makes sense, because it is written _for_ me, no matter what instructions i actually give it, bc of memories and a million other things. and i say this as a complete maximalist wrt. trying to use llms for absolutely every last thing they possibly can be used for, just to see what it’s like.
i guess i would say that it does very, very little to make the writing process meaningfully faster; it _can_ do _plenty_ to help make your output better though, which is definitely something— just isn’t the thing most people are looking for.
Style guides are a big tools people are missing out on. It isn't enough to say "write concisely and technically". Give them a sample. Yea verily these many 5 or 6 years ago, "style transfer" was a big thing that early LLM tech was doing. It's still very good at it.
That said I still tend to cut out at least 25% of the resulting verbiage and adding back another 10% or so more of my own original content even under those circumstances.
Where it has been really helpful is that I tend to want to write in a conversational style that doesn't seem to match most people's expectations of a technical document. The LLMs let me write my way and style-shift it into something closer to what people expect. And LLMs, since let's be honest they're the primary audience nowadays. Which I am not even upset about; I'd rather 5 LLMs read the architecture document than the amortized .2 or so humans I could expect in the same circumstance 5 years ago.
Which also implies, in many cases, I am feeding the AI as much text as I expect to come out, or in some cases, even more, as I am describing context, reasoning, and other things that may impact the writing but are not necessarily repeated in the final text. I factored out a lot of the context into my user-level CLAUDE.md which has helped cut that down a bit.
Also the word "verbiage" ticks me off as a LLM-ism on the level of "delve" or "smoking gun". Are its users aware that in addition to its usage as somewhat of a synonym of "idiolect", it more commonly means "a profusion of words usually of little or obscure content", i.e. slop?
As for "verbiage", I guess that's too bad, I'm going to continue using it anyway. Interesting factoid about the word though. I will say that "utilize" is one of my pet peeves (99% of the time the word you actually want is "use") and I've excised it from my spoken vocabulary, but I like(d) how verbiage rhymes with foliage so I'm more reticent to give that one up.
Sadly people lost the skill to write a standard form concise 3-paragraphs-of-3-sentences cover letter even before keyword-based ATS initiated the beginning of the end. It wasn’t supposed to be a summary of a CV and an essay about your lifelong passion for automated ad-bidding quality checks or some other shit employers delude themselves into believing someone might organically be really into.
I don't use a lot of "skills" so if there are others you've started using frequently I'd be curious to learn about them.
Writing with a LLM is a boon if you are trying to write while mastering a new field. Imagine a neuroscientist entering robotics or an electrical engineer trying to enter the field of neuromorphic systems. In these cases a human will need help from an LLM to think and write more deeply AND broadly. I sure do.
You could object that this form of human-LLM collaborative writing is just for one reader; the writer. True at the start but not true toward the end.
I do agree that even the frontier models do not touch a really strong human writer. Way too much gimmickry. They do not yet match individuals such as Leslie Lamport, Terry Winograd, John MacArthur, Richard Rorty, Humberto Maturana, William James, Rodney Brooks. They don’t have the depth of understanding to simplify. But give them a few more years and excursions in the real world and a concept if “now” and the situation may improve.
I often ask Fable and Opus to explain difficult topics to me as if I were the canonical dog Rover. It works.
New areas are where it is most dangerous to use an LLM, as a person can’t tell when it doesn’t align with their views (because they don’t have any yet), is misleading, or flat out wrong.
If I want to know what an LLM thinks, I’ll ask an LLM. If I’m reading a paper from a human, it’s, because I want to know what that human thinks. If they need the LLM to fill in the gaps in their knowledge, then they don’t know enough to write about the topic yet. Doing that work to write something themselves could help move them learn enough to write something that is hopefully worth reading.
Explaining difficult topics or summarizing to you is fine. But that's a world of difference from having an LLM write for you. You THINK the LLM is helping you write above your level, but that's because you are not in a position to tell it's bad.
Writing is a form of thinking and writing with an LLM makes the argumentation, discussion, and progress toward goals faster and often better in my situation.
I am stating an opinion based on 2.5 years of intense interactions and writing with Anthropic models on a difficult topic related to asynchronous computational systems.
Background: I am a neuroscientist and behavioral biologist (45 years of experience) who understands the biology of behaviors fairly well but I need to translate this background into hardware architectures that get beyond current asynchronous system (e.g., this recent neuromorphic paper in Nature Communication by Li et al., 2025, A deterministic neuromorphic architecture, doi.org/10.1038.s41467-025-65268-z).
Working AND writing with Opus and now Fable has been a blessing of sorts.
$ chatgpt make this comment more glib
Word chains these programs produce are inherently meaningless. They do not feel so, because WE people are amazing at bringing meaning to any old string of symbols: https://www.ubu.com/sound/lin.html
They are very useful machines indeed, but they do not help you write well. They mathematically cannot. They just try to guess what is the next tokern you might want to read.
Also if you don't understand what you are documenting enough to write the docs that is a problem.
Charitably, they can be thinking that it's fine to have LLMs organize information intended for LLMs.
Don't you think they're useful as fact checkers?
Hallucinations don't really work that way. They happen when you ask an LLM to generate new material. Not so much reviewing existing material.
But there's always the "ask a second LLM to get two opinions" technique which gets you pretty close to error free.
And... now that I think of it... even if it was terrible and missed 50% of the errors... wouldn't that still be better than using no LLM which catches 0% of the errors?
You can reduce the likelihood of incorrectness by adding the real facts to the context, which is what LLMs do. But that in and of itself is not a guarantee that the LLM will be correct. Because its fundamental process of reasoning is probabilistic word-by-word hallucination.
Well - that's not how the term is used.
And it would be an odd way to describe all the extremely useful things they discover. Are you suggesting they're taking "magic" mushrooms of some kind?
Are you not aware?
Do you just believe every bit of press release that you read? These companies have an enormous amount of pressure on them to make profit. They can and will lie to try and convince you that AI is better than it is.
Is there any evidence outside of these companies that AI is breaking new ground on literally anything? No. It can help a little with a lot of different types of work– none of it novel.
My friend, I use AI every day. I am fully aware of how fucking amazing it is.
When you can no longer clamp your ears and eyes shut, come find me. I'll love the apology.
I can see being lazy and skipping it, but refusing to use them even for more minimal advice purposes seems crazy to me.
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!
Does anyone know about "Style: Lessons in Clarity and Grace" ? Seems like an interesting book
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. ;-;
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.
"On Writing" is another book like "The Elements of Style", both are too prescriptive, albeit not useless.
Paying close attention to posts written by folks who are not using LLM's the comments here regularly complain (sometimes extensively) about the writing quality. Doesn't matter what people are saying or suggesting about avoiding LLM writing, the real lesson I am seeing is either to naturally be gifted at writing or somehow figure out how to have LLM's help with your writing. "Rule Number One: You may not use a single word an LLM suggests to you." isn't helpful for probably the majority.
Developing a voice comes naturally as you go through the pain of being a terrible writer to eventually being mediocre. Accepting the LLM's words is actively harmful to that process because it lets you skip over the part where you read over what you wrote and rewrite until it doesn't suck as much.
I mean, you do you. I'm not your dad.
Clearly I am not a gifted writer as my original comment failed to convey this very well. And now I am pondering, maybe next time I should put my comment in an LLM to review it first. This meta lesson is playing out over and over everywhere by anyone writing anything.
I am in no way saying LLM writing in general is good, but ignoring the reality of what is happening is silly. For good or bad the not great writers not only will, but are doing things like voice rambling into an LLM as a starting point and editing it from there.
However, I directionally agree though with your message: if you are bad at writing, often you can use LLMs to surgically improve the flow of your sentences for example. One should use a carefully crafted workflow for this though.
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.
In the extreme these can be annoying, but more often I find them kinda charming, as an expression of the author's personality.
If you want to improve your writing, read more.
This sentence specifically seems like an LLM. It's hard to quantify that. They just take that sort of tone when they're needlessly summarizing what they already told you. "This thing contrastive thing remains important, but we got XYZ benefit."
- 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.
I have but it was a house construction worker explaining its practice on youtube.
If not, it does risk planting the seed of doubt - have I been duped once again into reading tokenslop again?
Risky move. Surprised the LLM copy editor didn’t flag it.
[0] https://x.com/search?q=%22load%20bearing%22%20until%3A2020-0...
Other than that, it's a time waster.
This is gold
> Rule Number One: You may not use a single word an LLM suggests to you.
Over the last week-ish I turned around 2 product announcement blogposts within ~24hrs. Both written manually by someone who doesn't do product marketing for a living. It actually felt incredibly satisfying to focus and smash something out that was 100% human (despite praising the benefits of agentic systems, oh the irony).
My flow has been: Go for a walk, record all of the thoughts on the topic with speech to text, and have LLM organise my thoughts into what I call "facts" - simple sentences of fact, and point out open questions and inconsistencies ("you are a product marketing expert helping me organise the content for an upcoming feature announcement" ...).
After some back and forth of inconsistency bashing I have a large pile of short sentences in the vein of "Feature X is for Y because Z", that I use to manually build a narrative with headings, and then write the actual copy covering a subset of them.
I also agree that light style LLM edits are fine, and that I don't let LLM do anything other than ordering and filler words removal and provide structural comments.
> organise my thoughts into what I call "facts" - simple sentences of fact, and point out open questions and inconsistencies Would appreciate little more details of how you do this.
I think there's a use case where this rule can be ignored. Sometimes I ask LLM to help me pick a single word or short phrase in some language, that has specific meaning, and connotations. Or doesn't carry some connotations or associations. Especially when I have an idiom from some language in mind, but I'm writing in some other language and my brain struggles to find a good replacement.
LLMs are super handy as fancy vibe-based dictionary search engines. They're all about words that go with other words, so when I'm struggling for a good term, I see no harm in asking and picking an option that I feel closest to what I want to express. My choice, my voice, just a very fancy thesaurus machine.
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.
I really hope this was intentional.
To your comment, I wasn't aware of the eye rolling meta for the phrase [0]. I hadn't seen it overused to the extent people are claiming.
[0] https://www.reddit.com/r/ClaudeAI/comments/1upxk9r/loadbeari...
Pangram 4 says it's 100% human written, including the "load-bearing" sentence.
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.
When most people are consuming _their_ news (as opposed to _the_ news) in the form of a thumbnail in an algorithmic social media feed, those copy errors matter less.
I'd love to know. I'll buy it!
So far so good…
“Those rethinks are load-bearing parts of your voice.”
Oops.
edit: it was! https://news.ycombinator.com/item?id=49753395
I'm looking forward to trying out your editor. When will it be officially available? But I have another question: English isn't my first language. What kind of impact can this editor have for non-English ecosystems? and if we add translation and LLM models on top of that, how they would interact is something I care about.
It's not even the suggestions in isolation, but it tries to convince you that your writing is stellar and then slips in a word or two, and you accept it because otherwise you'd be skipping greatness. I guess in some way one has to have a bit of hubris about one's writing to tell the model to f off.
And OP, assuming this, from your prompt, is your language:
> "… First get the bones up…"
Congratulations, you are now doing LLM-speak :-)
Writing is something we really, really need to learn from other people, because it's with people we are trying to communicate. There should be no need for "writing with an LLM". None at all. If what you are writing is important, ask a friend to proof read it. If you can't write, learn.
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’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.
Emitting words is not really what "writing" is.
https://twinada.wordpress.com/wp-content/uploads/2012/09/708...
The act of agonizing over the words you put on the page is thinking. A lot of humans don't do it. The writers who talk about writing being thinking are in a bubble and don't realize that their process is not universal.
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 ?
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.
> I conjured a bit of software to manage this for me[...]
I don't understand how the annotation application shown helps with the stated goal of preventing the model from praising you. What's the actual software assisted workflow?
Workshop a piece with me. NO ENCOURAGEMENT. Encouragement is
useless; the only useful things are suggested corrections.
DO NOT WRITE COPY FOR ME. Any words you provide will be disqualified,
so if you come up with good words, I’m fucked because I can’t use
them. Tell me ABOUT what should change.
Assume I know my audience extremely well.
We’re going to do this in a series of passes. Keep your suggestions
locked in to the current pass we’re on.
Note typos, but DO NOT generate long structural critiques based on
those typos; note them and move on, assuming that I meant to write
properly.
Don't.
Learn to be a better writer, because a bot cannot and will not help you.
For docs I have used an LLM for the convenience of having them kept up to date.
Ok - but that's not very useful.
For anything beyond "good", we're still on our own and I love that for us.
Context isn't an issue. It's style.
The argument the author is making is that if you use an LLM to write for you, you end up with a homogeneous work that's just like everything else.
In journalism, that's sort of what you want.
In novels/function, that's the opposite of what you want.
However, there's degrees to these things. If you're looking at a sentence you wrote, thinking there's something "off" it can be helpful to have the LLM rewrite it just to see how it thinks it could be better worded.
I've done this a few times and I do find it helpful. Just like the author, I never just accept what it gives me. I simply use it as an OK-ish copyeditor that's good at spotting issues and suggesting solutions.
For objective/factual writing, asking for a fact check is a widely-used practice. And there's a useful counterpart for writing subjective things and opinionated hot takes, namely to ask the LLM: “Anticipate how this might be misunderstood in ways that might cause offence or trigger people.” Sometimes, you might want to say something that will cause offence, and there might be no way around it. In this case, making a conscious choice to let that be might be the right thing to do. But I find it surprising, how often an LLM will see an angle that I myself did not see, where something might cause offence in a way that I genuinely missed and genuinely want to fix. And then that gives me a chance to re-write before posting something.
---
and then
> 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.
---
Are they wrong in the first paragraph? Or are they simply not following the advice they dispense to others?
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.
And yet, I'm stuck having to write an important message to someone that has real stakes riding on it, so I use Claude to refine it, and I see myself succumbing to Claude's pressure to use its words instead of my own. I hate it, but my words aren't very good.
And finally Claude proclaims the message perfect and I let my wife take a look at it, and she rejects the entire thing, restoring some semblance of sanity again.
Today even if you don’t use LLM’s following certain formal style you end up sounding like ad copy or wannabe Atlantic -hack. The. You might read some kid on Tumblr writing beautiful prose with typos and errors and all.
First of all you should have something to say. I write a blog just to make tutorials for myself on stuff I need to regularly check out. I don’t really care if anyone reads it. I don’t use LLM to make the posts since, well, pride, but also because I learn better when I need to write the stuff down and check my facts several times. I don’t much care if I make a typo or grammar error here and there. If I want to write it as Jeeves from Wodehouse book, fuck it, I will.
If you have a story you have something to write about. It doesn’t have to be fiction, it can be a story on how multicast works. And it needs to be your story. Then the words come by themselves and writing is relatively easy. To have a story you have to live and think. LLM’s don’t help in that.
But if you've generally been doing drafts and want to keep writing like that, if part of your process is going to continue to be refining and sharpening, then I think you should know that we're at a point where automation can accelerate that process for you. You can be getting the same amount of work done with less effort, or allowing yourself to concentrate more on the creative thinking part and less on the tedium.
That's the point of the post.
Style is kind of a pretense. Communication in general is a sort of a mucks game of guessing how to get your point across across the medum, whether that of air, FM-radio, writing, internet, video... In academia, copywriting and so-on we create these patterns we call writing styles to agree on the rules on the game. But honestly, we should use sá lot more fuckin swear words when disproving the fuckin conjectures.
Yes I know automation makes things quicker. Sometimes even close to fast. But htere's things you should do slow.
Writing and thinking are amongst them. And hey, LLM's can help you here as well. But they should not speed you up, they should slow you down. Why did you say this? WHat is your premise? What is the source of this? KIM by Moonshot is really easy to take a system prompt that makes the models output a biggest asshole editor in the world. It doesn't help me at all, on the contrary, it creates friction.
I don't think more speed is what we need at this point of evolution.
I have personally decided, in advance, that one goal I have in writing is improved clarity. The model didn't tell me to do that; Joseph Williams did, decades ago.
In the service of that goal, I've written down a proofreading checklist. It includes things like "look for overused words", "eliminate throat-clearing filler like 'it's important to remember that'", "make sure every paragraph has a clear topic sentence", "make sure my real verbs aren't buried inside nouns". Stuff like that.
Those are stylistic choices, I suppose, but they're mostly the difference between bad style and a whole universe of good styles.
More importantly: they're technical. They're not really judgement calls. The subject of a sentence either is or is not a character taking actions in the story you're trying to tell. It is better, almost always, if the subject is a character, and not accidentally some random piece of the scenery of your point.
Models are very good at flagging these things. Humans are too, but they're not as thorough.
Many people on this thread have suggested that having a model flag these kinds of issues bleaches out personal style. I don't think that's true; what I think is happening is that they missed the point I'm making, and assumed that I was advocating for having a model make broad subjective calls about my writing.
Oh, we do. We can also detect hackeneyed, LLM-made sentences where just one word had been replaced. "That's where it's going to f*** you".
I'll get my writing tips somewhere else.
Don’t
This might work well for some tasks (coding), but for writing it will absolutely take reasonable text and turn it into a pile of incoherent garbage no human would ever write.
Maybe I’m the only one keeps trying this (more often than I’m willing to admit), but I suspect it’s a common cope engineers reach for when having to deal with the not-so-fun task of writing prose.
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 stopped reading at this point. All of the above may have been correct five years ago, when there was a "canon" of good writing, but now LLMs pack that canon as tight as teenagers in a boy band concert. So, go with a flow that reflects your state of mind--LLMs don't have that--and yield the words in fistfuls and with abandon. To further distance yourself from LLM output, your writing can be any or all of these things: sassy, prosaic, heathen.
Uses for LLMs: grammar checks, and to see if your idea comes through. However, depending on what you are writing, I recommend leaving in your writing a tiny doses of typos, grammar issues, and infuriating word order choices. Even some level of ambiguity is fine; your fellow humans and LLMs are not all of one mind, and if some of them are going to fight with knives over your exact meaning, well, it's them who are the extremists, not your problem.
"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'm personally furious that "load-bearing" has become an LLM tell, it's a really useful metaphor!
Pasteurising something is making it sterile. Safer, but bland and not as interesting. I guess that’s the intended metaphor.
Simply don't.
Don't.
TLDR: Don’t.