I don’t disagree, but this ignores some of what we have learned about critical masses and network effects. If everyone can just take their own copy with no interaction, many won’t join the community and it may never reach a critical mass despite and because of its openness.
Im really interested in seeing where the anti-ai crowd ends up in a few years.
Im sympathetic to the perspective in some ways but I think it's wildly impractical to actually try to keep AI out of software development.
I suppose an anti-ai optimist might hope for a future where AI written software proves too problematic over time and things return to their ideal vision of no AI in most software.
I really, really dont think that's in the range of possibilities. So what would be left?
- pockets of anti ai software
- these people slowly disappear or the window moves ("ai in software BUT ...")
1. All modern LLMs are (in my eyes) inherently morally problematic. They were trained on stolen information, and where they didn't break the law they corrupted social contracts. Their manner of operation is also problematic: disadvantaged communities are being exploited at a rapidly increasing disproportionate rate in the name of "more compute".
2. I am fundamentally disinterested in the productivity mindset that LLMs lean into. I don't care to "produce more"; the value of my work does not lie in the amount of money that can be gotten in exchange for it. This is an easier position to hold as someone who does not hold a typical corporate job, but even if I did I would still resist because an increase in my output would not correlate to an increase in my salary.
At the root of it all is people. I care about people (their existence, their happiness, my connections to them, their connections to each other), and any use of LLMs is (in my mind) anti-people, so I want no part of it.
>1. All modern LLMs are (in my eyes) inherently morally problematic.
Yes, I think this is a key point and I'm not surprised to hear it. It's a fundamental problem with AI. It's not something that will disappear as it becomes more useful or makes less mistakes.
How do you feel about a dependency that was written by AI? It seems like things will get progressively harder and harder to maintain a real boundary, outside of just writing everything yourself. And perhaps hosting.
I think this is a certainty. Future generations will be born into a world where education, technology, and jobs have corrected for AI's existence. They will never know a world without it or have the experience of the current growing pains.
We've seen this plenty of times with new tech in the past, and as disruptive as AI is, I can't imagine the cat going back in the bag.
Do you believe that we're in an AI bubble? If so what do you think will happen if & when it pops? What happens if Anthropic and OpenAI go under? Will Microsoft and Google continue investing? Or will local/open-weight LLMs be viable alternatives? In the latter case will there be new ones trained, and if so who's gonna build the AI data centers that turned to be unprofitable in absence of a frontier model to distill cheaply? Or do you think we can run current models and point them at the Internet and be done with it?
Is that all that relevant? Surely even if AI becomes financially unsustainable in the next two years, it's still rational to use it right now while it's VC-funded.
Give it a few years and nobody will want to run software that hasn't been debugged and hardened by AIs. It's going to be much more thorough than human auditors - it already is, to some extent, and faster anyway, but I'm talking about the future. So it's going to be the latter, complete normalization of AI. You can still write the program by hand, but you'll need AI to make it trustworthy and secure enough, if only because the attackers will also be mostly AI.
If AI changes how software is built by _that_ much, then there's no reason to believe that AI won't change how software is consumed by an equal or greater amount.
You need to build a thing first. Community doesn't matter if you don't have cool software. License doesn't matter if you don't have cool software. I'm seeing people forget this principle a lot recently.
The best thing you can do to promote free software is to make good free software. You have to use copyleft licenses or else the software quickly becomes unfree, if it's actually worth anything. This doesn't happen to most non-copyleft free software because it's not worth very much, not because it doesn't happen. I was recently invited to sign a European digital sovereignty petition (which I am not eligible to sign). Know what helps much more with that? Making European digital sovereign systems. Instead of petitioning the government to switch MS Office to LibreOffice someone would have to actually put in the effort to find out why the government doesn't use LibreOffice, and fix that. (It's probably not because it doesn't run on their computers!)
A community of users doesn't have much value. Any forum or Discord guild or mailing list is secondary to the actual software and it actually being used. Plenty of successful software has no "community" - sudo, for example. Plenty of software with a "community" is not successful. Some software is successful and has a community, like Blender.
A community of developers is only valuable if they're doing valuable things. The Fediverse chat group that shared the link to the European digital sovereignty petition - no useful Fediverse software seems to be coming out of the people in that group. It's all about ivory-tower protocols that won't possibly work in the real world, and stuff like that. No actually working software.
We also have to stop over-estimating the worth of free software just because it's free. Good free software is good software which is also free software. No software is bad, bad software is also bad. Some random code you wrote and published somewhere on the internet usually isn't very useful, and that's probably why it doesn't have a community. Don't try to create a community first - that won't make the software any better. You have to make good software, that can spread by word of mouth, first and then you can have a community after that. Think about the number of random little programs on the internet, now think how many are installed on your computer - not a very big percentage, but the ones that you do have installed, many people have installed. Those are the useful ones.
In most projects the original creators remain the main developers no matter how many other contributors join. Late contributors having a significant effect on a project is the exception, not the rule, and when it happens it's actually an undesirable power inversion since the original creator retains unearned influence - see the PolyMC incident. There is no way to become the chief developer of a project other than having created it. If you think you can rewrite the whole thing and make it better, don't "join the community" - make a fork (which makes you part of the community, but that is not the main goal!).
If you're actually doing the majority of the work, offering a dual license is not unethical. We are forced to live in capitalism whether we like it or not, and we should be trying to squeeze money out of capitalist entities - lord knows they're good at squeezing money out of us. To not play that game is to lose by default. This isn't incompatible with being nice to people who are not money-squeezers, if you dual-license. A community fork can legally continue as AGPL-only. (You don't necessarily need a full CLA either, you just need people to agree their code will be in your commercial version. Programmers treat licenses and CLAs like satisfying the borrow checker, but that's not how they actually work.)
> Does the project have an inclusive code of conduct?
If you want a project to be community driven, adding an agreement that empowers its leaders to punish and exclude any community member for any reason, including reasons that have nothing to do with the project itself, is the opposite of that.
Codes of conduct result in dictator-driven projects, not community-driven projects.
The entire post is heavily driven by this author's ideologies around what they believe to correct. I mean just look at the footer and tell me this is someone on a reasonable part of the political spectrum.
I notice this with ruby right now, where claude took over development.
But it is not confined to just ruby. AI slop eliminates human contributors.
Why would I want to even try to interact with AI slop meta-designing
software? If I go to another developer and ask "hey many, how did you add
this great feature #123"? I'd get the reply "claude slop". Wowsers. Now
tell me this is healthy to a community ...
I wonder how people work with omarchyslop (the linux distribution spam-slopped by DHH). Do they have any real community outside of paid fanboi accounts? If so, how do they interact? I feel that there is no real exchange of e. g. code since claude slop just autogenerates that. It must be strange for them.
The agentic development loop is just another tool at a developer's disposal. Whether it produces slop is still a matter of the driver's design, taste, risk tolerance, skill, etc. Even in your hypothetical, it implies liking a feature enough to reach out and ask about it.
One perspective that you might want to conside, and which is discussed by Oxide folks in this podcast episode [1], is that agents are a kind of non-deterministic compiler which is still, to some extent, capable of producing useful output. Consider this: even before agents, ruby source code might have been seen slop to someone writing a strongly typed language.
Lastly, there is a community for users of that distro if you care to look for it (I am not among them).
We really feel for the poor oppressed AI people who have no financial, political or corporate backing unlike Codeberg, which gets $0.5 trillion in investor money!
You are not allowed to say certain personalities who cover some of this stuff on Hacker News, albeit from a little bit of a theatrical position, but it still does get to the heart of some of this stuff. This sheer nonsense. What this person is talking about is not a community. They want a club.
More evidence of tribalism, combined with exceptionally precise confessions running as whataboutisms
Pelagicaustral is on the money.
It would just be nice if people understood the difference between wanting a club with a gate versus a community. I am impressed, though. I never thought I’d see a vomit diarrhea load of bumper stickers, but a digital form at the end of the article.
If I publish my code under an open source license, you are free to take it and build any community around it you want.
If I keep my code proprietary, you are beholden to the shape of the community I dictate.
The license is first, it precedes, any notion about community.
I don’t disagree, but this ignores some of what we have learned about critical masses and network effects. If everyone can just take their own copy with no interaction, many won’t join the community and it may never reach a critical mass despite and because of its openness.
Im really interested in seeing where the anti-ai crowd ends up in a few years.
Im sympathetic to the perspective in some ways but I think it's wildly impractical to actually try to keep AI out of software development.
I suppose an anti-ai optimist might hope for a future where AI written software proves too problematic over time and things return to their ideal vision of no AI in most software.
I really, really dont think that's in the range of possibilities. So what would be left?
- pockets of anti ai software
- these people slowly disappear or the window moves ("ai in software BUT ...")
- ai use is completely and thoroughly normalized
Two things are important to me:
1. All modern LLMs are (in my eyes) inherently morally problematic. They were trained on stolen information, and where they didn't break the law they corrupted social contracts. Their manner of operation is also problematic: disadvantaged communities are being exploited at a rapidly increasing disproportionate rate in the name of "more compute".
2. I am fundamentally disinterested in the productivity mindset that LLMs lean into. I don't care to "produce more"; the value of my work does not lie in the amount of money that can be gotten in exchange for it. This is an easier position to hold as someone who does not hold a typical corporate job, but even if I did I would still resist because an increase in my output would not correlate to an increase in my salary.
At the root of it all is people. I care about people (their existence, their happiness, my connections to them, their connections to each other), and any use of LLMs is (in my mind) anti-people, so I want no part of it.
>1. All modern LLMs are (in my eyes) inherently morally problematic.
Yes, I think this is a key point and I'm not surprised to hear it. It's a fundamental problem with AI. It's not something that will disappear as it becomes more useful or makes less mistakes.
How do you feel about a dependency that was written by AI? It seems like things will get progressively harder and harder to maintain a real boundary, outside of just writing everything yourself. And perhaps hosting.
> ai use is completely and thoroughly normalized
I think this is a certainty. Future generations will be born into a world where education, technology, and jobs have corrected for AI's existence. They will never know a world without it or have the experience of the current growing pains.
We've seen this plenty of times with new tech in the past, and as disruptive as AI is, I can't imagine the cat going back in the bag.
A lot of this comes and goes like fashion styles, I expect most of them to be doing the same bit about some other topic in a couple years.
Do you believe that we're in an AI bubble? If so what do you think will happen if & when it pops? What happens if Anthropic and OpenAI go under? Will Microsoft and Google continue investing? Or will local/open-weight LLMs be viable alternatives? In the latter case will there be new ones trained, and if so who's gonna build the AI data centers that turned to be unprofitable in absence of a frontier model to distill cheaply? Or do you think we can run current models and point them at the Internet and be done with it?
Is that all that relevant? Surely even if AI becomes financially unsustainable in the next two years, it's still rational to use it right now while it's VC-funded.
Give it a few years and nobody will want to run software that hasn't been debugged and hardened by AIs. It's going to be much more thorough than human auditors - it already is, to some extent, and faster anyway, but I'm talking about the future. So it's going to be the latter, complete normalization of AI. You can still write the program by hand, but you'll need AI to make it trustworthy and secure enough, if only because the attackers will also be mostly AI.
If AI changes how software is built by _that_ much, then there's no reason to believe that AI won't change how software is consumed by an equal or greater amount.
Several points:
You need to build a thing first. Community doesn't matter if you don't have cool software. License doesn't matter if you don't have cool software. I'm seeing people forget this principle a lot recently.
The best thing you can do to promote free software is to make good free software. You have to use copyleft licenses or else the software quickly becomes unfree, if it's actually worth anything. This doesn't happen to most non-copyleft free software because it's not worth very much, not because it doesn't happen. I was recently invited to sign a European digital sovereignty petition (which I am not eligible to sign). Know what helps much more with that? Making European digital sovereign systems. Instead of petitioning the government to switch MS Office to LibreOffice someone would have to actually put in the effort to find out why the government doesn't use LibreOffice, and fix that. (It's probably not because it doesn't run on their computers!)
A community of users doesn't have much value. Any forum or Discord guild or mailing list is secondary to the actual software and it actually being used. Plenty of successful software has no "community" - sudo, for example. Plenty of software with a "community" is not successful. Some software is successful and has a community, like Blender.
A community of developers is only valuable if they're doing valuable things. The Fediverse chat group that shared the link to the European digital sovereignty petition - no useful Fediverse software seems to be coming out of the people in that group. It's all about ivory-tower protocols that won't possibly work in the real world, and stuff like that. No actually working software.
We also have to stop over-estimating the worth of free software just because it's free. Good free software is good software which is also free software. No software is bad, bad software is also bad. Some random code you wrote and published somewhere on the internet usually isn't very useful, and that's probably why it doesn't have a community. Don't try to create a community first - that won't make the software any better. You have to make good software, that can spread by word of mouth, first and then you can have a community after that. Think about the number of random little programs on the internet, now think how many are installed on your computer - not a very big percentage, but the ones that you do have installed, many people have installed. Those are the useful ones.
In most projects the original creators remain the main developers no matter how many other contributors join. Late contributors having a significant effect on a project is the exception, not the rule, and when it happens it's actually an undesirable power inversion since the original creator retains unearned influence - see the PolyMC incident. There is no way to become the chief developer of a project other than having created it. If you think you can rewrite the whole thing and make it better, don't "join the community" - make a fork (which makes you part of the community, but that is not the main goal!).
If you're actually doing the majority of the work, offering a dual license is not unethical. We are forced to live in capitalism whether we like it or not, and we should be trying to squeeze money out of capitalist entities - lord knows they're good at squeezing money out of us. To not play that game is to lose by default. This isn't incompatible with being nice to people who are not money-squeezers, if you dual-license. A community fork can legally continue as AGPL-only. (You don't necessarily need a full CLA either, you just need people to agree their code will be in your commercial version. Programmers treat licenses and CLAs like satisfying the borrow checker, but that's not how they actually work.)
> Does the project have an inclusive code of conduct?
If you want a project to be community driven, adding an agreement that empowers its leaders to punish and exclude any community member for any reason, including reasons that have nothing to do with the project itself, is the opposite of that.
Codes of conduct result in dictator-driven projects, not community-driven projects.
The entire post is heavily driven by this author's ideologies around what they believe to correct. I mean just look at the footer and tell me this is someone on a reasonable part of the political spectrum.
That's literally what a blog post is, yes.
And inclusive of whom?
[flagged]
this wasn't flagged, you just wrote [flagged]
I notice this with ruby right now, where claude took over development. But it is not confined to just ruby. AI slop eliminates human contributors. Why would I want to even try to interact with AI slop meta-designing software? If I go to another developer and ask "hey many, how did you add this great feature #123"? I'd get the reply "claude slop". Wowsers. Now tell me this is healthy to a community ...
I wonder how people work with omarchyslop (the linux distribution spam-slopped by DHH). Do they have any real community outside of paid fanboi accounts? If so, how do they interact? I feel that there is no real exchange of e. g. code since claude slop just autogenerates that. It must be strange for them.
The agentic development loop is just another tool at a developer's disposal. Whether it produces slop is still a matter of the driver's design, taste, risk tolerance, skill, etc. Even in your hypothetical, it implies liking a feature enough to reach out and ask about it.
One perspective that you might want to conside, and which is discussed by Oxide folks in this podcast episode [1], is that agents are a kind of non-deterministic compiler which is still, to some extent, capable of producing useful output. Consider this: even before agents, ruby source code might have been seen slop to someone writing a strongly typed language.
Lastly, there is a community for users of that distro if you care to look for it (I am not among them).
[1] https://www.youtube.com/watch?v=iSOdBO9I6q8
> The importance of communities around software
Unless you use AI.
Same people, same mute message; every time, with no delay. "Accept me, then I'll see if I want to accept you".
Fuck off.
We really feel for the poor oppressed AI people who have no financial, political or corporate backing unlike Codeberg, which gets $0.5 trillion in investor money!
You are not allowed to say certain personalities who cover some of this stuff on Hacker News, albeit from a little bit of a theatrical position, but it still does get to the heart of some of this stuff. This sheer nonsense. What this person is talking about is not a community. They want a club.
More evidence of tribalism, combined with exceptionally precise confessions running as whataboutisms
Pelagicaustral is on the money.
It would just be nice if people understood the difference between wanting a club with a gate versus a community. I am impressed, though. I never thought I’d see a vomit diarrhea load of bumper stickers, but a digital form at the end of the article.
If it is truly free it should use Apache 2.0