It's such a shame that a list titled "AI-free FOSS Alternatives" ends up becoming the opposite of its stated goal, with the top level section being "Tainted Software" and occasional sporadic little notes listing alternatives.
I like the steadfast anti-AI people even though I'm not really one of them, and I want them to thrive and carve out a lasting space online. Putting more focus on the positives of the anti-AI vision and way less on the rejection and refusal stuff would probably help with that, and I think this is a really clear example where the negative side of the message has stolen the show unnecessarily. The "tainted" list is inside baseball stuff and belongs in the equivalent of a Wikipedia talk page on there.
Putting more focus on the positives of the anti-AI vision and way less on the rejection and refusal stuff would probably help with that,
As a general rule, I think that "focus on the positives" is a good rule of thumb.
As a blanket rule, it doesn't work. Sometimes it's really about the negatives. If you're trying to convince people, say, not to smoke or do heroin... really, "it's great not being addicted to things" is not as strong a message as "you'll get lung cancer if you smoke" or "this is what heroin addiction does to people". (Granted "just say no" was also terrible as a campaign, but for different reasons.)
I'm not saying that AI usage is equivalent to heroin usage, exactly, but I'm not sure that "the positives of the anti-AI vision" is as persuasive as the arguments against generative AI.
where the negative side of the message has stolen the show unnecessarily
Again, I don't really agree with this. There are some things that it makes sense to be against. This is especially true in the case of generative AI in my opinion because of the way it spreads and has basically overrun the industry in a couple of years. For some of us, it's the black mold or asbestos of software. "Gee, having a house without black mold is so awesome" isn't really quite the same as "here's what black mold is going to do to you if it infests your house".
Funny you mention smoking! Alan Carr's "Easy Way" book was a huge help to me in quitting, and one of its big central ideas is its focus on convincing you that life gets easier and better without smoking rather than browbeating you into increasing your willpower or scaring you with health statistics.
One nice example could be the text editor Helix, which I've used a ton lately. Its subreddit is almost rabid with anger about the slow pace of development progress but I see it as a selling point. The whole reason to use Helix instead of Vim or Neovim is that its design is distinctive due to the taste of the creators. Its reason to exist would be undermined if they adopted a feature factory mentality to maintenance just to appease the crowd, and it's great that they have the clarity of vision to avoid that trap and take the time they need to exercise their judgement about what goes into Helix. Plus it's a text editor and it's as good as finished anyway so what's the rush?
I think we could convince some of the more extreme AI maxis use cases exist where old fashioned coding by hand is optimal if stories like those get some reach. This would amount to a meaningful narrative shift in favour of the anti-AI people. But it can't happen in the case of Helix because the maintainers have coincidentally accepted a few PRs that contained some AI-generated stuff, and so the entire thing is shitcanned as "tainted software" by a page that ironically enough looks something Claude would churn out.
Alright, the AI version of that should be relatively easy to write:
Keeping and growing your coding skills seems like a useful thing to keep yourself employed in the long run. AI usage has been shown to lead to deskilling, and why would a business hire an expensive senior to push the buttons and make the AI do its thing when you can also get a cheap junior with the same skill set to do it?
Actually enjoying your work and understanding the code in your project can give more meaning to your work than rubber-stamping AI-generated PRs that come in by the firehose. And debugging code you actually understand when the shit hits the fan is easier than debugging someone else's code.
If you intend to stay in this business for longer, a deliberate pace can also keep you sane and prevent burn-out. The switching between longer bouts of code writing and code review can contribute to that. This beats quick task switching after spending a few minutes to set off a swarm of agents and getting back to the aforementioned firehose of PRs.
Regarding choosing software as a user: it's often been shown that people have trouble adjusting to change. If a project doesn't adopt AI, the changes that could be made from one release to the next are limited, as it's bottlenecked by humans doing the coding and reviewing. This leads to a more manageable pace of change that you can keep up with more easily. There's nothing so nice as software that's more or less "done", as you also point out in the case of Helix.
I strongly agree that it is better to fight for something good rather than against something bad. However, I find the list a very useful starting point. Studying it led me to the (startling to me) conclusion that the GNU/FSF vision is a positive one that encompasses the rejection of AI. There are probably others as well, but starting with GNU is not starting from scratch at least. Flipping positions on GNU was not on my personal philosophy bingo card for this year.
I think this is a common kind of unspoken ethical framework where if you're decently progressive and/or in any way marginalised then you're ontologically incapable of being abusive because your shitty behaviour is actually all just righteous anger and moral courage and any critique you receive is actually just so much privileged tone policing. I used to subscribe to this way of thinking too and have moved on from it but it's especially popular in the anti-AI activism sphere at the moment.
Being overtly nasty might make you feel alive for a moment but it’s in conflict with any meaningful real-world goals, and it’s so normalised in anti-AI spaces that I’ve come to the conclusion that the whole thing is more about venting collective grief about unwanted change than achieving anything concrete.
I find it interesting that a number of folks who seem to like this list are upset if folks don’t use an LLM analog of “sent from my iPhone” ad/disclaimer to disclose LLM use, but then this list comes with this note: “This list is not a resource to be used for the harassment of other open-source developers. If you wish to advocate for a "No AI" Policy from another project, we ask that it be done respectfully and constructively (and do not mention this list).”
I have been studying this list for the last few months. One of the main results is that I have gone back to Emacs. This is not my first go-round with it but I have found that nowadays it's much easier to get to a working IDE-like state than ever before. There are alternatives to most of the things I used to use within Emacs itself. I found that once I got over my initial reticence, it provides the kind of human-focused, malleable environment that I didn't realize I was missing.
I am having trouble understanding the article you linked, but I note that it links to a potential fork of Emacs that would satisfy my requirements, so I think you actually have good news for me.
By listing the last “untainted” commit and listing forks, this list seems to encourage using versions that have known security vulnerabilities. Upstream is likely to have LLM-discovered disclosed sec bugs. Is there any vision for security other than running software with known vulnerabilities to accompany this list?
Well, we didn't/couldn't really plan for this, right? Some projects are being forked, and are fixing bugs themselves, but there are just so many projects that are suddenly not available anymore for people wanting their software as ethically sourced as possible. Luckily, a lot of software simply runs locally, so if you only work on your own files you are pretty much safe from attacks and just have to live with a bug here or there. But it's heartbreaking.
A slice of the vulnocapyse firehose arrives on my plate to triage or fix, and from what I’ve seen I’m skeptical of anti-AI forks fixing security bugs themselves.
Empirically, humans haven’t found the bugs that have been present for a long time and that LLMs are finding now, so a project that takes the stance that the method of discovery makes knowledge of the bugs inadmissoble aren’t going to find the bugs independently, so they will eemain unfixed.
Projects that accept awareness of upstream bugs are going to run into ones where there is one reasonable fix (sometimes it really just is flipping the right boolean) is the one that came as an LLM suggestion along with the vulnerability report. Now you can’t prove if you wrote the same fix yourself or if you took the LLM fix. The list is going to go with the least charitable interpretation, so “$FORK_NAME is now slop”.
Well, I work there, but we are both a FOSS nonprofit and also have a firm no-gen-AI policy at the Spritely Institute, and do pretty cool decentralized network tech also
I honestly always liked Mercurial better over git, but sort of gave up at some point and moved all my stuff to git, feels weird to use both now especially since hg doesn't give me any portability perks over git.
CVS is for times and situations when you want to use the version control system from some older system, like old macintosh system7 or 9 (even though @jcs made a very nice tool for it - Amend and even amendhub.com!), or *DOS.
A few of us who want to avoid it, or at least be very selective about when we talk about it have gathered at imtiredboss.org. We're collecting some lists and links, and have a couple of discussion spaces that are kept free of generative AI.
You are aware that the resource you link was made with a site builder that mentions an AI in the second sentence of their product hero? The resource was handmade, but not far away from AI.
Rather than avoiding AI, why not opt for responsible usage of AI? LLMs speed things up, but it comes at a cost of losing correctness, which I do not think one can easily guarantee because of the way LLMs operate - statistically matching rather than logic/intuitive proofs which is what humans are good at. To get the best of both worlds, you want to use AI for smaller, more easily reviewed tasks, then use those (also w/ or w/out AI) to build a more complicated product.
I've no problem with not having to write out a bunch of code that can be written nearly identically by an AI. The QA and validation process is still a human endeavor and where you probably spend most of your time as a dev anyway. Actual coding amounts to no more than 20% of the time.
Nixing a tool like LLM is IMHO a bad idea, if it can help us save time and increase quality of output, then so be it.
I'm not looking to start a discussion on what's good/bad or (un)worthy about AI, or who's for/against, just links and/or communities promoting this sentiment/position.
Comments on posts about/for people who don't like and don't want to use AI which amount to "have you considered that AI isn't all bad actually?" are counterproductive and frankly really exhausting.
I just don't want to use it. It doesn't interest me. I don't want to use things that used it either. I don't care if it "works better" or is "more efficient". It's okay to be a luddite.
That's a fine enough opinion for you to have in your own space, but that's not what the OP is after and they've not asked for people to persuade them. Respect that.
nate30480 | 23 hours ago
https://codeberg.org/ethical-foss/ai-free-foss-alternatives
henrycatalinismith | 21 hours ago
It's such a shame that a list titled "AI-free FOSS Alternatives" ends up becoming the opposite of its stated goal, with the top level section being "Tainted Software" and occasional sporadic little notes listing alternatives.
I like the steadfast anti-AI people even though I'm not really one of them, and I want them to thrive and carve out a lasting space online. Putting more focus on the positives of the anti-AI vision and way less on the rejection and refusal stuff would probably help with that, and I think this is a really clear example where the negative side of the message has stolen the show unnecessarily. The "tainted" list is inside baseball stuff and belongs in the equivalent of a Wikipedia talk page on there.
jzb | 17 hours ago
As a general rule, I think that "focus on the positives" is a good rule of thumb.
As a blanket rule, it doesn't work. Sometimes it's really about the negatives. If you're trying to convince people, say, not to smoke or do heroin... really, "it's great not being addicted to things" is not as strong a message as "you'll get lung cancer if you smoke" or "this is what heroin addiction does to people". (Granted "just say no" was also terrible as a campaign, but for different reasons.)
I'm not saying that AI usage is equivalent to heroin usage, exactly, but I'm not sure that "the positives of the anti-AI vision" is as persuasive as the arguments against generative AI.
Again, I don't really agree with this. There are some things that it makes sense to be against. This is especially true in the case of generative AI in my opinion because of the way it spreads and has basically overrun the industry in a couple of years. For some of us, it's the black mold or asbestos of software. "Gee, having a house without black mold is so awesome" isn't really quite the same as "here's what black mold is going to do to you if it infests your house".
henrycatalinismith | 3 hours ago
Funny you mention smoking! Alan Carr's "Easy Way" book was a huge help to me in quitting, and one of its big central ideas is its focus on convincing you that life gets easier and better without smoking rather than browbeating you into increasing your willpower or scaring you with health statistics.
One nice example could be the text editor Helix, which I've used a ton lately. Its subreddit is almost rabid with anger about the slow pace of development progress but I see it as a selling point. The whole reason to use Helix instead of Vim or Neovim is that its design is distinctive due to the taste of the creators. Its reason to exist would be undermined if they adopted a feature factory mentality to maintenance just to appease the crowd, and it's great that they have the clarity of vision to avoid that trap and take the time they need to exercise their judgement about what goes into Helix. Plus it's a text editor and it's as good as finished anyway so what's the rush?
I think we could convince some of the more extreme AI maxis use cases exist where old fashioned coding by hand is optimal if stories like those get some reach. This would amount to a meaningful narrative shift in favour of the anti-AI people. But it can't happen in the case of Helix because the maintainers have coincidentally accepted a few PRs that contained some AI-generated stuff, and so the entire thing is shitcanned as "tainted software" by a page that ironically enough looks something Claude would churn out.
sjamaan | an hour ago
Alright, the AI version of that should be relatively easy to write:
Keeping and growing your coding skills seems like a useful thing to keep yourself employed in the long run. AI usage has been shown to lead to deskilling, and why would a business hire an expensive senior to push the buttons and make the AI do its thing when you can also get a cheap junior with the same skill set to do it?
Actually enjoying your work and understanding the code in your project can give more meaning to your work than rubber-stamping AI-generated PRs that come in by the firehose. And debugging code you actually understand when the shit hits the fan is easier than debugging someone else's code.
If you intend to stay in this business for longer, a deliberate pace can also keep you sane and prevent burn-out. The switching between longer bouts of code writing and code review can contribute to that. This beats quick task switching after spending a few minutes to set off a swarm of agents and getting back to the aforementioned firehose of PRs.
Regarding choosing software as a user: it's often been shown that people have trouble adjusting to change. If a project doesn't adopt AI, the changes that could be made from one release to the next are limited, as it's bottlenecked by humans doing the coding and reviewing. This leads to a more manageable pace of change that you can keep up with more easily. There's nothing so nice as software that's more or less "done", as you also point out in the case of Helix.
dkl | 21 hours ago
I strongly agree that it is better to fight for something good rather than against something bad. However, I find the list a very useful starting point. Studying it led me to the (startling to me) conclusion that the GNU/FSF vision is a positive one that encompasses the rejection of AI. There are probably others as well, but starting with GNU is not starting from scratch at least. Flipping positions on GNU was not on my personal philosophy bingo card for this year.
sunshowers | 17 hours ago
This list used to be called Open Slopware. I wrote about my personal relationship to it here: https://lobste.rs/s/unpaan/arborium_is_ai_slopware_should_not_be#c_twqzh3.
henrycatalinismith | 16 hours ago
Mmmm not terribly surprising unfortunately.
I think this is a common kind of unspoken ethical framework where if you're decently progressive and/or in any way marginalised then you're ontologically incapable of being abusive because your shitty behaviour is actually all just righteous anger and moral courage and any critique you receive is actually just so much privileged tone policing. I used to subscribe to this way of thinking too and have moved on from it but it's especially popular in the anti-AI activism sphere at the moment.
Wrote down a few of my thoughts about it the other day actually.
hsivonen | 5 hours ago
I find it interesting that a number of folks who seem to like this list are upset if folks don’t use an LLM analog of “sent from my iPhone” ad/disclaimer to disclose LLM use, but then this list comes with this note: “This list is not a resource to be used for the harassment of other open-source developers. If you wish to advocate for a "No AI" Policy from another project, we ask that it be done respectfully and constructively (and do not mention this list).”
[OP] brn | 23 hours ago
Thanks, I wasn't aware of this list, though it shows very few alternatives (and I'm not saying they exist, but it almost makes it seem more helpless).
nate30480 | 22 hours ago
It also links to other resources but I think https://noai.starlightnet.work/list.html could be mainly what you're looking for.
dkl | 21 hours ago
I have been studying this list for the last few months. One of the main results is that I have gone back to Emacs. This is not my first go-round with it but I have found that nowadays it's much easier to get to a working IDE-like state than ever before. There are alternatives to most of the things I used to use within Emacs itself. I found that once I got over my initial reticence, it provides the kind of human-focused, malleable environment that I didn't realize I was missing.
jzb | 17 hours ago
Oof. Do I have some bad news for you...
dkl | 15 hours ago
I am having trouble understanding the article you linked, but I note that it links to a potential fork of Emacs that would satisfy my requirements, so I think you actually have good news for me.
hsivonen | 5 hours ago
By listing the last “untainted” commit and listing forks, this list seems to encourage using versions that have known security vulnerabilities. Upstream is likely to have LLM-discovered disclosed sec bugs. Is there any vision for security other than running software with known vulnerabilities to accompany this list?
kolja | 4 hours ago
Well, we didn't/couldn't really plan for this, right? Some projects are being forked, and are fixing bugs themselves, but there are just so many projects that are suddenly not available anymore for people wanting their software as ethically sourced as possible. Luckily, a lot of software simply runs locally, so if you only work on your own files you are pretty much safe from attacks and just have to live with a bug here or there. But it's heartbreaking.
hsivonen | 34 minutes ago
A slice of the vulnocapyse firehose arrives on my plate to triage or fix, and from what I’ve seen I’m skeptical of anti-AI forks fixing security bugs themselves.
Empirically, humans haven’t found the bugs that have been present for a long time and that LLMs are finding now, so a project that takes the stance that the method of discovery makes knowledge of the bugs inadmissoble aren’t going to find the bugs independently, so they will eemain unfixed.
Projects that accept awareness of upstream bugs are going to run into ones where there is one reasonable fix (sometimes it really just is flipping the right boolean) is the one that came as an LLM suggestion along with the vulnerability report. Now you can’t prove if you wrote the same fix yourself or if you took the LLM fix. The list is going to go with the least charitable interpretation, so “$FORK_NAME is now slop”.
dustyweb | 22 hours ago
Well, I work there, but we are both a FOSS nonprofit and also have a firm no-gen-AI policy at the Spritely Institute, and do pretty cool decentralized network tech also
bsandro | 19 hours ago
Recently made https://braincoded.software for this - it is not "anti", just "no".
Not much there for now apart from IRC channel, but I'm working on tiny invite-only link aggregator + public git and cvs repositories for members.
stephenr | 15 hours ago
Git and Cvs is certainly a choice.
I understand offering alternatives to Git. I think mercurial deserves a lot more love than it gets.
But why Cvs specifically, over any number of other non-Git alternatives?
bsandro | 15 hours ago
I honestly always liked Mercurial better over git, but sort of gave up at some point and moved all my stuff to git, feels weird to use both now especially since hg doesn't give me any portability perks over git.
CVS is for times and situations when you want to use the version control system from some older system, like old macintosh system7 or 9 (even though @jcs made a very nice tool for it - Amend and even amendhub.com!), or *DOS.
hoistbypetard | 19 hours ago
A few of us who want to avoid it, or at least be very selective about when we talk about it have gathered at imtiredboss.org. We're collecting some lists and links, and have a couple of discussion spaces that are kept free of generative AI.
isagalaev | 18 hours ago
Another one: https://codeberg.org/brib/slopfree-software-index
bbbhltz | 22 hours ago
https://noailist.org/
dsschnau | 22 hours ago
The "No AI" Webring is related - https://baccyflap.com/noai/
[OP] brn | 21 hours ago
Thanks, my blog is a part of it! 😊
sjamaan | 22 hours ago
Similar to the rebel tech alliance, there's the trustworthy technology group. They have a Zulip chat you could join to discuss this further.
th0ma5 | 17 hours ago
This is an academic's collection of materials: https://olivia.science/ai/
jakehamilton | 14 hours ago
I'm putting together a software collective which is "anti-ai" as one of its core principles: https://conflagrate.software/
My idea is for it to be a safe, pleasant space for people to create and collaborate.
Zurga | 22 hours ago
You are aware that the resource you link was made with a site builder that mentions an AI in the second sentence of their product hero? The resource was handmade, but not far away from AI.
[OP] brn | 22 hours ago
Nice catch, but they used it before it was "AI-powered", which speaks a bit to the difficulty I'm facing trying to find alternatives to that.
Zurga | 21 hours ago
Yes, sloppification is everywhere. Things that were once normal products now have an AI...
hwj | 5 hours ago
Recently, I was shopping for an SQLite library for Go. It seems https://github.com/mattn/go-sqlite3/ doesn't use AI.
ItsFinishd | 14 hours ago
Rather than avoiding AI, why not opt for responsible usage of AI? LLMs speed things up, but it comes at a cost of losing correctness, which I do not think one can easily guarantee because of the way LLMs operate - statistically matching rather than logic/intuitive proofs which is what humans are good at. To get the best of both worlds, you want to use AI for smaller, more easily reviewed tasks, then use those (also w/ or w/out AI) to build a more complicated product.
I've no problem with not having to write out a bunch of code that can be written nearly identically by an AI. The QA and validation process is still a human endeavor and where you probably spend most of your time as a dev anyway. Actual coding amounts to no more than 20% of the time.
Nixing a tool like LLM is IMHO a bad idea, if it can help us save time and increase quality of output, then so be it.
seb | 13 hours ago
From the post you're commenting on:
Comments on posts about/for people who don't like and don't want to use AI which amount to "have you considered that AI isn't all bad actually?" are counterproductive and frankly really exhausting.
khairo | 13 hours ago
I just don't want to use it. It doesn't interest me. I don't want to use things that used it either. I don't care if it "works better" or is "more efficient". It's okay to be a luddite.
zesterer | 21 minutes ago
That's a fine enough opinion for you to have in your own space, but that's not what the OP is after and they've not asked for people to persuade them. Respect that.