This was so wordy I had to ask an LLM to tell me what the point is.
So you don't have to:
"you don’t have to embrace a trend, tool, or narrative simply because others say you should — especially if it doesn’t resonate with you or align with your values"
An important new twist to add to the great AI versus NO AI discussion.
> There were entire classes of Hacker News submissions that I refused to read the comments on. Including the comments about this article, should such comments ever materialize.
The author has made the correct call. There's a pretty deep irony that all the top-level comments at the time of this writing are about how the article is too long. It's quite clearly not trying to succinctly convince you of a point, it's meant to be a piece of genuinely human writing, and enjoyed (or not!) on the basis of that.
I loved the essay. If anyone didn't enjoy it, quit halfway, or decided not to read it, that's absolutely fine. There's plenty of thoughtful writing I don't enjoy or don’t feel like spending my time with. But it is well-executed.
But apparently a sizable percentage of today's HN user base can't get through it, and finds the very idea of being able to get through anything longer than an LLM summary objectionable.
If I wanted to be an ass, I would call it a "skill issue."
But I don't want to be an ass. It's just deeply sad.
Honestly I don't think so. An essay like this is more than just content, it's an experience for the reader. I value the time I got to spend with it and feel I came way with value that a summary or condensed version would just not have had.
And "if a clod be washed away by the C" which is a reference to John Donne's poem that introduced the phrases "No man is an island" and "for whom the bell tolls".
I accidently clicked on the article instead of the comments link for this one, a rare mistake as I usually glance at the comments before deciding to read, but I'm glad I did in this case.
I read it all, and found myself engaged throughout. Not to say that it was all riveting, there were certainly dryer spots than others, but it felt 'real'. Maybe they did use AI (I somehow doubt that given the content), but even if they did they went over everything in a way that retained a voice that felt authentic.
I hate many of the articles I read now all feel like they have the same half hearted attempt at trying to grab your attention without every actually clearly saying what they mean.
As for the content, I had actually just been told by management this last week that I need to become AI 'fluent' as part of future performance evaluations and I have been deeply conflicted about it. I do think AI has value to add, but I don't think it's something that should be forced and so this article resonated with me.
It's a long read, and not for everyone, but I recommend it as a way of hearing another humans opinion and deciding for yourself if it has value.
I see a lot of arguing over whether this is "good" or not. This seems like a subjective question. Some people enjoy it, if you don't, it wasn't written for you--don't read it.
Maybe the arguing is really over whether it's higher-status to enjoy longform content, or to criticize it for not being more efficient? By identifying the argument, I've revealed it as silly, and clearly proven myself to be higher status than either side. The arguing may stop now. You're welcome.
Maybe it’s just me, but I feel that same kind of treachery when somebody tries to pass off a piece of AI-generated work as if it were their own voice.
There's a flaw in the Milli Vanilli argument. The band had no input into their songs. They 'performed' them by lip-syncing on stage, but all of the music and lyrics were someone elses. Milli Vanilli had no part in the creative process.
That's not technically true of AI content. There's some tiny little seed of a creative starting point in the form of a prompt needed for AI. When someone makes something with Claude or Nano Banana it's based on their idea, with their prompt, and their taste selecting whether the output is an acceptable artefact of what they wanted to make. I don't think you can just disregard that. They might not have wielded the IDE or camera or whatever, and you might believe that prompting and selecting which output you like has no value, but you can't claim there's no input or creativity required from the author. There is.
I think this essay illustrates pretty well the value in indulging an experience not just for the sake of it but to try and truly know it emotionally.. and perhaps also given some of the responses, it is rightly counterbalancing a lack of appreciation and understanding for anyone doing just that.
I do wonder the prospects of any etsy-like outcome for largely hand crafted software though. While you can personally find stylistic expression in the craft i'm not sure how apparent the nuances of crafting code is to users of the product beyond the requirements of a UX design and vision. It's hard not to imagine generation industrializing a lot of this part of the craft of making software.
For me I think the important thing to not lose sight as we use generation more and more in software is our care for the work piece. It feels like care, and deep understanding are set up to become further valuable rarities in the future as we become less and less intimately involved and we have to be intentional about in order to keep.
I feel like there is some parallels here to industrial designers and their desire to hold on to obsessing about and understanding the details in the face of using industrialized tooling and being very much removed from the intimate feeling of crafting every millimetre. Deeply caring is still meaningful and valuable even if it isn't minimally required.
It’s a very principled and well reasoned stance. Think it underestimates the relentlessness of progress and capitalism though. Short of those that are independently wealthy and can do artisanal things for the sake of it I suspect most shortly won’t have a choice
> then I become a little pissed off at having my time and attention wasted by somebody who didn’t care enough about what they were doing to actually do that thing.
I remember people saying this about emails vs postal mail.
"I’m not arguing that this technology should be unilaterally destroyed; I am arguing that we are collectively using it in the dumbest possible way, causing the most self-inflicted injury, and maximizing the amount of angst and suffering we’ll all have to contend with. I am angry at generative AI because it seems to be making us think and act like complete idiots."
> It is a miracle of human ingenuity that we can etch 100 billion transistors onto a piece of rock we dug out of the ground.
I know this is probably a deliberate simplification as part of a rhetorical flourish but one of my favourite parts about semiconductors is the fact that we don't dig up the rocks, we grow them to order. The though fills me with childlike wonder...
This really resonated with me, thank you for writing it <3
> Companies value velocity and new launches and shipping first at all costs because of course they do; it’s table stakes. Speed of delivery is basically the number one corporate value of every organization whether they admit to it or not.
Yeah this one is again one of the causes of where we are today (alongside profit extraction, or perhaps because of it). It used to be the case that you would find companies that would offer quality at a slightly higher price, and people would be more than willing to pay for it. Now the feeling is that this is all marketing driven and there is no 'higher quality' because everyone gave up and went after speed of delivery. And well, as the old saying goes, that's valuable when you're catching fleas.
I feel this article genuinely represents the gold standard for all such posts in this current genre and zeitgeist. It really does feel like a Spirit Bomb that Goku asked the entire planet to lend their energy to build, a collection of all the hopes and dreams against a mind-rending and incalculable opponent that seeks to eradicate the last remaining few shreds of the craft of software development that feel noble and decent.
I find myself wondering what to do with this article and the incredibly well-condensed collection of sentiments therein. Do I drop a furtive link in that #ai-enthusiasts Slack channel, in full view of everybody, including that CTO of mine who is REALLY excited about this stuff, and isn't quite ready to issue an AI usage mandate? Or do I keep this secreted away, like a bible in my left jacket pocket in case one of "their" sharpshooters attempts to assassinate me with an AI silver bullet, as though I were some wild and reckless heathen that seeks to abandon all good sense in the selfish pursuit of personal pride?
I don't mean to be hyperbolic, but I never envisioned this moment of feeling bizarre, feckless and iconoclast for wanting to do what feels like the reasonable thing. The circular finances propping this stuff up don't add up. Do I really want to be left dependent on a thing that only really feels like it just flashed into precarious existence, that will almost without a doubt be propped up, classed as "too big to fail" by one particular nation-state, whose actual intent and interests are to subjugate its subjects into a quivering oblivion when it's finally acknowledged the whole thing is in arrears? No, goddammit, I shouldn't have to accept this barrage against and wholesale weakening of my strongest assets: My mind and my discipline. At least that's how I feel about it anyway.
You wanna know why this article is great? I can't quote it. There isn't a single, gold nugget line in this post that can be copy pasted into any possible form of short form content, without it losing some important aspect of the original message. Every idea is presented in conjunction with important supporting details that, if you take the time to digest it, you will finally get it. Why we recoil at AI generated content. Why code quality IS product quality. What "craftsmanship" argument is actually about. And like 12 other nuanced ideas we've all heard before, but may not have fully understood. I have nothing but immense praise for the author.
Great article, but a couple of things jarred a bit.
I'm in a technical but non-IT industry that currently rents its software at commercial scale. The software in question that I use is terrible, and it is probably the worst on the market. The industry is such that it is not an exaggeration to say that people have died on account of its flaws. All other solutions in the domain are better but not good.
I have had a slowly-congealing dream/blueprint in my head for over a decade about how the system I use should work, and in the last 6 months I have been accelerated enough by AI (significantly more so by Opus 4.5/6) that I have built a version of the software that I am now using in production at in my job, and it is the most satisfied I have been in my career in the last decade.
Point being (and it was almost made in the article) that the software doesn't actually matter (no-one's reading the assembly either way) but the function and what it enables does. If it doesn't, no end user cares whether it was 25k SLoC or 25M.
I agree with the article's general thrust: use AI if you want to, don't use it if you don't want to, whether or not you'd want to will probably change as AI continues to evolve, and most people seem to be being pushed to use AI in the dumbest ways imaginable.
----
I rather strongly disagree with the framing around the environmental impacts, though; the article would make a much stronger point if it resisted the urge to peddle the same “muh water and electricity” disinformation that gets parroted all over the place by people who can't be bothered to put numbers into the context of other numbers.
For example:
> A single DGX B200 AI server is rated to consume 14,300 watts of electrical power at peak. You can cram about four of these on a rack if you like to live on the edge, and these four units might draw something like 200 amps of current combined. For a point of comparison, a typical single-family home in the United States will have wires from the utility company that are thick enough to provide a 200 amp service.
Cool, and how many households' worth of AI queries would that single B200 (let alone the rack of 'em) be able to handle? Probably a lot more than any individual household could ever hope to produce per second, even assuming a household consisting entirely of hardcore AI stans (let alone someone like the author, or like myself, who uses AI sparingly). Each of those servers is handling requests from thousands upon thousands of users; those power and water requirements get amortized over such a large quantity of requests (and people making them) that if you've ever eaten a single hamburger in your entire life then you've done more harm to the environment than hundreds (if not thousands) of those AI queries.
This all comes after this quip in the margins:
> You think you’re just gonna self-host an open weight model like GLM-5 on your personal hardware and cut out the hosting costs? Well, alright, hope you have 1,727 GB of VRAM lying around.
and like… the author does understand that not everyone needs such a large model with such a large VRAM requirement, right? Or that VRAM itself ain't even strictly necessary (it just happens to make things faster — which is more important for a server handling requests from thousands of users than it is for my laptop handling requests from exactly one user: me)? That's indeed part of the issue the author correctly identifies with people using AI in seemingly the dumbest way possible: that dumbness includes the demand for instantaneous responses, and the consequent demand for throwing more and more VRAM and SSDs at the problem, when “just make a cup of coffee while the LLM ‘thinks’ about what you asked of it” is a perfectly workable approach. As I'm typing out this comment, I've got Olmo 3.1 on this same exact machine doing a bunch of thinking about how to respond to me asking it “How much wood would a woodchuck chuck if a woodchuck could chuck wood?”¹, and it's totally fine that it's taking multiple minutes because there are other things I can do while I wait.
This all ain't to say that we shouldn't care about AI's power and water usage. We should absolutely be pushing for better efficiency. That includes acknowledging that there are options besides “throw more and more VRAM at it and hope for the best”; the article instead prefers to assume that the big beefy servers are the only option, dismissing the notion of self-hosting with little thought, and that dismissal does the article's broader point a disservice.
----
The discussion around AI being considered a “tool” also rubbed me the wrong way a bit:
> This unlocks a common refrain from the booster class: “A true craftsperson uses every tool at their disposal!” Which, if you think about it for more than three seconds, is ridiculous on its face. Gotta dig some holes for fence posts? Okay! Bring along every shovel on the truck, the Ditch Wit...
28 comments
[ 2.9 ms ] story [ 39.1 ms ] threadOver sixteen thousand words about how the author doesn’t really use language models very much but might in the future
So you don't have to:
"you don’t have to embrace a trend, tool, or narrative simply because others say you should — especially if it doesn’t resonate with you or align with your values"
An important new twist to add to the great AI versus NO AI discussion.
The author has made the correct call. There's a pretty deep irony that all the top-level comments at the time of this writing are about how the article is too long. It's quite clearly not trying to succinctly convince you of a point, it's meant to be a piece of genuinely human writing, and enjoyed (or not!) on the basis of that.
I loved the essay. If anyone didn't enjoy it, quit halfway, or decided not to read it, that's absolutely fine. There's plenty of thoughtful writing I don't enjoy or don’t feel like spending my time with. But it is well-executed.
But apparently a sizable percentage of today's HN user base can't get through it, and finds the very idea of being able to get through anything longer than an LLM summary objectionable.
If I wanted to be an ass, I would call it a "skill issue."
But I don't want to be an ass. It's just deeply sad.
Honestly I don't think so. An essay like this is more than just content, it's an experience for the reader. I value the time I got to spend with it and feel I came way with value that a summary or condensed version would just not have had.
I read it all, and found myself engaged throughout. Not to say that it was all riveting, there were certainly dryer spots than others, but it felt 'real'. Maybe they did use AI (I somehow doubt that given the content), but even if they did they went over everything in a way that retained a voice that felt authentic.
I hate many of the articles I read now all feel like they have the same half hearted attempt at trying to grab your attention without every actually clearly saying what they mean.
As for the content, I had actually just been told by management this last week that I need to become AI 'fluent' as part of future performance evaluations and I have been deeply conflicted about it. I do think AI has value to add, but I don't think it's something that should be forced and so this article resonated with me.
It's a long read, and not for everyone, but I recommend it as a way of hearing another humans opinion and deciding for yourself if it has value.
Maybe the arguing is really over whether it's higher-status to enjoy longform content, or to criticize it for not being more efficient? By identifying the argument, I've revealed it as silly, and clearly proven myself to be higher status than either side. The arguing may stop now. You're welcome.
There's a flaw in the Milli Vanilli argument. The band had no input into their songs. They 'performed' them by lip-syncing on stage, but all of the music and lyrics were someone elses. Milli Vanilli had no part in the creative process.
That's not technically true of AI content. There's some tiny little seed of a creative starting point in the form of a prompt needed for AI. When someone makes something with Claude or Nano Banana it's based on their idea, with their prompt, and their taste selecting whether the output is an acceptable artefact of what they wanted to make. I don't think you can just disregard that. They might not have wielded the IDE or camera or whatever, and you might believe that prompting and selecting which output you like has no value, but you can't claim there's no input or creativity required from the author. There is.
I do wonder the prospects of any etsy-like outcome for largely hand crafted software though. While you can personally find stylistic expression in the craft i'm not sure how apparent the nuances of crafting code is to users of the product beyond the requirements of a UX design and vision. It's hard not to imagine generation industrializing a lot of this part of the craft of making software.
For me I think the important thing to not lose sight as we use generation more and more in software is our care for the work piece. It feels like care, and deep understanding are set up to become further valuable rarities in the future as we become less and less intimately involved and we have to be intentional about in order to keep.
I feel like there is some parallels here to industrial designers and their desire to hold on to obsessing about and understanding the details in the face of using industrialized tooling and being very much removed from the intimate feeling of crafting every millimetre. Deeply caring is still meaningful and valuable even if it isn't minimally required.
I remember people saying this about emails vs postal mail.
"I’m not arguing that this technology should be unilaterally destroyed; I am arguing that we are collectively using it in the dumbest possible way, causing the most self-inflicted injury, and maximizing the amount of angst and suffering we’ll all have to contend with. I am angry at generative AI because it seems to be making us think and act like complete idiots."
Made me smile.
I know this is probably a deliberate simplification as part of a rhetorical flourish but one of my favourite parts about semiconductors is the fact that we don't dig up the rocks, we grow them to order. The though fills me with childlike wonder...
> Companies value velocity and new launches and shipping first at all costs because of course they do; it’s table stakes. Speed of delivery is basically the number one corporate value of every organization whether they admit to it or not.
Yeah this one is again one of the causes of where we are today (alongside profit extraction, or perhaps because of it). It used to be the case that you would find companies that would offer quality at a slightly higher price, and people would be more than willing to pay for it. Now the feeling is that this is all marketing driven and there is no 'higher quality' because everyone gave up and went after speed of delivery. And well, as the old saying goes, that's valuable when you're catching fleas.
I find myself wondering what to do with this article and the incredibly well-condensed collection of sentiments therein. Do I drop a furtive link in that #ai-enthusiasts Slack channel, in full view of everybody, including that CTO of mine who is REALLY excited about this stuff, and isn't quite ready to issue an AI usage mandate? Or do I keep this secreted away, like a bible in my left jacket pocket in case one of "their" sharpshooters attempts to assassinate me with an AI silver bullet, as though I were some wild and reckless heathen that seeks to abandon all good sense in the selfish pursuit of personal pride?
I don't mean to be hyperbolic, but I never envisioned this moment of feeling bizarre, feckless and iconoclast for wanting to do what feels like the reasonable thing. The circular finances propping this stuff up don't add up. Do I really want to be left dependent on a thing that only really feels like it just flashed into precarious existence, that will almost without a doubt be propped up, classed as "too big to fail" by one particular nation-state, whose actual intent and interests are to subjugate its subjects into a quivering oblivion when it's finally acknowledged the whole thing is in arrears? No, goddammit, I shouldn't have to accept this barrage against and wholesale weakening of my strongest assets: My mind and my discipline. At least that's how I feel about it anyway.
Don't skim this one.
I'm in a technical but non-IT industry that currently rents its software at commercial scale. The software in question that I use is terrible, and it is probably the worst on the market. The industry is such that it is not an exaggeration to say that people have died on account of its flaws. All other solutions in the domain are better but not good.
I have had a slowly-congealing dream/blueprint in my head for over a decade about how the system I use should work, and in the last 6 months I have been accelerated enough by AI (significantly more so by Opus 4.5/6) that I have built a version of the software that I am now using in production at in my job, and it is the most satisfied I have been in my career in the last decade.
Point being (and it was almost made in the article) that the software doesn't actually matter (no-one's reading the assembly either way) but the function and what it enables does. If it doesn't, no end user cares whether it was 25k SLoC or 25M.
----
I rather strongly disagree with the framing around the environmental impacts, though; the article would make a much stronger point if it resisted the urge to peddle the same “muh water and electricity” disinformation that gets parroted all over the place by people who can't be bothered to put numbers into the context of other numbers.
For example:
> A single DGX B200 AI server is rated to consume 14,300 watts of electrical power at peak. You can cram about four of these on a rack if you like to live on the edge, and these four units might draw something like 200 amps of current combined. For a point of comparison, a typical single-family home in the United States will have wires from the utility company that are thick enough to provide a 200 amp service.
Cool, and how many households' worth of AI queries would that single B200 (let alone the rack of 'em) be able to handle? Probably a lot more than any individual household could ever hope to produce per second, even assuming a household consisting entirely of hardcore AI stans (let alone someone like the author, or like myself, who uses AI sparingly). Each of those servers is handling requests from thousands upon thousands of users; those power and water requirements get amortized over such a large quantity of requests (and people making them) that if you've ever eaten a single hamburger in your entire life then you've done more harm to the environment than hundreds (if not thousands) of those AI queries.
This all comes after this quip in the margins:
> You think you’re just gonna self-host an open weight model like GLM-5 on your personal hardware and cut out the hosting costs? Well, alright, hope you have 1,727 GB of VRAM lying around.
and like… the author does understand that not everyone needs such a large model with such a large VRAM requirement, right? Or that VRAM itself ain't even strictly necessary (it just happens to make things faster — which is more important for a server handling requests from thousands of users than it is for my laptop handling requests from exactly one user: me)? That's indeed part of the issue the author correctly identifies with people using AI in seemingly the dumbest way possible: that dumbness includes the demand for instantaneous responses, and the consequent demand for throwing more and more VRAM and SSDs at the problem, when “just make a cup of coffee while the LLM ‘thinks’ about what you asked of it” is a perfectly workable approach. As I'm typing out this comment, I've got Olmo 3.1 on this same exact machine doing a bunch of thinking about how to respond to me asking it “How much wood would a woodchuck chuck if a woodchuck could chuck wood?”¹, and it's totally fine that it's taking multiple minutes because there are other things I can do while I wait.
This all ain't to say that we shouldn't care about AI's power and water usage. We should absolutely be pushing for better efficiency. That includes acknowledging that there are options besides “throw more and more VRAM at it and hope for the best”; the article instead prefers to assume that the big beefy servers are the only option, dismissing the notion of self-hosting with little thought, and that dismissal does the article's broader point a disservice.
----
The discussion around AI being considered a “tool” also rubbed me the wrong way a bit:
> This unlocks a common refrain from the booster class: “A true craftsperson uses every tool at their disposal!” Which, if you think about it for more than three seconds, is ridiculous on its face. Gotta dig some holes for fence posts? Okay! Bring along every shovel on the truck, the Ditch Wit...