Interesting to watch the rapid fragmentation of FOSS into “hard” anti-AI, AI-neutral-to-discouraged, AI-supplemented, and AI-driven camps, akin to the early GPL vs BSD wars. Suddenly I don’t see as many complaints about licensing, and some of the die-hard GPL stans are silently dropping that position as Linux adopts AI-supplemented development.
On one hand it’s unfortunate, because it hypothetically dilutes developer efforts, but on the other hand the most extreme groups already had incompatible personalities and weren’t working well together. This self-sorting may actually result in a better cadence and less energy wasted on arguing.
The downside is that a lot of loud and annoying people seem to be converging on BSD land. I am AI-nonaligned (I just care about Quality) but I’m very much in favor of OpenBSD’s heads-down focused dev culture. I hope that doesn’t change.
The reason for NetBSD rejecting any use of AI is the concern that a future legal case will rule that the output of LLMs trained on GPL code is itself subject to the GPL due to the viral clause in the licence.
I think it's a good thing. In a couple of years we'll have a clear way to measure the true productivity and impact (or lack thereof) of AI on software dev. It's the perfect "natural experiment".
We have ways to measure productivity now. AI doesn't change what to measure unless you just want to flex. Every time I see "100k lines of code, all AI," I think they probably could have done the same thing with a fraction of that. Anytime productivity is measured that way, you will always fail. As a CTO, my mantra was "I pay for outcomes, not hours." AI coding tools are just that, tools. My outlook on that hasn't changed. I enjoy working with LLM's especially locally, to handle quite a bit of the heavy lifting. I'm not paying for tokens, I'm paying for outcomes.
projected ideology doesn't really match reality, so not really.
detecting AI written code without a minder chained to your arm watching your every movement is still a topic of research...
...and as someone who has been to the last 10 miscellaneous FSF/GNU celebrations I can tell you that community has just as many license cheaters and scoundrels as the rest of the computing world, they just tend to bias towards the good stuff, similarly to how every microsoft employee isn't required to drink the blood of infants (it just looks good on quarterlies.)
No. If an open source project allows small amounts of LLM slop to please their foundation donors, is all productivity in that project due to LLMs? Of course not.
If 10 of 20 developers work for corporations and have to pretend that their code is AI written to please the corporations, is that AI productivity?
The advertising and social pressure machinery is so AI heavy right now because most techies are cowards and only speak up on issues (including DEI as we see since Trump) if HR tells them to.
Then of course many projects like Linux or git are entrenched. You can't copy Unix and bitkeeper like Linus did and be immediately successful, because the whole culture is against you (unlike in the 1990s, when the culture was anti-Microsoft).
So yeah, if Gen-Alpha declares OpenAI and Anthropic the new Microsoft and builds up a new counter-culture, then in 10 years you might have some clue.
Otherwise, we have to stick to the current situation and observe that no software has gotten better since AI.
So if software is developed with AI and that's better then we should discount AI's contribution, but if software is developed without AI is better then obviously it's because AI had no productivity boost?
It’s possible that they have different interpretations of how generated code affects licensing, but I think in general everyone is now abandoning old positions on licensing. It’s an exciting time for those of us who dislike copyright, something is bound to shift soon.
Whatever its stance towards the GPL, Linux has become the default home of GPL land with the exception of the few dedicated souls still working on Hurd. (I hope they keep making steady progress.)
They are not onboard with the innovation that was GPLv3. That innovation was mostly about software patents iirc. But the result was also that big corps liked GPLv3 even less than GPLv2.
On top of that it was going to be very hard (near impossible) to get all devs to agree to a v2->v3 license bump.
I think what you're finding is that people who don't care about low quality of LLM outputs also don't care about licensing. If you're in the faction that supports all of this, you don't see a lot of people around you still caring about licensing or quality.
I am in neither faction, as I already indicated. I’m not an LLM user except for search engine summaries.
I also already mentioned that some formerly GPL die-hards have been jumping ship from Linux to BSDs simply because of the respective stances on AI. That’s enough for me to say this works both ways.
I expect the overall copyright regime to be weakened over time due to the effects of AI. It’s clear that the intent is to replace copyright gatekeeping with model/compute gatekeeping. This is a battle between two groups I greatly dislike, and I don’t really side with either, but there may be a small benefit for the average person who just wants to “download a car”.
You're being pretty, I'm not sure the right word, scaredycat?, about licensing. Licensing doesn't matter unless somebody sues you, which they won't in most cases. The AI era has exposed this truth that you're extremely unlikely to get sued about copyright violation - so unless you're the kind of person who avoids car trips because you could crash, you can ignore it.
> The prospect of being sued because the LLM outputted training code you're not allowed to use, or to relicense, is exciting?
The problem with this is that if someone sues you for this they have to somehow prove that you actually infringed. They can't just point to the code and say "See, mah, infringement!" The legal system does recognize that it is very possible for two people, completely independently, to arrive at a nearly identical solution.
Of course we're talking of code for which that's implausible.
I think in most cases twenty lines of identical code are more than enough for that to be the case (at least if it contains a few non-single-letter identifiers).
I mean, if it uses the exact same identifiers and maybe even comments, then sure, that's pretty good evidence. However, 99.99999999999 percent of the time this will not be the case. As in, the super-super-super-majority of LLM-generated code is customized for the specific use case it's being employed in. As such, claiming any kind of infringement, let alone proving it, will be insanely difficult.
> Why are proprietary software developers entirely unbothered by this?
Because the chance of someone auditing the actual source code is near zero, but for open source software it's near 100% that someone or, worse something aka an AI bot looking for a revenue stream, will have a look at your code. And EVEN IF you get caught stealing by whatever random incident (say, your code repository being 0wned) - it's a matter of a settlement payment. If you are not a company, but an individual person... well, just look at what happened to Aaron Swartz RIP.
And for the WINE/ReactOS/console homebrew folks, it's vital to be able to claim no code contributor ever got exposed, even indirectly, to proprietary code.
On the other hand it could be very expensive if caught. It could be very difficult to remove the code too.
Your point about being caught is valid, but it depends on AIs producing identifiable chunks of code, and that is incompatible with your license. AI training is going to lean towards open source, because its available to train on.
I also think the widespread use of AI will mean accidental copying will become more acceptable. Even as it is, its going to be regarded as an accident.
I heard many cases of LLMs outputting their training data verbatim; they sure don't check that such data can be legally used by you before doing it.
As others said, for closed source software it's a lot harder to detect plagiarism.
And it's in general very hard to detect plagiarized code; there's some product that does it, but it of course has only a partial inventory of the existing open-source code.
I think that some big company does use those products, though.
> Linux, for one, for what I can tell never cared too much about the GPL.
The reason that Linux actually succeeded where BSD didn't is because of the GPL. MIT licensed stuff(if it's any good) just ends up being incorporated into someone else's commercial product, with no contributions.
That's just obviously not true. Check out the commit history on FreeBSD and look at all the contributions from Netflix. Off the top of my head they contributed kTLS and asynchronous sendfile which made it such that a single syscall could send the full contents of a file over a socket even with TLS.
Grepping through the git logs for the "Sponsored by" lines shows these top results:
15925 The FreeBSD Foundation
3738 Netflix
2084 EMC / Isilon Storage Division
1922 Rubicon Communications, LLC ("Netgate")
1733 Dell EMC Isilon
1550 Chelsio Communications
1446 DARPA, AFRL
1421 Mellanox Technologies
1262 Klara, Inc.
784 Arm Ltd
729 iXsystems, Inc.
707 Solarflare Communications, Inc.
552 Netflix, Inc.
477 Beckhoff Automation GmbH & Co. KG
473 Microsoft
421 Juniper Networks, Inc.
411 DARPA, NAI Labs
386 Stormshield
385 Microsoft OSTC
384 DARPA
375 NetApp, Inc.
358 Innovate UK
355 DARPA, Network Associates Laboratories
317 Yandex LLC
299 NVIDIA Networking
One company that isn't in that list but shows up as an author is Datto which gave us native encryption on ZFS (shout out to Tom for writing that!)
The existence of some contributions to MIT code does not disprove the assertion.
If you're fit to even be in this conversation then you have enough logic to know this, which makes the comment not merely in error but knowingly disingenuous.
What are the equivalent numbers for everything that was ever licenced GPL vs MIT? Neither of us knows but I defy you to even slightly attempt to hint that the world did not benefit from gpl code about a million times more than mit code as a direct result of their fundamental natures.
The assertion I was addressing is that copyfree code "just ends up being incorporated into someone else's commercial product, with no contributions".
My post proved that copyfree code bases actually do receive contributions back from the commercial users of their code. Certainly not 100% of users, but the same is true for the GPL. Every company I've worked for after college has used Linux extensively and none of them have contributed back to Linux in either patches or dollars.
Why do you both keep incorrectly referring to the license as "MIT"? The BSDs are under the various BSD licenses, not the MIT license. If you're trying to refer to the whole category then I would suggest using the words "copyleft" for GPL-style licenses and "copyfree" for BSD-style licenses.
> If you're fit to even be in this conversation then you have enough logic to know this, which makes the comment not merely in error but knowingly disingenuous.
It did not prove any such thing except in the meaningless pedantic technicality way. No reasonably rational person read that comment and thought they meant literally 100% not a single example. The existence of exceptions does not disprove anything. Everything has exceptions.
Numbers prove or disprove it. Which neither of us has. The fact that everything you showed even fits in a little comment like that compared to the scope of all software actually proves what a joke it was to even try to present this as any sort of argument.
"why do you keep saying MIT?" Oh FFS really? MIT, or BSD, is perfectly understandable shorthand for any of the mit/bsd/x11/etc in a general discussion like this. They are all of a class. How do you know? Because no one has attempted to discuss any fine points here that would make any of those any different from any of the others. You have to go out of your way to say they are different, and then cannot say in what way that matters or changes anything. So why do you yet again pretend an insignificant detail in any way affects the thrust of the argument?
Go ahead and substitute "MIT" with "BSD" or any of the specific "various BSD licences". Ok now that you did that, what changed? The argument was valid or invalid before, and the opposite now?
So far I'm the only one here who has provided any numbers. All you and soperj have provided are unsubstantiated claims. Why don't you do some analysis on the "scope of all software" and get back to me? Given the scale, I imagine it will take you at least a couple of years.
Linux success has nothing to do with GPL and everything to do with a lawsuit and timing. Linus himself said if it wasn't for the lawsuit Linux probably would not exist.
Linus doesn't agree. He once said in an interview that making Linux GPL'd was definitely the best thing I ever did, [0] and has said that I love GPLv2. I really think the license has been one of the defining factors in the success of Linux because it enforced that you have to give back. [1]
> The reason that Linux actually succeeded where BSD didn't is because of the GPL.
Linux succeeded where BSD was tied-up in a lawsuit[0] from AT&T and the fear of “picking the wrong horse” w BSD under threat allowed Linux to slide into public consciousness and installs.
Literally nothing is less interesting than yet another "discussion" about the relative merits of software licenses, or generalizations about the kinds of people that prefer license A to license B.
It's more akin to copy-pasting from stack-overflow versus people writing code, or code written by engineers vs code written by outsourced "code monkey" farms/contractors/consultants?
It's not about freedom, which is what GPL vs BSD is fundamentally about.
Except in this case the monkeys are pretty good at coding as long as you have enough bananas (tokens). They can follow directions pretty accurately and work independently for at least short periods and incrementally test the work even if you’re not present onsite monkey-handling all the time. I don’t understand this hard anti-ai stance from open-source groups except as a submission volume issue. If and when nearly the entire professional industry changes to LLM-based code in most cases, what are they going to do to find those relatively few people who are still willing to write everything manually and spend their spare time maintaining these projects?
Especially when it's the corporate plagiarism machine owned by the amoral billionaire class that is profiting wildly off the well-paid legislature's shocking sudden disinterest in protecting the rights and privacy of the people after the capture of the first, second, and third branches alongside the fourth estate.
People are tired of exploding costs, loss of privacy, intrusive technology, pervasive advertising, malignant anti-intellectualism, endless rent-seeking products, and fascist leadership. Adding economic crisis from job loss is a fast track to violence.
I use AI and see it's value and even I'm furious about how it's going.
Perhaps one day we’ll finally understand the value of simpler software that moves ever-closer to “complete” (simple maintenance only), rather than constantly reinventing Zawinsky’s Law.
It’s ok for software to be replaced when something better comes along for a new generation, but what I wouldn’t give for more well-maintained tools that are feature complete.
What is the modern version of Zawinsky’s Law, anyway? Every program expands until it contains a web browser (Electron), notifications, and a chatbot?
These types of projects tend to have a flash of appeal, especially to people used to Linux, because it lessens one degree of unfamiliarity, but they don't seem to stick around for to me obvious reasons.
It's hard to get involved with something like a BSD project because you have to learn a lot and put in enough effort to get committer rights. And then if you want to move the needle you have to do even more work and build consensus without getting overly emotional when people disagree with you. But it is the path I would recommend to have a lasting impact, and I am speaking from experience as a FreeBSD and NetBSD src committer.
Also, pkgsrc is one of the strong points of NetBSD, coupled with pbulk and pkgin. Wrapping that all up in an opinionated installation (like Omarchy recently did to Arch) would be a more survivable effort. But as my friend Larry McVoy would say, it means you don't always get to drive the big red fire truck by yourself.
The BSDs do have comparatively miserable contributor experiences, though I understand why. At least from my very, very limited exposure.
Several years ago now, I ran into a problem with playing audio on NetBSD/sparc. I had an RDI PowerLite laptop, which is 80% of a SPARCstation 5 clone stuffed into a laptop chassis. Part of the 20% different was the audio chip, which was the same AMD ISDN chip used in the SPARCstation 1/2, doing mu-law audio.
Whenever I'd play audio, the laptop would instantly reset. Huh. That's weird. I ended up figuring out that someone had recently massively overhauled the audio driver subsystem in the kernel, and basically the relevant audio driver was missing some mutex operation, and an assert() was panic'ing the kernel.
For a non SMP system, the lock wasn't necessary (IIRC) but the new code was being more rigorous and causing a kernel panic.
The patch was pretty easy, just adding a few locks/unlocks. But I could never figure out what I was supposed to do with the patch after.
I'm probably just stupid. I don't want to imply there's no path to do this, I just never figured it out. I didn't try for long either, since getting 8-bit mu law audio working was just something goofy I was doing for fun.
But for comparison, when I found a bug in the signal handler for 32-bit PowerPC, my memory was that it was pretty straightforward to produce a patch: find the relevant maintainer, open a dialogue, and eventually submit a patch: https://marc.info/?l=linuxppc-embedded&m=160963760120837
Neither fix was ultimately very important, and neither would have impacted more than a couple dozen people on the entire planet. But Linux got a patch, while NetBSD didn't...
Just out of curiosity, why the Sparc laptop? While I'd love to find one today to tinker with, I assume anyone who had one back then was mainly working on sun machine's / unix and needed some at home familiarity.
I bought mine back in 2019, when I was ... 22. I was going through my "I want a real UNIX machine!" phase, and that's why I went through the motions of trying to make NetBSD "genuinely usable" on it.
Nowadays it runs SunOS 4.1 with a port of KDE 1 on it, and netboots Sprite for fun.
Of course, after porting the Sprite kernel to said laptop, I learned why the Linux kernel is littered with comments like this, and why it was never quite the beautiful, elegant, UNIX :tm: machine I imagined it was.
/*
* Are you now convinced that the Swift is one of the
* biggest VLSI abortions of all time? Bravo Fujitsu!
* Fujitsu, the !#?!%$'d up processor people. I bet if
* you examined the microcode of the Swift you'd find
* XXX's all over the place.
*/
Or, maybe, it was even more UNIX-y than I had thought. :-)
(of course, the laptop itself is lovely -- the plastic shell has a nice rounded edge for your wrists, and the keyboard uses proper mechanical switches that feel much nicer than actual Sun keyboards of that era)
Ahh a modern purchase! I get temptations every time I've seen a sparc powered laptop end up on Ebay, but never for a low enough price to justify the the tinkering urges.
> Wrapping that all up in an opinionated installation (like Omarchy recently did to Arch) would be a more survivable effort.
Well we currently have fallout in FreeBSD land because of Jordan Hubbard slopping together Daemonarchy.
Going about as well as you expect when you even loosely tie yourself to a crazy white nationalist like DHH. Even in what is barely a reference to his slop dotfiles masquerading as a distro.
Most residents of Greenland are indigenous Inuits. A boot-licking fascist would be happy to betray his own country and support selling off the territory.
> AlpineBSD is not a fork of Alpine Linux, its package collection is original (though many packages are ultimately based on aports).
> Alpine Linux uses the OpenRC init system. Currently, AlpineBSD uses s6 and we intend to add OpenRC as an alternative.
I'm not sure where exactly the line should be, but that feels like it's not an Alpine* then. Like, if you're starting your package collection separately and use a different service manager, you're further from Alpine Linux than Adélie Linux is.
(To be clear, I'm a big fan of making a new OS from a *BSD core and a new userspace. I'm just quibbling about naming semantics)
I don't like it when unrelated projects squat on the name of some existing project. Trademark isn't just a commercially useful thing, it helps users understand what they're getting and who they're getting it from.
It's shitty for a new project to ride on the coattails of a trusted project with years of reputation. It misleads and confuses folks.
Take inspiration anywhere you like, borrow the code all you like, but don't steal the damned name.
I'm not the OP, but I would personally say that not wanting to work with LLM-generated output (including that submitted via PRs and issues) is reasonable, but not wanting to use LLM-generated code from well-regarded established projects (not vibe-coded slop) is veering closer to tribalism.
Because to me, it's entirely fair to care about the process when you're the one working with it, but if you've never looked at the code and it works well, it seems odd to care how 10% of it was produced.
I still wouldn't quite use the loaded "tribalism" term here myself though, because there are still a few valid reasons to be skeptical of LLM-generated code, but for many people, their hard anti-LLM stance appears to be more of a shibboleth than a principled stance from my perspective.
> Maintainers get their FOSS projects bombed with AI slop PRs and issues, and you call their position to not accept slop a tribalist
Sure, and that's the motte.
This is a direct quote from the linked website:
> Any software whose code, documentation, binaries, or other assets include, in any part, the output of a generative AI tool such as an LLM, is considered “tainted” for our purposes.
Ok so it's "AI free". But why? It doesn't justify the decision but just assumes you will understand the reasoning. I have to say: I don't. For example they are pinning binutils, gcc, and gdb to the last version they know does not have LLM generated code. I assume this means they will not be able to accept any upstream patches, but instead have to take over maintenance of those packages themselves for the forseeable future, effectively meaning development on those will stop for the purposes of AlpineBSD. Seems stupid, to me.
It's a losing battle. Give it time and this attitude will go away by the sheer weight of AI-coded open source they really cannot say no to just because there is no human-coded alternative available that is good enough.
I can respect this. Looks like AI auditing code is fine and I can't imagine why auditing would ever be a con. As for the lack of reasons given I guess they leave that for the users which seems completely fine to me. If you are chimping out over not getting a justification then you should probably do some soul-searching.
> NOTICE: The NetBSD kernel has spotty hardware support, much less so than Linux. You cannot necessarily expect AlpineBSD to work on any random computer you have lying around.
Huh. I feel like the author has their BSDs mixed up! NetBSD's main claim to fame is that it runs on practically everything, even lots of platforms no longer supported by mainstream Linux distributions. The page links to the laptop wiki, which is about AMD64 laptops and their finicky peripherals and driver support. A better place to start for hardware support is https://wiki.netbsd.org/ports/.
> NetBSD's main claim to fame is that it runs on practically everything
That means available number of architectures. NetBSD still support some pretty esoteric stuff, but though I haven't actually counted I would expect Linux to run on a comparable number of architectures.
Now, the kernel will probably boot successfully on a given machine, but whether hardware (i.e. driver) support will be decent enough for everyday use is a completely different matter! For example, NetBSD "works" on this laptop of mine... except for Ethernet card, Wi-Fi card, touchpad, ACPI suspension, external monitors via the docking station, ... :)
I'm trying (for fun) to bridge some of these gaps, but you can see how this isn't appealing to the average user.
The impression I get is that NetBSD has just enough code to get it up and running on a machine but there is not enough development and testing to get it to run well on a machine.
That was true 20 years ago, but not really today. Today's hardware is much more complicated and I wouldn't expect NetBSD to work well on anything but the most basic hardware.
After reading the web page, I am left with some questions unanswered,
* Is it a goal for AlpineBSD to run on native hardware, or is it intended for containers (a major use-case for Alpine itself)?
* For users who want to avoid AI-generated code, why run this rather than netbsd itself?
If the team want to attract interest, they should give more context.
The mission is unclear. The team could address this by stating goals and milestones on the front page.
There is a link to a tool s6 on the page. The description of that tool opens with a circular definition. It would be clearer if it stated why it exists in the opening paragraph. Near the bottom it talks about "One of the design goals of s6-frontend". It would be better to state all of the design goals.
I just reviewed version 11. It's a pain to install, a pain to configure, a pain to add a proper desktop, etc. If you do manage to install a desktop environment, the ones in the NetBSD repos are mostly long out of date.
The community is great. The mailing list tried very hard to help me to install and run CDE.
After 3 days of trying, I gave up. For reference, this was my 2nd or 3rd try.
No, the NetBSD userland is not good, and the docs are not great either. There are an absolute tonne of implicit assumptions of knowledge that newcomers won't have.
The recommended way is "boot from install medium, install over the top". Note that the handbook does not actually say this, and goes on to detail an alternative method called "sysupgrade" which looks much like the OpenBSD method and resembles conventional Linux system upgrade methods.
So I tried that. It irretrievably destroyed my only bare metal install.
To forestall an inevitable comment: it's a testbed, so no, I did not have any backups. It contains -- well, contained -- no data at all.
One thing I'm cautiously watching is NixBSD. With nix you can trivially compile all your software from source and modify it (both of which I do on the Linux-based NixOS). So I think that would cover the main selling point of Gentoo, though naturally it adds the Nix layer that you may either love or hate. The only reason I haven't dived into it yet is my FreeBSD machines are all foundational parts of my infrastructure so I don't want to take a gamble on such a young project. So for now I just load it up into a virtual machine and poke at it longingly from time to time.
I never got the chance to try it, but yeah. I've looked at the Gentoo/FreeBSD and Gentoo/OpenBSD pages a few times just yearning to try them. Portage and all other Gentoo tools being a fairly kernel agnostic toolkit is just the coolest. Even though OpenBSD and NetBSD are in my favorites with Gentoo and Void Linux, I don't think I could ever primarily use anything else unless it had Portage. This rising BSD popularity is welcome news for anyone leaning more UNIX among Linux users.
Neat project. One thing I don't see covered: is everything going to be statically linked like it is on Alpine Linux? Is statically linking libc advisable on a NetBSD kernel? I know (unlike Linux) the FreeBSD kernel ABI can change every major release, but I'm not familiar with NetBSD's approach.
Alpine Linux is not statically linked. It provides a musl environment which makes it easy to statically link your own builds, but software shipped as part of the distro is usually dynamically linked.
Woah, TIL! My Alpine Linux world view has just been shattered. Static linking was frankly *the* selling point of Alpine Linux for me and it turns out I was just plain wrong.
Alpine selling point was that it's small, which was selling people only for people that either don't understand docker layers or can't make Dockerfile correctly.
Alpine’s connection to static linking is mainly that it provides a convenient musl build environment. A fully static binary can run in `FROM scratch`.
> A fully static binary can run in `FROM scratch`.
Not just `FROM scratch` but also in other contexts. For example:
I wrote my own password manager which lives on a USB drive. Since I compiled a static binary, I can execute it on any Linux machine.
Running static binaries in FreeBSD's Linuxulator is easier. The Linuxulator is a separate set of syscalls that get used when you attempt to run a Linux binary, so a static binary will "just work" for the most part (unless the program needs special filesystems like /proc or /sys). A dynamically linked binary means I need to make sure I have any necessary Linux libraries in the proper location and the Linuxulator's `ld` working properly. Updates to these libraries can then make something that used to work, suddenly stop working. Certainly not an insurmountable task, but there's so many fewer places where things can break when you're just running a static binary.
Well, `FROM scratch` implies that if there is something that works with linux syscalls (i.e. Linuxulator) then it generally runs. Unless, like you said, it needs special filesystems.
> which was selling people only for people that either don't understand docker layers or can't make Dockerfile correctly.
I don't think that's fair. Alpine-derived images are generally smaller, all things held equal. Of course if you do everything else right then it will matter less, but a proper multistage Dockerfile should still be smaller than the same thing built out of another distro (most of the time).
Regardless of how you feel about using LLMs or using products that are developed with LLMs, it is a good idea to have an ecosystem of open source projects with differing levels of LLM involvement, including pristine software written entirely by humans.
The pragmatic argument for hard no-LLM projects: if models are to be useful, they need good input. Hand-written software that has been carefully audited is free from the kinds of feedback errors that will eventually weaken LLM code generation. Otherwise, a bad pattern gets copied. In the next generation, this bad pattern is seen a bit more often as stochastic models converge on it and more open source code adopts it. This is reinforced in the next generation and it becomes a common pattern. Then, in the next generation, this pattern is seen all over. This could be inert "non-coding DNA" like weird naming conventions or pointless comments. Or, it could be bad edge cases that become commonly exploitable.
I've been using AI with alpine and had tremendous results. fixing legacy bugs, reducing resources, improving performance. with the right supervision, the results are great.
AI can fix a lot of issues that lingered due to lack of time & patience for proper testing and experimentation.
For example, I had VM instances crashing because of a commercial vendor's agents (telemetry, auth etc). AI identified why the crash occurred, and rewrote the buggy vendor agent with a lighter and more resilient agent.
I brought the issue up with the vendor, and they were stuck with their approach for legacy reasons. Fine! we can both operate happily now. I don't have to use their software.
>A small amount of essential software is pinned to the last version which is not tainted. We pin software as a last resort, preferring to find alternatives or move the package into the tainted repository. The current list of pinned software in “base” and “main” is:
I don't think NetBSD has ever had a reputation for what they mean by "spotty hardware support" here, which is apparently uh... power management on laptops. Their links take you here: https://wiki.netbsd.org/laptops/
while sitting on a cab a few weeks ago, i was watching other large trucks passing through, and i was wondering about what it would be like if these trucks were entirely made by bots, would travelling on roads feel less safer?
when you know that the existing old stuff is made by humans and most humans who are good at their work genuinely put in effort and care in their craft, can we say the same about work generated by clankers?
Really got me wondering, still split on if all this AI hype will be a net good in the long run.
Sure you're more productive and end up spitting out more code, as long as its non critical SaaS stuff, who cares if things go wrong right?
for crucial, mission critical stuff, relying only on AI autonomously, especially when they are still prone to make errors, seems like a very dystopian reality.
Just today claude code hallucinated the first name of a person because in the username the first name was one letter.
I don't think they do. They just care about owning shiny things.
If you care about software (or math, or art, or literature, or any other aspect of human creativity), AI is at best like hiring someone to solve jigsaw puzzles for you.
You complain an the shit future we're building. Someone says fuck you to your dystopia and starts explicitly weeding it out, and you call it backwards.
I'm a hardcore LLM booster but I agree that there's a place for LLM-free projects. In fact human generated code is still superior to even frontier LLM generated code, as long as you're willing take a massive productivity penalty.
I'm of the opinion that improved developer productivity has been a net negative for at least a decade. Regardless of my opinion of LLMs, I don't want it any easier for people to add more antifeatures and pointless changes.
As far as I can tell, either LLMs naturally fail to be viable, we collectively decide to put a stop to it, or we end up in one of several dystopias.
My bet is that we end up building an awful outcome for nearly humans, which means that trying to put a stop to it is a duty, even if it's likely to fail. I feel the same about climate change -- it's likely that we're going to fail to keep much of earth habitable, but it's not morally acceptable to stop trying.
I now feel like the efforts put into trying to stop it are wasted and the people wasting these efforts are smart enough that they could instead be trying to steer the whole thing in a better direction.
Ignorance won't help us. I agree with the negatives, I just feel like this approach is backwards.
> AlpineBSD is an operating system based on the NetBSD kernel with a userspace inspired by Alpine Linux and the objective of making it possible to use as little AI-generated software as possible.
It is really very telling that the AI bros, after flagging bad Google earnings from the front page, need to descend on a tiny new project and flag or downvote all comments that sting.
They are really very insecure people, professionally and socially.
EDIT: It took one second for an AI patrol to downvote this one.
Don't waste your time here. The front page needs patrolling: There is an article that Google AI is cagey that must be flagged. I cannot do it, you need more karma.
Interesting to watch the rapid fragmentation of FOSS into “hard” anti-AI, AI-neutral-to-discouraged, AI-supplemented, and AI-driven camps, akin to the early GPL vs BSD wars. Suddenly I don’t see as many complaints about licensing, and some of the die-hard GPL stans are silently dropping that position as Linux adopts AI-supplemented development.
On one hand it’s unfortunate, because it hypothetically dilutes developer efforts, but on the other hand the most extreme groups already had incompatible personalities and weren’t working well together. This self-sorting may actually result in a better cadence and less energy wasted on arguing.
The downside is that a lot of loud and annoying people seem to be converging on BSD land. I am AI-nonaligned (I just care about Quality) but I’m very much in favor of OpenBSD’s heads-down focused dev culture. I hope that doesn’t change.
The reason for NetBSD rejecting any use of AI is the concern that a future legal case will rule that the output of LLMs trained on GPL code is itself subject to the GPL due to the viral clause in the licence.
I think it's a good thing. In a couple of years we'll have a clear way to measure the true productivity and impact (or lack thereof) of AI on software dev. It's the perfect "natural experiment".
We have ways to measure productivity now. AI doesn't change what to measure unless you just want to flex. Every time I see "100k lines of code, all AI," I think they probably could have done the same thing with a fraction of that. Anytime productivity is measured that way, you will always fail. As a CTO, my mantra was "I pay for outcomes, not hours." AI coding tools are just that, tools. My outlook on that hasn't changed. I enjoy working with LLM's especially locally, to handle quite a bit of the heavy lifting. I'm not paying for tokens, I'm paying for outcomes.
projected ideology doesn't really match reality, so not really.
detecting AI written code without a minder chained to your arm watching your every movement is still a topic of research...
...and as someone who has been to the last 10 miscellaneous FSF/GNU celebrations I can tell you that community has just as many license cheaters and scoundrels as the rest of the computing world, they just tend to bias towards the good stuff, similarly to how every microsoft employee isn't required to drink the blood of infants (it just looks good on quarterlies.)
It doesn't need to be 0% ai vs 100% ai for the comparison to work.
No. If an open source project allows small amounts of LLM slop to please their foundation donors, is all productivity in that project due to LLMs? Of course not.
If 10 of 20 developers work for corporations and have to pretend that their code is AI written to please the corporations, is that AI productivity?
The advertising and social pressure machinery is so AI heavy right now because most techies are cowards and only speak up on issues (including DEI as we see since Trump) if HR tells them to.
Then of course many projects like Linux or git are entrenched. You can't copy Unix and bitkeeper like Linus did and be immediately successful, because the whole culture is against you (unlike in the 1990s, when the culture was anti-Microsoft).
So yeah, if Gen-Alpha declares OpenAI and Anthropic the new Microsoft and builds up a new counter-culture, then in 10 years you might have some clue.
Otherwise, we have to stick to the current situation and observe that no software has gotten better since AI.
So if software is developed with AI and that's better then we should discount AI's contribution, but if software is developed without AI is better then obviously it's because AI had no productivity boost?
Ask fricking LLM to explain the comment to you if you are too dumb.
Projects that decided to welcome AI of course don't care of its possible licensing problems.
Linux, for one, for what I can tell never cared too much about the GPL.
It’s possible that they have different interpretations of how generated code affects licensing, but I think in general everyone is now abandoning old positions on licensing. It’s an exciting time for those of us who dislike copyright, something is bound to shift soon.
Whatever its stance towards the GPL, Linux has become the default home of GPL land with the exception of the few dedicated souls still working on Hurd. (I hope they keep making steady progress.)
Linux (Linus T himself iirc) is very GPLv2.
They are not onboard with the innovation that was GPLv3. That innovation was mostly about software patents iirc. But the result was also that big corps liked GPLv3 even less than GPLv2.
On top of that it was going to be very hard (near impossible) to get all devs to agree to a v2->v3 license bump.
I think what you're finding is that people who don't care about low quality of LLM outputs also don't care about licensing. If you're in the faction that supports all of this, you don't see a lot of people around you still caring about licensing or quality.
I am in neither faction, as I already indicated. I’m not an LLM user except for search engine summaries.
I also already mentioned that some formerly GPL die-hards have been jumping ship from Linux to BSDs simply because of the respective stances on AI. That’s enough for me to say this works both ways.
The prospect of being sued because the LLM outputted training code you're not allowed to use, or to relicense, is exciting?
Or to be forced to change license to large portions of your project?
With Linux I meant the Linux kernel
Copyright is based on laws, and laws can be changed or clarified. Given the amount of overinvestment in AI, I expect them to change. The EU is already making moves in this direction: https://noyb.eu/en/ai-eu-member-states-approve-digital-expro...
I expect the overall copyright regime to be weakened over time due to the effects of AI. It’s clear that the intent is to replace copyright gatekeeping with model/compute gatekeeping. This is a battle between two groups I greatly dislike, and I don’t really side with either, but there may be a small benefit for the average person who just wants to “download a car”.
It could be, but:
- For now the laws are in force, and I don't even know of discussions to change them (that EU proposal only wants to allow training on personal data)
- changing intellectual property rights law would typically require changing international treaties
- however much companies are in love with AI now, I'm not sure they'd support abolishing legal protections on which their business relies
You're being pretty, I'm not sure the right word, scaredycat?, about licensing. Licensing doesn't matter unless somebody sues you, which they won't in most cases. The AI era has exposed this truth that you're extremely unlikely to get sued about copyright violation - so unless you're the kind of person who avoids car trips because you could crash, you can ignore it.
Uhm, you're being pretty, careless?
I'm not sure that that's how legal departments think.
> The prospect of being sued because the LLM outputted training code you're not allowed to use, or to relicense, is exciting?
The problem with this is that if someone sues you for this they have to somehow prove that you actually infringed. They can't just point to the code and say "See, mah, infringement!" The legal system does recognize that it is very possible for two people, completely independently, to arrive at a nearly identical solution.
Of course we're talking of code for which that's implausible.
I think in most cases twenty lines of identical code are more than enough for that to be the case (at least if it contains a few non-single-letter identifiers).
I mean, if it uses the exact same identifiers and maybe even comments, then sure, that's pretty good evidence. However, 99.99999999999 percent of the time this will not be the case. As in, the super-super-super-majority of LLM-generated code is customized for the specific use case it's being employed in. As such, claiming any kind of infringement, let alone proving it, will be insanely difficult.
We're talking of the case when it happens (and it's detected).
I don't know if that's in the 0.000000000001 percent of the time, but it can happen.
How many cases of an LLM outputting training code have you heard of? let alone training code that you are not allowed to use or relicense?
Why are proprietary software developers entirely unbothered by this?
They did in 2023 and the AI thieves built logic to rephrase the plagiarism so it isn't immediately detectable.
Proprietary developers don't care because their code wasn't trained on. Is that a trick question?
They do care now that reverse engineering is possible.
Can you explain further? What was the code affected?
> Why are proprietary software developers entirely unbothered by this?
Because the chance of someone auditing the actual source code is near zero, but for open source software it's near 100% that someone or, worse something aka an AI bot looking for a revenue stream, will have a look at your code. And EVEN IF you get caught stealing by whatever random incident (say, your code repository being 0wned) - it's a matter of a settlement payment. If you are not a company, but an individual person... well, just look at what happened to Aaron Swartz RIP.
And for the WINE/ReactOS/console homebrew folks, it's vital to be able to claim no code contributor ever got exposed, even indirectly, to proprietary code.
On the other hand it could be very expensive if caught. It could be very difficult to remove the code too.
Your point about being caught is valid, but it depends on AIs producing identifiable chunks of code, and that is incompatible with your license. AI training is going to lean towards open source, because its available to train on.
I also think the widespread use of AI will mean accidental copying will become more acceptable. Even as it is, its going to be regarded as an accident.
I heard many cases of LLMs outputting their training data verbatim; they sure don't check that such data can be legally used by you before doing it.
As others said, for closed source software it's a lot harder to detect plagiarism.
And it's in general very hard to detect plagiarized code; there's some product that does it, but it of course has only a partial inventory of the existing open-source code.
I think that some big company does use those products, though.
Interestingly, Linux is GPLv2 and NetBSD is permissive, so I assume AlpineBSD is permissive as well?
> Linux, for one, for what I can tell never cared too much about the GPL.
The reason that Linux actually succeeded where BSD didn't is because of the GPL. MIT licensed stuff(if it's any good) just ends up being incorporated into someone else's commercial product, with no contributions.
> no contributions
That's just obviously not true. Check out the commit history on FreeBSD and look at all the contributions from Netflix. Off the top of my head they contributed kTLS and asynchronous sendfile which made it such that a single syscall could send the full contents of a file over a socket even with TLS.
Grepping through the git logs for the "Sponsored by" lines shows these top results:
One company that isn't in that list but shows up as an author is Datto which gave us native encryption on ZFS (shout out to Tom for writing that!)So interesting to see Darpa represented in different labs. I wouldn't have bother to look for that interesting tidbit otherwise, so thanks for that.
The existence of some contributions to MIT code does not disprove the assertion.
If you're fit to even be in this conversation then you have enough logic to know this, which makes the comment not merely in error but knowingly disingenuous.
What are the equivalent numbers for everything that was ever licenced GPL vs MIT? Neither of us knows but I defy you to even slightly attempt to hint that the world did not benefit from gpl code about a million times more than mit code as a direct result of their fundamental natures.
The assertion I was addressing is that copyfree code "just ends up being incorporated into someone else's commercial product, with no contributions".
My post proved that copyfree code bases actually do receive contributions back from the commercial users of their code. Certainly not 100% of users, but the same is true for the GPL. Every company I've worked for after college has used Linux extensively and none of them have contributed back to Linux in either patches or dollars.
Why do you both keep incorrectly referring to the license as "MIT"? The BSDs are under the various BSD licenses, not the MIT license. If you're trying to refer to the whole category then I would suggest using the words "copyleft" for GPL-style licenses and "copyfree" for BSD-style licenses.
> If you're fit to even be in this conversation then you have enough logic to know this, which makes the comment not merely in error but knowingly disingenuous.
I would like to direct you to the HN guidelines since this is a clear violation of them: https://news.ycombinator.com/newsguidelines.html
It did not prove any such thing except in the meaningless pedantic technicality way. No reasonably rational person read that comment and thought they meant literally 100% not a single example. The existence of exceptions does not disprove anything. Everything has exceptions.
Numbers prove or disprove it. Which neither of us has. The fact that everything you showed even fits in a little comment like that compared to the scope of all software actually proves what a joke it was to even try to present this as any sort of argument.
"why do you keep saying MIT?" Oh FFS really? MIT, or BSD, is perfectly understandable shorthand for any of the mit/bsd/x11/etc in a general discussion like this. They are all of a class. How do you know? Because no one has attempted to discuss any fine points here that would make any of those any different from any of the others. You have to go out of your way to say they are different, and then cannot say in what way that matters or changes anything. So why do you yet again pretend an insignificant detail in any way affects the thrust of the argument?
Go ahead and substitute "MIT" with "BSD" or any of the specific "various BSD licences". Ok now that you did that, what changed? The argument was valid or invalid before, and the opposite now?
So far I'm the only one here who has provided any numbers. All you and soperj have provided are unsubstantiated claims. Why don't you do some analysis on the "scope of all software" and get back to me? Given the scale, I imagine it will take you at least a couple of years.
Linux success has nothing to do with GPL and everything to do with a lawsuit and timing. Linus himself said if it wasn't for the lawsuit Linux probably would not exist.
> Linux success has nothing to do with GPL
Linus doesn't agree. He once said in an interview that making Linux GPL'd was definitely the best thing I ever did, [0] and has said that I love GPLv2. I really think the license has been one of the defining factors in the success of Linux because it enforced that you have to give back. [1]
[0] https://web.archive.org/web/20070826212454/http:/www.tlug.jp...
[1] https://www.zdnet.com/article/linus-torvaldss-love-hate-rela...
> The reason that Linux actually succeeded where BSD didn't is because of the GPL.
Linux succeeded where BSD was tied-up in a lawsuit[0] from AT&T and the fear of “picking the wrong horse” w BSD under threat allowed Linux to slide into public consciousness and installs.
[0] https://en.wikipedia.org/wiki/UNIX_System_Laboratories,_Inc.....
BSD was tied up in lawsuits for years, giving Linux a head start.
Also why did you mention MIT licensed, it is obviously BSD licensed.
Literally nothing is less interesting than yet another "discussion" about the relative merits of software licenses, or generalizations about the kinds of people that prefer license A to license B.
I have flagged this comment.
I'll flag yours, because it was a completely senseless flagging.
It wasn't even a discussion about that, I wonder if you read the two lines of the comment.
> akin to the early GPL vs BSD wars
It's more akin to copy-pasting from stack-overflow versus people writing code, or code written by engineers vs code written by outsourced "code monkey" farms/contractors/consultants?
It's not about freedom, which is what GPL vs BSD is fundamentally about.
Most of what you use is built by "code monkeys" and is on the way to being vibe coded.
Yeah. We noticed. We've been complaining about the quality for a while.
I wasn’t referring to the specifics of the disagreement, I meant the level of vitriol.
Except in this case the monkeys are pretty good at coding as long as you have enough bananas (tokens). They can follow directions pretty accurately and work independently for at least short periods and incrementally test the work even if you’re not present onsite monkey-handling all the time. I don’t understand this hard anti-ai stance from open-source groups except as a submission volume issue. If and when nearly the entire professional industry changes to LLM-based code in most cases, what are they going to do to find those relatively few people who are still willing to write everything manually and spend their spare time maintaining these projects?
>I don’t understand this hard anti-ai stance from open-source groups except as a submission volume issue.
plagiarism machine can not be trusted.
therefore plagiarism machine must not be allowed to make decisions.
Especially when it's the corporate plagiarism machine owned by the amoral billionaire class that is profiting wildly off the well-paid legislature's shocking sudden disinterest in protecting the rights and privacy of the people after the capture of the first, second, and third branches alongside the fourth estate.
People are tired of exploding costs, loss of privacy, intrusive technology, pervasive advertising, malignant anti-intellectualism, endless rent-seeking products, and fascist leadership. Adding economic crisis from job loss is a fast track to violence.
I use AI and see it's value and even I'm furious about how it's going.
Perhaps one day we’ll finally understand the value of simpler software that moves ever-closer to “complete” (simple maintenance only), rather than constantly reinventing Zawinsky’s Law.
It’s ok for software to be replaced when something better comes along for a new generation, but what I wouldn’t give for more well-maintained tools that are feature complete.
What is the modern version of Zawinsky’s Law, anyway? Every program expands until it contains a web browser (Electron), notifications, and a chatbot?
These types of projects tend to have a flash of appeal, especially to people used to Linux, because it lessens one degree of unfamiliarity, but they don't seem to stick around for to me obvious reasons.
It's hard to get involved with something like a BSD project because you have to learn a lot and put in enough effort to get committer rights. And then if you want to move the needle you have to do even more work and build consensus without getting overly emotional when people disagree with you. But it is the path I would recommend to have a lasting impact, and I am speaking from experience as a FreeBSD and NetBSD src committer.
Also, pkgsrc is one of the strong points of NetBSD, coupled with pbulk and pkgin. Wrapping that all up in an opinionated installation (like Omarchy recently did to Arch) would be a more survivable effort. But as my friend Larry McVoy would say, it means you don't always get to drive the big red fire truck by yourself.
The BSDs do have comparatively miserable contributor experiences, though I understand why. At least from my very, very limited exposure.
Several years ago now, I ran into a problem with playing audio on NetBSD/sparc. I had an RDI PowerLite laptop, which is 80% of a SPARCstation 5 clone stuffed into a laptop chassis. Part of the 20% different was the audio chip, which was the same AMD ISDN chip used in the SPARCstation 1/2, doing mu-law audio.
Whenever I'd play audio, the laptop would instantly reset. Huh. That's weird. I ended up figuring out that someone had recently massively overhauled the audio driver subsystem in the kernel, and basically the relevant audio driver was missing some mutex operation, and an assert() was panic'ing the kernel.
For a non SMP system, the lock wasn't necessary (IIRC) but the new code was being more rigorous and causing a kernel panic.
The patch was pretty easy, just adding a few locks/unlocks. But I could never figure out what I was supposed to do with the patch after.
I'm probably just stupid. I don't want to imply there's no path to do this, I just never figured it out. I didn't try for long either, since getting 8-bit mu law audio working was just something goofy I was doing for fun.
But for comparison, when I found a bug in the signal handler for 32-bit PowerPC, my memory was that it was pretty straightforward to produce a patch: find the relevant maintainer, open a dialogue, and eventually submit a patch: https://marc.info/?l=linuxppc-embedded&m=160963760120837
Neither fix was ultimately very important, and neither would have impacted more than a couple dozen people on the entire planet. But Linux got a patch, while NetBSD didn't...
> But I could never figure out what I was supposed to do with the patch after.
Send an email to the dev list with a patch. Same as Linux, Git itself, and a lot of other projects.
>I had an RDI PowerLite laptop
Just out of curiosity, why the Sparc laptop? While I'd love to find one today to tinker with, I assume anyone who had one back then was mainly working on sun machine's / unix and needed some at home familiarity.
I bought mine back in 2019, when I was ... 22. I was going through my "I want a real UNIX machine!" phase, and that's why I went through the motions of trying to make NetBSD "genuinely usable" on it.
Nowadays it runs SunOS 4.1 with a port of KDE 1 on it, and netboots Sprite for fun.
Of course, after porting the Sprite kernel to said laptop, I learned why the Linux kernel is littered with comments like this, and why it was never quite the beautiful, elegant, UNIX :tm: machine I imagined it was.
Or, maybe, it was even more UNIX-y than I had thought. :-)(of course, the laptop itself is lovely -- the plastic shell has a nice rounded edge for your wrists, and the keyboard uses proper mechanical switches that feel much nicer than actual Sun keyboards of that era)
Ahh a modern purchase! I get temptations every time I've seen a sparc powered laptop end up on Ebay, but never for a low enough price to justify the the tinkering urges.
> Wrapping that all up in an opinionated installation (like Omarchy recently did to Arch) would be a more survivable effort.
Well we currently have fallout in FreeBSD land because of Jordan Hubbard slopping together Daemonarchy.
Going about as well as you expect when you even loosely tie yourself to a crazy white nationalist like DHH. Even in what is barely a reference to his slop dotfiles masquerading as a distro.
As DHH is a Danish guy partly or mostly living in the US, it would be interesting to know this nationalist's view of the whole Greenland thing.
Most residents of Greenland are indigenous Inuits. A boot-licking fascist would be happy to betray his own country and support selling off the territory.
> AlpineBSD is not a fork of Alpine Linux, its package collection is original (though many packages are ultimately based on aports).
> Alpine Linux uses the OpenRC init system. Currently, AlpineBSD uses s6 and we intend to add OpenRC as an alternative.
I'm not sure where exactly the line should be, but that feels like it's not an Alpine* then. Like, if you're starting your package collection separately and use a different service manager, you're further from Alpine Linux than Adélie Linux is.
(To be clear, I'm a big fan of making a new OS from a *BSD core and a new userspace. I'm just quibbling about naming semantics)
I don't like it when unrelated projects squat on the name of some existing project. Trademark isn't just a commercially useful thing, it helps users understand what they're getting and who they're getting it from.
It's shitty for a new project to ride on the coattails of a trusted project with years of reputation. It misleads and confuses folks.
Take inspiration anywhere you like, borrow the code all you like, but don't steal the damned name.
The whole "as little AI code as possible" is tribalism.
I want software that is brilliant and well audited, no matter who or what produced it.
No AI guarantees nothing, especially if it's a C codebase like netbsd, why not also join the rust tribe then?
It's more like the Amish not using certain technology we all take for granted.
I see this much broader: opting out of certain technologies.
As a vegan I opt out of all animal domestication/husbandry and wild catch/hunt.
And I try to stay away from refined sugar and refined starches.
Abortion is also a technology people feel strongly about.
And now we have the use of AI.
Did we also have people that did not want to use an electronic calculator when it just came out? Probably.
What makes their position tribialism and not yours?
I think rational tool use doesn't have to be tribalist. Right tool for the right job.
It becomes tribalism when there is an enemy. Us and them, mac vs pc, AI vs Humans, the left vs the right.
Maintainers get their FOSS projects bombed with AI slop PRs and issues, and you call their position to not accept slop a tribalist position?
I'm not the OP, but I would personally say that not wanting to work with LLM-generated output (including that submitted via PRs and issues) is reasonable, but not wanting to use LLM-generated code from well-regarded established projects (not vibe-coded slop) is veering closer to tribalism.
Because to me, it's entirely fair to care about the process when you're the one working with it, but if you've never looked at the code and it works well, it seems odd to care how 10% of it was produced.
I still wouldn't quite use the loaded "tribalism" term here myself though, because there are still a few valid reasons to be skeptical of LLM-generated code, but for many people, their hard anti-LLM stance appears to be more of a shibboleth than a principled stance from my perspective.
> Maintainers get their FOSS projects bombed with AI slop PRs and issues, and you call their position to not accept slop a tribalist
Sure, and that's the motte.
This is a direct quote from the linked website:
> Any software whose code, documentation, binaries, or other assets include, in any part, the output of a generative AI tool such as an LLM, is considered “tainted” for our purposes.
... and that's the bailey.
Pragmatism is the opposite of tribalism. GP is a pragmatist, therefore not a tribalist.
Tribalism ↔ Pragmatism: prioritising group loyalty and identity versus prioritising practical outcomes and what works.
I think saying all code that was touched by an LLM is tainted is not pragmatic.
> I want software that is brilliant and well audited, no matter who or what produced it.
Not caring about ethics is selfish.
Alpine is also the codename of the first iOS, hence the default root pwd ‘alpine’. https://mashable.com/archive/apple-ios-codenames#9.CLuUuywuq...
Back in my day, we didn't have artificial intelligence. We had actual ignorance. And we liked it that way.
For those who won't notice, this is a gag from the OpenBSD 8.0 release song.
Love OpenBSD, but the song thing is creepy.
Ok so it's "AI free". But why? It doesn't justify the decision but just assumes you will understand the reasoning. I have to say: I don't. For example they are pinning binutils, gcc, and gdb to the last version they know does not have LLM generated code. I assume this means they will not be able to accept any upstream patches, but instead have to take over maintenance of those packages themselves for the forseeable future, effectively meaning development on those will stop for the purposes of AlpineBSD. Seems stupid, to me.
It's a losing battle. Give it time and this attitude will go away by the sheer weight of AI-coded open source they really cannot say no to just because there is no human-coded alternative available that is good enough.
I can respect this. Looks like AI auditing code is fine and I can't imagine why auditing would ever be a con. As for the lack of reasons given I guess they leave that for the users which seems completely fine to me. If you are chimping out over not getting a justification then you should probably do some soul-searching.
> NOTICE: The NetBSD kernel has spotty hardware support, much less so than Linux. You cannot necessarily expect AlpineBSD to work on any random computer you have lying around.
Huh. I feel like the author has their BSDs mixed up! NetBSD's main claim to fame is that it runs on practically everything, even lots of platforms no longer supported by mainstream Linux distributions. The page links to the laptop wiki, which is about AMD64 laptops and their finicky peripherals and driver support. A better place to start for hardware support is https://wiki.netbsd.org/ports/.
> NetBSD's main claim to fame is that it runs on practically everything
That means available number of architectures. NetBSD still support some pretty esoteric stuff, but though I haven't actually counted I would expect Linux to run on a comparable number of architectures.
Now, the kernel will probably boot successfully on a given machine, but whether hardware (i.e. driver) support will be decent enough for everyday use is a completely different matter! For example, NetBSD "works" on this laptop of mine... except for Ethernet card, Wi-Fi card, touchpad, ACPI suspension, external monitors via the docking station, ... :)
I'm trying (for fun) to bridge some of these gaps, but you can see how this isn't appealing to the average user.
The impression I get is that NetBSD has just enough code to get it up and running on a machine but there is not enough development and testing to get it to run well on a machine.
As a NetBSD developer, I wouldn't claim that it works on everything.
Depends on what you mean with "it runs". Does it boot? Probably. Can you use your WiFi adapter or graphics card? Probably not
That was true 20 years ago, but not really today. Today's hardware is much more complicated and I wouldn't expect NetBSD to work well on anything but the most basic hardware.
"runs on practically everything" I take to mean the main CPU and general machine architecture, not every possible peripheral driver one might need.
yeah. it'll boot. want wifi? you need a virtual machine running linux to contain the driver lol
After reading the web page, I am left with some questions unanswered,
* Is it a goal for AlpineBSD to run on native hardware, or is it intended for containers (a major use-case for Alpine itself)?
* For users who want to avoid AI-generated code, why run this rather than netbsd itself?
If the team want to attract interest, they should give more context.
The mission is unclear. The team could address this by stating goals and milestones on the front page.
There is a link to a tool s6 on the page. The description of that tool opens with a circular definition. It would be clearer if it stated why it exists in the opening paragraph. Near the bottom it talks about "One of the design goals of s6-frontend". It would be better to state all of the design goals.
Alpine is really nice host OS for daily drive. Simple and small. Given that Drew is leading it my guess would be it is meant for native hardware.
> For users who want to avoid AI-generated code, why run this rather than netbsd itself?
Alpine userland is nice.
> Alpine userland is nice
NetBSD userland is not?
> NetBSD userland is not?
Honestly, no.
I just reviewed version 11. It's a pain to install, a pain to configure, a pain to add a proper desktop, etc. If you do manage to install a desktop environment, the ones in the NetBSD repos are mostly long out of date.
https://www.theregister.com/os-platforms/2026/08/20/netbsd-1...
The community is great. The mailing list tried very hard to help me to install and run CDE.
After 3 days of trying, I gave up. For reference, this was my 2nd or 3rd try.
No, the NetBSD userland is not good, and the docs are not great either. There are an absolute tonne of implicit assumptions of knowledge that newcomers won't have.
An example:
https://www.netbsd.org/docs/guide/en/chap-upgrading.html
The recommended way is "boot from install medium, install over the top". Note that the handbook does not actually say this, and goes on to detail an alternative method called "sysupgrade" which looks much like the OpenBSD method and resembles conventional Linux system upgrade methods.
So I tried that. It irretrievably destroyed my only bare metal install.
To forestall an inevitable comment: it's a testbed, so no, I did not have any backups. It contains -- well, contained -- no data at all.
> * For users who want to avoid AI-generated code, why run this rather than netbsd itself?
To start a community.
But the NetBSD community already exists?
If this were based on OpenBSD or HardenedBSD I could get into it.
I so wish Gentoo/BSD still was a thing.
One thing I'm cautiously watching is NixBSD. With nix you can trivially compile all your software from source and modify it (both of which I do on the Linux-based NixOS). So I think that would cover the main selling point of Gentoo, though naturally it adds the Nix layer that you may either love or hate. The only reason I haven't dived into it yet is my FreeBSD machines are all foundational parts of my infrastructure so I don't want to take a gamble on such a young project. So for now I just load it up into a virtual machine and poke at it longingly from time to time.
https://github.com/nix-community/nixbsd
I never got the chance to try it, but yeah. I've looked at the Gentoo/FreeBSD and Gentoo/OpenBSD pages a few times just yearning to try them. Portage and all other Gentoo tools being a fairly kernel agnostic toolkit is just the coolest. Even though OpenBSD and NetBSD are in my favorites with Gentoo and Void Linux, I don't think I could ever primarily use anything else unless it had Portage. This rising BSD popularity is welcome news for anyone leaning more UNIX among Linux users.
Neat project. One thing I don't see covered: is everything going to be statically linked like it is on Alpine Linux? Is statically linking libc advisable on a NetBSD kernel? I know (unlike Linux) the FreeBSD kernel ABI can change every major release, but I'm not familiar with NetBSD's approach.
Alpine Linux is not statically linked. It provides a musl environment which makes it easy to statically link your own builds, but software shipped as part of the distro is usually dynamically linked.
Woah, TIL! My Alpine Linux world view has just been shattered. Static linking was frankly *the* selling point of Alpine Linux for me and it turns out I was just plain wrong.
Alpine selling point was that it's small, which was selling people only for people that either don't understand docker layers or can't make Dockerfile correctly.
Alpine’s connection to static linking is mainly that it provides a convenient musl build environment. A fully static binary can run in `FROM scratch`.
> A fully static binary can run in `FROM scratch`.
Not just `FROM scratch` but also in other contexts. For example:
I wrote my own password manager which lives on a USB drive. Since I compiled a static binary, I can execute it on any Linux machine.
Running static binaries in FreeBSD's Linuxulator is easier. The Linuxulator is a separate set of syscalls that get used when you attempt to run a Linux binary, so a static binary will "just work" for the most part (unless the program needs special filesystems like /proc or /sys). A dynamically linked binary means I need to make sure I have any necessary Linux libraries in the proper location and the Linuxulator's `ld` working properly. Updates to these libraries can then make something that used to work, suddenly stop working. Certainly not an insurmountable task, but there's so many fewer places where things can break when you're just running a static binary.
Well, `FROM scratch` implies that if there is something that works with linux syscalls (i.e. Linuxulator) then it generally runs. Unless, like you said, it needs special filesystems.
> which was selling people only for people that either don't understand docker layers or can't make Dockerfile correctly.
I don't think that's fair. Alpine-derived images are generally smaller, all things held equal. Of course if you do everything else right then it will matter less, but a proper multistage Dockerfile should still be smaller than the same thing built out of another distro (most of the time).
NetBSD retains the old ABI as well if anything changes to a syscall.
Regardless of how you feel about using LLMs or using products that are developed with LLMs, it is a good idea to have an ecosystem of open source projects with differing levels of LLM involvement, including pristine software written entirely by humans.
The pragmatic argument for hard no-LLM projects: if models are to be useful, they need good input. Hand-written software that has been carefully audited is free from the kinds of feedback errors that will eventually weaken LLM code generation. Otherwise, a bad pattern gets copied. In the next generation, this bad pattern is seen a bit more often as stochastic models converge on it and more open source code adopts it. This is reinforced in the next generation and it becomes a common pattern. Then, in the next generation, this pattern is seen all over. This could be inert "non-coding DNA" like weird naming conventions or pointless comments. Or, it could be bad edge cases that become commonly exploitable.
Hmm I'm not sure that this project was born out of corcern for the well-being of future LLMs.
This is an argument, anyhow, but it's very far from being the only (pragmatic) one.
I've been using AI with alpine and had tremendous results. fixing legacy bugs, reducing resources, improving performance. with the right supervision, the results are great.
AI can fix a lot of issues that lingered due to lack of time & patience for proper testing and experimentation.
For example, I had VM instances crashing because of a commercial vendor's agents (telemetry, auth etc). AI identified why the crash occurred, and rewrote the buggy vendor agent with a lighter and more resilient agent.
I brought the issue up with the vendor, and they were stuck with their approach for legacy reasons. Fine! we can both operate happily now. I don't have to use their software.
>A small amount of essential software is pinned to the last version which is not tainted. We pin software as a last resort, preferring to find alternatives or move the package into the tainted repository. The current list of pinned software in “base” and “main” is:
>binutils gcc gdb libffi perl xorriso
Good luck replacing those ...
Well, several of those have reasonable alternatives (e.g. clang). They're just, you know, probably also using AI
> NetBSD kernel has spotty hardware support
Oh, how the mighty have fallen. Still remember the times when the meme was that the NetBSD was running on absolutely anything.
yes! "of course it runs netbsd" just checked, it's gone on their website. probably gone for 10 yrs alread, that's roughly when i checked last
I don't think NetBSD has ever had a reputation for what they mean by "spotty hardware support" here, which is apparently uh... power management on laptops. Their links take you here: https://wiki.netbsd.org/laptops/
"Oh, how the mighty have fallen."
This suggests the commenter once thought that NetBSD was mighty
Perhaps he never had such thoughts
If so, then this comment is disingenuous
There is an obvious difference between "hardware (feature) support" and "runs NetBSD"
For example I don't need X11, ACPI, audio or WiFi to "run NetSBD". I want a server not a "desktop"
Great website, great AI policy. The two are probably related.
while sitting on a cab a few weeks ago, i was watching other large trucks passing through, and i was wondering about what it would be like if these trucks were entirely made by bots, would travelling on roads feel less safer? when you know that the existing old stuff is made by humans and most humans who are good at their work genuinely put in effort and care in their craft, can we say the same about work generated by clankers? Really got me wondering, still split on if all this AI hype will be a net good in the long run. Sure you're more productive and end up spitting out more code, as long as its non critical SaaS stuff, who cares if things go wrong right? for crucial, mission critical stuff, relying only on AI autonomously, especially when they are still prone to make errors, seems like a very dystopian reality. Just today claude code hallucinated the first name of a person because in the username the first name was one letter.
This AI policy is so utterly backwards and stupid I don't even have a nicer way of saying it.
You could ask your favorite LLM to come up with a nicer way of saying it.
Just vibecode your own OS no?
Why do vihecoders still care of other people's software?
> Just vibecode your own OS no?
Why stop at just one?
Why do they still care of software at all?
I don't think they do. They just care about owning shiny things.
If you care about software (or math, or art, or literature, or any other aspect of human creativity), AI is at best like hiring someone to solve jigsaw puzzles for you.
Tbh it makes sense if you want a truly LLM-output free project. I can respect it.
Interesting response in light of your previous post, https://news.ycombinator.com/item?id=48197191
You complain an the shit future we're building. Someone says fuck you to your dystopia and starts explicitly weeding it out, and you call it backwards.
I'm a hardcore LLM booster but I agree that there's a place for LLM-free projects. In fact human generated code is still superior to even frontier LLM generated code, as long as you're willing take a massive productivity penalty.
I'm of the opinion that improved developer productivity has been a net negative for at least a decade. Regardless of my opinion of LLMs, I don't want it any easier for people to add more antifeatures and pointless changes.
Gotta give the AIs some new original material to learn from, eh?
I see why you see it as contradiction. I guess we've all been through a lot of contradictory feelings these past one or two years.
I just feel like eliminating LLM output is (1) not feasible (2) not the right solution to the problem.
As far as I can tell, either LLMs naturally fail to be viable, we collectively decide to put a stop to it, or we end up in one of several dystopias.
My bet is that we end up building an awful outcome for nearly humans, which means that trying to put a stop to it is a duty, even if it's likely to fail. I feel the same about climate change -- it's likely that we're going to fail to keep much of earth habitable, but it's not morally acceptable to stop trying.
I now feel like the efforts put into trying to stop it are wasted and the people wasting these efforts are smart enough that they could instead be trying to steer the whole thing in a better direction.
Ignorance won't help us. I agree with the negatives, I just feel like this approach is backwards.
In which direction would you steer a house fire to lead to a better outcome?
There's no good endpoint for people if AI succeeds, it's that simple. It doesn't matter how many people get to roast marshmallows as the place burns.
> AlpineBSD is an operating system based on the NetBSD kernel with a userspace inspired by Alpine Linux and the objective of making it possible to use as little AI-generated software as possible.
Fucking get over yourself.
It is really very telling that the AI bros, after flagging bad Google earnings from the front page, need to descend on a tiny new project and flag or downvote all comments that sting.
They are really very insecure people, professionally and socially.
EDIT: It took one second for an AI patrol to downvote this one.
Like you?
Don't waste your time here. The front page needs patrolling: There is an article that Google AI is cagey that must be flagged. I cannot do it, you need more karma.