I also think he is hinting at the core reason why this shift in the industry is basically inevitable. It's the economics of it. Let's assume that today's models can't deliver software of the same quality as a skilled developer. That I think is fairly uncontroversial. The economics of getting software at 80% the quality level for a fraction of the cost just makes this the sane path economically. If it bugs out 20% of the time, you can likely afford to pay someone to put a bandaid on it for a fraction of the cost. The question is if the software is good enough, not if the software is good. In many cases the answer will be that it is good enough even if it is a bugridden mess that is impossible to maintain.
As for myself... I agree that there is a reckoning happening in the industry and I have felt a bit of an identity crisis looking at what these models are capable of. Well, I'll stick around for a while longer. I enjoy the people I work with and have fairly large degrees of freedom, but I'm not sure if I will stick around till the end of my career.
Honestly we’ve all been doing this for many years already by hiring contractors to build entire platforms by themselves with very little oversight.
It doesn’t matter what country they’re in, the company makes the decision to hire like 100 contractors to build the thing and like 5 devs to oversee it and the whole project is a massive dumpster fire and somehow makes it to prod anyway and then the company is stuck with it for years until they decide to rewrite it and rinse and repeat.
Now it’s just 5 devs overseeing the AI in those companies that do things like that.
Not all companies do it that way, though. My company hired full time workers in other countries so that devs are always producing code 24 hrs a day and we are definitely dealing with slop, but we were before, and on some teams the AI has actually helped enforce code standards.
The massive issue we’ve having is that it often produces code that looks right, but isn’t. It’s so easy to pass on your own misunderstandings to the agent that sometimes we end up with features that, for example, don’t take into account established scope and permissions requirements. This seems like something that will be solved with time, though.
We already had years and years of overall software quality becoming a trash fire. I don't think anyone is excited about OS updates anymore, and practically every application has something that people feel breaks after an update. Software is getting larger and larger and consuming more resources to do the same things, as heavy frameworks are stacked up and companies rely on JavaScript crude for desktop and mobile applications. Things are slow on expensive (and getting more expensive) hardware when they don't need to be.
And that's just from the user-facing side. The industry is full of people who don't know what they're doing and software robustness is a constant fight between the people who care about quality and the people who just want to say LGTM and merge. And that attitude stacks up technical debt that makes work harder in the future or bites you with a production incident.
Now that LLM usage is everywhere, that fight is getting harder. People throw PRs down faster, apparently look at and test them less, and more people are just rubber stamping them without looking at the chances. This is essentially shifting a massive amount of work onto more senior developers who are baby sitting more.
Just the other day I caught someone submitting a config file change with a unit test testing the configuration, and multiple other people approved the PR before I blocked it. Not testing the configuration loader/parser, which already had through tests; Claude made a new test file to test that the configuration changes were set as it expected. Seriously, you don't write code to test runtime configuration (which already has full coverage for the functionality the config drives).
Software quality is cooked, and the experience for end users is going to get much worse if this madness doesn't die down. And probably kill people as this affects cars, planes, heavy machinery, infrastructure, etc..
I'd hope medical device companies learned from THERAC-25, but hey, it's the future now. Let's just let Claude worry about it. /s
I'd hope medical device companies learned from THERAC-25
Any one of the major data breaches or privacy violations should’ve been treated as the Software Engineering equivalent of the Ford Pinto or Hyatt Regency Walkway collapse, and the fact that they aren’t sickens me. (I got my major in university in Computer Science and Engineering, but did a minor in Philosophy and took multiple engineering ethics courses. Of course, most tech CEOs apparently did not).
The more fully agentic-driven apps there are, the more worried I am about this, especially for single-developer apps. Security & authentication are HARD problems, given how many places have screwed it up, even before LLMs generating code (I can’t find a link about the exact one I’m thinking of right now, but it was either personal information, SSNs, or something else being leaked because a website used incrementing IDs as account numbers in the URL to expose the personal information).
If someone is vibe coding an app, they don’t really have a good way of verifying the security posture of it, and sure, maybe they’ll throw an agent at it to check for security flaws…but if the security-focused ones cost too much, or if they don’t think it’s important…
I’d like to get out too… but what are the options? I mean yes I could do anything else but I’d be giving up 15 years of experience and the income that goes with it. I could plant potatoes in a field but it’s backbreaking labour that pays nothing. Where do we go from here?
I wonder if we will see a flip of wages where technical work becomes nearly worthless and manual labour becomes very expensive. As long as Western countries are using cheap imported labour or even illegal immigrants to do all the manual labour that seems unlikely. There is a common argument in Canada that 'Kids these days don't want to do manual labour and work the fields'. They are somewhat right, but I don't think it's the work that is the problem, it's the wage. I would have gone to work the fields in my youth, hell I might even do it now, but it's brutal work so I expect at least $100/hr and I want to be able to retire before 50.
No, if anything a hypothetical drop in demand for skilled knowledge work would increase the labor pool of people for manual labor and drive the pay down even more.
Technical work pays more because a limited number of people can do it and it takes a large upfront investment in time to be able to do it professionally. Unskilled manual labor is something that anyone who's able-bodied can do.
Things like agriculture don't have more of an automation push because people are already paid next to nothing and developing the automation would cost more. Robots largely took over heavy manufacturing decades ago—it's largely large machinery building cars, for example. (Except for cheap commodities where it's cheaper to pay people in a heavily-outsourced area, like sewing shirts.) People do the dexterous in-between jobs, but that's the next big push. There's a reason why Hyundai/Boston Dynamics is working on hands for Atlas right now.
Preferably to something that isn't next in line to be gobbled up by agents+harnesses. Hard also to find a field that has the same benefits (salary levels + other benefits). Something that allows for focus at the same level is kind of rare I think. Let me know if you find it!
Personally I've been thinking about being an indie game developer. It tickles the main reason I got into software in the first place, which is because I found it fun basically, and it tickles the creative itch.
That is kind of a moonshot in terms of being a long term career replacement though. So yes, let me know if you or @archevel find something out.
I’m gonna stick with it until they lay me off and then I’ll go for a masters in whatever looks good at the time. I’ve got leadership skills and so far they’re just making me do both development and project management so thats kept my income safe for now.
Ctrl-F didn't turn up any results for "union" and only an off-target hit for "organize". RIP to an entire industry which never learned that, in order to negotiate effectively, you need to come to the bargaining table with leverage (and that leverage evaporates when you become trivially replaceable). Oh well.
Hopefully this is a lesson to every other industry that AI is being forced into. The best time to set up sustainable industries which exclude employer-driven exploitation is yesterday, the next best time is now.
It's relatively easy to start a new software company compared to other industries and I think software developers with ambition are more likely to do that. Why form a union to fight with management when you can replace management with your own?
And even people who don't want to be management might join a new company with their friends, or just switch companies. The grass is always greener.
It's relatively easy to start a new software company [...]
Relatively easy != easy.
Why form a union to fight with management when you can replace management with your own?
Existing customer base, proven business model, coworkers as a known factor.
And even people who don't want to be management might join a new company with their friends, or just switch companies. The grass is always greener.
Switching companies doesn't change that employers negotiate as a bloc, and employees negotiate as individuals. The mentality of leaving everyone to fend for themselves has gotten us here. For people who're fine with this state of affairs, cool, more power to ya. This blog post examines people who aren't, so my comment is targeted at them, and others who could relate to that situation.
As long as you can find more overworked software developers, and push them hard until they burn out trying to maintain a minimum standard of quality, you can keep the “AI”-driven labour alienation going.
This hits nail on the head. Not sure did I use idiom correctly, but that's my main issue with all these AI staff. More responsibility, more velocity, more automation, and you still responsible for the resulting mess.
I honestly stopped to point errors in AI generated documentation made by other people, why bother. Maybe I'm just burn out.
I went through a bit of a doomer phase like last year or the year before but then my company started paying for me to use AI to develop and I actually kinda like it. I get things done in hours that used to take me days, and our teams cadence hasn’t caught up so right now that leaves me time for completing things on my own years-long backlog like refactoring/cleanup/new features to make releasing easier.
In the background Ive always been kinda doomer, I left a job in healthcare where I was actually helping people to be at this job where no one needs me to do the work that I do but it pays like 4x was I was making in healthcare so I’m just taking their money while I can cause I’m sure that this isn’t sustainable and eventually our society will need everyone to do useful work.
I think that attitude has kinda helped me a little, like, I was always working a useless job, AI has just made it so less of my brain power is spent on said useless job.
Software development wasn’t ever really my life’s calling, I was always just decent at it, and even more decent at leading projects, so AI making it such that I don’t have to write code much any more doesn’t really bother me. Its actually fine, cause people interrupting me with meetings all day made writing code really difficult, and I cant keep people from doing that, so its nice I don’t have to write code any more. This is much less stressful than having to write code and not having the time to do so.
I think the vast majority of us will have to try to shift towards more design/architecture-focused roles - i.e. not just being a programmer, but also a product lead, taking charge and driving decisions around the products that we develop, if we want to have a decent shot at remaining in the industry long-term. Lots of us (and I'm not saying I'm safe) will probably not make the cut, which is unfortunate.
If models keep getting better, companies will eventually flat out refuse to let us keep writing code by hand as models could output more of it faster, and our jobs will be prompting, reviewing, and dealing with the corporate tedium that models can't handle.
As much as I enjoy manual coding sessions, I can't deny that I can ship significantly more with LLMs than without them, so I've been picking and choosing my battles and dedicating the manual coding time to things I strongly care about.
At the same time, I'm considering trying to develop some products outside of my job and monetizing them as a side-gig; AI makes the time investment required a lot smaller, so I can justify spending a bit of my free time every day on them, and I'm confident enough in my QA skills to ensure a reasonable amount of quality. It's still a gamble, though, and I do need to find some good unexplored niches, which gets harder and harder as time goes on. But I'm definitely not banking on my job lasting forever so I need to find a solution that works for me.
This is what software development will be like, now and forever.
Forever is a long time and I wouldn't care to predict where we are in two years. I don't think things will go back the way they were, but given that we're living in a time of change and uncertainty, I think a better bet would be that software development will keep changing for a while as people figure out better ways to work.
That might even include new companies that don't work the same way as the old ones.
The fewer people you have working for you, the fewer stock options you have to pay out, the less resistance there will be to watering down the product, collaborating with the military-industrial complex, or helping authoritarian governments control and murder their citizens. Over the past two decades, one of the more effective forces for keeping tech companies in line have been the tech workers themselves who have tried resist corrupt and unethical behaviour. It hasn’t always worked. There wasn’t enough of it. But what little there was it added meaningful friction at several important moments in tech history.
Have they though?
Fifteen years ago, the industry was desperate to hire coders. Landing a software engineering gig in a Big Tech firm was the dream. Their salaries were huge, their offices seemed luxurious and they were based in some of the most prestigious places in the world. And out of all of them, only Amazon really had a bad reputation for how they treated their workers pre-COVID (and it was really in their warehouse and logistics divisions.)
Since then it's been a race to shittify the entire industry.
archevel | 7 hours ago
Jupp. There is definitely skill atrophy going on.
I also think he is hinting at the core reason why this shift in the industry is basically inevitable. It's the economics of it. Let's assume that today's models can't deliver software of the same quality as a skilled developer. That I think is fairly uncontroversial. The economics of getting software at 80% the quality level for a fraction of the cost just makes this the sane path economically. If it bugs out 20% of the time, you can likely afford to pay someone to put a bandaid on it for a fraction of the cost. The question is if the software is good enough, not if the software is good. In many cases the answer will be that it is good enough even if it is a bugridden mess that is impossible to maintain.
As for myself... I agree that there is a reckoning happening in the industry and I have felt a bit of an identity crisis looking at what these models are capable of. Well, I'll stick around for a while longer. I enjoy the people I work with and have fairly large degrees of freedom, but I'm not sure if I will stick around till the end of my career.
snake_case | 5 hours ago
Honestly we’ve all been doing this for many years already by hiring contractors to build entire platforms by themselves with very little oversight.
It doesn’t matter what country they’re in, the company makes the decision to hire like 100 contractors to build the thing and like 5 devs to oversee it and the whole project is a massive dumpster fire and somehow makes it to prod anyway and then the company is stuck with it for years until they decide to rewrite it and rinse and repeat.
Now it’s just 5 devs overseeing the AI in those companies that do things like that.
Not all companies do it that way, though. My company hired full time workers in other countries so that devs are always producing code 24 hrs a day and we are definitely dealing with slop, but we were before, and on some teams the AI has actually helped enforce code standards.
The massive issue we’ve having is that it often produces code that looks right, but isn’t. It’s so easy to pass on your own misunderstandings to the agent that sometimes we end up with features that, for example, don’t take into account established scope and permissions requirements. This seems like something that will be solved with time, though.
redwall_hp | 4 hours ago
We already had years and years of overall software quality becoming a trash fire. I don't think anyone is excited about OS updates anymore, and practically every application has something that people feel breaks after an update. Software is getting larger and larger and consuming more resources to do the same things, as heavy frameworks are stacked up and companies rely on JavaScript crude for desktop and mobile applications. Things are slow on expensive (and getting more expensive) hardware when they don't need to be.
And that's just from the user-facing side. The industry is full of people who don't know what they're doing and software robustness is a constant fight between the people who care about quality and the people who just want to say LGTM and merge. And that attitude stacks up technical debt that makes work harder in the future or bites you with a production incident.
Now that LLM usage is everywhere, that fight is getting harder. People throw PRs down faster, apparently look at and test them less, and more people are just rubber stamping them without looking at the chances. This is essentially shifting a massive amount of work onto more senior developers who are baby sitting more.
Just the other day I caught someone submitting a config file change with a unit test testing the configuration, and multiple other people approved the PR before I blocked it. Not testing the configuration loader/parser, which already had through tests; Claude made a new test file to test that the configuration changes were set as it expected. Seriously, you don't write code to test runtime configuration (which already has full coverage for the functionality the config drives).
Software quality is cooked, and the experience for end users is going to get much worse if this madness doesn't die down. And probably kill people as this affects cars, planes, heavy machinery, infrastructure, etc..
I'd hope medical device companies learned from THERAC-25, but hey, it's the future now. Let's just let Claude worry about it. /s
Zorind | 3 hours ago
Any one of the major data breaches or privacy violations should’ve been treated as the Software Engineering equivalent of the Ford Pinto or Hyatt Regency Walkway collapse, and the fact that they aren’t sickens me. (I got my major in university in Computer Science and Engineering, but did a minor in Philosophy and took multiple engineering ethics courses. Of course, most tech CEOs apparently did not).
The more fully agentic-driven apps there are, the more worried I am about this, especially for single-developer apps. Security & authentication are HARD problems, given how many places have screwed it up, even before LLMs generating code (I can’t find a link about the exact one I’m thinking of right now, but it was either personal information, SSNs, or something else being leaked because a website used incrementing IDs as account numbers in the URL to expose the personal information).
If someone is vibe coding an app, they don’t really have a good way of verifying the security posture of it, and sure, maybe they’ll throw an agent at it to check for security flaws…but if the security-focused ones cost too much, or if they don’t think it’s important…
preposterous | 7 hours ago
I’d like to get out too… but what are the options? I mean yes I could do anything else but I’d be giving up 15 years of experience and the income that goes with it. I could plant potatoes in a field but it’s backbreaking labour that pays nothing. Where do we go from here?
plutonic | 6 hours ago
I wonder if we will see a flip of wages where technical work becomes nearly worthless and manual labour becomes very expensive. As long as Western countries are using cheap imported labour or even illegal immigrants to do all the manual labour that seems unlikely. There is a common argument in Canada that 'Kids these days don't want to do manual labour and work the fields'. They are somewhat right, but I don't think it's the work that is the problem, it's the wage. I would have gone to work the fields in my youth, hell I might even do it now, but it's brutal work so I expect at least $100/hr and I want to be able to retire before 50.
redwall_hp | 4 hours ago
No, if anything a hypothetical drop in demand for skilled knowledge work would increase the labor pool of people for manual labor and drive the pay down even more.
Technical work pays more because a limited number of people can do it and it takes a large upfront investment in time to be able to do it professionally. Unskilled manual labor is something that anyone who's able-bodied can do.
Things like agriculture don't have more of an automation push because people are already paid next to nothing and developing the automation would cost more. Robots largely took over heavy manufacturing decades ago—it's largely large machinery building cars, for example. (Except for cheap commodities where it's cheaper to pay people in a heavily-outsourced area, like sewing shirts.) People do the dexterous in-between jobs, but that's the next big push. There's a reason why Hyundai/Boston Dynamics is working on hands for Atlas right now.
archevel | 7 hours ago
Preferably to something that isn't next in line to be gobbled up by agents+harnesses. Hard also to find a field that has the same benefits (salary levels + other benefits). Something that allows for focus at the same level is kind of rare I think. Let me know if you find it!
chroma | 6 hours ago
Personally I've been thinking about being an indie game developer. It tickles the main reason I got into software in the first place, which is because I found it fun basically, and it tickles the creative itch.
That is kind of a moonshot in terms of being a long term career replacement though. So yes, let me know if you or @archevel find something out.
snake_case | 5 hours ago
I’m gonna stick with it until they lay me off and then I’ll go for a masters in whatever looks good at the time. I’ve got leadership skills and so far they’re just making me do both development and project management so thats kept my income safe for now.
kacey | 2 hours ago
Ctrl-F didn't turn up any results for "union" and only an off-target hit for "organize". RIP to an entire industry which never learned that, in order to negotiate effectively, you need to come to the bargaining table with leverage (and that leverage evaporates when you become trivially replaceable). Oh well.
Hopefully this is a lesson to every other industry that AI is being forced into. The best time to set up sustainable industries which exclude employer-driven exploitation is yesterday, the next best time is now.
skybrian | an hour ago
It's relatively easy to start a new software company compared to other industries and I think software developers with ambition are more likely to do that. Why form a union to fight with management when you can replace management with your own?
And even people who don't want to be management might join a new company with their friends, or just switch companies. The grass is always greener.
kacey | an hour ago
Relatively easy != easy.
Existing customer base, proven business model, coworkers as a known factor.
Switching companies doesn't change that employers negotiate as a bloc, and employees negotiate as individuals. The mentality of leaving everyone to fend for themselves has gotten us here. For people who're fine with this state of affairs, cool, more power to ya. This blog post examines people who aren't, so my comment is targeted at them, and others who could relate to that situation.
Deely | 4 hours ago
Ufff.
This hits nail on the head. Not sure did I use idiom correctly, but that's my main issue with all these AI staff. More responsibility, more velocity, more automation, and you still responsible for the resulting mess.
I honestly stopped to point errors in AI generated documentation made by other people, why bother. Maybe I'm just burn out.
snake_case | 4 hours ago
I went through a bit of a doomer phase like last year or the year before but then my company started paying for me to use AI to develop and I actually kinda like it. I get things done in hours that used to take me days, and our teams cadence hasn’t caught up so right now that leaves me time for completing things on my own years-long backlog like refactoring/cleanup/new features to make releasing easier.
In the background Ive always been kinda doomer, I left a job in healthcare where I was actually helping people to be at this job where no one needs me to do the work that I do but it pays like 4x was I was making in healthcare so I’m just taking their money while I can cause I’m sure that this isn’t sustainable and eventually our society will need everyone to do useful work.
I think that attitude has kinda helped me a little, like, I was always working a useless job, AI has just made it so less of my brain power is spent on said useless job.
Software development wasn’t ever really my life’s calling, I was always just decent at it, and even more decent at leading projects, so AI making it such that I don’t have to write code much any more doesn’t really bother me. Its actually fine, cause people interrupting me with meetings all day made writing code really difficult, and I cant keep people from doing that, so its nice I don’t have to write code any more. This is much less stressful than having to write code and not having the time to do so.
arqalite | 3 hours ago
I think the vast majority of us will have to try to shift towards more design/architecture-focused roles - i.e. not just being a programmer, but also a product lead, taking charge and driving decisions around the products that we develop, if we want to have a decent shot at remaining in the industry long-term. Lots of us (and I'm not saying I'm safe) will probably not make the cut, which is unfortunate.
If models keep getting better, companies will eventually flat out refuse to let us keep writing code by hand as models could output more of it faster, and our jobs will be prompting, reviewing, and dealing with the corporate tedium that models can't handle.
As much as I enjoy manual coding sessions, I can't deny that I can ship significantly more with LLMs than without them, so I've been picking and choosing my battles and dedicating the manual coding time to things I strongly care about.
At the same time, I'm considering trying to develop some products outside of my job and monetizing them as a side-gig; AI makes the time investment required a lot smaller, so I can justify spending a bit of my free time every day on them, and I'm confident enough in my QA skills to ensure a reasonable amount of quality. It's still a gamble, though, and I do need to find some good unexplored niches, which gets harder and harder as time goes on. But I'm definitely not banking on my job lasting forever so I need to find a solution that works for me.
skybrian | 3 hours ago
This bit sticks out:
Forever is a long time and I wouldn't care to predict where we are in two years. I don't think things will go back the way they were, but given that we're living in a time of change and uncertainty, I think a better bet would be that software development will keep changing for a while as people figure out better ways to work.
That might even include new companies that don't work the same way as the old ones.
Bullmaestro | 7 minutes ago
Have they though?
Fifteen years ago, the industry was desperate to hire coders. Landing a software engineering gig in a Big Tech firm was the dream. Their salaries were huge, their offices seemed luxurious and they were based in some of the most prestigious places in the world. And out of all of them, only Amazon really had a bad reputation for how they treated their workers pre-COVID (and it was really in their warehouse and logistics divisions.)
Since then it's been a race to shittify the entire industry.