I have been thinking on this a lot.
I know ai certainly reduced the number of available development roles, but concerns me is that there is no value right now in software. Not because it got cheaper (we don't know how much of that is true yet) but because we lost the sense of competitiveness.
The most trouble someone has to make a simple version of a competitor feature is to record a video or describe it well enough.
Note, I am not saying clones are good. Time goes on and something gives the feeling of "this was totally ai site" although sometimes it is hard to say what exactly but sometimes, the speed of animation, the kind of bugs and small behaviour quirks.
Let's stop beating around the bush here. The issue is that Asahi people do not want their open source code to be used by people they don't like. This is essentially a knee jerk reaction to Omarchy, an open source distro. GPL is no solution for this. They need a new license with their made up restrictions baked in.
Very true. At this point it's too late though and nobody is going to care about Asahi in a year from now compared to the other projects that have forked it in the last 6 months and actually added compatibility for latest gen macs.
If the LLM has ‘seen’ the GPL-licensed code, it is debatable whether a reimplementation by that LLM is from scratch.
It doesn’t help the “yes it is; the LLM learned from reading that code, but didn’t copy it” camp that there have been cases where an LLM output blocks of code that are very similar to parts of GPL-licensed source code.
Question for copyright to apply then is whether those blocks are copyrightable and if so, whether the ‘copied’ text is similar enough to the source text.
But a lot of these reverse engineered projects claim “clean lab.” If that LLM was only given documents and didn’t have web search or access to the copy-left repo, cloning it is reverse engineering according to the courts.
Even if that model has been trained on that copy left code.
> If that LLM […] didn’t have web search or access to the copy-left repo, […]
> Even if that model has been trained on that copy left code
How can an LLM be trained on code that it doesn’t have access to?
> cloning it is reverse engineering according to the courts.
What courts? Has there been any court case on this w.r.t. LLMs?
AFAIK, learning from copyrighted material is fine, copying copyrighted material (with some exceptions) isn’t. Where one ends and the other starts is a gray area, but at the far end of that scale lies learning by head and then reproducing verbatim. That definitely is copyright infringement.
Also, claiming “clean lab” isn’t worth much. Question is whether that holds up in court.
I would guess/hope/think one need not go as far as Phoenix technologies went to make a BIOS (https://en.wikipedia.org/wiki/Clean-room_design), but I can see why they did things that way to be on the safe side.
By that logic, anybody who's learned operating systems development via the book Operating Systems: Design and Implementation (Tanenbaum and Woodhull) would be legally unable to write an operating system, ever. That book had Minix source code, which at the time was, nevermind copyleft, proprietary.
AIUI this includes Linus Torvalds, and would have affected Linux. Which, by the way, is heavily written by LLM these days, including by Linus himself using Anthropic's tools.
An LLM reading some code and learning from it (at any point in time, in training or as used through agentic programming tools) is in this regard no different from an human being doing the same thing.
Nothing stops me from reading Asahi's source code and then writing my own from what I learned. My agent can do the same thing faster and better. It would also be less likely to resemble Asahi's, as an LLM has seen and learned from a far larger body of code than I have in my entire 30+ years of programming.
If anything, code is becoming no longer a product but a commodity, which brings us closer to the sort of freedom GNU and the FSF aimed for; Something to celebrate.
No, there's no way the genie is going back into the bottle. The past is not coming back. Reality is best accepted and worked with, not fought.
Much of this is far from settled law. All it takes is one law declaring code produced by LLMs to be a derivative work of training data for the entire house of cards to collapse.
Not saying this is likely, but it's possible, and the combined market power of many industries dwarfs the market power of the fledgeling LLM industry, so it's not necessarily unlikely, either.
>Much of this is far from settled law. All it takes is one law declaring code produced by LLMs to be a derivative work of training data for the entire house of cards to collapse.
This is, realistically, not happening. We aren't going back to the stone age because of some dumb judge.
Especially not worldwide. Even if some country is left behind momentarily, this won't prevent other countries from continuing to push the frontier forward.
The site guidelines require using the title from the article, and I wanted to emphasize the license change for this post, so I looked for an article that was specific to that point. I'm sorry you don't like the source. I use an ad-blocker so I did not notice any ads.
Are you implying it's impossible to submit a story with a headline specific to an issue that was originally mentioned in a different article with a different/unrelated title? Effectively making it impossible to bring notice to a single issue? I don't interpret the guidelines that way.
I wonder what's so special about Mac audio that necessitates changes to core ALSA.
I also think it's interesting that beamforming is a thing that exists on such a generic device as a laptop. Is that how the mic tries to pick up only things in front of the laptop? Must be!
Might have something to do with how Apple expects the software to model the temperature of the speakers based on the power it has sent them, and limit volume to keep them from being blown. Other laptops do this in hardware/chipset. https://asahilinux.org/2022/11/november-2022-report/
> code was being lifted straight out of our personal forks and published without attribution or lazily rewritten with LLM assistance to obfuscate its origin. The MIT licence explicitly allows this
Um, no... MIT requires that the license text be included. That's attribution, because you normally put your name or the name of your organization in there. You can't just take MIT code and have AI rewrite it without attribution, at least, you need to do some kind of air gap or something, and lawyers might poke holes in that too. I'm sure people are doing it though. But GPL won't save you from that.
But I do understand why they want to move to GPL, to block out scammers who want to make closed-source forks and charge money for them.
Rumor is that the change was made because of Omarchy. There is controversy over the use of AI-written code though I suspect the real reason is political and they don't like the people behind the Omarchy project.
Either way, I don't presume any good faith actor will have an issue with the GPL license. Just seems like pointless posturing. The usual Linux drama.
Omarchy.. That free and completely open source project that has brought tens of thousands of new people to Linux in the last year and has received millions of dollars in sponsorships to just make good project?
Omarchy isn't contributing to Linux development, though. Donating to Omarchy doesn't help put code in the Linux kernel, it just creates more tech debt for their distro.
Viral licenses like the GPL are a shot across the bow for anyone foisting downstream code and laundering it as their own features. Omarchy can still use Asahi's upstream code, but they generally shouldn't take credit for downstream work.
Trivial has to be pretty trivial. Google was found by a jury to have infringed Oracle's copyright on 9 lines of source code (the RangeCheck function). Although that fact did get lost in surrounding rulings and appeals relating to a further 11,500 lines of API declarations. The legal criteria for what's trivial is subjective and not well defined at all. And the legal costs for finding out what is or is not trivial can easily run into millions of dollars.
If I were you, I would not copy even truly "trivial" amounts of code.
I am no lawyer but I am law-curious, so I went looking for more info on rangeCheck. As we all know AI erases liability, but I'm a poor old human still trying to compete with AI and I do have legal liability so I guess I still have to think about what the rules mean, eh?
Here's my thing: Oracle v Google was a fight that ended up being about two specific things, one was whether API-declaring interface code could be a subject of copyright law. Once it was found that interfaces could be copyrighted the lawsuit turned its focus to thing two: whether Google's use was fair use.
The part of copyright law, as written, that protects your ability to copy a 9-line RangeCheck function, is NOT, emphatically NOT fair use. It is convergence, the least-known of all the copyright principles.
Convergence says that if many people not coordinating could be expected to all land on the same results, then logically the activity must not be creative expression and the result must not be possible to copyright. This is why nobody owns math equations like e=mc^2. It's not that no creativity was required to derive the equation, it's that anyone working at it would end up with a result of that form.
In programming this would protect you when you do something like write a parser for a programming language, or implement a common helper function. Anything where you may, through now fault of your own, do work that creates a result similar to work that someone else has already done.
I don't know that any significant court case has ever even had arguments based on convergence even though it's the most practical shield most of us have most of the time.
Doubt it. If you needed an LLM to launder it it's surely not trivial. Trivial is any well known code pattern for example.
"for(int j=0; j<ARRAY_SIZE(a); j++) total += a[j];" - you could copy paste this and more out of leaked Windows source code and it still wouldn't be a copyright violation. But beware that a sum of a large number of trivialities can be not trivial.
I would be worried about transformation via LLM being considered a "trivial transformation". Trivial transformations do not remove the original copyright.
All completely unsettled law at this point. And, as a general guiding principle, you do NOT want to be one of the people who pays millions in legal fees and court costs to find out, whether you are right or wrong.
I have been thinking on this a lot. I know ai certainly reduced the number of available development roles, but concerns me is that there is no value right now in software. Not because it got cheaper (we don't know how much of that is true yet) but because we lost the sense of competitiveness. The most trouble someone has to make a simple version of a competitor feature is to record a video or describe it well enough.
Note, I am not saying clones are good. Time goes on and something gives the feeling of "this was totally ai site" although sometimes it is hard to say what exactly but sometimes, the speed of animation, the kind of bugs and small behaviour quirks.
Discussion on source (411 points, 2 days ago, 387 comments) https://news.ycombinator.com/item?id=50401867
Let's stop beating around the bush here. The issue is that Asahi people do not want their open source code to be used by people they don't like. This is essentially a knee jerk reaction to Omarchy, an open source distro. GPL is no solution for this. They need a new license with their made up restrictions baked in.
Very true. At this point it's too late though and nobody is going to care about Asahi in a year from now compared to the other projects that have forked it in the last 6 months and actually added compatibility for latest gen macs.
> … that have forked it in the last 6 months and actually added compatibility for latest gen macs.
like what ? examples please.
For example MacBook Neo, with GPU acceleration support.
You can try the installer here: https://omarchy.org/#install M4-M6 work is very advanced as well, but not yet end-user installer friendly.
There will still be AI rewrites of their work, but they will no longer give Asahi any attribution.
Or they'll just GPL the rewrite? I don't see how this really deters anyone.
A less permissive license raises the threshold of working with, and makes recreating instead look more attractive.
This has always been true, but recreating has become easier through agentic programming tools, upsetting the balance.
I don't get your point.
"makes recreating instead look more attractive"
They're already recreating and they have zero intention of sharing.
... which matches the definition of lincense laundering and is independent from the original license.
Expand on "license laundering". How is a from-scratch re-implementation connected in any way? There's no code in common.
And the hardware the drivers are for is Apple's, not Asahi's, to begin with.
If the LLM has ‘seen’ the GPL-licensed code, it is debatable whether a reimplementation by that LLM is from scratch.
It doesn’t help the “yes it is; the LLM learned from reading that code, but didn’t copy it” camp that there have been cases where an LLM output blocks of code that are very similar to parts of GPL-licensed source code.
Question for copyright to apply then is whether those blocks are copyrightable and if so, whether the ‘copied’ text is similar enough to the source text.
But a lot of these reverse engineered projects claim “clean lab.” If that LLM was only given documents and didn’t have web search or access to the copy-left repo, cloning it is reverse engineering according to the courts.
Even if that model has been trained on that copy left code.
> If that LLM […] didn’t have web search or access to the copy-left repo, […]
> Even if that model has been trained on that copy left code
How can an LLM be trained on code that it doesn’t have access to?
> cloning it is reverse engineering according to the courts.
What courts? Has there been any court case on this w.r.t. LLMs?
AFAIK, learning from copyrighted material is fine, copying copyrighted material (with some exceptions) isn’t. Where one ends and the other starts is a gray area, but at the far end of that scale lies learning by head and then reproducing verbatim. That definitely is copyright infringement.
Also, claiming “clean lab” isn’t worth much. Question is whether that holds up in court.
I would guess/hope/think one need not go as far as Phoenix technologies went to make a BIOS (https://en.wikipedia.org/wiki/Clean-room_design), but I can see why they did things that way to be on the safe side.
By that logic, anybody who's learned operating systems development via the book Operating Systems: Design and Implementation (Tanenbaum and Woodhull) would be legally unable to write an operating system, ever. That book had Minix source code, which at the time was, nevermind copyleft, proprietary.
AIUI this includes Linus Torvalds, and would have affected Linux. Which, by the way, is heavily written by LLM these days, including by Linus himself using Anthropic's tools.
An LLM reading some code and learning from it (at any point in time, in training or as used through agentic programming tools) is in this regard no different from an human being doing the same thing.
Nothing stops me from reading Asahi's source code and then writing my own from what I learned. My agent can do the same thing faster and better. It would also be less likely to resemble Asahi's, as an LLM has seen and learned from a far larger body of code than I have in my entire 30+ years of programming.
If anything, code is becoming no longer a product but a commodity, which brings us closer to the sort of freedom GNU and the FSF aimed for; Something to celebrate.
No, there's no way the genie is going back into the bottle. The past is not coming back. Reality is best accepted and worked with, not fought.
Much of this is far from settled law. All it takes is one law declaring code produced by LLMs to be a derivative work of training data for the entire house of cards to collapse.
Not saying this is likely, but it's possible, and the combined market power of many industries dwarfs the market power of the fledgeling LLM industry, so it's not necessarily unlikely, either.
In other words, it's a significant risk.
>Much of this is far from settled law. All it takes is one law declaring code produced by LLMs to be a derivative work of training data for the entire house of cards to collapse.
This is, realistically, not happening. We aren't going back to the stone age because of some dumb judge.
Especially not worldwide. Even if some country is left behind momentarily, this won't prevent other countries from continuing to push the frontier forward.
When did AI ever give attribution?
This is a terrible source, riddled with ads and unreadable.
Use the correct source instead: https://asahilinux.org/2026/11/progress-report-7-3/
The site guidelines require using the title from the article, and I wanted to emphasize the license change for this post, so I looked for an article that was specific to that point. I'm sorry you don't like the source. I use an ad-blocker so I did not notice any ads.
> The site guidelines require using the title from the article
They also require using the official source for submissions. [0]
>> Please submit the original source. If a post reports on something found on another site, submit the latter.
[0] https://news.ycombinator.com/newsguidelines.html
Are you implying it's impossible to submit a story with a headline specific to an issue that was originally mentioned in a different article with a different/unrelated title? Effectively making it impossible to bring notice to a single issue? I don't interpret the guidelines that way.
I wonder what's so special about Mac audio that necessitates changes to core ALSA.
I also think it's interesting that beamforming is a thing that exists on such a generic device as a laptop. Is that how the mic tries to pick up only things in front of the laptop? Must be!
Might have something to do with how Apple expects the software to model the temperature of the speakers based on the power it has sent them, and limit volume to keep them from being blown. Other laptops do this in hardware/chipset. https://asahilinux.org/2022/11/november-2022-report/
> code was being lifted straight out of our personal forks and published without attribution or lazily rewritten with LLM assistance to obfuscate its origin. The MIT licence explicitly allows this
Um, no... MIT requires that the license text be included. That's attribution, because you normally put your name or the name of your organization in there. You can't just take MIT code and have AI rewrite it without attribution, at least, you need to do some kind of air gap or something, and lawyers might poke holes in that too. I'm sure people are doing it though. But GPL won't save you from that.
But I do understand why they want to move to GPL, to block out scammers who want to make closed-source forks and charge money for them.
scammers who want to make closed-source forks and charge money for them
I've yet to see evidence that this is real.
Amazon with ElasticSearch?
Sorry, I meant for Asahi specifically.
Rumor is that the change was made because of Omarchy. There is controversy over the use of AI-written code though I suspect the real reason is political and they don't like the people behind the Omarchy project.
Either way, I don't presume any good faith actor will have an issue with the GPL license. Just seems like pointless posturing. The usual Linux drama.
Omarchy.. That free and completely open source project that has brought tens of thousands of new people to Linux in the last year and has received millions of dollars in sponsorships to just make good project?
Omarchy isn't contributing to Linux development, though. Donating to Omarchy doesn't help put code in the Linux kernel, it just creates more tech debt for their distro.
Viral licenses like the GPL are a shot across the bow for anyone foisting downstream code and laundering it as their own features. Omarchy can still use Asahi's upstream code, but they generally shouldn't take credit for downstream work.
If the code is trivial, not copyrightable, you do not have to follow the license.
Trivial has to be pretty trivial. Google was found by a jury to have infringed Oracle's copyright on 9 lines of source code (the RangeCheck function). Although that fact did get lost in surrounding rulings and appeals relating to a further 11,500 lines of API declarations. The legal criteria for what's trivial is subjective and not well defined at all. And the legal costs for finding out what is or is not trivial can easily run into millions of dollars.
If I were you, I would not copy even truly "trivial" amounts of code.
I am no lawyer but I am law-curious, so I went looking for more info on rangeCheck. As we all know AI erases liability, but I'm a poor old human still trying to compete with AI and I do have legal liability so I guess I still have to think about what the rules mean, eh?
Here's my thing: Oracle v Google was a fight that ended up being about two specific things, one was whether API-declaring interface code could be a subject of copyright law. Once it was found that interfaces could be copyrighted the lawsuit turned its focus to thing two: whether Google's use was fair use.
The part of copyright law, as written, that protects your ability to copy a 9-line RangeCheck function, is NOT, emphatically NOT fair use. It is convergence, the least-known of all the copyright principles.
Convergence says that if many people not coordinating could be expected to all land on the same results, then logically the activity must not be creative expression and the result must not be possible to copyright. This is why nobody owns math equations like e=mc^2. It's not that no creativity was required to derive the equation, it's that anyone working at it would end up with a result of that form.
In programming this would protect you when you do something like write a parser for a programming language, or implement a common helper function. Anything where you may, through now fault of your own, do work that creates a result similar to work that someone else has already done.
I don't know that any significant court case has ever even had arguments based on convergence even though it's the most practical shield most of us have most of the time.
Doesn't the concept of what is trivial shift with what is possible with LLMs?
Doubt it. If you needed an LLM to launder it it's surely not trivial. Trivial is any well known code pattern for example.
"for(int j=0; j<ARRAY_SIZE(a); j++) total += a[j];" - you could copy paste this and more out of leaked Windows source code and it still wouldn't be a copyright violation. But beware that a sum of a large number of trivialities can be not trivial.
I would be worried about transformation via LLM being considered a "trivial transformation". Trivial transformations do not remove the original copyright.
All completely unsettled law at this point. And, as a general guiding principle, you do NOT want to be one of the people who pays millions in legal fees and court costs to find out, whether you are right or wrong.