I'm really surprised VKD3D works out of the box. So many abstraction layers stacked on top of each other feel like a jenga tower on the brink of falling.
Correct, those are the versions in the coming release.
For those not familiar with OpenBSD, it’s had an accelerated amdgpu driver since 2019, and inteldrm/radeondrm for a few years longer, generally kept up to date with Linux stable throughout that time.
Another interesting story is that on FreeBSD, Nvidia actually provides a port of their proprietary binary driver. The user mode piece does not support cuda, but I successfully ran a Linux emulator container that can run cuda with a FreeBSD kernel. I use it to run llms on FreeBSD.
But that is freebsd, not openbsd. I couldn't imagine it on OpenBSD because the priorities are very different
This is actually really easy, I've had an abliterated Qwen 3.8 FN working on it.
Current Wine just needs a couple of fixes from upstream libxml2 to be able to use the Retail C2R (Click-To-Run) installer, then the apps themselves work fine with a few Direct2D patches (see the effort to get Paint.NET running).
The real issue is that getting activation (even using MAS tooling) working requires a real Windows kernel running for the Warbird-protected activation system, after two weeks I gave up, 99% of my MSO needs are provided by Artcraft now anyway...
For most users it is enough. I know a number of companies that nowadays only provides the basic office365 license to their employees (which disable access to desktop apps) for cost saving reasons. They only provide the full license to a small subset of administrative users who are totally irrelevant in an OpenBSD related discussion as they would use windows or MacOS desktop anyway.
The typical Office installer is a small bootstrapper that requires an Internet connection to download required files, but applications themselves are native and work offline (some features and applications use web tech for UI, of course).
Subscription-based licensing requires periodic check-ins, but perpetual product key-based licensing is available, though compared to the subscription versions, the release cadence for new features is slow.
The online component of traditional apps like Excel works best as a means for sharing documents, with the web UIs useful for reading and light editing, and the same documents fully editable in the native apps.
There was a whole era of mid '90s Microsoft games where they were still competing with "Exit to DOS to get the best game experience" so they went out of their way to make sure you were aware you were running in Windows while playing their games. Aside from the early Age of Empires games I also recall the "Madness" family of racing games had different takes on the same idea from a variety of different developers. I actually just reinstalled Monster Truck Madness 1 the other day and have been working my way through trying to bodge it in to being an acceptable experience for this exact reason. The game itself runs fine but the UI does a lot of weird things that may have worked on Windows 9x with a low-res CRT but are very annoying on a modern system.
I'm having a blast with AoE I on Linux, in this case the Definitive Edition. I don't think it looks like Office though ;) But what is true, it's tricky to get running on WINE. With Proton which I use the configuration works by default though IIRC.
> I doubt much will be merged upstream. Many updates we have can technically
run on other OSes (%gs stubs won't be built on Linux because %gs is correctly set, SIGUSR1 will work but less optimally, etc). However I picked an approach where everything that is done for OpenBSD is just marked as OpenBSD only. I did that purely to simplify my life while I was working on it and clearly mark the changes.
> When I was talking to Wine maintainers a month ago, they were hesitant to accept even patching Wine so it compiles for OpenBSD, mostly because it's another OS they will have to support and keep operational, while the %gs stub fix is something they definitely cannot take upstream. They are very hesitant to take anything that is changing the "emulated" application code.
Yes, unfortunately OpenBSD maintainers will have to keep this running for now. NetBSD is apparently in a similar situation. Perhaps their efforts can be pooled at some point if the patches don’t diverge too much.
"Can't even match GNU's coreutils"? Care to elaborate? Last time I checked OpenBSD is pretty much in line with Posix and there are plenty of us that consider many of the GNU extensions to be anti-features.
I use a Mac, so the CLIs are the BSD versions. I don't have any opinion on which is better, but it's just annoying that it's different from GNU, so a lot of my work ends up being on a Linux VM or server instead. POSIX compliance doesn't seem to buy me anything. I like the BSD license though.
That's a brand new excuse, Apple refused to distribute GPLv3 when GPLv3 was first published, 20ish years ago. How is GPLv3 incompatible with their system volume, yet GPLv2 is still fine?
GPLv3 added language which specifically requires that GPLv3 software embedded in products be distributed with "Installation Information" explaining how modified versions of that software can be installed. It is not at all clear how this would apply to software in an immutable, signed system volume. This language was not present in GPLv2.
You're entitled to your opinion, but that doesn't change the fact that the situation is "Apple can't distribute them", not "Apple doesn't want to distribute them".
Of course, but any script meant for Mac is going to assume Mac core utils, even assume zsh instead of bash. You don't want to be the one person on the team with a non-stock setup.
> The GNU versions work fine in macOS. Apple doesn’t want to distribute them, but that’s mean you can’t use them.
We also test GNU coreutils on macOS. It is a bit tricky since the other maintainers and I don't own any macOS machines for testing. But we seem to do well enough with a shared system that Adélie Linux kindly allows the compile farm [1] to use, along with some additional CI machines.
I'm not trying to. And it's not even like GNU/Linux/whatever is a superset of POSIX and I'm using the extensions. There are mutual differences in basic functionality.
I don't know how compatible my Mac will really be with a Free/OpenBSD machine. Last time I tested this theory, some C code I was building on Mac clang didn't build on FreeBSD clang cause I was unknowingly using some extension enabled by default. But that was about 12 years ago.
"[Doesn't] get the job done"? Well, it gets my job done. You are welcome to use whatever you subjectively prefer (I use GNU's date(1) for a few scripts, but would prefer the English date grammar parsing as a tool of its own), but if you are unable to talk about anything in concrete terms there is not point in trying to discuss in good faith and I will just roll my eyes and move on.
I am not the parent commenter, but I have the same sense that the GNU ones are better.
But the only concrete example I can think of is that `cat -n a b`, if `a` doesn't end in a newline, produces something that is obviously not what the user wanted. GNU cat does it right.
I didn't report it to the openbsd people because I didn't want them to yell at me that `-n` is nonstandard anyway with no "right" behavior, so I should STFU. Which would be what I'd expect from having reported problems in the past.
It is, but an already niche operating system that refuses to leverage the advantages of LLMs is only going to fall further and further behind. I say that as someone who has actively contributed to the BSDs since the 90s and would hate to see that happen.
The slop problem is very real. People are spewing reams of bad code that generates entropy faster than humans or LLMs can clean it up. LLM best practices are not yet obvious, and we're still figuring out how best to leverage them while maintaining code quality and human understanding.
... but LLMs are coming whether we want them or not. Models keep consistently improving, the slop keeps getting better, and human-only development is going to struggle to keep up.
In my day job, I'm driving an enormous number of tokens at Astra+Ultramax solving problems in Lean 4 that previously we never would have attempted simply due to the incredible human engineering cost; we review the high-level lemmas and let it prove the rest.
LLMs are the only solution I've seen over these decades that could finally close the feature/usability gap between mainstream and more niche operating systems -- software support, features, drivers, and other broad usability barriers. All of that has historically been far too much work for a much smaller developer community to reasonably undertake.
Think about how far Linux has/hasn't come in the last 10 years. The user experience is nicer, drivers "just work" more, there's more support for more stuff...but at the end of the day, the actual computing experience is mostly the same today as when I installed Debian back in 2016. It's polish, interoperability, etc that's improved.
Maybe VR takes off and takes over the desktop, and then openBSD can't keep up with that, but I don't think that's likely. So what are they trying to "keep up" with?
The number of OpenBSD contributors, since 2017 is going down. For all the praise the system gets here, the developer team is small. And feature-wise the system is a toy compared to Linux. It's like comparing DigitalOcean with AWS.
Ppl praise the stance of OpenBSD/NetBSD and what-not. That's commendable, but then close the tab, open their spotify and get back to agentic programming for their proprietary company source. But God-forbid if OpenBSD turns anything more than a museum OS, full of purity but with a user-base of five. Because the GPU doesn't work properly, the sound drive is clunjy, wifi might work, etc.
LLMs could port huge amount of drivers to both system with a fraction of the effort. Specs can be implemented easily and properly. Then of course, you need to test and that takes time, but it's technically easier and more ppl could contribute.
I get the feeling, but they're not OpenBSD a favor, if anything else I feel the exact opposite is happening.
In this case, reading the link you're replying under, it's being able to mount a disk formatted with a popular filesystem. The alternative here is waiting for a human to type this relatively "solved" problem, as a procedural requirement, more than anything.
So, it can easily mean "things won't work in this OS". But, that's to be expected with FreeBSD, for legitimately good (from the perspective of FreeBSD) reasons.
Bugfixes, feature implementation, and hardware support will not be able to be added to OpenBSD with the speed that they are added to Linux.
Whether you like this new reality or not, it is an objective fact that LLMs are the silver bullet Fred Brooks told us didn't exist—multiplying development speed by 10x or more—and that if you're not using them you are hobbling yourself. This especially goes for a "security-focused" OS like OpenBSD, because AI is more diligent about finding vulnerabilities and can do so without rest. Meaning that if you're serious about security, you are using AI to harden your software... and if not, someone else is using it to hack you.
I'm sitting here watching astra do in a day what would have taken me at least 3-6 months. The code is verified correct. The model finds, reads, understands, and then integrates the latest literature into its designs and plans. It implements the code, proves it correct, and has come up with its own novel solutions on top of the state of the art. It then refines its implementation to hit the performance goals I give it. This is the reality today.
As one example I can think of, there are a ton of devices that only work on Windows (or Windows and MacOS). The audio interface I have is one example: its "official drivers" are Windows and MacOS only. I know that someone is trying to make Linux drivers for it, using LLMs, because if there's one thing LLMs are very good at, it's RE and figuring out how hardware works. One of my other friends made an accessible equivalent to the Razer app for his Razer headphones because the official app is complete shit.
honestly LLMs could probably write drivers in CompCert C or some other formally verified dialect. having AI discharge proof obligations seems like a terrific use for all this newly-developed mathematical process.
you could have the driver code prove it implements a spec, that there's no UB, that hardware invariants are maintained, that routines take a fixed time to execute.
the formal driver spec must still be written by sober, well-caffeinated humans of course.
Well, they were always slow. But they made a stand
against AI slop, which is great. Look at how AI slop
infiltrated the ruby core team. You even get slop-spammed
down when you try to interact with e. g. Hiroshi Shibata,
who suddenly has ... perfect english language skills. A
few years ago his command over the english language was rather
weak (which is fine by the way, I much prefer broken
english from a real human than AI slop spam text as filter acting in between humans).
If you don't break the law you do indeed fall further behind people who do.
Nothing in that post is against using an LLM. It's all about copyright:
>Since every file in OpenBSD has been continually verified to ensure
correct Copyright, and we've even deleted code which has incorrect
Copyright, the chances of us accepting such new code with such a suspicious
Copyright situation is zero.
Once the current bubble pops we're going to see a lot of litigation over copyright enfringement by LLMs. Currently no one wants to rock the boat because everyone hopes to make a trillion dollars from AI, Oracle included, once they see they can't the knives are going to come out and I wouldn't want to be Linux at that point: https://en.wikipedia.org/wiki/Google_LLC_v._Oracle_America
Also my personal opinion about copyright _in the FOSS domain_ is that I simply dont care. Copyleft as the GNU folks like to call it is just using the copyright laws to achieves certain free software objectives. If that doesnt work anymore there will be another fix
So long as there's big money interests around copyright, I assure you it will not die. And all the predictions about the knives coming out after the AI bubble pops are spot on.
It took a while for the DMCA to happen after Napster. But it indeed happened, and we all know the saying about history repeating itself.
> It took a while for the DMCA to happen after Napster.
By the time they came for Napster technology had already moved on. I can still download entire albums and movies in minutes 2 decades later. Heck, Usenet is still a bastion of piracy and that's been around for 46 years.
The DMCA was passed in 1998. Napster didn't really hit big until 1999.
The 90s young people are familiar with are... way different from the 90s I remember. Their understanding of it is like... if someone asked to AI-gen the decade.
I think it's the complete saturation of the platform with members who make it feel like HN is a platform for people to boast about how many tokens they burnt to rewrite something that already worked fine.
It polarizes people to either join the "the LLM is my god" crowd or the "if AI touches the repo, it's all slop" crowd.
The careful balance is rarely discussed, probably because that's the part that still takes work.
It really depends on the post anymore. Have had a lot of really thoughtful discussions about the nuance that came out of HN convos. Seems like most businesses and people are still just trying to figure out.
I think "careful balance" is the status quo even if people are unaware of it being so.
Agents and LLM-assisted coding currently work within the confines of already working hand-coded OSes, language frameworks, test harnesses and even CLI tooling. Pull one of those out, and agents become unhinged and they lose direction.
Plenty of issues that have been patched recently found using such tools as well. OpenBSD's policy remains largely the same as always: You need to understand the patch you propose and it needs to be maintainable. The link elsewhere in this thread is mainly discussing large patches where it is unclear if the LLM drew too heavily on code elsewhere.
OpenBSD's errata list has recently started growing rapidly, and the scale is a real departure from the past. I have taken to running syspatch much more regularly than in the past. Anyway, it's clear that the project's stance on AI is "as long as reports are detailed and high-quality (and probably from a person with a record of such work) the project will act on it." See slide 17 here: https://events.eurobsdcon.org/media/2026/submissions/ZGDXQR/...
Talked to bluhm@ after that talk in Brussels and the general feeling (from what I can remember from the conversation) so far is that there is a large increase in overall reports (good and bad), but that the severity of what is reported has not really increased. This also mirrors my own experience from reading source-changes@. In essence, it seems like we have known how to build secure software for the last thirty or so years, just that most do not seem to have bothered to do so.
In that thread I suggested he submit it to the ports tree, I'm not sure if he took my advice or came up with the idea on his own. At any rate this is cool.
I never hid the fact that I iterated with AI. It's pretty hard to commit to an approach if you don't even know if it will succeed. And boy oh boy this was a shot in the dark.
I also respect Wine and OpenBSD and their rules to not have any LLM code in the repo/ports. So my approach was to iterate over solutions, find approach that works, ask AI to "just see if my idea even works", learn what/how/why of that approach, generate some supporting manuals for myself (for example how to decode %gs instructions and how to reencode them) and then write my own code based just on decoding tables (and not LLM code).
I did not know how to work with REX/SIB/ModRM, how to detect prefixes, how to parse every single instruction that contains %gs, how offsets work, how to encode jmp rels, why I want to use JMP *0(%RIP) instead of JMP 32rel on return, why I want to leave stub on SIGUSR (or any other signal). But I do know now.
I was incredibly lucky that patching just these %gs instructions were enough to launch Factorio and play, with broken sound, but anyways! Holy molly this was and is so much FUN!
The approach still feels like cheating, and if it feels the same for you, I perfectly understand.
this strikes me as the responsible way to use AI period - test it out, see what works then clean it up until its something you understand & can feel good about.
Yes, treating it as an occasionally flawed/wrong personal tutor (instead of a code genie) seems like a good use of it. Nothing wrong with learning by whatever means necessary IMO.
Yes, he said the same on the mailing list[0] when he asking about a related thing 3 weeks ago:
> I had some success in porting Wine to amd64.OpenBSD. Was able to run a few =
games. But that required in memory patching of %gs commands (LLM's suggesti=
on and 1st implementation, then I re-did it from scratch - I'm not that sma=
rt with amd64 assembly code)
At that time, that LLM built solution was like 3+ weeks old if not more. It's just me and Fabien for now though. Rest are working on helping us turn it into a nice Port Wine
If "slopcode" means code where the author consulted AI at all, yeah. But it's pretty easy to tell if the author AI-spammed without even reviewing all of it, or even if they did review it but didn't hand-type it. The email says "you did not write them, you said an AI did," so the author did admit to using AI at the same time they claimed copyright over the code.
That has a static import from 13 hours ago. Anyone know if this wine port has its own repository somewhere else tracking the history of the development effort?
"Still uses CVS", I mean, CVS worked twenty years ago and it still works today for all the platforms OpenBSD supports. If you prefer something else, that is all right, but here is emulators/wine on CVSWeb:
They mirror to GitHub regularly, but development still happens in CVS. They are building “got” as a potential replacement that allows git-driven development while following OpenBSD’s idiosyncratic approach: https://gameoftrees.org/goals.html
My guess is they based this on the similar NetBSD port but I don’t know where the repo is.
How is that ironic? If you know how to do this, ensure that the license of the code is compatible [1], and maintain the code, then send the patch to tech@ for review and maybe it gets in.
You can download the attachment and take a look at the patches. Wine already supports FreeBSD and macOS, so it's not tooooo crazy hard to add another BSD/Unix to the pile. Looks like there's a few Unix APIs that have drifted, so the patches add some #if OpenBSD style stuff to avoid using some symbols that don't exist on OpenBSD. The gs register stuff is by far the most complex bit. IIRC that register is used for thread-local storage on Windows, but for other purposes on other OS's, so handling access from win32 binaries to the gs register is a big problem when porting to a new OS.
On the whole it looks well written to me, especially with how complex & low-level these areas of Wine are. Good work to the author. I'd encourage them to try upstreaming it if they're interested. Some of it looks a little hacky, which upstream will object to, but maybe they could get some help cleaning it up, which will avoid having to maintain a diff and result in an even better port.
There are some basic OpenBSD-isms like ucontext, stack mapping, mmap, fix trap context so rest of the %gs patching logic even knows where to work.
Rest is basically ~~two~~ three main issues
1. %gs/gsbase - we trap on %gs instructions because %gs=0 gsbase=0 so all of them illegally point at page zero. So we take the instruction, decode, turn it into %gs-less instruction, and stow it somewhere else. To do conversion we need a scratch register (now two, thanks MINGW32-GCC, you c*nt), we replace whatever gsbase should've been with our TEB pointer, and gogogo (with restore of the scratch register state {or two}). Old instruction (if wide enough) becomes jmp rel32 (I picked it because it is 5 bytes, and short jump branches are too... short). If it isn't, we are just constantly trap on that so we just create a hash between trap location and stub location. Those cases are thankfully rare.
Just this one patch gives us Notepad, Notepad++, Factorio, and many many other games
2. SIGUSR1. Basically Windows lets threads chat thru their thread API, and Electron apps (I really wanted Eve Online to work) and Chromium apps constantly chat and interrogate each other with that thread API. Wine decided to emulate some of the APIs with unix signals. Now the main issue here is that OpenBSD, does not allow one process to signal a thread in another process. So we catch it with the currently active thread, and reemit it now inside of the process as SIGUSR2. Read on SuspendThread, GetThreadContext, ReadProcessMemory, and why Mojo/Crashpad works how they work, very fun! Makes you really enjoy the security OpenBSD gives you. Another dev is making this code better and handle other SIGnals.
3. ReadProcessMemory / WriteProcessMemory (NtReadVirtualMemory / NtWriteVirtualMemory) - this is also for Electron/Chromium but mostly Steam. We only needed it majorly when we started pushing towards steam working. Steamwebhelper, steam overlay, and steam watching you play and tracking your game time, are completely against OpenBSD's security model. Wine processes are not descendants of wineserver, so wineserver cannot ptrace them unless you enable kern.global_ptrace=1. We were trying to keep kernel as stock as possible so...
2 and 3 are very much together and if you ever seen on windows how chrome looks in the Task Manager and how many chrome processes there are, well now you know my pain, haha
I wouldn’t call either Wine or OpenBSD “obscure”. Most people haven’t heard of them, sure, but they’ve each been going for three decades and should be known (at the very least by name) by any person with an interest in computing for more than making money.
But of course, no one is born knowing, so here you go.
Now the BSDs are not so popular compared to Linux, but
I think this is great news as more eyes help here. Let's
have EVERYTHING from windows portable towards linux.
(I struggled with the 32-bit versus 64-bit stuff; this
should also be a prime goal for wine to make this a
trivial super-easy to solve thing for linux in general.
Back then when I ran into this issue, it was not trivial
to solve.)
I remember reading once how far away something like this way to work - for all the time that BSD has kept servers and networks safer than other options, it's nice to see things like WINE making it there as well.
Its hard to not think of the possibility that BSD might offer new or different advantages to keeping legacy windows applications safe until there is a willingness to adopt or develop alternatives.
Semi-related, but I was just thinking today I wonder what sort of performance improvements, if any, one could see by having one of the frontier models build a operating system from scratch that targets the subset of imports WINE uses as the standard library? Basically, build whatever syscalls (or shims) WINE and the open source vulkan drivers need and nothing more.
Probably outside the scope of what an individual could pull off on thier own in a reasonable time, but I could totally see for example Valve considering that approach for a later iteration of SteamOS.
Disco Elysium https://video.gamerstavern.online/w/qYjXr8dBauynzvEquJ7cP4
Hollow Knight https://video.gamerstavern.online/w/8tjjFRDtGJQZwxfE4ojVoC
Prey 2017 https://video.gamerstavern.online/w/vNxVThRcpijbT9FAQPcKdz
Death Stranding https://video.gamerstavern.online/w/rzz2TLCxcmneHsydukAFLU
Witcher 3 Remastered https://video.gamerstavern.online/w/oZcTRKL5JMiU4M4FqGKcNH
According to the reddit post:
AMD Ryzen 9 9900X, 16GB RAM, AMD Radeon RX 9070
https://bsd.network/@brynet/117407035545900963
For those not familiar with OpenBSD, it’s had an accelerated amdgpu driver since 2019, and inteldrm/radeondrm for a few years longer, generally kept up to date with Linux stable throughout that time.
Since many years longer, including the nv(4) driver for old Nvidia GPUs up to the earliest GTX models.
There is no kms drm or 3D acceleration for nvidia on OpenBSD.
Over time there can be periodic drift and out of date drivers until they re-sync, but it usually works pretty well.
But that is freebsd, not openbsd. I couldn't imagine it on OpenBSD because the priorities are very different
Current Wine just needs a couple of fixes from upstream libxml2 to be able to use the Retail C2R (Click-To-Run) installer, then the apps themselves work fine with a few Direct2D patches (see the effort to get Paint.NET running).
The real issue is that getting activation (even using MAS tooling) working requires a real Windows kernel running for the Warbird-protected activation system, after two weeks I gave up, 99% of my MSO needs are provided by Artcraft now anyway...
Subscription-based licensing requires periodic check-ins, but perpetual product key-based licensing is available, though compared to the subscription versions, the release cadence for new features is slow.
The online component of traditional apps like Excel works best as a means for sharing documents, with the web UIs useful for reading and light editing, and the same documents fully editable in the native apps.
(Its presentation is so stiff that I refused to play it - it's supposedly a very good game, but I just couldn't bear playing what felt like MS Office)
> I doubt much will be merged upstream. Many updates we have can technically run on other OSes (%gs stubs won't be built on Linux because %gs is correctly set, SIGUSR1 will work but less optimally, etc). However I picked an approach where everything that is done for OpenBSD is just marked as OpenBSD only. I did that purely to simplify my life while I was working on it and clearly mark the changes.
> When I was talking to Wine maintainers a month ago, they were hesitant to accept even patching Wine so it compiles for OpenBSD, mostly because it's another OS they will have to support and keep operational, while the %gs stub fix is something they definitely cannot take upstream. They are very hesitant to take anything that is changing the "emulated" application code.
https://marc.info/?l=openbsd-tech&m=177411620801633&w=2
Apple can't distribute them - coreutils is GPLv3, which is incompatible with Apple's immutable system volume.
https://www.gnu.org/licenses/gpl-3.0.en.html#section6
(There are other issues with GPLv3 which make it a dubious choice for companies like Apple, but this one is particularly problematic.)
We also test GNU coreutils on macOS. It is a bit tricky since the other maintainers and I don't own any macOS machines for testing. But we seem to do well enough with a shared system that Adélie Linux kindly allows the compile farm [1] to use, along with some additional CI machines.
[1] https://portal.cfarm.net/
Posix is more an art form documenting the far gone past rather than a practical tool
But the only concrete example I can think of is that `cat -n a b`, if `a` doesn't end in a newline, produces something that is obviously not what the user wanted. GNU cat does it right.
I didn't report it to the openbsd people because I didn't want them to yell at me that `-n` is nonstandard anyway with no "right" behavior, so I should STFU. Which would be what I'd expect from having reported problems in the past.
The slop problem is very real. People are spewing reams of bad code that generates entropy faster than humans or LLMs can clean it up. LLM best practices are not yet obvious, and we're still figuring out how best to leverage them while maintaining code quality and human understanding.
... but LLMs are coming whether we want them or not. Models keep consistently improving, the slop keeps getting better, and human-only development is going to struggle to keep up.
In my day job, I'm driving an enormous number of tokens at Astra+Ultramax solving problems in Lean 4 that previously we never would have attempted simply due to the incredible human engineering cost; we review the high-level lemmas and let it prove the rest.
LLMs are the only solution I've seen over these decades that could finally close the feature/usability gap between mainstream and more niche operating systems -- software support, features, drivers, and other broad usability barriers. All of that has historically been far too much work for a much smaller developer community to reasonably undertake.
Hating AI isn't going to make it go away.
Think about how far Linux has/hasn't come in the last 10 years. The user experience is nicer, drivers "just work" more, there's more support for more stuff...but at the end of the day, the actual computing experience is mostly the same today as when I installed Debian back in 2016. It's polish, interoperability, etc that's improved.
Maybe VR takes off and takes over the desktop, and then openBSD can't keep up with that, but I don't think that's likely. So what are they trying to "keep up" with?
Ppl praise the stance of OpenBSD/NetBSD and what-not. That's commendable, but then close the tab, open their spotify and get back to agentic programming for their proprietary company source. But God-forbid if OpenBSD turns anything more than a museum OS, full of purity but with a user-base of five. Because the GPU doesn't work properly, the sound drive is clunjy, wifi might work, etc.
LLMs could port huge amount of drivers to both system with a fraction of the effort. Specs can be implemented easily and properly. Then of course, you need to test and that takes time, but it's technically easier and more ppl could contribute.
I get the feeling, but they're not OpenBSD a favor, if anything else I feel the exact opposite is happening.
In this case, reading the link you're replying under, it's being able to mount a disk formatted with a popular filesystem. The alternative here is waiting for a human to type this relatively "solved" problem, as a procedural requirement, more than anything.
So, it can easily mean "things won't work in this OS". But, that's to be expected with FreeBSD, for legitimately good (from the perspective of FreeBSD) reasons.
Whether you like this new reality or not, it is an objective fact that LLMs are the silver bullet Fred Brooks told us didn't exist—multiplying development speed by 10x or more—and that if you're not using them you are hobbling yourself. This especially goes for a "security-focused" OS like OpenBSD, because AI is more diligent about finding vulnerabilities and can do so without rest. Meaning that if you're serious about security, you are using AI to harden your software... and if not, someone else is using it to hack you.
That is not remotely an objective fact. It's an opinion, and it isn't even an accurate assessment of the tech at that.
I'm sitting here watching astra do in a day what would have taken me at least 3-6 months. The code is verified correct. The model finds, reads, understands, and then integrates the latest literature into its designs and plans. It implements the code, proves it correct, and has come up with its own novel solutions on top of the state of the art. It then refines its implementation to hit the performance goals I give it. This is the reality today.
I've got some bad news for you.
[[citation needed]]
Theo de Raadt would spin in his grave — after the security implications of this to OpenBSD gives him a fatal aneurysm.
Either that, or he'll surprise-launch OpenLLM, a model bundled in the next OpenBSD release that produces driver code free of security bugs.
Isn't he still alive?
you could have the driver code prove it implements a spec, that there's no UB, that hardware invariants are maintained, that routines take a fixed time to execute.
the formal driver spec must still be written by sober, well-caffeinated humans of course.
Nothing in that post is against using an LLM. It's all about copyright:
>Since every file in OpenBSD has been continually verified to ensure correct Copyright, and we've even deleted code which has incorrect Copyright, the chances of us accepting such new code with such a suspicious Copyright situation is zero.
Once the current bubble pops we're going to see a lot of litigation over copyright enfringement by LLMs. Currently no one wants to rock the boat because everyone hopes to make a trillion dollars from AI, Oracle included, once they see they can't the knives are going to come out and I wouldn't want to be Linux at that point: https://en.wikipedia.org/wiki/Google_LLC_v._Oracle_America
Also my personal opinion about copyright _in the FOSS domain_ is that I simply dont care. Copyleft as the GNU folks like to call it is just using the copyright laws to achieves certain free software objectives. If that doesnt work anymore there will be another fix
The BSD licences are not copyleft, or indeed part of the GNU copyright regime.
Copyright is dead.
I remember hearing that in 1996 too.
It took a while for the DMCA to happen after Napster. But it indeed happened, and we all know the saying about history repeating itself.
By the time they came for Napster technology had already moved on. I can still download entire albums and movies in minutes 2 decades later. Heck, Usenet is still a bastion of piracy and that's been around for 46 years.
The 90s young people are familiar with are... way different from the 90s I remember. Their understanding of it is like... if someone asked to AI-gen the decade.
You may not care, but if the OS you prefer to run is the defendant in a copyright case, that's likely to impact you in a negative way.
"Those that do not learn history" and so forth.
It polarizes people to either join the "the LLM is my god" crowd or the "if AI touches the repo, it's all slop" crowd.
The careful balance is rarely discussed, probably because that's the part that still takes work.
Agents and LLM-assisted coding currently work within the confines of already working hand-coded OSes, language frameworks, test harnesses and even CLI tooling. Pull one of those out, and agents become unhinged and they lose direction.
https://events.eurobsdcon.org/media/2026/submissions/ZGDXQR/...
Talked to bluhm@ after that talk in Brussels and the general feeling (from what I can remember from the conversation) so far is that there is a large increase in overall reports (good and bad), but that the severity of what is reported has not really increased. This also mirrors my own experience from reading source-changes@. In essence, it seems like we have known how to build secure software for the last thirty or so years, just that most do not seem to have bothered to do so.
https://www.reddit.com/r/openbsd/s/k0dz2WrkCo
In that thread I suggested he submit it to the ports tree, I'm not sure if he took my advice or came up with the idea on his own. At any rate this is cool.
I also respect Wine and OpenBSD and their rules to not have any LLM code in the repo/ports. So my approach was to iterate over solutions, find approach that works, ask AI to "just see if my idea even works", learn what/how/why of that approach, generate some supporting manuals for myself (for example how to decode %gs instructions and how to reencode them) and then write my own code based just on decoding tables (and not LLM code).
I did not know how to work with REX/SIB/ModRM, how to detect prefixes, how to parse every single instruction that contains %gs, how offsets work, how to encode jmp rels, why I want to use JMP *0(%RIP) instead of JMP 32rel on return, why I want to leave stub on SIGUSR (or any other signal). But I do know now.
I was incredibly lucky that patching just these %gs instructions were enough to launch Factorio and play, with broken sound, but anyways! Holy molly this was and is so much FUN!
The approach still feels like cheating, and if it feels the same for you, I perfectly understand.
> I had some success in porting Wine to amd64.OpenBSD. Was able to run a few = games. But that required in memory patching of %gs commands (LLM's suggesti= on and 1st implementation, then I re-did it from scratch - I'm not that sma= rt with amd64 assembly code)
Also, it's not a single author. Other people contributed significantly and their contributions don't seem written by AI: https://github.com/NethIafin/wine/commits/openbsd-master/
[0]: https://marc.info/?l=openbsd-tech&m=179070255546118&w=2
That has a static import from 13 hours ago. Anyone know if this wine port has its own repository somewhere else tracking the history of the development effort?
https://cvsweb.openbsd.org/ports/emulators/wine
My guess is they based this on the similar NetBSD port but I don’t know where the repo is.
Edit: sibling commenter found it
[1] https://github.com/NetBSD/pkgsrc/tree/trunk/emulators/wine
[1]: https://man.openbsd.org/nv
[2]: https://man.openbsd.org/amdgpu
[3]: https://man.openbsd.org/intel
It is verifiable and has a clear exit criteria. Functional and performance parity.
[1]: https://www.openbsd.org/policy.html
On the whole it looks well written to me, especially with how complex & low-level these areas of Wine are. Good work to the author. I'd encourage them to try upstreaming it if they're interested. Some of it looks a little hacky, which upstream will object to, but maybe they could get some help cleaning it up, which will avoid having to maintain a diff and result in an even better port.
Edit: Oh, looks like they already talked with upstream https://marc.info/?l=openbsd-ports&m=179241414029592 It's not a huge diff to maintain, so maybe that is the better route. Great work either way.
Rest is basically ~~two~~ three main issues 1. %gs/gsbase - we trap on %gs instructions because %gs=0 gsbase=0 so all of them illegally point at page zero. So we take the instruction, decode, turn it into %gs-less instruction, and stow it somewhere else. To do conversion we need a scratch register (now two, thanks MINGW32-GCC, you c*nt), we replace whatever gsbase should've been with our TEB pointer, and gogogo (with restore of the scratch register state {or two}). Old instruction (if wide enough) becomes jmp rel32 (I picked it because it is 5 bytes, and short jump branches are too... short). If it isn't, we are just constantly trap on that so we just create a hash between trap location and stub location. Those cases are thankfully rare.
Just this one patch gives us Notepad, Notepad++, Factorio, and many many other games
2. SIGUSR1. Basically Windows lets threads chat thru their thread API, and Electron apps (I really wanted Eve Online to work) and Chromium apps constantly chat and interrogate each other with that thread API. Wine decided to emulate some of the APIs with unix signals. Now the main issue here is that OpenBSD, does not allow one process to signal a thread in another process. So we catch it with the currently active thread, and reemit it now inside of the process as SIGUSR2. Read on SuspendThread, GetThreadContext, ReadProcessMemory, and why Mojo/Crashpad works how they work, very fun! Makes you really enjoy the security OpenBSD gives you. Another dev is making this code better and handle other SIGnals.
3. ReadProcessMemory / WriteProcessMemory (NtReadVirtualMemory / NtWriteVirtualMemory) - this is also for Electron/Chromium but mostly Steam. We only needed it majorly when we started pushing towards steam working. Steamwebhelper, steam overlay, and steam watching you play and tracking your game time, are completely against OpenBSD's security model. Wine processes are not descendants of wineserver, so wineserver cannot ptrace them unless you enable kern.global_ptrace=1. We were trying to keep kernel as stock as possible so...
2 and 3 are very much together and if you ever seen on windows how chrome looks in the Task Manager and how many chrome processes there are, well now you know my pain, haha
“what is {insert obscure tech thing}”
“why do people typically use {insert obscure tech thing}”
But of course, no one is born knowing, so here you go.
https://en.wikipedia.org/wiki/Wine_(software)
https://en.wikipedia.org/wiki/OpenBSD
Todd Howard probably
Its hard to not think of the possibility that BSD might offer new or different advantages to keeping legacy windows applications safe until there is a willingness to adopt or develop alternatives.
There's even the remaster of The Witcher running! If I had money I would simply gift to those guys.
Probably outside the scope of what an individual could pull off on thier own in a reasonable time, but I could totally see for example Valve considering that approach for a later iteration of SteamOS.