9 ms·

One thing that really bothered me that the author glossed over (perhaps they don't care, given the tone of the article) is where they said: > Does an intern co

One thing that really bothered me that the author glossed over (perhaps they don't care, given the tone of the article) is where they said: > Does an intern cost $20/month? Because that’s what Cursor.ai costs. > Part of being a senior developer is making less-able coders productive, be they fleshly or algebraic. But do you know what another part of being a senior developer is? Not just making them more productive, but also guiding the junior developers into becoming better, independent, self-tasking, senior coders. And that feedback loop doesn't exist here. We're robbing ourselves of good future developers, because we aren't even thinking about the fact that the junior devs are actively learning from the small tasks we give them. Will AI completely replace devs before we all retire? Maybe. Maybe not. But long before that, the future coders who aren't being hired and trained because a senior dev doesn't understand that the junior devs become senior devs (and that's an important pipeline) and would rather pay $20/month for an LLM, are going to become a major loss/ brain drain domestically.

paffdragon 1y ago
I imagine it like this. Juniors will be taught by LLMs on some things, but seniors will still be there, they will still assist, pair program, code review, etc. but they will have another party, the LLM, like a smarter calculator.
esskay 1y ago
It's a bit misleading to compare $20/month with an actual human person. The junior dev wont get half way through the day and tell you they've used up all their coding time for the month and will now respond with jibberish. Cursor is a heck of a lot more than $20/month if you actually want it working for a full work day, every day.
- falcor84 1y ago
  > The junior dev wont get half way through the day and tell you they've used up all their coding time for the month and will now respond with jibberish. This issue manifests a bit differently in people, but I've definitely worked with people (not only juniors) who only have a few productive hours a month in them. And for what it's worth, some of those people were sufficiently productive in those few hours that it was rational for the company to keep them.
- 4hg4ufxhy 1y ago
  My coworkers are burning 10k/month on cursor.
  - kfajdsl 1y ago
    how?? I consider myself pretty heavy handed with letting Gemini fill up its 1M token context window for individual tasks, and I don't exceed $20 per day. Do they just let the agent spin in a loop for 4 hours?
- patrickmay 1y ago
  I worked on a team where a new hire from a prestigious school told his manager "That work is boring and mundane and I'm not going to do it." He didn't last long.
  - esskay 1y ago
    Yeah, so you fire them and replace them with another human, thats still vastly cheaper than a person plus a per-token ai fee.
  - norir 1y ago
    And it wouldn't surprise me if he is now the boss of people like his prior boss.
- Closi 1y ago
  Maybe - but it will still probably be less than a junior dev. You could probably hammer the most expensive cursor API all-day every-day and it would still be a fraction of the cost of a junior dev.
ignoramous 1y ago
How many senior developers understand the minute, intimate details of the frameworks, libraries, languages they use? How many understand the databases they use? TFA says, many (but not all) don't have to care as long as the product ships. That's exactly how code written by LLMs is meant to be tested and evaluated. And if you set up a good enough build/test environment, TFA argues that you can automate most of the schelp away.
- deltaburnt 1y ago
  In my experience, senior engineers without the curiosity to occasionally dig deeper into their frameworks are significantly worse engineers. No framework, library, or language is perfect. A big part of your job is understanding how your code will execute. Sometimes that even requires, at a very high level, imagining how that code will compile down to assembly once you strip away all the abstractions. Eventually you will get a memory leak even in a GCd language. Eventually there will be some incredibly obscure, unreported bug in a library. Eventually you will find an issue in unmaintained code you depend on. Eventually there will be performance problems caused by too many layers of abstraction. You either need to know, roughly, how your dependencies work by occasionally digging into their code/reading the documentation. Or you need intuition to know how it probably works, but you usually build that intuition by actually writing/reading code.
  - kfajdsl 1y ago
    The people who want to look under the hood even if they have no immediate reason to do so will always exist, and the people who don't care and just learn the bare minimum to produce app features that usually work will always exist. All LLMs do is decrease the skill requirement to be developer B, but they also make it easier to learn what you need to be developer A.
    - deltaburnt 1y ago
      I think we will end up with significantly more developer B and significantly less developer A. I can't see the future, I don't know how that would play out long term, maybe it will all even out. But I think in the medium term it will be a lot of headaches for someone like me who spends a significant amount of time reviewing code and steering architecture. Also, I spend a lot of time mentoring, and I'd like to think A will grow to be more like B over time. But now it feels like it's wasted effort to try mentoring those skills if it won't be valued.
raddan 1y ago
I think what is going to happen is that junior devs will develop a strong reliance on AI tools to be able to do anything. I cynically think this was OpenAI’s aim when they made ChatGPT free for students. I had a rather depressing experience this semester in my office hours with two students who had painted themselves in a corner with code that was clearly generated. They came to me for help, but were incapable of explaining why they had written what was on their screens. I decided to find where they had lost the thread of the class and discovered that they were essentially unable to write a helloworld program. In other words, they lost the thread on day one. Up until this point, both students had nearly perfect homework grades while failing every in-class quiz. From one perspective I understand the business case for pushing these technologies. But from another perspective, the long term health of the profession, it’s pretty shortsighted. Who knows, in the end maybe this will kill off the group of students who enroll in CS courses “because mom and dad think it’s a good job,” and maybe that will leave me with the group that really wants to be there. In the meantime, I will remind students that there is a difference between programming and computer science and that you really need a strong grasp of the latter to be an effective coder. Especially if you use AI tools.
- hyperbovine 1y ago
  > Who knows, in the end maybe this will kill off the group of students who enroll in CS courses “because mom and dad think it’s a good job,” I see this so much. “Data science major” became the 2020s version of law school. It’s such a double edged sword. It’s led to a huge increase in enrollment and the creation of multiple professional masters programs, so the college loves us. We hire every year and there’s always money for just about anything. On the other hand, class sizes are huge, which is not fun, and worse a large fraction of the students appear to have minimal intrinsic interest in coding or analyzing data. They’re there because it’s where the jobs are. I totally get that, in some sense college has always been that way, but it does make me look back fondly on the days when classes were 1/4 as big and filled with people who were genuinely interested in the subject. Unfortunately I think I may get my wish. AI is going to eliminate a lot of those jobs and so the future of our field looks a bit bleak. Worse, it’s the very students who are going to become redundant the quickest that are the least willing to learn. I’d be happy to teach them basic analysis and coding skills, but they are dead set on punching everything into ChatGPT.
xmodem 1y ago
Further, Cursor might cost $20/month today, but to what degree is that subsidized by VC investment? All the information we have points to frontier models just not being profitable to run at those types of prices, and those investors are going to want a return at some point.
- modzu 1y ago
  the market will indeed balance this out. remember when a taxi was $20 and an uber $5? now an uber is $25. nobody is going to go back to humans with all their wet meat sack problems, we will get more value for it, but it aint gona stay $5 if those putting up all this capital have anything to do with it. then again, we might get cheap, self hostable local copies (unless theyre made illegal for "safety" or some bullshit)
  - captainbland 1y ago
    I think the most likely thing is the cheap self hostable copies will broadly stop improving significantly. It'll be too costly for a community project to distill a bleeding edge cloud model and companies will stop releasing them. What's free now will remain free, we might even get another gen or 2 of improvements (possibly with diminishing returns) on free/cheap local models but those days are numbered.
    - awkward 1y ago
      What's the force that you see keeping those days numbered? Many ollama models are runnable on a local laptop, and models like deepseek even more so. Aside from open source interest, there are many large state and corporate actors who would like to see models commoditized.
      - captainbland 1y ago
        The only real wild card to me is state sponsorship. Everywhere else I expect such a capital intensive tech industry to approach winner (or maybe duopoly) takes all over time. My view is that right now, because of the willingness of corporations and other investors to swallow short term (but massive) losses on this, we're basically in AI fiscal fantasy land. The question we should be asking is how do we get access to these local models in the first place? It's all based on the work of these hyper expensive base models as the best small models are quantisations and distills. None of this works as soon as the profit motive comes into play and companies start gatekeeping effectively, which they will.
samyar 1y ago
Well i hope that happens, you go apply for a job already 10k applied. What kind of job market is this.
- ang_cire 1y ago
  > What kind of job market is this. One which is reeling from layoffs caused by short-term profit-maximizing bandwagoning in the tech sector. CEO see, CEO do. We didn't just magically have tens of thousands of new programmers/ IT pros. pop out of nowhere, they were laid off. There is no situation in which this brain-drain will benefit them. They can't wait around for 10 years for it to get really bad, in the hopes that at that point they'll be back in high demand.
empath75 1y ago
> But do you know what another part of being a senior developer is? Not just making them more productive, but also guiding the junior developers into becoming better, independent, self-tasking, senior coders. And that feedback loop doesn't exist here. Almost every senior developer I know is spending that time making LLM's more productive and useful instead. Whatever you think the job is of the senior developer, it will not be "coding". I think people need to stop thinking of themselves as computer programmers and start thinking of themselves as _engineers_. Your job isn't writing programs, your job is _using the technology you have available to solve problems_. Maybe that is through writing code, but maybe it's orchestrating LLM's to write code for you. The important part is solving the problem.
- ang_cire 1y ago
  > Almost every senior developer I know is spending that time making LLM's more productive and useful instead. LLMs may become more productive/ accurate/ useful, but they're not self-tasking or independent. > I think people need to stop thinking of themselves as computer programmers and start thinking of themselves as _engineers_. Your job isn't writing programs, your job is _using the technology you have available to solve problems_. There is a progression of skill required to master any profession, starting with fundamentals, and progressing and developing until you are an expert/ senior at that profession. How is a senior sw dev supposed to become that without writing code? Just reading LLM code and bugfixing isn't the same level or kind of experience. You're going to have devs who can't code by themselves, and that's a bad place to be in. There are already too many people in IT using tools that they don't understand the workings of (and thus can't troubleshoot, can't replace, can't customize to their env, etc), and this will just exacerbate that x100. MMW there is going to be a very bad skill deficit in IT in 20 years, which is going to cause an innovation deficit.
ReptileMan 1y ago
>We're robbing ourselves of good future developers You call it robbing ourselves of good future developers, I call it hourly consultancy rate increase.
elric 1y ago
This is something that's been simmering in the back of my mind for a while. Using an AI Agent instead of talking to your human colleagues deprives both of you from learning opportunities. There are probably short term gains in many cases, but I fear there will be long term losses over time.
- causal 1y ago
  I agree, and think that organizations that figure out how to use AI well in a collaborative way will succeed in the long-term. Developer community is still where the important growth happens.
- throw8393494 1y ago
  Is it possible to talk to coworkers? What if you voted for wrong party? Are closeted gay/trans/qeer? Radical femimist?! Or dog atrack survivor, and they really _REALLY_ like dogs! Talking to colleagues at work is a chore, and huge risk! Not opportunity! At least AI respects my privacy, and will not get my fired!
  - throwaway29812 1y ago
    [dead]

paffdragon 1y ago

I imagine it like this. Juniors will be taught by LLMs on some things, but seniors will still be there, they will still assist, pair program, code review, etc. but they will have another party, the LLM, like a smarter calculator.

esskay 1y ago

It's a bit misleading to compare $20/month with an actual human person. The junior dev wont get half way through the day and tell you they've used up all their coding time for the month and will now respond with jibberish. Cursor is a heck of a lot more than $20/month if you actually want it working for a full work day, every day.

falcor84 1y ago
> The junior dev wont get half way through the day and tell you they've used up all their coding time for the month and will now respond with jibberish. This issue manifests a bit differently in people, but I've definitely worked with people (not only juniors) who only have a few productive hours a month in them. And for what it's worth, some of those people were sufficiently productive in those few hours that it was rational for the company to keep them.
4hg4ufxhy 1y ago
My coworkers are burning 10k/month on cursor.
- kfajdsl 1y ago
  how?? I consider myself pretty heavy handed with letting Gemini fill up its 1M token context window for individual tasks, and I don't exceed $20 per day. Do they just let the agent spin in a loop for 4 hours?
patrickmay 1y ago
I worked on a team where a new hire from a prestigious school told his manager "That work is boring and mundane and I'm not going to do it." He didn't last long.
- esskay 1y ago
  Yeah, so you fire them and replace them with another human, thats still vastly cheaper than a person plus a per-token ai fee.
- norir 1y ago
  And it wouldn't surprise me if he is now the boss of people like his prior boss.
Closi 1y ago
Maybe - but it will still probably be less than a junior dev. You could probably hammer the most expensive cursor API all-day every-day and it would still be a fraction of the cost of a junior dev.

ignoramous 1y ago

How many senior developers understand the minute, intimate details of the frameworks, libraries, languages they use? How many understand the databases they use? TFA says, many (but not all) don't have to care as long as the product ships. That's exactly how code written by LLMs is meant to be tested and evaluated. And if you set up a good enough build/test environment, TFA argues that you can automate most of the schelp away.

deltaburnt 1y ago
In my experience, senior engineers without the curiosity to occasionally dig deeper into their frameworks are significantly worse engineers. No framework, library, or language is perfect. A big part of your job is understanding how your code will execute. Sometimes that even requires, at a very high level, imagining how that code will compile down to assembly once you strip away all the abstractions. Eventually you will get a memory leak even in a GCd language. Eventually there will be some incredibly obscure, unreported bug in a library. Eventually you will find an issue in unmaintained code you depend on. Eventually there will be performance problems caused by too many layers of abstraction. You either need to know, roughly, how your dependencies work by occasionally digging into their code/reading the documentation. Or you need intuition to know how it probably works, but you usually build that intuition by actually writing/reading code.
- kfajdsl 1y ago
  The people who want to look under the hood even if they have no immediate reason to do so will always exist, and the people who don't care and just learn the bare minimum to produce app features that usually work will always exist. All LLMs do is decrease the skill requirement to be developer B, but they also make it easier to learn what you need to be developer A.
  - deltaburnt 1y ago
    I think we will end up with significantly more developer B and significantly less developer A. I can't see the future, I don't know how that would play out long term, maybe it will all even out. But I think in the medium term it will be a lot of headaches for someone like me who spends a significant amount of time reviewing code and steering architecture. Also, I spend a lot of time mentoring, and I'd like to think A will grow to be more like B over time. But now it feels like it's wasted effort to try mentoring those skills if it won't be valued.

raddan 1y ago

I think what is going to happen is that junior devs will develop a strong reliance on AI tools to be able to do anything. I cynically think this was OpenAI’s aim when they made ChatGPT free for students. I had a rather depressing experience this semester in my office hours with two students who had painted themselves in a corner with code that was clearly generated. They came to me for help, but were incapable of explaining why they had written what was on their screens. I decided to find where they had lost the thread of the class and discovered that they were essentially unable to write a helloworld program. In other words, they lost the thread on day one. Up until this point, both students had nearly perfect homework grades while failing every in-class quiz. From one perspective I understand the business case for pushing these technologies. But from another perspective, the long term health of the profession, it’s pretty shortsighted. Who knows, in the end maybe this will kill off the group of students who enroll in CS courses “because mom and dad think it’s a good job,” and maybe that will leave me with the group that really wants to be there. In the meantime, I will remind students that there is a difference between programming and computer science and that you really need a strong grasp of the latter to be an effective coder. Especially if you use AI tools.

hyperbovine 1y ago
> Who knows, in the end maybe this will kill off the group of students who enroll in CS courses “because mom and dad think it’s a good job,” I see this so much. “Data science major” became the 2020s version of law school. It’s such a double edged sword. It’s led to a huge increase in enrollment and the creation of multiple professional masters programs, so the college loves us. We hire every year and there’s always money for just about anything. On the other hand, class sizes are huge, which is not fun, and worse a large fraction of the students appear to have minimal intrinsic interest in coding or analyzing data. They’re there because it’s where the jobs are. I totally get that, in some sense college has always been that way, but it does make me look back fondly on the days when classes were 1/4 as big and filled with people who were genuinely interested in the subject. Unfortunately I think I may get my wish. AI is going to eliminate a lot of those jobs and so the future of our field looks a bit bleak. Worse, it’s the very students who are going to become redundant the quickest that are the least willing to learn. I’d be happy to teach them basic analysis and coding skills, but they are dead set on punching everything into ChatGPT.

xmodem 1y ago

Further, Cursor might cost $20/month today, but to what degree is that subsidized by VC investment? All the information we have points to frontier models just not being profitable to run at those types of prices, and those investors are going to want a return at some point.

modzu 1y ago
the market will indeed balance this out. remember when a taxi was $20 and an uber $5? now an uber is $25. nobody is going to go back to humans with all their wet meat sack problems, we will get more value for it, but it aint gona stay $5 if those putting up all this capital have anything to do with it. then again, we might get cheap, self hostable local copies (unless theyre made illegal for "safety" or some bullshit)
- captainbland 1y ago
  I think the most likely thing is the cheap self hostable copies will broadly stop improving significantly. It'll be too costly for a community project to distill a bleeding edge cloud model and companies will stop releasing them. What's free now will remain free, we might even get another gen or 2 of improvements (possibly with diminishing returns) on free/cheap local models but those days are numbered.
  - awkward 1y ago
    What's the force that you see keeping those days numbered? Many ollama models are runnable on a local laptop, and models like deepseek even more so. Aside from open source interest, there are many large state and corporate actors who would like to see models commoditized.
    - captainbland 1y ago
      The only real wild card to me is state sponsorship. Everywhere else I expect such a capital intensive tech industry to approach winner (or maybe duopoly) takes all over time. My view is that right now, because of the willingness of corporations and other investors to swallow short term (but massive) losses on this, we're basically in AI fiscal fantasy land. The question we should be asking is how do we get access to these local models in the first place? It's all based on the work of these hyper expensive base models as the best small models are quantisations and distills. None of this works as soon as the profit motive comes into play and companies start gatekeeping effectively, which they will.

samyar 1y ago

Well i hope that happens, you go apply for a job already 10k applied. What kind of job market is this.

ang_cire 1y ago
> What kind of job market is this. One which is reeling from layoffs caused by short-term profit-maximizing bandwagoning in the tech sector. CEO see, CEO do. We didn't just magically have tens of thousands of new programmers/ IT pros. pop out of nowhere, they were laid off. There is no situation in which this brain-drain will benefit them. They can't wait around for 10 years for it to get really bad, in the hopes that at that point they'll be back in high demand.

empath75 1y ago

> But do you know what another part of being a senior developer is? Not just making them more productive, but also guiding the junior developers into becoming better, independent, self-tasking, senior coders. And that feedback loop doesn't exist here. Almost every senior developer I know is spending that time making LLM's more productive and useful instead. Whatever you think the job is of the senior developer, it will not be "coding". I think people need to stop thinking of themselves as computer programmers and start thinking of themselves as _engineers_. Your job isn't writing programs, your job is _using the technology you have available to solve problems_. Maybe that is through writing code, but maybe it's orchestrating LLM's to write code for you. The important part is solving the problem.

ang_cire 1y ago
> Almost every senior developer I know is spending that time making LLM's more productive and useful instead. LLMs may become more productive/ accurate/ useful, but they're not self-tasking or independent. > I think people need to stop thinking of themselves as computer programmers and start thinking of themselves as _engineers_. Your job isn't writing programs, your job is _using the technology you have available to solve problems_. There is a progression of skill required to master any profession, starting with fundamentals, and progressing and developing until you are an expert/ senior at that profession. How is a senior sw dev supposed to become that without writing code? Just reading LLM code and bugfixing isn't the same level or kind of experience. You're going to have devs who can't code by themselves, and that's a bad place to be in. There are already too many people in IT using tools that they don't understand the workings of (and thus can't troubleshoot, can't replace, can't customize to their env, etc), and this will just exacerbate that x100. MMW there is going to be a very bad skill deficit in IT in 20 years, which is going to cause an innovation deficit.

ReptileMan 1y ago

>We're robbing ourselves of good future developers You call it robbing ourselves of good future developers, I call it hourly consultancy rate increase.

elric 1y ago

This is something that's been simmering in the back of my mind for a while. Using an AI Agent instead of talking to your human colleagues deprives both of you from learning opportunities. There are probably short term gains in many cases, but I fear there will be long term losses over time.

causal 1y ago
I agree, and think that organizations that figure out how to use AI well in a collaborative way will succeed in the long-term. Developer community is still where the important growth happens.
throw8393494 1y ago
Is it possible to talk to coworkers? What if you voted for wrong party? Are closeted gay/trans/qeer? Radical femimist?! Or dog atrack survivor, and they really _REALLY_ like dogs! Talking to colleagues at work is a chore, and huge risk! Not opportunity! At least AI respects my privacy, and will not get my fired!
- throwaway29812 1y ago
  [dead]