5 ms·

The irony is that the vast deskilling that's happening because of this means that most "software engineers" will become incapable of understanding, let alone fi

by titzer 7mo ago

The irony is that the vast deskilling that's happening because of this means that most "software engineers" will become incapable of understanding, let alone fixing or even building new versions of the systems that they are utterly dependent on. There should be thousands or tens of thousands people worldwide that can build the operating systems, virtual machines, libraries, containers, and applications that AI is built on. But the number will dwindle and we'll ironically be unable to build what our ancestors did, utterly dependent on the AI artifacts to do it for us. God I hope it doesn't all crash at once.

nicksergeant 7mo ago
I feel I've upskilled in so many directions (not just "ability to prompt LLMs") since going all in on LLM coding. So many tools, techniques, systems, and new areas of research I'd never have had the time to fully learn in the past. I have a hard time believing any tenured developer is not actually learning things when using LLMs to build. They make interesting choices that are repeatable (new CLIs I didn't even know existed, writing scripts to churn through tricky data, using specific languages for specific tasks like Go for concurrently working through large numerous tasks, etc.) Anyone not learning things via LLM coding right now either doesn't care at all about the underlying code/systems, or they had no foundational knowledge or interest in programming to begin with (which is also a valid way to use these tools, but they don't work very well without guidance for too long [yet]).
- Waterluvian 7mo ago
  I think there's a considerable difference in its ability to help with breadth vs. depth of expertise.
- anovikov 7mo ago
  This. I never had patience to figure how to build a from-scratch iOS app because it required too much boilerplate work. Now i do, and i got to enjoy Swift as a language, and learned a lot of iOS (and Mac) APIs.
  - JustResign 7mo ago
    But it isn't "from scratch", is it? It's "from Claude".
    - nicksergeant 7mo ago
      If you build a house from scratch but you didn't mill the lumber, did you build it from scratch? If you make a pizza from scratch but you used canned sauce was it from scratch? What if you used store bought dough? What if you made the sauce and the dough but you didn't grow the tomato?
- zozbot234 7mo ago
  What do you mean by "LLM coding"? That's not a very meaningful term, it covers everything from 100% vibe coded projects, to using the LLM to gradually flesh out a careful initial design and then verifying that the implementation is done correctly at every step with meticulous human review and checking.
  - nicksergeant 7mo ago
    The latter.
- tripledry 7mo ago
  For me both are true at the same time. I vividly remember understanding how calculus works after watching some 3blue1brown videos on youtube, but once I looked at some exercises I quickly realized I was not able to solve them. Similar thing happens with LLMs and programming. Sure I understand the code but I'm not intimately familiar with it like if I programmed it "old school". So yes, I do learn more but I can't shake the feeling that there is some dunning kruger effect going on. In essence I think that "banging my head against the wall" while learning is a key part of the learning process. Or maybe it's just me :D
  - mwigdahl 7mo ago
    It's not just you. I feel the same thing, and I saw it in practice helping my son study for a chemistry test just last night. He had worked through a bunch of problems by following the steps in his notes and got the right answers, but couldn't solve them without the notes because his comprehension of why he was taking all the steps wasn't solid. Once we addressed that, he did great solo. Working the mechanics of the problems with the notes helped, but it was getting independent understanding of the reason for each step that put everything together for him.
- agentultra 7mo ago
  > Anyone not learning things via LLM coding right now either doesn't care at all about the underlying code/systems How many bytes is a pointer in C? How many bytes is a shared pointer in C++? What does sysctl do? What about fsync? What is a mutex lock? How is it different from a spin lock? You want to find the n nearest points to a given point on a 2-D Cartesian plane. Could you write the code to solve that on your own? Can you answer any of these questions without searching for the answer? I don't use LLMs and I learn things fine. Always have. For several decades. I care deeply about the underlying code and systems. It annoys me when people say they do and they cannot even understand how the computer works. I'm fine with people having domain-specific knowledge of programming: maybe you've only been interested in web development and scripting DOM elements. But don't pretend that your expertise in that area means you understand how to write an operating system. Or worse: that it prevents you from learning how to write an operating system. You can do that without an LLM. There's no royal road. You have to understand the theory, read the books, read the code, write the code, make mistakes, fix mistakes, read papers, talk to other people with more experience than you... and just write code. And rewrite it. And do it all again. I find the opposite is true: those who use LLM coding exclusively never enjoyed programming to begin with, only learned as much as they needed to, and want the end results.
  - nicksergeant 7mo ago
    Agree with pretty much everything you wrote here, I guess with the addendum that LLMs can be a part of the learning experience you're describing. It's as easy as telling the LLM "don't write a single line of code nor command, I want to do everything, your goal is to help me understand what we're doing here." There are always going to be people who just want the end result. The only difference now is that LLM tools allow them to get much closer to the end result than they previously were able to. And on the other side, there are always going to be people who want to _understand_ what's happening, and LLMs can help accelerate that. I use LLMs as a personalized guide to learning new things.
    - agentultra 7mo ago
      Best of luck in your journey! To those reading this thread though, be wary of the answers LLMs generate: they're plausible sounding and the LLM's are designed to be sycophants. Be wary, double check their answers to your queries against credible sources. And read the source!
- titzer 7mo ago
  Learning calculus by watching the professor solve integrals on the board for an hour doesn't result in the same level and depth of understanding as working through homeworks every week for a semester. If you ran off to your TA to solve every problem in your homework, you just won't learn calculus. I've vibe coded plenty. I mostly don't look at the crap coming out. Don't want to. When I do I absorb a tiny bit, but not enough to recreate the thing from scratch. I might have a modicum more surface-level knowledge, but I don't have deep understanding and I don't have skills. To the extent that I've fixed or tweaked AI-generated code, it's not been to restructure, rearchitecture, or refactor. If this is all I did day in and day out, my entire skillset would atrophy.
  - nicksergeant 7mo ago
    "I mostly don't look at the crap coming out." This is pretty much my point. I use LLMs to code _and_ to learn. I read everything that comes out. Half of it is wrong or incomplete. The other half saved me a bunch of time and taught me things.
kingkawn 7mo ago
If a catastrophic failure occurs we will have to return to first principles and re-derive the solutions. Not so bad, probably enlivening even to get to spin up the mind again after a break.
- cdetrio 7mo ago
  We found 500 zero-days in ten year old widely used open-source projects. Was that not a demonstration of the catastrophic failure of human debugging capability?
  - kingkawn 7mo ago
    And yet the world keeps turning we’ll figure it out
hnthrow0287345 7mo ago
>But the number will dwindle and we'll ironically be unable to build what our ancestors did, utterly dependent on the AI artifacts to do it for us. That's only a brief moment in time. We learned it once, we can learn it again if we have to. People will tinker with those things as hobbies and they'll broadcast that out too. Worst case we hobble along until we get better at it. And if we have to hobble along and it's important, someone's going to be paying well for learning all of that stuff from zero, so the motivation will be there. Why do people worry about a potential, temporary loss of skill?
- Waterluvian 7mo ago
  I imagine it being a "does anybody know COBOL?!" but much sooner than sixty years rom now.
  - RhysU 7mo ago
    COBOL also came to mind. The COBOL thing seems to be working out just fine last I heard. Today a small number of people get paid well to know COBOL's depths and legacy platforms/software. The world moved on, where possible, to lower cost labor and tools. Arguably, that outcome was the right creative destruction. Market economics doesn't long-term incentivize any other outcomes. We'll see the arc of COBOL play out again with LLM coding.
    - jerf 7mo ago
      I've been waiting for the article talking about how AI is affecting COBOL. Preferably with quotes from actual COBOL programmers since I can already theorize as well as the next guy but I'm interested in the reports from the field. While LLMs have become pretty good at generating code, I think some of their other capabilities are still undersold and poorly understood, and one of them is that they are very good at porting. AI may offer the way out for porting COBOL finally. You definitely can't just blindly point it at one code base and tell it to convert to another. The LLMs do "blur" the code, I find, just sort of deciding that maybe this little clause wasn't important and dropping it. (Though in some cases I've encountered this, I sometimes understand where it is coming from, when the old code was twisty and full of indirection I often as a human have a hard time being sure what is and is not used just by reading the code too...) But the process is still way, way faster than the old days of typing the new code in one line at a time by staring at the old code. It's definitely way cheaper to port a code base into a new language in 2026 than it was in 2020. In 2020 it was so expensive it was almost always not even an option. I think a lot of people have not caught up with the cost reductions in such porting actions now, and are not correctly calculating that into their costs. It is easier than ever to get out of a language that has some fundamental issue that is hard to overcome (performance, general lack of capability like COBOL) and into something more modern that doesn't have that flaw.
tuvang 7mo ago
There is a deadly game of chicken going on. Junior recruiting already stopped for the most part. Only way this doesn’t end in a catastrophe is if AI becomes genuinely as good as the most skilled developers before we run out of them. Which I doubt very much but don’t find completely impossible.
- theshrike79 7mo ago
  And the irony is that AI usage should make onboarding juniors easier. Before it was "hey $senior_programmer where's the $thing defined in this project?", which either required a dedicated person onboarding or someone's flow was interrupted - an expected cost of bringing up juniors. Now a properly configured AI Agent can answer that question in 60 seconds, unblocking the Junior to work on something. And no, it doesn't mean Juniors or anyone else get to make 10k line PRs of code they haven't read nor understand. That's a very different issue that can be solved by slapping people over the head.
  - bragr 7mo ago
    The problem is that juniors given access to AI don't seem to learn as much. AI just gives them fish over and over instead of learning how to fish.
    - andrekandre 7mo ago
      > The problem is that juniors given access to AI don't seem to learn as much. i see this first-hand; they don't even know what they don't know so they circle over and over with ai leading them down rabbit holes and code that breaks in weird ways they cant even guess how to fix... stuff that if you were a real programmer you would have wrote in a few minutes let alone hours or days...
    - theshrike79 7mo ago
      Yea, giving people a blank Claude with no setup will get you that. What you could do is encourage (or force with IT's assistance) them to use a prompt (or hook or whatever) that refuses to do work for them, but instead telling them where to change and what without actually doing the work.
turlockmike 7mo ago
How many kernel devs does the world need? A dozen or two? It will be the same with software. AI will be writing and consuming most software. We will be utilizing experiences built on top of that, probably generated in real time for hyper personalization. Every app on your phone will be replaced by one app. (Except maybe games, at least for a short while longer). Everyone's treating writing code as this reverent thing. No one wrote code 100 years ago. Very few today write assembly. It will become lost because the economic neccesity is gone. It's the end of an era, but also the beginning of a new one. Building agentic systems is really hard, a hard enough problem that we need a ton of people building those systems. AI hardware devices have barely been registered, we need engineers who can build and integrate all sorts of systems. Engineering as a discipline will be the last job to be automated, since who do you think is going to build all the worlds automation?
- vdqtp3 7mo ago
  > How many kernel devs does the world need? A dozen or two? You're low by several orders of magnitude. "The 2025 development cycle saw 2,134 developers contribute to [Linux] kernel 6.18" [1] [1] https://commandlinux.com/statistics/linux-kernel-contributors-lines-of-code-statistics/ https://commandlinux.com/statistics/linux-kernel-contributor...
- rafterydj 7mo ago
  How wildly dismissive of the foundation of the X$ billion dollar software industry. You think humans just stumbled into writing code by accident or something? How does building agentic systems, a "really hard" problem, not just end up a "regular code" problem? Because that is what it is. A distributed systems problem with non-deterministic run lengths. How do you switch agent contexts? Similar to how you solve regular program context switching. How do you search tool capabilities and verify them? How do you effectively manage scheduled tasks? Oh, look, you've just invented the operating system kernel. Suddenly, those 'dozen or two' experts don't seem so archaic after all!
- oblio 7mo ago
  Does it even make sense to build everything on top of machines that are 70% reliable? The sheer orchestration and validation overhead at scale risks being more expensive than just keeping most software engineers and having them manage a few AI agents. Also, 200 years ago we didn't have bike mechanics. Car mechanics. Boat mechanics. Plumbers. Electricians. Not all new professions fade away.
qsera 7mo ago
Trust me. All those people do it for the love of doing it, so I don't think they will outsource the jobs to some automation.... I have been coding long before internet and before there were huge demand for software devs..and I would be coding even after there is no demand for the same.
anon291 7mo ago
I mean there should be. But there's not. Despite the millions of CS grads produced many people could not reasonably be expected to produce many 'standard' parts of a software stack
joegibbs 7mo ago
"What if the AI goes away" seems to be a common argument but it's just not ever going to happen without, say, a solar flare wiping out all electronics, which will be an issue which skilled programmers can't help anyway. The same thing happened when high-level languages came around, not many people can hand-write assembly anymore or work with punch cards but society hasn't collapsed.

by titzer 7mo ago

nicksergeant 7mo ago

I feel I've upskilled in so many directions (not just "ability to prompt LLMs") since going all in on LLM coding. So many tools, techniques, systems, and new areas of research I'd never have had the time to fully learn in the past. I have a hard time believing any tenured developer is not actually learning things when using LLMs to build. They make interesting choices that are repeatable (new CLIs I didn't even know existed, writing scripts to churn through tricky data, using specific languages for specific tasks like Go for concurrently working through large numerous tasks, etc.) Anyone not learning things via LLM coding right now either doesn't care at all about the underlying code/systems, or they had no foundational knowledge or interest in programming to begin with (which is also a valid way to use these tools, but they don't work very well without guidance for too long [yet]).

Waterluvian 7mo ago
I think there's a considerable difference in its ability to help with breadth vs. depth of expertise.
anovikov 7mo ago
This. I never had patience to figure how to build a from-scratch iOS app because it required too much boilerplate work. Now i do, and i got to enjoy Swift as a language, and learned a lot of iOS (and Mac) APIs.
- JustResign 7mo ago
  But it isn't "from scratch", is it? It's "from Claude".
  - nicksergeant 7mo ago
    If you build a house from scratch but you didn't mill the lumber, did you build it from scratch? If you make a pizza from scratch but you used canned sauce was it from scratch? What if you used store bought dough? What if you made the sauce and the dough but you didn't grow the tomato?
zozbot234 7mo ago
What do you mean by "LLM coding"? That's not a very meaningful term, it covers everything from 100% vibe coded projects, to using the LLM to gradually flesh out a careful initial design and then verifying that the implementation is done correctly at every step with meticulous human review and checking.
- nicksergeant 7mo ago
  The latter.
tripledry 7mo ago
For me both are true at the same time. I vividly remember understanding how calculus works after watching some 3blue1brown videos on youtube, but once I looked at some exercises I quickly realized I was not able to solve them. Similar thing happens with LLMs and programming. Sure I understand the code but I'm not intimately familiar with it like if I programmed it "old school". So yes, I do learn more but I can't shake the feeling that there is some dunning kruger effect going on. In essence I think that "banging my head against the wall" while learning is a key part of the learning process. Or maybe it's just me :D
- mwigdahl 7mo ago
  It's not just you. I feel the same thing, and I saw it in practice helping my son study for a chemistry test just last night. He had worked through a bunch of problems by following the steps in his notes and got the right answers, but couldn't solve them without the notes because his comprehension of why he was taking all the steps wasn't solid. Once we addressed that, he did great solo. Working the mechanics of the problems with the notes helped, but it was getting independent understanding of the reason for each step that put everything together for him.
agentultra 7mo ago
> Anyone not learning things via LLM coding right now either doesn't care at all about the underlying code/systems How many bytes is a pointer in C? How many bytes is a shared pointer in C++? What does sysctl do? What about fsync? What is a mutex lock? How is it different from a spin lock? You want to find the n nearest points to a given point on a 2-D Cartesian plane. Could you write the code to solve that on your own? Can you answer any of these questions without searching for the answer? I don't use LLMs and I learn things fine. Always have. For several decades. I care deeply about the underlying code and systems. It annoys me when people say they do and they cannot even understand how the computer works. I'm fine with people having domain-specific knowledge of programming: maybe you've only been interested in web development and scripting DOM elements. But don't pretend that your expertise in that area means you understand how to write an operating system. Or worse: that it prevents you from learning how to write an operating system. You can do that without an LLM. There's no royal road. You have to understand the theory, read the books, read the code, write the code, make mistakes, fix mistakes, read papers, talk to other people with more experience than you... and just write code. And rewrite it. And do it all again. I find the opposite is true: those who use LLM coding exclusively never enjoyed programming to begin with, only learned as much as they needed to, and want the end results.
- nicksergeant 7mo ago
  Agree with pretty much everything you wrote here, I guess with the addendum that LLMs can be a part of the learning experience you're describing. It's as easy as telling the LLM "don't write a single line of code nor command, I want to do everything, your goal is to help me understand what we're doing here." There are always going to be people who just want the end result. The only difference now is that LLM tools allow them to get much closer to the end result than they previously were able to. And on the other side, there are always going to be people who want to _understand_ what's happening, and LLMs can help accelerate that. I use LLMs as a personalized guide to learning new things.
  - agentultra 7mo ago
    Best of luck in your journey! To those reading this thread though, be wary of the answers LLMs generate: they're plausible sounding and the LLM's are designed to be sycophants. Be wary, double check their answers to your queries against credible sources. And read the source!
titzer 7mo ago
Learning calculus by watching the professor solve integrals on the board for an hour doesn't result in the same level and depth of understanding as working through homeworks every week for a semester. If you ran off to your TA to solve every problem in your homework, you just won't learn calculus. I've vibe coded plenty. I mostly don't look at the crap coming out. Don't want to. When I do I absorb a tiny bit, but not enough to recreate the thing from scratch. I might have a modicum more surface-level knowledge, but I don't have deep understanding and I don't have skills. To the extent that I've fixed or tweaked AI-generated code, it's not been to restructure, rearchitecture, or refactor. If this is all I did day in and day out, my entire skillset would atrophy.
- nicksergeant 7mo ago
  "I mostly don't look at the crap coming out." This is pretty much my point. I use LLMs to code _and_ to learn. I read everything that comes out. Half of it is wrong or incomplete. The other half saved me a bunch of time and taught me things.

kingkawn 7mo ago

If a catastrophic failure occurs we will have to return to first principles and re-derive the solutions. Not so bad, probably enlivening even to get to spin up the mind again after a break.

cdetrio 7mo ago
We found 500 zero-days in ten year old widely used open-source projects. Was that not a demonstration of the catastrophic failure of human debugging capability?
- kingkawn 7mo ago
  And yet the world keeps turning we’ll figure it out

hnthrow0287345 7mo ago

>But the number will dwindle and we'll ironically be unable to build what our ancestors did, utterly dependent on the AI artifacts to do it for us. That's only a brief moment in time. We learned it once, we can learn it again if we have to. People will tinker with those things as hobbies and they'll broadcast that out too. Worst case we hobble along until we get better at it. And if we have to hobble along and it's important, someone's going to be paying well for learning all of that stuff from zero, so the motivation will be there. Why do people worry about a potential, temporary loss of skill?

Waterluvian 7mo ago
I imagine it being a "does anybody know COBOL?!" but much sooner than sixty years rom now.
- RhysU 7mo ago
  COBOL also came to mind. The COBOL thing seems to be working out just fine last I heard. Today a small number of people get paid well to know COBOL's depths and legacy platforms/software. The world moved on, where possible, to lower cost labor and tools. Arguably, that outcome was the right creative destruction. Market economics doesn't long-term incentivize any other outcomes. We'll see the arc of COBOL play out again with LLM coding.
  - jerf 7mo ago
    I've been waiting for the article talking about how AI is affecting COBOL. Preferably with quotes from actual COBOL programmers since I can already theorize as well as the next guy but I'm interested in the reports from the field. While LLMs have become pretty good at generating code, I think some of their other capabilities are still undersold and poorly understood, and one of them is that they are very good at porting. AI may offer the way out for porting COBOL finally. You definitely can't just blindly point it at one code base and tell it to convert to another. The LLMs do "blur" the code, I find, just sort of deciding that maybe this little clause wasn't important and dropping it. (Though in some cases I've encountered this, I sometimes understand where it is coming from, when the old code was twisty and full of indirection I often as a human have a hard time being sure what is and is not used just by reading the code too...) But the process is still way, way faster than the old days of typing the new code in one line at a time by staring at the old code. It's definitely way cheaper to port a code base into a new language in 2026 than it was in 2020. In 2020 it was so expensive it was almost always not even an option. I think a lot of people have not caught up with the cost reductions in such porting actions now, and are not correctly calculating that into their costs. It is easier than ever to get out of a language that has some fundamental issue that is hard to overcome (performance, general lack of capability like COBOL) and into something more modern that doesn't have that flaw.

tuvang 7mo ago

There is a deadly game of chicken going on. Junior recruiting already stopped for the most part. Only way this doesn’t end in a catastrophe is if AI becomes genuinely as good as the most skilled developers before we run out of them. Which I doubt very much but don’t find completely impossible.

theshrike79 7mo ago
And the irony is that AI usage should make onboarding juniors easier. Before it was "hey $senior_programmer where's the $thing defined in this project?", which either required a dedicated person onboarding or someone's flow was interrupted - an expected cost of bringing up juniors. Now a properly configured AI Agent can answer that question in 60 seconds, unblocking the Junior to work on something. And no, it doesn't mean Juniors or anyone else get to make 10k line PRs of code they haven't read nor understand. That's a very different issue that can be solved by slapping people over the head.
- bragr 7mo ago
  The problem is that juniors given access to AI don't seem to learn as much. AI just gives them fish over and over instead of learning how to fish.
  - andrekandre 7mo ago
    > The problem is that juniors given access to AI don't seem to learn as much. i see this first-hand; they don't even know what they don't know so they circle over and over with ai leading them down rabbit holes and code that breaks in weird ways they cant even guess how to fix... stuff that if you were a real programmer you would have wrote in a few minutes let alone hours or days...
  - theshrike79 7mo ago
    Yea, giving people a blank Claude with no setup will get you that. What you could do is encourage (or force with IT's assistance) them to use a prompt (or hook or whatever) that refuses to do work for them, but instead telling them where to change and what without actually doing the work.

turlockmike 7mo ago

How many kernel devs does the world need? A dozen or two? It will be the same with software. AI will be writing and consuming most software. We will be utilizing experiences built on top of that, probably generated in real time for hyper personalization. Every app on your phone will be replaced by one app. (Except maybe games, at least for a short while longer). Everyone's treating writing code as this reverent thing. No one wrote code 100 years ago. Very few today write assembly. It will become lost because the economic neccesity is gone. It's the end of an era, but also the beginning of a new one. Building agentic systems is really hard, a hard enough problem that we need a ton of people building those systems. AI hardware devices have barely been registered, we need engineers who can build and integrate all sorts of systems. Engineering as a discipline will be the last job to be automated, since who do you think is going to build all the worlds automation?

vdqtp3 7mo ago
> How many kernel devs does the world need? A dozen or two? You're low by several orders of magnitude. "The 2025 development cycle saw 2,134 developers contribute to [Linux] kernel 6.18" [1] [1] https://commandlinux.com/statistics/linux-kernel-contributors-lines-of-code-statistics/ https://commandlinux.com/statistics/linux-kernel-contributor...
rafterydj 7mo ago
How wildly dismissive of the foundation of the X$ billion dollar software industry. You think humans just stumbled into writing code by accident or something? How does building agentic systems, a "really hard" problem, not just end up a "regular code" problem? Because that is what it is. A distributed systems problem with non-deterministic run lengths. How do you switch agent contexts? Similar to how you solve regular program context switching. How do you search tool capabilities and verify them? How do you effectively manage scheduled tasks? Oh, look, you've just invented the operating system kernel. Suddenly, those 'dozen or two' experts don't seem so archaic after all!
oblio 7mo ago
Does it even make sense to build everything on top of machines that are 70% reliable? The sheer orchestration and validation overhead at scale risks being more expensive than just keeping most software engineers and having them manage a few AI agents. Also, 200 years ago we didn't have bike mechanics. Car mechanics. Boat mechanics. Plumbers. Electricians. Not all new professions fade away.

qsera 7mo ago

Trust me. All those people do it for the love of doing it, so I don't think they will outsource the jobs to some automation.... I have been coding long before internet and before there were huge demand for software devs..and I would be coding even after there is no demand for the same.

anon291 7mo ago

I mean there should be. But there's not. Despite the millions of CS grads produced many people could not reasonably be expected to produce many 'standard' parts of a software stack

joegibbs 7mo ago

"What if the AI goes away" seems to be a common argument but it's just not ever going to happen without, say, a solar flare wiping out all electronics, which will be an issue which skilled programmers can't help anyway. The same thing happened when high-level languages came around, not many people can hand-write assembly anymore or work with punch cards but society hasn't collapsed.