I really don't know what to feel about AI generated code. It's clear that it improves velocity dramatically in code output. It has some value to it. But what is the cost?
While I code without the help of AI, sometimes I feel like I am being left behind.
Ohh you wrote only 400 lines today? I shipped a whole app while being in the gym... hits hard.
AI-assisted coding helps me a lot to churn through boilerplate, straightforward implementations, test writing, exploration of unfamiliar codebases, and other ancillary tasks in coding. It does not replace tacit knowledge/domain knowledge required to design a good solution, no matter how much of the architecture is described as documentation it cannot predict the future to have judgement about a proposed design.
In general AI is an assistance, it doesn't replace the intellectual work that is needed for a good implementation.
In my opinion, approach it like you'd approach any other tool: there are good use-cases for it while being definitely not the right tool for other cases.
If that makes you feel better, the guy who vibecoded an app in one sitting will probably have 0 clue on how the code works, the app itself will look like the average claude eyesore and the docs will be unreadable by humans without stimulants
Sadly, not. I trying to work with doctor's offices to improve their productivity with IT support and better structure for workers to work with patients.
You don't want to know how often I heard: "But I saw on Instagram, with AI it can be done cheaper and faster".
The people that would pay me don't care for good docs, eyesore, or if the programmer actually knows how the app works. They care for how fast are you done? And how much does it cost?
The code itself indeed has become cheap but the question still is out whether it solves the right problem. I've shifted my work significantly toward designing more high-level portions where multiple things would interact in somewhat unpredictable manner (read as failure modes) that's where I found the models being quite poor, they do suggest many things to solve these or that problems but in a unity these suggestions miss second order effects. I was quite surprised that models can't hold all of the connected pieces simultaneously to reason about. But once I have a complete design after a week of deliberations and various attempts, the models can produce lots of correct code to cover that. And then another pass starts because they still deviate in implementation of various pieces and once you start questioning those the models start self-correcting. Judging the output is still way faster than writing the code yourself and indeed models still show some cases that slipped once mind. It's a definitive increase in terms of quality and value of the end code no doubt. Whether it can one-shot a complex problem, not really.
Code is merely the means to an end. It does not matter who writes it as long as it does what you want. Correctly.
Stands to reason though that someone who knows programming AND has domain knowledge can get LLMs to produce much better output compared to someone who does not.
I never used to have time to make all the stuff I needed or was interested in. With LLMs, I can.
This "shipped a whole app while being in the gym" does not work for me though. It takes me a couple of days to a week to produce solid, functional software (~10KLOC). Simple tools (3-400LOC)? Yeah, those you can produce in 30-60 minutes.
As is with any engineering… a key skill is knowing which tool to use when and for what.
There will always be people who push the extreme limits of either way. “I will not use AI for anything!” Vs “I will vibe code this while I’m at the gym”.
The winner is the one that knows when and how to use AI as a _tool_ and focus on the better outcome.
Please don’t do this if your livelihood depends on it. You’re just going to make your life more painful.
As a professional programmer, you’re selling a service. The people buying your service have fully bought into AIs and accepted their shortcomings. You can’t fight this.
You can still scratch your itch with hobby apps, or contributing to human-first projects like Zig.
All you need to us actually look around, has software quality dramatically increased over the last year? Have useful features been coming out at an unprecedented rate?
Because after all this talk about being left behind and I see nothing to show for it.
Within my own company, yes. We’ve shipped a lot more features this year with the same number of people. They’re also better features. We get from 0 to 90% in an hour and then still put a lot of effort into that last 10%. The projects have better test coverage, docs, logging, etc. then they did 2 years ago. Stretch goals are reached much more frequently. What used to be 3 week projects are closer to 3 days. Our commit history is cleaner, easier to understand, has has better commit messages because we have Claude rewrite the history before opening the PRs. Our product is faster, more stable, and has nicer UX.
But this is anecdotal. It’s a team of senior devs who already knew the codebase well. We have a ton of detailed guidelines that make sure it’s writing code we deem acceptable. We review each others code, and get LLMs to review, a LOT.
I code because the software i need is not there yet and these tools still fail at doing those.
Not sure why the next xy (already done) app has any economic value outside being a cheaper clone.
Still there is nothing wrong with having an AI point out problems which i assume is the case here and someone still is responsible and knowledgeable about the patched (!) code.
The reality is that you will be left behind, you need to decide if thats a bad thing or not. If you want software development to be your career it probably is, if you enjoy it as a hobby than probably not. You're the 2020s version of the 1990s developer sticking with ASM.
This argument of LLMs vs manual coding being compared to higher level languages vs assembly really never holds well. To be honest people who know assembly are still definitely in demand for various open source projects as well as proprietary codebases in companies like within embedded systems, linux development, compiler engineering, etc. Whereas LLMs are a whole new class of problems and discussions with very little to almost no similarities
The same feeling, do AI knows the hardware? Does it have the context of the changes? Did it read Linus decisions why something should be the way he thinks is correct? I would prefer such approach where AI knows so much context as Linus and the most important persons in the kernel team
This is what I mean, its not a low number far from it. But I know people that only value that number, so previously they churn out a big LOC, higher ups praise. Now its just big number good, bigger number even better. And LLMs do exactly that.
Now LOC comes down to how much tokens you are willing to spend for a given day.
Why would you code without the help of AI, genuinely asking? You can have quite strict "guardrails" that would maintain code consistency/quality that can be close to guaranteed nowadays.
It's not only about the pace, but also about the quality, most softwares have proven that not getting AI-reviewed is risky in its own, at least in the field of security.
In my case, it would be to make sure I have a good mental model of the problem at hand instead of offloading it to a LLM. I'd also rather not put myself in a position where I'm reliable on a LLM to do my job.
Oh and there are also the huge moral and ethical issues with the companies that offer LLMs.
That's a good question. I don't have anything to critique regarding consistency or quality. You are right to say that the newest frontier models are quite good on that front. Especially when you are doing a task you have little experience in.
I like programming, it fun! Using AI feels like I am optimizing the fun out of it. I learn so much about different programming languages and topics I would simply not learn when using AI. I also like learning things!
The cost is similar to doing any other task in a rushed, careless manner, with little regard for the consequences.
> sometimes I feel like I am being left behind.
And why does that bother you? “Forward” does not equal “good” or “desirable”. If a van is driving towards a cliff at high speed, would you not rather be left behind? “Being left behind” is a talking point to cause you FOMO, don’t fall for it.
You are left behind.
That's like saying: "I am using punch cards for programming and I feel left behind as others are using real-time IDEs"
You don't need to vibe code (and even this is only a matter of time as llms become better and better), but not using AI at all will definitely leave you behind.
As an AI user you're not that different from the punch cards programmer.
You're more like a cobbler who adopted some new electrical tools for their craft, and feel smug against backwards cobblers using manual tools, while in the meantime the newer versions of those tools in the form of factory automation move in to make the whole craft obsolete.
But what is that cobbler supposed to do during that time? He can keep his head above water using the machine, suspecting the factory is coming for him too or ... he can refuse to modernise and fail now as others outperform him?
"Stop the factory being built" is not within his power.
At the very least, AI can review all the code you have written. You can still reject every suggestion, but in the worst it will just slow you down. No citation needed for that. If an AI were to truly confuse you sufficiently to accept a wrong suggestion based on phony logic, then it just shows you didn't grok the logic deeply in the first place. (Which is a helpful finding for the second iteration.)
You didn't say if you did, but just in case if you truly never had a serious look at LLM tools, you really need to step up. Never even trying something like Claude Code borders on negligence nowadays.
AI is still a bubble, but pretty much every software developer having a hard-dependency on these companies nowadays, is was sustains it. It sucks, but currently there is no realistic way around it.
Exceptions apply for data protection reasons, obviously don't upload all your personal data and business-critical intellectual property into some 3rd-party AI tool.
Linux kernel is a huge project - with a lot of people volunteering. You’d be burning through a crazy amount of input tokens to gain enough context to get any work done. I doubt many volunteers pay for their own top-tier subscription plans that would be able to do so cost-effectively.
My follow up would be - how many are unemployed vs employed. From what I do know, most linux devs also work someplace, so either directly or indirectly have the means for a 20-100 dollar subscription a month.
More than that, I've not seen people do agentic coding in one part of their work, and manual coding in another. Is there any first hand account of someone doing this? It kind of rewires your brain, and makes it hard to do both at the same time.
Embedded code and systems code tends to have a lot of extra constraints the model doesn't know about and aren't necessarily written down where it can see. This produces a higher rate of "superficially plausible but doesn't actually work" code.
(There are probably wider benchmarks, but this was a definite finding in our company wide evaluation of effectiveness. LLMs aren't great at RTL either)
I would assume that linux devs are relatively older. Furthermore they are probably very knowledgeable in that area, and one thing I've noted on here is people who are knowledgeable not liking the quality of code output from LLMs, if they have stringent quality requirements it might be easier for them to write the code themselves.
The pressure to use LLMs is high in the companies that sell shovels in this gold rush. It would look bad if they did not use them. Linux developers work at those companies.
in your version of the conspiracy theory, the developers are pressured to use AI for their patches in Linux from the company. not that they want to use it because they want to? Like everyone else?
Just thinking loudly here, I guess Co-Author in commits is a hint. PRs where LLMs have been involved is probably another measurement. Github Copoilot has its own account too.
I really don't know what to feel about AI generated code. It's clear that it improves velocity dramatically in code output. It has some value to it. But what is the cost?
While I code without the help of AI, sometimes I feel like I am being left behind.
Ohh you wrote only 400 lines today? I shipped a whole app while being in the gym... hits hard.
AI-assisted coding helps me a lot to churn through boilerplate, straightforward implementations, test writing, exploration of unfamiliar codebases, and other ancillary tasks in coding. It does not replace tacit knowledge/domain knowledge required to design a good solution, no matter how much of the architecture is described as documentation it cannot predict the future to have judgement about a proposed design.
In general AI is an assistance, it doesn't replace the intellectual work that is needed for a good implementation.
In my opinion, approach it like you'd approach any other tool: there are good use-cases for it while being definitely not the right tool for other cases.
I think if you review the code and keep a good mental model of the codebase, the costs are pretty minimal but the value is huge. I'm not going back...
If that makes you feel better, the guy who vibecoded an app in one sitting will probably have 0 clue on how the code works, the app itself will look like the average claude eyesore and the docs will be unreadable by humans without stimulants
Sadly, not. I trying to work with doctor's offices to improve their productivity with IT support and better structure for workers to work with patients.
You don't want to know how often I heard: "But I saw on Instagram, with AI it can be done cheaper and faster".
The people that would pay me don't care for good docs, eyesore, or if the programmer actually knows how the app works. They care for how fast are you done? And how much does it cost?
> "But I saw on Instagram, with AI it can be done cheaper and faster".
You’re a doctor. What’s your reaction when a patient rejects your diagnosis due to something they saw on Instagram?
> The people that would pay me don't care for good docs, eyesore, or if the programmer actually knows how the app works.
Do they care for liability?
Leave your card to fix the mess later.
well even if the results are crap, it still has big economic influence:
"since I can use AI to create this app this much, I'll pay you less"
"although current AI made this crap, it'll outpace you eventually, so I'll pay you less"
The code itself indeed has become cheap but the question still is out whether it solves the right problem. I've shifted my work significantly toward designing more high-level portions where multiple things would interact in somewhat unpredictable manner (read as failure modes) that's where I found the models being quite poor, they do suggest many things to solve these or that problems but in a unity these suggestions miss second order effects. I was quite surprised that models can't hold all of the connected pieces simultaneously to reason about. But once I have a complete design after a week of deliberations and various attempts, the models can produce lots of correct code to cover that. And then another pass starts because they still deviate in implementation of various pieces and once you start questioning those the models start self-correcting. Judging the output is still way faster than writing the code yourself and indeed models still show some cases that slipped once mind. It's a definitive increase in terms of quality and value of the end code no doubt. Whether it can one-shot a complex problem, not really.
Code is merely the means to an end. It does not matter who writes it as long as it does what you want. Correctly.
Stands to reason though that someone who knows programming AND has domain knowledge can get LLMs to produce much better output compared to someone who does not.
I never used to have time to make all the stuff I needed or was interested in. With LLMs, I can.
This "shipped a whole app while being in the gym" does not work for me though. It takes me a couple of days to a week to produce solid, functional software (~10KLOC). Simple tools (3-400LOC)? Yeah, those you can produce in 30-60 minutes.
Building a business on vibe coded software feels like building your house on sand.
Sooner or later you'll end up troubleshooting something you just don't understand.
And the tricky part about AI is that it only has to make one bad decision confidently, to then base all future decisions on that one bad one.
As is with any engineering… a key skill is knowing which tool to use when and for what.
There will always be people who push the extreme limits of either way. “I will not use AI for anything!” Vs “I will vibe code this while I’m at the gym”.
The winner is the one that knows when and how to use AI as a _tool_ and focus on the better outcome.
Please don’t do this if your livelihood depends on it. You’re just going to make your life more painful.
As a professional programmer, you’re selling a service. The people buying your service have fully bought into AIs and accepted their shortcomings. You can’t fight this.
You can still scratch your itch with hobby apps, or contributing to human-first projects like Zig.
All you need to us actually look around, has software quality dramatically increased over the last year? Have useful features been coming out at an unprecedented rate?
Because after all this talk about being left behind and I see nothing to show for it.
Within my own company, yes. We’ve shipped a lot more features this year with the same number of people. They’re also better features. We get from 0 to 90% in an hour and then still put a lot of effort into that last 10%. The projects have better test coverage, docs, logging, etc. then they did 2 years ago. Stretch goals are reached much more frequently. What used to be 3 week projects are closer to 3 days. Our commit history is cleaner, easier to understand, has has better commit messages because we have Claude rewrite the history before opening the PRs. Our product is faster, more stable, and has nicer UX.
But this is anecdotal. It’s a team of senior devs who already knew the codebase well. We have a ton of detailed guidelines that make sure it’s writing code we deem acceptable. We review each others code, and get LLMs to review, a LOT.
So I'm guessing your user satisfaction has gone up a bunch and revenue now reflects your increased productivity?
I code because the software i need is not there yet and these tools still fail at doing those.
Not sure why the next xy (already done) app has any economic value outside being a cheaper clone.
Still there is nothing wrong with having an AI point out problems which i assume is the case here and someone still is responsible and knowledgeable about the patched (!) code.
The reality is that you will be left behind, you need to decide if thats a bad thing or not. If you want software development to be your career it probably is, if you enjoy it as a hobby than probably not. You're the 2020s version of the 1990s developer sticking with ASM.
This argument of LLMs vs manual coding being compared to higher level languages vs assembly really never holds well. To be honest people who know assembly are still definitely in demand for various open source projects as well as proprietary codebases in companies like within embedded systems, linux development, compiler engineering, etc. Whereas LLMs are a whole new class of problems and discussions with very little to almost no similarities
[delayed]
The same feeling, do AI knows the hardware? Does it have the context of the changes? Did it read Linus decisions why something should be the way he thinks is correct? I would prefer such approach where AI knows so much context as Linus and the most important persons in the kernel team
400 locs is a very exhausting day for me, I think I rarely did so much work a day in my entire 15 years career
This is what I mean, its not a low number far from it. But I know people that only value that number, so previously they churn out a big LOC, higher ups praise. Now its just big number good, bigger number even better. And LLMs do exactly that.
Now LOC comes down to how much tokens you are willing to spend for a given day.
Depends on the kind. 400 locs JS or Java or something is barely worth mentioning. Can do more, a lot of it is boilerplate bs.
400 locs of critical C code, sure.
Why would you code without the help of AI, genuinely asking? You can have quite strict "guardrails" that would maintain code consistency/quality that can be close to guaranteed nowadays.
It's not only about the pace, but also about the quality, most softwares have proven that not getting AI-reviewed is risky in its own, at least in the field of security.
In my case, it would be to make sure I have a good mental model of the problem at hand instead of offloading it to a LLM. I'd also rather not put myself in a position where I'm reliable on a LLM to do my job.
Oh and there are also the huge moral and ethical issues with the companies that offer LLMs.
That's a good question. I don't have anything to critique regarding consistency or quality. You are right to say that the newest frontier models are quite good on that front. Especially when you are doing a task you have little experience in.
I like programming, it fun! Using AI feels like I am optimizing the fun out of it. I learn so much about different programming languages and topics I would simply not learn when using AI. I also like learning things!
> But what is the cost?
The cost is similar to doing any other task in a rushed, careless manner, with little regard for the consequences.
> sometimes I feel like I am being left behind.
And why does that bother you? “Forward” does not equal “good” or “desirable”. If a van is driving towards a cliff at high speed, would you not rather be left behind? “Being left behind” is a talking point to cause you FOMO, don’t fall for it.
You are left behind. That's like saying: "I am using punch cards for programming and I feel left behind as others are using real-time IDEs" You don't need to vibe code (and even this is only a matter of time as llms become better and better), but not using AI at all will definitely leave you behind.
As an AI user you're not that different from the punch cards programmer.
You're more like a cobbler who adopted some new electrical tools for their craft, and feel smug against backwards cobblers using manual tools, while in the meantime the newer versions of those tools in the form of factory automation move in to make the whole craft obsolete.
This analogy may well be very apposite.
But what is that cobbler supposed to do during that time? He can keep his head above water using the machine, suspecting the factory is coming for him too or ... he can refuse to modernise and fail now as others outperform him?
"Stop the factory being built" is not within his power.
"Becoming an engineering manager will leave you unemployed"
Citation needed?
Sure you can churn out more code with a LLM, but is it good code? Is it correct? If being left behind means I don't ship slop, I'll gladly bow out.
At the very least, AI can review all the code you have written. You can still reject every suggestion, but in the worst it will just slow you down. No citation needed for that. If an AI were to truly confuse you sufficiently to accept a wrong suggestion based on phony logic, then it just shows you didn't grok the logic deeply in the first place. (Which is a helpful finding for the second iteration.)
You didn't say if you did, but just in case if you truly never had a serious look at LLM tools, you really need to step up. Never even trying something like Claude Code borders on negligence nowadays.
AI is still a bubble, but pretty much every software developer having a hard-dependency on these companies nowadays, is was sustains it. It sucks, but currently there is no realistic way around it.
Exceptions apply for data protection reasons, obviously don't upload all your personal data and business-critical intellectual property into some 3rd-party AI tool.
Seems low. 80% people are manually typing in things? In my circle it's close to 0% code written manually.
Do linux devs live in a completely different microcosm?
My hypothesis:
Linux kernel is a huge project - with a lot of people volunteering. You’d be burning through a crazy amount of input tokens to gain enough context to get any work done. I doubt many volunteers pay for their own top-tier subscription plans that would be able to do so cost-effectively.
My follow up would be - how many are unemployed vs employed. From what I do know, most linux devs also work someplace, so either directly or indirectly have the means for a 20-100 dollar subscription a month.
More than that, I've not seen people do agentic coding in one part of their work, and manual coding in another. Is there any first hand account of someone doing this? It kind of rewires your brain, and makes it hard to do both at the same time.
Embedded code and systems code tends to have a lot of extra constraints the model doesn't know about and aren't necessarily written down where it can see. This produces a higher rate of "superficially plausible but doesn't actually work" code.
(There are probably wider benchmarks, but this was a definite finding in our company wide evaluation of effectiveness. LLMs aren't great at RTL either)
I would assume that linux devs are relatively older. Furthermore they are probably very knowledgeable in that area, and one thing I've noted on here is people who are knowledgeable not liking the quality of code output from LLMs, if they have stringent quality requirements it might be easier for them to write the code themselves.
Small changes can have a huge impact.. far beyond your context window ;-)
How big is your circle? What industries does it span? Maybe the people writing 0% code manually are outliers and your circle is relatively small?
The pressure to use LLMs is high in the companies that sell shovels in this gold rush. It would look bad if they did not use them. Linux developers work at those companies.
in your version of the conspiracy theory, the developers are pressured to use AI for their patches in Linux from the company. not that they want to use it because they want to? Like everyone else?
> In September, AI generated code has made up 17.25% of all Linux Kernel patches
I'm curious ... how would they even know?
Just thinking loudly here, I guess Co-Author in commits is a hint. PRs where LLMs have been involved is probably another measurement. Github Copoilot has its own account too.
They require disclosure of what tool was used if submitting AI-helped code.
(Which probably means this is slightly underreported from people not disclosing.)
So the Linux kernel has now just become by definition "slop"? I mean, there is a definition here which says it is now "tainted". [0]
Does that mean we should stop using it?
[0] https://codeberg.org/ethical-foss/open-slopware
Huh.
Wait until it blows.
And I say this as someone who uses AI (for fun) and to check things I may havbe overlooked.
There's a place and time for everything. Using AI where you shouldn't is gonna backfire, big time.
Also, just please show me ONE single app/product that has been generated 100% by AI--that works and competes with human-written ones?
I'll be waiting, right here.