This is well put and something I think a lot of us struggle with.
The internet itself seems filled with the AI maximalists who have no end of hype for the latest prompt 'technique' or AI doomers that would skin you alive if you mention you used AI to make a photo once.
Quietly among peers I talk to there seems to be a very different group that exists: those that recognize AI is super valuable for getting a lot of work done but also feel that suddenly the work is so much less exciting than it was before.
Sure, there's the obvious "I enjoy the mechanics of programming" component that's gone but then there's also this general feeling that is hard to shake which feels a little deeper. I've wondered about it. Is it all the layoffs? The suddenly shaky economics for a lot of our companies? The uncertainty of the career's future in general?
I think the author has put their thumb on it well to describe it as a lack of intellectual community. That feels like at least part of it.
The issue may be simpler than that. The technical cohort that's on the chopping block of automation right now uses seemingly sophisticated words to basically describe what the coal miners and human calculators felt in the past. When you compound this with the general struggle on the social skills dimension that defines the majority of this group, you can see why the idea of stepping out of their respective narrow field and getting a hobby would seem ludicrous.
Yea it’s probably deep seated horror at slowly being replaced. I’m not sure what I feel but ordering a clanker around everyday is exhausting both emotionally and intellectually. I would rather not even talk to the thing, but they fucking suck at software architecture no matter what anyone says. You can’t just leave them alone with a task because you get the worst designs 7/10 times. The other 3/10 scare me enough that I’ve started learning how to install solar and battery systems online.
The whole John Donne meditation the title is referencing:
No man is an island,
Entire of itself;
Every man is a piece of the continent,
A part of the main.
If a clod be washed away by the sea,
Europe is the less,
As well as if a promontory were:
As well as if a manor of thy friend's
Or of thine own were.
Any man's death diminishes me,
Because I am involved in mankind.
And therefore never send to know for whom the bell tolls;
It tolls for thee
(had to memorize it in High School, and its one of my favorites. It was written over 400 years ago and has aged quite well.)
Hemingway also borrowed the penultimate line in For Whom The Bell Tolls. I also had to memorize this back in the day but I still stop to read it whenever I see it somewhere.
I really liked this. I've felt the exact thing the author is describing - I used to care a lot about my craft (iOS apps), and that has been blown apart by AI. Not that craft and taste isn't still a differentiator - I think it's just so much less satisfying to create something perfect over weeks, if you can get 80% of the way in an afternoon.
"Private intellectual activity that is sustained, complex, and long-term requires an external intellectual community to provide material and motivation, like fuel and oxidizer. That private activity, in turn, sustains the community..."
I think a lot of the greatest art has been produced by this kind of "scene". Music for example isn't created only by an individual, it comes from listening and watching and playing with others. Later, you sit down and all of that input turns into something brand new.
I think at the same time I started getting "out of coding", I started getting deeper into music. And if it was just me, that would've been a short lived thing. But having people in real life to play with and bounce off makes it about 100x better and keeps you going long after you would've fallen off on your own.
I'd encourage anyone who feels this way to find something similar. The act of creation is still as essential as before, you just need to direct your energy toward something more purely human. Get into writing, drawing, woodworking... even as these things are instantly mass-producible, we'll still value and pay for humans to do them.
I've been overhauling my parent's backyard. What a great displacement for my creative energy with the added benefit of being physically demanding.
I'd add to your comment that AI is the thousand-ton cherry on top of the perfect storm. Covid destroyed our professional communities. Work doesn't and won't look like how it was pre-2020. AI just makes everything super weirder. It's hard, as a human, to cope.
Yes. I've told a few friends and acquaintances that somehow AI is the least interesting technology in my lifetime - the boosters use AI to write about how they used AI to build new tools to help them AI their AI, the haters write (or sometimes, shockingly, use AI to write) about how AI will AI all of us. Have you tried this new skill? Read my research paper, it shows that AI can do good work unsupervised? Oh, yeah, that's interesting, what do you think about how it did X, Y, and Z? I dunno man, I didn't spend that much time on it.
It's getting lonely out there, even if or especially if you've converged on "I don't care much about AI".
> even if or especially if you've converged on "I don't care much about AI"
I don't care much about AI but hot damn it makes working on the things I do care about easier.
We use it a lot at work (soft eng) and my job has slowly morphed into "Okay how do we continue shipping sane software when we effectively 10x'd our org size in six months as agents became reliable?". That has been pretty fun. I care about creating good software.
And I care about budgeting and finance projections for home finances, especially as a new baby entered the room. But I am very bad at spreadsheet fiddling. AI has been super helpful in doing the fiddling so I can focus on the thinking.
If it had no other function than a better than Google search engine that is able to teach me topics I'm interested in at exactly my speed it would be the most interesting technology to come along in a while. The fact that it does all the other stuff it's doing is hard to even come to grips with in the time it's been around.
It seems we (now more than ever) have to fight for fun, fight for a shared meaningful reality. Arriving again at the idea that enlightenment is a group activity.
I did my Masters in ML and a decade ago it seemed AI was so exciting. You could build relatively good models at home. Our university had a cluster that could train SOTA models. It seemed like every year we were discovering a breakthrough new neural net structure.
Nowadays all the development is done at a bigtech lab, and you aren't going to be able to tell anyone about it. Everyone else is relegated to interacting with slightly larger and larger LLMs over an API. On one hand it's cool that the tech is now far more accessible and impactful, but on the other AI has become immensely boring.
It's a little uncharitable take as I know some people are doing some cool innovative stuff still, and not every tech needs to be new forever anyway. I would feel less strongly if it wasn't impossible to not keep hearing about it :)
I mean, to me, this is very similar to how it was when "the web" came into its current pervasive form. You could write exactly this comment about how people were writing about how they liked the web, on the web, and also about how they didn't like it, maybe shockingly, also on the web, etc. etc.
To me, this is just how it is when a new paradigm hits the scene. Whether it's good or bad, whether you care for it or not, it is pervasive.
But you could still focus on backend stuff and desktop or CLI programs, think and discuss about programming languages and software engineering on the code level in all the non-web related ways with a large community. You could opt-out of most of the web-related things (which is what I did) and still have a wide community to interact with. Whereas AI appears to be dissolving almost all that community. I hope that will turn out to be a false appearance, but the thesis is that it won’t.
There's several things I disagree with in this post (a matter of perspective, I think):
> The discourse around software engineering has gotten dumber. It’s like everyone in the industry lost 30 IQ points. People used to talk about compilers, type systems, logic. Now they talk about “prompts”, “harnesses”, “loops”. The discourse is narrower, shallower, and more repetitive. There’s only so many times I can hear about “agentic harnesses” before I lose my mind.
I agree with the frustration here :)
> You design a revolutionary new programming language: who cares? Maybe Claude cares, for what that’s worth. But humans don’t write or even read the code anymore. The programming language is an implementation detail the humans no longer have to care about.
Perhaps the people who weren't inclined to think about compilers, type systems, and logic before coding agents will never think about those things. If anything, I think about those systems significantly more now, because I view them as ways to describe the search space for the agent ... and perhaps, in the future, even tools to structure agentic latent representations.
So I propose rotating your viewport: yes, perhaps entertain the idea that it doesn't make sense to think about these things for human consumption ... but instead, what if we think of programming language and compiler design as being a tool to control the search space, and suddenly there's lots of new ways to think about these things.
Is this a valuable way to think? I'm not sure, but we should try and find out.
I agree with you and disagree with the OP in similar ways.
I work on my own "agentically-engineered" projects and the pursuit of certain ideas for what you're calling intrinsic motivation, and I don't particularly care if others find it interesting or even find it at all or not, because the ideas themselves are cool to me.
It's kind of like I can suddenly tinker at 2,000% efficiency and share what I'm doing, _and if someone else finds it interesting, that's cool too_
Amazing read. AI indeed completely annihilated every other topic. You can go to any tech conference these days and anything that isn't related to AI just feels so irrelevant. It wasn't like this at all for crypto, NFT or any other hype topic of the day.
LLMs actually _improved_ this for me, because they can be the sparring partner and the endless lexicon of interconnected thoughts, most other humans simply lack the ability to.
They do have their limits too, but I'd genuinely, objectively say that for me and how my brain is wired, they've given me the "community" context that the world could not really provide.
I do feel like I'm approaching the limits of what they can do there though.
It's worse as soon as the things see the prompts of most public-facing harnesses, but even without that, they can only do so much.
Real humans would have a better intellectual community to offer, but for that, you actually need to find them first and hope that your schedules and energy budgets both align.
___
I think as much as we're seeing consequences of AI there, we're even more seeing the consequences of how society itself has been strip mined by big corps and big tech.
Human connection is dead not because of AI but because of Social Media.
Motivation is dead not because of AI or weakness but because spaces and contexts that would foster that have been replaced with corporate-approved standardized gray.
Globalization plays a role here. Atomization. Urbanization.
The expectation to hop around the nation for your job, nuking any ability at building a local community.
Context collapse glueing massive amounts over overhead on every single communication transmission.
This really speaks to me, and Terence Tao literally wrote something very similar few days ago [1] where he advocates for a more holistic approach to intellectual fields, and the value of community building.
On a personal level I had so many to-dos, like learning Rust "properly", but I feel incredibly demotivated now, because I can just get Opus or Sol to do it. But maybe I just need to block out one day a week to be my "no AI day", and just do it.
Here's an analogy: People slowly developed skills on killing things; methods learned over thousands of hours and it is seen as a lifelong skill. These methods remain very useful for understanding the discipline of killing all of the various living things that need to be killed to sustain our modern world.
AI is a gun. Pulling the prompt trigger takes 99% of the effort out of the daily grind - point, shoot, reload.
Once the gun arrives, everyone that needs to kill is getting at least one gun. Use of the old knife, arrow, sword, etc. is downgraded from life skill to sport. The discipline remains intact, but just the parts needed to understand how it all is supposed to work remain useful.
Very well written. I tend to forget how important the people around me are, and as you say, force myself to act with the "high status" intrinsic motivation.
Recently, I'm in a context with extremely talented people I respect and want to impress to some extent. The difference is night and day
On AI and communication with others: before AI when joining a company, the first weeks was all about talking to team mates asking about the systems we own as a team, how things work, how to do X, how to get unstuck…
Nowadays you ask all of that to codex/claude. And it works. It’s a regression. I don’t mind spending 2h helping a colleague debugging an issue. It’s fun, it creates bonds and we both benefit from it (ofc my employer would think it’s a waste of time, but I couldn’t care less about what executive millionaires think)
The end result of this is either AI annihilates human value, and we become pets or are annihilated ourselves, or AI is restricted to never pass a certain level.
I used to get calls often from my friends to ask me questions about my area of expertise. Not urgent things, just friendly "hey--what if..." kinds of questions.
Without the question, they might have never called. But because they called, we usually ended up just talking about other things too. Life. Girls. Work.
I rarely get calls like that anymore. It breaks my heart.
For the last few years, me and my friend have been working on a toy problem together. It's a problem that the most recent wave of LLMs can almost certainly solve and probably even one-shot. We now realize that no one's going to read our paper if we ever decide to publish our work. But we still use our weekly meetings to talk about things and make little progress here and there, come up with new questions, etc. It's bittersweet.
This sounds like a bummer. But honestly I can't relate to this at all. I've never gotten calls like that. It's just not what my relationships with friends and family have ever been about. (Nobody cares about my expertise, and to the extent they have, they always just wanted me to fix their virus interested windows xp machine, because that's the limit of their understanding of what "computer people" know about, and I always hated that!)
I agree that prompting is a much shallower skill than engineering, and I really hate the term “prompt engineering” since I feel it puts something that (by design) is easy to learn on the same level as something relatively hard to learn.
There is admittedly a bit of skill to use these tools, but they’re really not hard to use. For the most part Claude can decipher my typo’d-to-hell prompt and figure out what I want, and even if it gets it wrong the first time I can usually correct it. The hardest part is determining some objective test for Claude Code to test against and “anchor” it, but I still think that’s easier than writing a bunch of unit and integration tests.
I like these tools; I think Claude Code is very useful and even if I didn’t like it I would still use it because it’s clear that’s where the world is headed and I want to stay competitive.
For personal projects I actually care about, I still write code by hand. I do think I write better code than Claude but I would still do it by hand even if I didn’t, just because I realized I was losing my intuition about software because I was outsourcing to Claude for everything.
There are plenty of things that I am honestly fine losing my intuition with. I don’t really derive much enjoyment from reading dmesg logs, and I am more than happy to let a robot have its fun with that, but it can take the mathey shit from my cold dead hands now.
The skill isn't "prompting", it is figuring out what to do, validating that it works the way you want, and thinking about how to best support future iterations of things you might want to do.
Which, not coincidentally, is pretty much the same description as what the job was already about.
I know, but even given that I really don’t think it’s that hard, and certainly not as hard as regular pre-AI software engineering.
We all like to act like “the hard part was figuring out what we want”, but I don’t think I really buy that anymore. Even when I really half-ass Claude prompts, it still usually figures out what I want well enough, and when it doesn’t it usually does once I give it correction.
I think this actually gets at a big part of what is causing this existential crisis. You're right, I don't think it was ever true that "the hard part was figuring out what we want". I think there have always been two hard parts: 1. Figuring out what is useful, and 2. Implementing a useful thing well. The second of these has gotten a lot easier, which means the first has become more important. But software engineers have long had a tendency to underrate that first part and consider it easy. I have long perceived a (to me) baffling arrogance that product management is an easier and less important job than software engineering. But that was never an accurate perspective, imo. And now that the "implementing things" job has gotten easier and the "figuring out what to implement" job has become more important, people with this perspective are really being thrown for a loop.
I think there is high uncertainty of how the world will change. My personal plan has two arms.
1) Be at the top of applying AI. The frontier overestimates of how able the average person is of applying what’s possible now. I would hope that with my software I can make money for a couple of more years. I constantly test my ideas on how far I can scale them through AI and what angles could be not as easily replicated. I’m aware that the average knowledge is catching up and I’m not faster. One could argue that also the expectation of how a software works gets higher. So users have lower attention spans want customized landing pages and UX needs to be more visual intuitive. So part of the gain in downstream benefitting users.
2) You can argue very long about 1) and I have arguments for both sides. So far it works for me and the numbers are on my side, however on the other side my plan B or 2) in this case is getting ready for life outside. I’m going in front of the camera I’m improving my talking skills I’m try to learn to build things. If you want if all knowledge work is gone. It’s literally entertainment, work with your body, serving or selling. I personally like entertainment most of that.
I’ve been trying to predict and I’ve changed my mind often, so now I’m just saying I don’t really know how the world’s gonna change.
> People used to talk about compilers, type systems, logic. Now they talk about “prompts”, “harnesses”, “loops”. The discourse is narrower, shallower, and more repetitive.
I feel like you can write a parody of this for every lower->higher level language change that's happened throughout history. I know LLMs are bigger change than going from ASM->C but it rhymes, at least to me. And I've never liked C, I've used pointers a handful of times in college and then thankfully never had to think about it again. I work in higher-level languages and that's not a thing I need to concern myself with.
I'm sure that statement will encourage name calling by some people or gasps in horror by people who work in C or similar languages but I've never enjoyed them. I find technology, computers, and program languages all fascinating and enjoy learning the history of it. I enjoy things like the Crash Bandicoot blog series talking about all the tricks they had to do to get the game to run on underpowered hardware and a SDK that didn't know what it wanted to be (my own re-wording) but I don't pine to go back to those days.
I see LLMs as just a step up the stack and making it easier for more people to create. Will it create a loss of parts of the art of programming? Undoubtedly. But I imagine there is a long list of people who used to be really good at some aspect of programming that a higher-level language removed the need for, that happens. This is happening on a _much_ larger scale, no doubt but it's the same concept.
> We care, and we should care, about contributing to the world. And if technology makes our contributions superfluous, then what is left?
It's made a specific aspect of our contributions (the actual writing of the code) superfluous but it hasn't removed all the other contributions. Maybe you all are using better models than I am (doubtful, I'm using Claude too) but "Idea -> Implementation" is much faster but "Implementation -> Polished Version" still takes a human (for now), at least in my experience. The LLM still writes the code to get to a polished state but there are tons of little UI/UX things it does badly and needs my guidance on. And, more importantly, writing code was never the hardest part of my job really, it was things like taking requests from the business side and turning them working code. The LLM makes parts of that easier for sure. I love being able to brainstorm and kick ideas around so easily or knock out a POC in an afternoon. However, there is still plenty of "human work" that needs done. I'm sure this will shift further, nothing is ever "settled", but when that happens I'll be looking at the next layer of problems I can solve.
I like solving problems, and yes, the LLM has taken away a whole slew of problems that I used to have to solve myself but I still see an unending ocean of problems to address that will satisfy me the same way writing code used it.
Semi-related: I believe so many people are writing "Software Factories" because it's the "next" problem to solve. Many of us looked up one level in the stack and said, "cool, now that the code is handled, how can I automate/improve the pipelines around that". I feel like as an industry there are many of us who effectively have been working to automate our jobs further, at least as a "background process" in our day-to-day, all along. Now we have a different part of that workflow to automate.
Incredible blog post! I've been very much disagreeing with most mathematicians on this topic and also getting kind of annoyed at how ineloquent they are at getting to the point.
This on the other hand has actually kind of swayed me over? I think it's still a massive net-benefit for society but I resonated a lot with the author's feelings and I want to understand how to cope with them myself, perhaps by joining groups of people who are still interested in pure software (which AI is a part of now).
The -30IQ line in particular reminded me of DHH's ramblings on Twitter the past few days.
I find this perspective to be narrow and uncurious. AI is disrupting community -- it is not utterly dissolving it. I see no evidence of that. So many, including the OP, see AI as separator, an entombing force, somehow a black hole of intellect. In reality it is an evolution of the search engine which connects, navigates and further constructs digital civilization.
Doesn’t that always seem to be the problem with the booster rhetoric. The upside is always somewhere in the hypothetical future while the downsides are already being acutely felt.
> So many, including the OP, see AI as separator, an entombing force, somehow a black hole of intellect.
Because the ultra-rich people who are building and controlling most of it are using it to do exactly those things to society, because they aren't capable of sharing this planet with the rest of us equitably. Were it being used to empower humanity (instead of just a small handful of the worst of us), I'm pretty sure the "AI haters" would be a tiny minority at most (if they existed at all).
In that vein, "Swapping Money For Expertise" [0] suggests that there's a political/power-dynamic at play for why this particular kind of AI is so appealing to your boss' boss:
> The fact that a theory-free AI might demand far more cash to accomplish a task (even a "solved" one like playing chess) than an expert [helping with a symbolic AI] would charge is beside the point. Billionaires have money, they don't have expertise. Theory-free inference is a bid to substitute one for the other: the beauty and terror of being able to manipulate the world without studying or understanding it is that it can be done with money alone. No experts needed.
> In a world in thrall to financial power, expertise is the only substantial form of power that can reliably contest the power of wealth. [...] This is why AI bosses are so violently allergic to the idea of hybridizing AI with symbolic systems that operate on models of the world.
> Prompting is not a skill, at least, it’s a much shallower skill than software engineering.
Right now, that's the case, but I wouldn't put a Kalshi bet on it remaining that way.
I "grew up," from the days of pure Machine Code.
In Ye Days of Yore, all programmers were expected to understand a core dump, which is a printout (yes, a printout) of thousands of hex numbers, representing the memory of the server, when it crashed.
It's tempting to use this as some kind of bar of competence, but I am constantly astounded at the things that are produced by people that would absolutely freak out at a core dump.
In the early days of compiled/interpreted languages, even top-level programmers wrote in FORTRAN or BASIC. ASM was just getting its feet under it.
I think that we'll be seeing some marvelous stuff.
...as long as Skynet doesn't consider us a threat...
This is well put and something I think a lot of us struggle with.
The internet itself seems filled with the AI maximalists who have no end of hype for the latest prompt 'technique' or AI doomers that would skin you alive if you mention you used AI to make a photo once.
Quietly among peers I talk to there seems to be a very different group that exists: those that recognize AI is super valuable for getting a lot of work done but also feel that suddenly the work is so much less exciting than it was before.
Sure, there's the obvious "I enjoy the mechanics of programming" component that's gone but then there's also this general feeling that is hard to shake which feels a little deeper. I've wondered about it. Is it all the layoffs? The suddenly shaky economics for a lot of our companies? The uncertainty of the career's future in general?
I think the author has put their thumb on it well to describe it as a lack of intellectual community. That feels like at least part of it.
> a lack of intellectual community.
The issue may be simpler than that. The technical cohort that's on the chopping block of automation right now uses seemingly sophisticated words to basically describe what the coal miners and human calculators felt in the past. When you compound this with the general struggle on the social skills dimension that defines the majority of this group, you can see why the idea of stepping out of their respective narrow field and getting a hobby would seem ludicrous.
Yea it’s probably deep seated horror at slowly being replaced. I’m not sure what I feel but ordering a clanker around everyday is exhausting both emotionally and intellectually. I would rather not even talk to the thing, but they fucking suck at software architecture no matter what anyone says. You can’t just leave them alone with a task because you get the worst designs 7/10 times. The other 3/10 scare me enough that I’ve started learning how to install solar and battery systems online.
The whole John Donne meditation the title is referencing:
(had to memorize it in High School, and its one of my favorites. It was written over 400 years ago and has aged quite well.)Hemingway also borrowed the penultimate line in For Whom The Bell Tolls. I also had to memorize this back in the day but I still stop to read it whenever I see it somewhere.
Some say it was just Peninsulaman at high tide.
I really liked this. I've felt the exact thing the author is describing - I used to care a lot about my craft (iOS apps), and that has been blown apart by AI. Not that craft and taste isn't still a differentiator - I think it's just so much less satisfying to create something perfect over weeks, if you can get 80% of the way in an afternoon.
"Private intellectual activity that is sustained, complex, and long-term requires an external intellectual community to provide material and motivation, like fuel and oxidizer. That private activity, in turn, sustains the community..."
I think a lot of the greatest art has been produced by this kind of "scene". Music for example isn't created only by an individual, it comes from listening and watching and playing with others. Later, you sit down and all of that input turns into something brand new.
I think at the same time I started getting "out of coding", I started getting deeper into music. And if it was just me, that would've been a short lived thing. But having people in real life to play with and bounce off makes it about 100x better and keeps you going long after you would've fallen off on your own.
I'd encourage anyone who feels this way to find something similar. The act of creation is still as essential as before, you just need to direct your energy toward something more purely human. Get into writing, drawing, woodworking... even as these things are instantly mass-producible, we'll still value and pay for humans to do them.
I think this has really softened the blow for me!
I've been overhauling my parent's backyard. What a great displacement for my creative energy with the added benefit of being physically demanding.
I'd add to your comment that AI is the thousand-ton cherry on top of the perfect storm. Covid destroyed our professional communities. Work doesn't and won't look like how it was pre-2020. AI just makes everything super weirder. It's hard, as a human, to cope.
Yes. I've told a few friends and acquaintances that somehow AI is the least interesting technology in my lifetime - the boosters use AI to write about how they used AI to build new tools to help them AI their AI, the haters write (or sometimes, shockingly, use AI to write) about how AI will AI all of us. Have you tried this new skill? Read my research paper, it shows that AI can do good work unsupervised? Oh, yeah, that's interesting, what do you think about how it did X, Y, and Z? I dunno man, I didn't spend that much time on it.
It's getting lonely out there, even if or especially if you've converged on "I don't care much about AI".
> even if or especially if you've converged on "I don't care much about AI"
I don't care much about AI but hot damn it makes working on the things I do care about easier.
We use it a lot at work (soft eng) and my job has slowly morphed into "Okay how do we continue shipping sane software when we effectively 10x'd our org size in six months as agents became reliable?". That has been pretty fun. I care about creating good software.
And I care about budgeting and finance projections for home finances, especially as a new baby entered the room. But I am very bad at spreadsheet fiddling. AI has been super helpful in doing the fiddling so I can focus on the thinking.
If it had no other function than a better than Google search engine that is able to teach me topics I'm interested in at exactly my speed it would be the most interesting technology to come along in a while. The fact that it does all the other stuff it's doing is hard to even come to grips with in the time it's been around.
It seems we (now more than ever) have to fight for fun, fight for a shared meaningful reality. Arriving again at the idea that enlightenment is a group activity.
I did my Masters in ML and a decade ago it seemed AI was so exciting. You could build relatively good models at home. Our university had a cluster that could train SOTA models. It seemed like every year we were discovering a breakthrough new neural net structure.
Nowadays all the development is done at a bigtech lab, and you aren't going to be able to tell anyone about it. Everyone else is relegated to interacting with slightly larger and larger LLMs over an API. On one hand it's cool that the tech is now far more accessible and impactful, but on the other AI has become immensely boring.
It's a little uncharitable take as I know some people are doing some cool innovative stuff still, and not every tech needs to be new forever anyway. I would feel less strongly if it wasn't impossible to not keep hearing about it :)
I mean, to me, this is very similar to how it was when "the web" came into its current pervasive form. You could write exactly this comment about how people were writing about how they liked the web, on the web, and also about how they didn't like it, maybe shockingly, also on the web, etc. etc.
To me, this is just how it is when a new paradigm hits the scene. Whether it's good or bad, whether you care for it or not, it is pervasive.
But you could still focus on backend stuff and desktop or CLI programs, think and discuss about programming languages and software engineering on the code level in all the non-web related ways with a large community. You could opt-out of most of the web-related things (which is what I did) and still have a wide community to interact with. Whereas AI appears to be dissolving almost all that community. I hope that will turn out to be a false appearance, but the thesis is that it won’t.
There's several things I disagree with in this post (a matter of perspective, I think):
> The discourse around software engineering has gotten dumber. It’s like everyone in the industry lost 30 IQ points. People used to talk about compilers, type systems, logic. Now they talk about “prompts”, “harnesses”, “loops”. The discourse is narrower, shallower, and more repetitive. There’s only so many times I can hear about “agentic harnesses” before I lose my mind.
I agree with the frustration here :)
> You design a revolutionary new programming language: who cares? Maybe Claude cares, for what that’s worth. But humans don’t write or even read the code anymore. The programming language is an implementation detail the humans no longer have to care about.
Perhaps the people who weren't inclined to think about compilers, type systems, and logic before coding agents will never think about those things. If anything, I think about those systems significantly more now, because I view them as ways to describe the search space for the agent ... and perhaps, in the future, even tools to structure agentic latent representations.
So I propose rotating your viewport: yes, perhaps entertain the idea that it doesn't make sense to think about these things for human consumption ... but instead, what if we think of programming language and compiler design as being a tool to control the search space, and suddenly there's lots of new ways to think about these things.
Is this a valuable way to think? I'm not sure, but we should try and find out.
I agree with you and disagree with the OP in similar ways.
I work on my own "agentically-engineered" projects and the pursuit of certain ideas for what you're calling intrinsic motivation, and I don't particularly care if others find it interesting or even find it at all or not, because the ideas themselves are cool to me.
It's kind of like I can suddenly tinker at 2,000% efficiency and share what I'm doing, _and if someone else finds it interesting, that's cool too_
Amazing read. AI indeed completely annihilated every other topic. You can go to any tech conference these days and anything that isn't related to AI just feels so irrelevant. It wasn't like this at all for crypto, NFT or any other hype topic of the day.
LLMs actually _improved_ this for me, because they can be the sparring partner and the endless lexicon of interconnected thoughts, most other humans simply lack the ability to.
They do have their limits too, but I'd genuinely, objectively say that for me and how my brain is wired, they've given me the "community" context that the world could not really provide.
I do feel like I'm approaching the limits of what they can do there though. It's worse as soon as the things see the prompts of most public-facing harnesses, but even without that, they can only do so much.
Real humans would have a better intellectual community to offer, but for that, you actually need to find them first and hope that your schedules and energy budgets both align.
___
I think as much as we're seeing consequences of AI there, we're even more seeing the consequences of how society itself has been strip mined by big corps and big tech.
Human connection is dead not because of AI but because of Social Media.
Motivation is dead not because of AI or weakness but because spaces and contexts that would foster that have been replaced with corporate-approved standardized gray.
Globalization plays a role here. Atomization. Urbanization.
The expectation to hop around the nation for your job, nuking any ability at building a local community.
Context collapse glueing massive amounts over overhead on every single communication transmission.
Etc etc.
AI is often not a cause but just an accelerator.
This really speaks to me, and Terence Tao literally wrote something very similar few days ago [1] where he advocates for a more holistic approach to intellectual fields, and the value of community building.
On a personal level I had so many to-dos, like learning Rust "properly", but I feel incredibly demotivated now, because I can just get Opus or Sol to do it. But maybe I just need to block out one day a week to be my "no AI day", and just do it.
[1] https://mathstodon.xyz/@tao/117395269325940185
Here's an analogy: People slowly developed skills on killing things; methods learned over thousands of hours and it is seen as a lifelong skill. These methods remain very useful for understanding the discipline of killing all of the various living things that need to be killed to sustain our modern world.
AI is a gun. Pulling the prompt trigger takes 99% of the effort out of the daily grind - point, shoot, reload.
Once the gun arrives, everyone that needs to kill is getting at least one gun. Use of the old knife, arrow, sword, etc. is downgraded from life skill to sport. The discipline remains intact, but just the parts needed to understand how it all is supposed to work remain useful.
I don't disagree that the community of software engineering will diminish. What was once a skilled trade is now an unskilled trade.
But I think the intellectual activity will just migrate somewhere else, once all the dust has settled.
People who are intrinsically motivated will continue to be intrinsically motivated and new communities will pop up around new domains.
Coding only came into being 70 years ago. It's not some fundamental human institution.
We're all just going to have to adapt. The wheel turns.
Very well written. I tend to forget how important the people around me are, and as you say, force myself to act with the "high status" intrinsic motivation.
Recently, I'm in a context with extremely talented people I respect and want to impress to some extent. The difference is night and day
What about all the intellectual communities springing up around AI use?
On AI and communication with others: before AI when joining a company, the first weeks was all about talking to team mates asking about the systems we own as a team, how things work, how to do X, how to get unstuck…
Nowadays you ask all of that to codex/claude. And it works. It’s a regression. I don’t mind spending 2h helping a colleague debugging an issue. It’s fun, it creates bonds and we both benefit from it (ofc my employer would think it’s a waste of time, but I couldn’t care less about what executive millionaires think)
>ofc my employer would think it’s a waste of time, but I couldn’t care less about what executive millionaires think)
Do you care about the ROI of the shares in your 401k or other retirement savings? People usually tend to pick the shares and funds that earn the most.
The end result of this is either AI annihilates human value, and we become pets or are annihilated ourselves, or AI is restricted to never pass a certain level.
I used to get calls often from my friends to ask me questions about my area of expertise. Not urgent things, just friendly "hey--what if..." kinds of questions.
Without the question, they might have never called. But because they called, we usually ended up just talking about other things too. Life. Girls. Work.
I rarely get calls like that anymore. It breaks my heart.
For the last few years, me and my friend have been working on a toy problem together. It's a problem that the most recent wave of LLMs can almost certainly solve and probably even one-shot. We now realize that no one's going to read our paper if we ever decide to publish our work. But we still use our weekly meetings to talk about things and make little progress here and there, come up with new questions, etc. It's bittersweet.
This sounds like a bummer. But honestly I can't relate to this at all. I've never gotten calls like that. It's just not what my relationships with friends and family have ever been about. (Nobody cares about my expertise, and to the extent they have, they always just wanted me to fix their virus interested windows xp machine, because that's the limit of their understanding of what "computer people" know about, and I always hated that!)
I agree that prompting is a much shallower skill than engineering, and I really hate the term “prompt engineering” since I feel it puts something that (by design) is easy to learn on the same level as something relatively hard to learn.
There is admittedly a bit of skill to use these tools, but they’re really not hard to use. For the most part Claude can decipher my typo’d-to-hell prompt and figure out what I want, and even if it gets it wrong the first time I can usually correct it. The hardest part is determining some objective test for Claude Code to test against and “anchor” it, but I still think that’s easier than writing a bunch of unit and integration tests.
I like these tools; I think Claude Code is very useful and even if I didn’t like it I would still use it because it’s clear that’s where the world is headed and I want to stay competitive.
For personal projects I actually care about, I still write code by hand. I do think I write better code than Claude but I would still do it by hand even if I didn’t, just because I realized I was losing my intuition about software because I was outsourcing to Claude for everything.
There are plenty of things that I am honestly fine losing my intuition with. I don’t really derive much enjoyment from reading dmesg logs, and I am more than happy to let a robot have its fun with that, but it can take the mathey shit from my cold dead hands now.
The skill isn't "prompting", it is figuring out what to do, validating that it works the way you want, and thinking about how to best support future iterations of things you might want to do.
Which, not coincidentally, is pretty much the same description as what the job was already about.
I know, but even given that I really don’t think it’s that hard, and certainly not as hard as regular pre-AI software engineering.
We all like to act like “the hard part was figuring out what we want”, but I don’t think I really buy that anymore. Even when I really half-ass Claude prompts, it still usually figures out what I want well enough, and when it doesn’t it usually does once I give it correction.
I think this actually gets at a big part of what is causing this existential crisis. You're right, I don't think it was ever true that "the hard part was figuring out what we want". I think there have always been two hard parts: 1. Figuring out what is useful, and 2. Implementing a useful thing well. The second of these has gotten a lot easier, which means the first has become more important. But software engineers have long had a tendency to underrate that first part and consider it easy. I have long perceived a (to me) baffling arrogance that product management is an easier and less important job than software engineering. But that was never an accurate perspective, imo. And now that the "implementing things" job has gotten easier and the "figuring out what to implement" job has become more important, people with this perspective are really being thrown for a loop.
I think there is high uncertainty of how the world will change. My personal plan has two arms.
1) Be at the top of applying AI. The frontier overestimates of how able the average person is of applying what’s possible now. I would hope that with my software I can make money for a couple of more years. I constantly test my ideas on how far I can scale them through AI and what angles could be not as easily replicated. I’m aware that the average knowledge is catching up and I’m not faster. One could argue that also the expectation of how a software works gets higher. So users have lower attention spans want customized landing pages and UX needs to be more visual intuitive. So part of the gain in downstream benefitting users.
2) You can argue very long about 1) and I have arguments for both sides. So far it works for me and the numbers are on my side, however on the other side my plan B or 2) in this case is getting ready for life outside. I’m going in front of the camera I’m improving my talking skills I’m try to learn to build things. If you want if all knowledge work is gone. It’s literally entertainment, work with your body, serving or selling. I personally like entertainment most of that.
I’ve been trying to predict and I’ve changed my mind often, so now I’m just saying I don’t really know how the world’s gonna change.
> People used to talk about compilers, type systems, logic. Now they talk about “prompts”, “harnesses”, “loops”. The discourse is narrower, shallower, and more repetitive.
I feel like you can write a parody of this for every lower->higher level language change that's happened throughout history. I know LLMs are bigger change than going from ASM->C but it rhymes, at least to me. And I've never liked C, I've used pointers a handful of times in college and then thankfully never had to think about it again. I work in higher-level languages and that's not a thing I need to concern myself with.
I'm sure that statement will encourage name calling by some people or gasps in horror by people who work in C or similar languages but I've never enjoyed them. I find technology, computers, and program languages all fascinating and enjoy learning the history of it. I enjoy things like the Crash Bandicoot blog series talking about all the tricks they had to do to get the game to run on underpowered hardware and a SDK that didn't know what it wanted to be (my own re-wording) but I don't pine to go back to those days.
I see LLMs as just a step up the stack and making it easier for more people to create. Will it create a loss of parts of the art of programming? Undoubtedly. But I imagine there is a long list of people who used to be really good at some aspect of programming that a higher-level language removed the need for, that happens. This is happening on a _much_ larger scale, no doubt but it's the same concept.
> We care, and we should care, about contributing to the world. And if technology makes our contributions superfluous, then what is left?
It's made a specific aspect of our contributions (the actual writing of the code) superfluous but it hasn't removed all the other contributions. Maybe you all are using better models than I am (doubtful, I'm using Claude too) but "Idea -> Implementation" is much faster but "Implementation -> Polished Version" still takes a human (for now), at least in my experience. The LLM still writes the code to get to a polished state but there are tons of little UI/UX things it does badly and needs my guidance on. And, more importantly, writing code was never the hardest part of my job really, it was things like taking requests from the business side and turning them working code. The LLM makes parts of that easier for sure. I love being able to brainstorm and kick ideas around so easily or knock out a POC in an afternoon. However, there is still plenty of "human work" that needs done. I'm sure this will shift further, nothing is ever "settled", but when that happens I'll be looking at the next layer of problems I can solve.
I like solving problems, and yes, the LLM has taken away a whole slew of problems that I used to have to solve myself but I still see an unending ocean of problems to address that will satisfy me the same way writing code used it.
Semi-related: I believe so many people are writing "Software Factories" because it's the "next" problem to solve. Many of us looked up one level in the stack and said, "cool, now that the code is handled, how can I automate/improve the pipelines around that". I feel like as an industry there are many of us who effectively have been working to automate our jobs further, at least as a "background process" in our day-to-day, all along. Now we have a different part of that workflow to automate.
Incredible blog post! I've been very much disagreeing with most mathematicians on this topic and also getting kind of annoyed at how ineloquent they are at getting to the point.
This on the other hand has actually kind of swayed me over? I think it's still a massive net-benefit for society but I resonated a lot with the author's feelings and I want to understand how to cope with them myself, perhaps by joining groups of people who are still interested in pure software (which AI is a part of now).
The -30IQ line in particular reminded me of DHH's ramblings on Twitter the past few days.
I find this perspective to be narrow and uncurious. AI is disrupting community -- it is not utterly dissolving it. I see no evidence of that. So many, including the OP, see AI as separator, an entombing force, somehow a black hole of intellect. In reality it is an evolution of the search engine which connects, navigates and further constructs digital civilization.
This assertion would look good beside some concrete examples.
Doesn’t that always seem to be the problem with the booster rhetoric. The upside is always somewhere in the hypothetical future while the downsides are already being acutely felt.
> So many, including the OP, see AI as separator, an entombing force, somehow a black hole of intellect.
Because the ultra-rich people who are building and controlling most of it are using it to do exactly those things to society, because they aren't capable of sharing this planet with the rest of us equitably. Were it being used to empower humanity (instead of just a small handful of the worst of us), I'm pretty sure the "AI haters" would be a tiny minority at most (if they existed at all).
In that vein, "Swapping Money For Expertise" [0] suggests that there's a political/power-dynamic at play for why this particular kind of AI is so appealing to your boss' boss:
> The fact that a theory-free AI might demand far more cash to accomplish a task (even a "solved" one like playing chess) than an expert [helping with a symbolic AI] would charge is beside the point. Billionaires have money, they don't have expertise. Theory-free inference is a bid to substitute one for the other: the beauty and terror of being able to manipulate the world without studying or understanding it is that it can be done with money alone. No experts needed.
> In a world in thrall to financial power, expertise is the only substantial form of power that can reliably contest the power of wealth. [...] This is why AI bosses are so violently allergic to the idea of hybridizing AI with symbolic systems that operate on models of the world.
[0] https://pluralistic.net/2026/10/06/nonfungible/
> Prompting is not a skill, at least, it’s a much shallower skill than software engineering.
Right now, that's the case, but I wouldn't put a Kalshi bet on it remaining that way.
I "grew up," from the days of pure Machine Code.
In Ye Days of Yore, all programmers were expected to understand a core dump, which is a printout (yes, a printout) of thousands of hex numbers, representing the memory of the server, when it crashed.
It's tempting to use this as some kind of bar of competence, but I am constantly astounded at the things that are produced by people that would absolutely freak out at a core dump.
In the early days of compiled/interpreted languages, even top-level programmers wrote in FORTRAN or BASIC. ASM was just getting its feet under it.
I think that we'll be seeing some marvelous stuff.
...as long as Skynet doesn't consider us a threat...