I've seen a few people use ai to rewrite things, and the change from their writing style to a more "polished" generic LLM style feels very strange. A great averaging and evening out of future writing seems like a bad outcome to me.
That may be, but it's also exposing a lot of gatekeeping; the implication that what was interesting about a "Show HN" post was that someone had the technical competence to put something together, regardless of how intrinsically interesting that thing is; it wasn't the idea that was interesting, it was, well, the hazing ritual of having to bloody your forehead of getting it to work.
AI for actual prose writing, no question. Don't let a single word an LLM generates land in your document; even if you like it, kill it.
I agree with the sentiment in the first part of your post. But:
> AI for actual prose writing, no question. Don't let a single word an LLM generates land in your document; even if you like it, kill it.
Why? What's the difference? I'm genuinely curious about your perspective on this. Lots of people can't articulate themselves well, especially if they don't natively speak the language they're writing in. I have my problems with LLM generated text, but you seem to be taking an extreme approach here.
I have not seen many of these hypothetical "intrinsically interesting" Show HN posts generated by unskilled users.
The single one I can think of is someone who (I quote) "accidentally created the fastest CSV parser ever using SIMD". This person had no interest in researching prior art themselves, and thus incorrectly claimed credit for "coming up" with this approach - and they didn't even do that.
It's not only the prose that's the problem if submitters are determined not to think.
I mean, can't you just… prompt engineer your way out of this? A writer friend of mine literally just vibes with the model differently and gets genuinely interesting output.
The more interesting question is whether AI use causes the shallowness, or whether shallow people simply reach for AI more readily because deep engagement was never their thing to begin with.
Most ideas people have are not original, I have epiphanies multiple times a day, the chance that they are something no one has come up with before are basically 0. They are original to me, and that feels like an insightful moment, and thats about it. There is a huge case for having good taste to drive the LLMs toward a good result, and original voice is quite valuable, but I would say most people don't hit those 2 things in a meaningful way(with or without LLMs).
AI writing will make people who write worse than average, better writers. It'll also make people who write better than average, worse writers. Know where you stand, and have the taste to use wisely.
EDIT: also, just like creating AGENT.md files to help AI write code your way for your projects, etc. If you're going to be doing much writing, you should have your own prompt that can help with your voice and style. Don't be lazy, just because you're leaning on LLMs.
If you spent 3 hours on a show HN before, people most likely wouldn't appreciate it, as it's honestly not much to show. The fact that you now can have a more polished product in the same timeframe thanks to AI doesn't really change that. It just changes the baseline for what's expected. This goes for other things as well, like writings or art. If you normally spent 2 hours on a blog post, and you now can do it in 5 minutes, that most likely means it's a boring post to read. Spend 2 hours still, just with the help of AI it should now be better.
I've seen people say something along the lines of "I am not interested in reading something that you could not be bothered to actually write" and I think that pretty much sums it up. Writing and programming are both a form of working at a problem through text and when it goes well other practitioners of the form can appreciate its shape and direction. With AI you can get a lot of 'function' on the page (so to speak) but it's inelegant and boring. I do think AI is great at allowing you not to write the dumb boiler plate we all could crank out if we needed to but don't want to. It just won't help you do the innovative thing because it is not innovative itself.
I keep hearing the "boilerplate" argument but it never made sense to me. Text editors have had snippets for half a century now and they're strictly better than an engine that generates plausible-enough boilerplate.
It's not worth my time to read something that was not done to a high standard, where that standard has a definition with some basis in rigor rather than opinion, where even the notion of good taste is in some way attached to experience of the distribution from which examples of good and bad taste are drawn.
It is not about the author and it is in not about the effort. It is about the quality.
> AI is great at allowing you not to write the dumb boiler plate we all could crank...
I've actually started having a different view on this. After getting over the "glancing instead of reading llm suggestions" phase I started noticing that even for simple or boilerplate tasks, LLMs all too often produce quite wasteful results regardless the setting or your subscription. They are OK to get you going but in the last weeks I haven't accepted one Claude, devstral or gpt suggestion verbatim. Nevertheless, I often throw them boilerplate tasks even though I now know that typically I'll end up coding 6 out of 10 myself and only use the other four as skeletons. But just seeing the "naive" or "generic" implementation and deciding I don't like it is a plus as it seems to compress the time of thinking about it by a good part.
I've actually ran into few blogs that were incredibly shallow while sounding profound.
I think when people use AI to ex: compare docker to k8s and don't use k8s is how you get horrible articles that sound great, but to anyone that has experience with both are complete nonsense.
The headline should be qualified: Maybe it makes you boring compared to the counterfactual world where you somehow would have developed into an interesting auteur or craftsman instead, which few people in practice would do.
As someone who is fairly boring, conversing with AI models and thinking things through with them certainly decreased my blandness and made me tackle more interesting thoughts or projects. To have such a conversation partner at hand in the first place is already amazing - isn't it always said that you should surround yourself with people smarter than yourself to rise in ambition?
I actually have high hopes for AI. A good one, properly aligned, can definitely help with self-actualization and expression. Cynics will say that AI will all be tuned to keep us trapped in the slop zone, but when even mainstream labs like Anthropic speak a lot about AI for the betterment of humanity, I am still hopeful. (If you are a cynic who simply doesn't belief such statements by the firms, there's not much to say to convince you anyway.)
> The cool part about pre-AI show HN is you got to talk to someone who had thought about a problem for way longer than you had
Honestly, I agree, but the rash of "check out my vibe coded solution for perceived $problem I have no expertise in whatever and built in an afternoon" and the flurry of domain experts responding like "wtf, no one needs this" is kind of schadenfreude, but I feel guilty a little for enjoying it.
>and the flurry of domain experts responding like "wtf, no one needs this"
People have been saying this about Show HNs for time eternal. There have been an insane number of poorly thought out, poorly considered, often Get-Rich-Quick type of creations, long before AI. Things where the submitter clearly doesn't understand the industry they're targeting, doesn't provide any sort of solution, etc. Really strange if people actually think this is a new phenomenon.
Its subject is "Everything was Already AI", the point being that everyone is quantizing and simplifying and reflecting everyone else and the consensus, in such a fashion that people acting like AI ruined everything...yeah, it was already ruined. We already have furry artists drawing furry art just like countless other furry artists, declaring it an outrage that someone used AI to draw furry art, and so on. As the video covers, the whole idea of genres is basically people just cloning each other.
Be right back, going to put on a cowboy hat and denim and sing in a drawl about pickups and exes.
I’ve been partaking in my fair share, but more and more I’m just feeling sad for my fellow coders ‘cause a lot of what I’m hearing is about bad local choices and burdensome tech stacks.
Sure, it’s kinda hilarious watching a bunch of fashion obsessed front-end devs discover bash, TDD, and that, like, specifications, like, can really be useful, you know, for building stuff or whatever.
But then I think about a version of me who came up a bit later, bit into some reasonable sounding orthodoxy about React or Node as my first production language and who would be having the same ‘profound’ revelations. I never would have learned better. I wouldn’t be as empowered from having these system programming concepts hammered into me. LLMs would be more ‘magic’, I’d extrapolate more readily…
I’ve found myself thinking a lot of thoughts tantamount to “why don’t you dummies just use Haskell, or Lisp, or OCaml, or F#, or Kotlin for that?!”, and from their PoV I’m seeing a broken ladder. A ladder that was orthodoxy and well-documented when I was coming up.
LLMs should ideally bring SICP and Knuth and emacs to the masses. Fingers crossed.
It used to be that all bad writing was uniquely bad, in that a clear line could be drawn from the work to the author. Similarly, good writing has a unique style that typically identifies the author within a few lines of prose.
Now all bad writing will look like something generated by an LLM, grammatically correct (hopefully!) but very generic, lacking all punch and personality.
The silver lining is that good authors could also use LLMs to hide their identity while making controversial opinions. In an internet that's increasingly deanonymized, a potentially new privacy enhancing technique for public discourse is a welcome addition.
We don't know if the causality flows that way. It could be that AI makes you boring, but it could also be that boring people were too lazy to make blogs and Show HNs and such before, and AI simply lets a new cohort of people produce boring content more lazily.
I was onboard with the author until this paragraph:
> AI models are extremely bad at original thinking, so any thinking that is offloaded to a LLM is as a result usually not very original, even if they’re very good at treating your inputs to the discussion as amazing genius level insights.
The author comes off as dismissive of the potential benefits of the interactions between users and LLMs rather than open-minded. This is a degree of myopia which causes me to retroactively question the rest of his conclusions.
There's an argument to be made that rubber ducking and just having a mirror to help you navigate your thoughts is ultimately more productive and provides more useful thinking than just operating in a vacuum. LLMs are particularly good at telling you when your own ideas are un-original because they are good at doing research (and also have median of ideas already baked into their weights).
They also strawman usage of LLMs:
> The way human beings tend to have original ideas is to immerse in a problem for a long period of time, which is something that flat out doesn’t happen when LLMs do the thinking. You get shallow, surface-level ideas instead.
Who says you aren't spending time thinking about a problem with LLMs? The same users that don't spend time thinking about problems before LLMs will not spend time thinking about problems after LLMs, and the inverse is similarly true.
I think everybody is bad at original thinking, because most thinking is not original. And that's something LLMs actually help with.
> AI models are extremely bad at original thinking, so any thinking that is offloaded to a LLM is as a result usually not very original, even if they’re very good at treating your inputs to the discussion as amazing genius level insights.
This is repeated all the time now, but it's not true. It's not particularly difficult to pose a question to an LLM and to get it to genuinely evaluate the pros and cons of your ideas. I've used an LLM to convince myself that an idea I had was not very good.
> The way human beings tend to have original ideas is to immerse in a problem for a long period of time, which is something that flat out doesn’t happen when LLMs do the thinking. You get shallow, surface-level ideas instead.
Thinking about a problem for a long period of time doesn't bring you any closer to understanding the solution. Expertise is highly overrated. The Wright Brothers didn't have physics degrees. They did not even graduate from high school, let alone attend college. Their process for developing the first airplanes was much closer to vibe coding from a shallow surface-level understanding than from deeply contemplating the problem.
I think it's simpler than that. AI, like the internet, just makes it easier to communicate boring thoughts.
Boring thoughts always existed, but they generally stayed in your home or community. Then Facebook came along, and we were able to share them worldwide. And now AI makes it possible to quickly make and share your boring tools.
Real creativity is out there, and plenty of people are doing incredibly creative things with AI. But AI is not making people boring—that was a preexisting condition.
Just earlier I received a spew of LLM slop from my manager as "requirements". He clearly hadn't even spent two minutes reviewing whether any of it made sense, was achievable or even desirable. I ignored it. We're all fed up with this productivity theatre.
And the irony is it tries to make you feel like a genius while you're using it. No matter how dull your idea is, it's "absolutely the right next thing to be doing!"
We are going to have to find new ways to correct for low-effort work.
I have a report that I made with AI on how customers leave our firm…The first pass looked great but was basically nonsense. After eight hours of iteration, the resulting report is better than I could’ve made on my own, by a lot. But it got there because I brought a lot of emotional energy to the AI party.
As workers, we need to develop instincts for “plausible but incomplete” and as managers we need to find filters that get rid of the low-effort crap.
Based on a lot of real world experience, I'm convinced LLM-generated documentation is worse than nothing. It's a complete waste of everybody's time.
The number of people who I see having E-mail conversations where person A uses an LLM to turn two sentences into ten paragraphs, and person B uses an LLM to summarize the ten paragraphs into two sentences, is becoming genuinely alarming to me.
Yesterday my manager sent LLM-generated code that did a thing. Of course I didn't read it, I only read Claude's summary of it. Then I died a little inside.
It was especially unfortunate because to do its thing, the code required a third party's own personal user credentials including MFA, which is a complete non-starter in server-side code, but apparently the manager's LLM wasn't aware enough to know that.
LLM-generated documentation is great for LLMs to read so they can code better and/or more efficiently. You can write it manually, but as I've discovered over the decades, humans rarely read documentation anyway. So you'll be spending a lot of time writing good for the bots.
131 comments
[ 2.3 ms ] story [ 60.9 ms ] threadAI for actual prose writing, no question. Don't let a single word an LLM generates land in your document; even if you like it, kill it.
> AI for actual prose writing, no question. Don't let a single word an LLM generates land in your document; even if you like it, kill it.
Why? What's the difference? I'm genuinely curious about your perspective on this. Lots of people can't articulate themselves well, especially if they don't natively speak the language they're writing in. I have my problems with LLM generated text, but you seem to be taking an extreme approach here.
The single one I can think of is someone who (I quote) "accidentally created the fastest CSV parser ever using SIMD". This person had no interest in researching prior art themselves, and thus incorrectly claimed credit for "coming up" with this approach - and they didn't even do that.
It's not only the prose that's the problem if submitters are determined not to think.
EDIT: also, just like creating AGENT.md files to help AI write code your way for your projects, etc. If you're going to be doing much writing, you should have your own prompt that can help with your voice and style. Don't be lazy, just because you're leaning on LLMs.
It is not about the author and it is in not about the effort. It is about the quality.
And what are you basing that claim on? What are your sources? Your arguments?
I've actually started having a different view on this. After getting over the "glancing instead of reading llm suggestions" phase I started noticing that even for simple or boilerplate tasks, LLMs all too often produce quite wasteful results regardless the setting or your subscription. They are OK to get you going but in the last weeks I haven't accepted one Claude, devstral or gpt suggestion verbatim. Nevertheless, I often throw them boilerplate tasks even though I now know that typically I'll end up coding 6 out of 10 myself and only use the other four as skeletons. But just seeing the "naive" or "generic" implementation and deciding I don't like it is a plus as it seems to compress the time of thinking about it by a good part.
I think when people use AI to ex: compare docker to k8s and don't use k8s is how you get horrible articles that sound great, but to anyone that has experience with both are complete nonsense.
As someone who is fairly boring, conversing with AI models and thinking things through with them certainly decreased my blandness and made me tackle more interesting thoughts or projects. To have such a conversation partner at hand in the first place is already amazing - isn't it always said that you should surround yourself with people smarter than yourself to rise in ambition?
I actually have high hopes for AI. A good one, properly aligned, can definitely help with self-actualization and expression. Cynics will say that AI will all be tuned to keep us trapped in the slop zone, but when even mainstream labs like Anthropic speak a lot about AI for the betterment of humanity, I am still hopeful. (If you are a cynic who simply doesn't belief such statements by the firms, there's not much to say to convince you anyway.)
Honestly, I agree, but the rash of "check out my vibe coded solution for perceived $problem I have no expertise in whatever and built in an afternoon" and the flurry of domain experts responding like "wtf, no one needs this" is kind of schadenfreude, but I feel guilty a little for enjoying it.
People have been saying this about Show HNs for time eternal. There have been an insane number of poorly thought out, poorly considered, often Get-Rich-Quick type of creations, long before AI. Things where the submitter clearly doesn't understand the industry they're targeting, doesn't provide any sort of solution, etc. Really strange if people actually think this is a new phenomenon.
Indeed, a recent video that I rather loved touches on this - https://www.youtube.com/watch?v=Km2bn0HvUwg
Its subject is "Everything was Already AI", the point being that everyone is quantizing and simplifying and reflecting everyone else and the consensus, in such a fashion that people acting like AI ruined everything...yeah, it was already ruined. We already have furry artists drawing furry art just like countless other furry artists, declaring it an outrage that someone used AI to draw furry art, and so on. As the video covers, the whole idea of genres is basically people just cloning each other.
Be right back, going to put on a cowboy hat and denim and sing in a drawl about pickups and exes.
I’ve been partaking in my fair share, but more and more I’m just feeling sad for my fellow coders ‘cause a lot of what I’m hearing is about bad local choices and burdensome tech stacks.
Sure, it’s kinda hilarious watching a bunch of fashion obsessed front-end devs discover bash, TDD, and that, like, specifications, like, can really be useful, you know, for building stuff or whatever.
But then I think about a version of me who came up a bit later, bit into some reasonable sounding orthodoxy about React or Node as my first production language and who would be having the same ‘profound’ revelations. I never would have learned better. I wouldn’t be as empowered from having these system programming concepts hammered into me. LLMs would be more ‘magic’, I’d extrapolate more readily…
I’ve found myself thinking a lot of thoughts tantamount to “why don’t you dummies just use Haskell, or Lisp, or OCaml, or F#, or Kotlin for that?!”, and from their PoV I’m seeing a broken ladder. A ladder that was orthodoxy and well-documented when I was coming up.
LLMs should ideally bring SICP and Knuth and emacs to the masses. Fingers crossed.
Now all bad writing will look like something generated by an LLM, grammatically correct (hopefully!) but very generic, lacking all punch and personality.
The silver lining is that good authors could also use LLMs to hide their identity while making controversial opinions. In an internet that's increasingly deanonymized, a potentially new privacy enhancing technique for public discourse is a welcome addition.
> AI models are extremely bad at original thinking, so any thinking that is offloaded to a LLM is as a result usually not very original, even if they’re very good at treating your inputs to the discussion as amazing genius level insights.
The author comes off as dismissive of the potential benefits of the interactions between users and LLMs rather than open-minded. This is a degree of myopia which causes me to retroactively question the rest of his conclusions.
There's an argument to be made that rubber ducking and just having a mirror to help you navigate your thoughts is ultimately more productive and provides more useful thinking than just operating in a vacuum. LLMs are particularly good at telling you when your own ideas are un-original because they are good at doing research (and also have median of ideas already baked into their weights).
They also strawman usage of LLMs:
> The way human beings tend to have original ideas is to immerse in a problem for a long period of time, which is something that flat out doesn’t happen when LLMs do the thinking. You get shallow, surface-level ideas instead.
Who says you aren't spending time thinking about a problem with LLMs? The same users that don't spend time thinking about problems before LLMs will not spend time thinking about problems after LLMs, and the inverse is similarly true.
I think everybody is bad at original thinking, because most thinking is not original. And that's something LLMs actually help with.
This is repeated all the time now, but it's not true. It's not particularly difficult to pose a question to an LLM and to get it to genuinely evaluate the pros and cons of your ideas. I've used an LLM to convince myself that an idea I had was not very good.
> The way human beings tend to have original ideas is to immerse in a problem for a long period of time, which is something that flat out doesn’t happen when LLMs do the thinking. You get shallow, surface-level ideas instead.
Thinking about a problem for a long period of time doesn't bring you any closer to understanding the solution. Expertise is highly overrated. The Wright Brothers didn't have physics degrees. They did not even graduate from high school, let alone attend college. Their process for developing the first airplanes was much closer to vibe coding from a shallow surface-level understanding than from deeply contemplating the problem.
Boring thoughts always existed, but they generally stayed in your home or community. Then Facebook came along, and we were able to share them worldwide. And now AI makes it possible to quickly make and share your boring tools.
Real creativity is out there, and plenty of people are doing incredibly creative things with AI. But AI is not making people boring—that was a preexisting condition.
I have a report that I made with AI on how customers leave our firm…The first pass looked great but was basically nonsense. After eight hours of iteration, the resulting report is better than I could’ve made on my own, by a lot. But it got there because I brought a lot of emotional energy to the AI party.
As workers, we need to develop instincts for “plausible but incomplete” and as managers we need to find filters that get rid of the low-effort crap.
The number of people who I see having E-mail conversations where person A uses an LLM to turn two sentences into ten paragraphs, and person B uses an LLM to summarize the ten paragraphs into two sentences, is becoming genuinely alarming to me.
It was especially unfortunate because to do its thing, the code required a third party's own personal user credentials including MFA, which is a complete non-starter in server-side code, but apparently the manager's LLM wasn't aware enough to know that.