"GPL has become so complicated that only a programmer with relatively advanced legal knowledge and ability to read and properly understand “legalese” will be able to decipher it to know what they are able and unable to do with it"
Have you ever seen legal terms in easy to understand form in any software, service or agreements?
It's a failure because AI beat the FSF to the punch: it made software effectively free, as in freedom. You can have whatever software you like, now. The FSF never came close to that.
You can still create free software, and use free software, and even charge money for support for free software. The relative success of commercial software and the rise of AI- assisted software doesn't affect the idea and practice of free software. If you don't like that AI companies train on the code, well, the whole point was that you can't really control what people do with the code you send them other than directly copying it without the same license, because that makes it nonfree.
If your problem is that someone you don't like can benefit from your work, well, that was always the case. How many Russian drones run Linux?
>If you don't like that AI companies train on the code, well, the whole point was that you can't really control what people do with the code you send them other than directly copying it without the same license, because that makes it nonfree.
Which is exactly what they do. LLMs don't respect licenses or attribution. So your assertion that the rise of AI assisted software doesn't affect free software is incorrect - it completely disregards and undermines it.
It also adversely affects the ability of people to exercise freedom over that software. LLMs serve as a proprietary middleman controlling the management, generation and distribution of code and humans are constantly being pushed out of the loop, having no direct access to that code, and being unable to learn from it or edit it without an LLM.
Yes you can still create and use free software but it becomes a problem when everyone is only ever training on your code and no human ever contributes anything back.
I think the failure came with the erosion of copyleft like GPL in favor of BSD, MIT, and other "free for all" licenses that companies exploited to the detriment of the community instead of being forced to participate as equals.
I can’t think of an example of what you are talking about. When those have been exploited, it was generally to the detriment of a business built by the core team. When communities were split it’s generally been due to non-free licenses trying to prevent that.
The failure comes exactly from GPL because it wants to tie well-meaning people to a hard contract which they are generally not willing to agree with.
Sometimes people just want to make a little money on the side and not conquer the world. If a BSD open source project can help people sell a proprietary product, there's now a new stream of income in the community that could be partially reinvested in the project's core. GPL is excessively ideological and only makes sense for a narrow set of scenarios.
Please note that I'm not against GPL at all. Without it, I don't think we would have Linux and many other projects. What I'm against is promoting GPL to every scenario, which essentially kneecaps the community to protect against a hypothetical greedy large corporate entity.
GPL shouldn't be the default license when someone thinks of open sourcing their project. They should first consider MIT/BSD/etc and only pick GPL for reasons that are specific to that project or industry.
This rant seems to be all over the place. It mostly seems to be the usual BSD is more free than GPL, but then he comes back that Redhat source is only available to people who paid for the binary (the only way to get the binary). You wouldn't even be entitle to the source under the BSD license.
Maybe I read some of it wrong, but it was painful to get through with the constant telling me how bad things are before telling me what those things are. I truly despise this writing style. Set things up first, then discuss them, please. Then come to the conclusion of how bad it is with the reader.
> As soon as you leave this club and want to pull code from a GPL-licensed project into your MIT-licensed project, well, sorry, too bad for you.
So what? The GPL seems like a decently fair way of saying that "I'll make a contribution, as long as everyone else using it keeps making that same open contribution".
Personally, if I wanted to release open source code, I wouldn't want it to fall under an MIT license, because then some corp could simply take it and not make the same contribution back.
The author claims this "harms developers of non-GPL-licensed FOSS software" - give me a break. Does it break their legs? Give them the flu? It doesn't harm anyone. Maybe just inconveniences them. Cry me a river.
> rewrites of coreutils to be faster and cleaner (e.g. Busybox or various Rust projects)
Note that busybox is under the GPL.
As someone involved with coreutils, I can’t say anything specific without examples, but speaking generally I would caution against putting too much weight into benchmarks.
"Free Software" is objectively a failure in practice and economically in 2026 in three simple reasons:
1) The GPL (and many other permissible software licenses) are not enforceable unless the injured party has a lot of money (in the millions).
2) To 90% of people, it meant "Its free as in $0, I don't need to pay". With free loaders and increased AI slop.
3) It has been completely been trampled all over in the AI era.
The FSF has no economic solution to the "maintenance and support" for human developers. When AI gets involved and mass discovers vulnerabilities in free and open source software, it and has left (unpaid) volunteers to clean up the mess.
Stallman doesn't care and doesn't need to either. (Makes money via speaking fees instead of building free software) and it has created a culture of entitlement where new projects now have to compete against hundreds of forks assisted with AI and has made it worse for everyone else who is not hired to maintain a large open source project that is heavily funded such as Linux.
The license issue failed from the start, and AI models were the final nail in the coffin for "Free Software".
> In this post, I use the term "Free Software" to mean software licensed with Copyleft licenses, and to explicitly differentiate it from permissively-licensed "Open Source" software (which is what the Free Software Foundation and Co. also do).
That's not what "FSF and Co." do at all. I'd expect someone writing an article about whether Free Software failed or not to understand that if they want to be considered seriously.
16 comments
[ 4.5 ms ] story [ 35.7 ms ] threadHave you ever seen legal terms in easy to understand form in any software, service or agreements?
If your problem is that someone you don't like can benefit from your work, well, that was always the case. How many Russian drones run Linux?
Which is exactly what they do. LLMs don't respect licenses or attribution. So your assertion that the rise of AI assisted software doesn't affect free software is incorrect - it completely disregards and undermines it.
It also adversely affects the ability of people to exercise freedom over that software. LLMs serve as a proprietary middleman controlling the management, generation and distribution of code and humans are constantly being pushed out of the loop, having no direct access to that code, and being unable to learn from it or edit it without an LLM.
Yes you can still create and use free software but it becomes a problem when everyone is only ever training on your code and no human ever contributes anything back.
Sometimes people just want to make a little money on the side and not conquer the world. If a BSD open source project can help people sell a proprietary product, there's now a new stream of income in the community that could be partially reinvested in the project's core. GPL is excessively ideological and only makes sense for a narrow set of scenarios.
Please note that I'm not against GPL at all. Without it, I don't think we would have Linux and many other projects. What I'm against is promoting GPL to every scenario, which essentially kneecaps the community to protect against a hypothetical greedy large corporate entity.
GPL shouldn't be the default license when someone thinks of open sourcing their project. They should first consider MIT/BSD/etc and only pick GPL for reasons that are specific to that project or industry.
Maybe I read some of it wrong, but it was painful to get through with the constant telling me how bad things are before telling me what those things are. I truly despise this writing style. Set things up first, then discuss them, please. Then come to the conclusion of how bad it is with the reader.
> As soon as you leave this club and want to pull code from a GPL-licensed project into your MIT-licensed project, well, sorry, too bad for you.
So what? The GPL seems like a decently fair way of saying that "I'll make a contribution, as long as everyone else using it keeps making that same open contribution".
Personally, if I wanted to release open source code, I wouldn't want it to fall under an MIT license, because then some corp could simply take it and not make the same contribution back.
The author claims this "harms developers of non-GPL-licensed FOSS software" - give me a break. Does it break their legs? Give them the flu? It doesn't harm anyone. Maybe just inconveniences them. Cry me a river.
Note that busybox is under the GPL.
As someone involved with coreutils, I can’t say anything specific without examples, but speaking generally I would caution against putting too much weight into benchmarks.
1) The GPL (and many other permissible software licenses) are not enforceable unless the injured party has a lot of money (in the millions).
2) To 90% of people, it meant "Its free as in $0, I don't need to pay". With free loaders and increased AI slop.
3) It has been completely been trampled all over in the AI era.
The FSF has no economic solution to the "maintenance and support" for human developers. When AI gets involved and mass discovers vulnerabilities in free and open source software, it and has left (unpaid) volunteers to clean up the mess.
Stallman doesn't care and doesn't need to either. (Makes money via speaking fees instead of building free software) and it has created a culture of entitlement where new projects now have to compete against hundreds of forks assisted with AI and has made it worse for everyone else who is not hired to maintain a large open source project that is heavily funded such as Linux.
The license issue failed from the start, and AI models were the final nail in the coffin for "Free Software".
That's not what "FSF and Co." do at all. I'd expect someone writing an article about whether Free Software failed or not to understand that if they want to be considered seriously.