Installing *BSD in 2025 part 0b - Modern myths: AI ramblings (NetBSD and OpenBSD)
In part A of this article I covered answers from several Large Language Models to questions about DragonFly BSD and FreeBSD. The intent there was to get an idea about how much *BSD knowledge was included in their training data. As a comparison, the same questions were asked about Linux. While all models answered correctly regarding the penguin OS, they made some wrong and weird claims about the BSDs, especially the more niche DragonFly.
Installing *BSD in 2025 part 0a – Modern myths: AI ramblings (DragonFly BSD and FreeBSD)
Let's see what the situation looks like with NetBSD and OpenBSD in this second part.
The same two prompts as last time were used:
Again, these are the models:
Like in the first part we're going to take a look at how the operating systems are summarized by the models.
Llama 3.3 responded like this:
"The NetBSD project is a free and open-source operating system that, like Linux, is a Unix-like OS, but it's based on the Berkeley Software Distribution (BSD) codebase, with a focus on portability, stability, and security across a wide range of hardware platforms. Unlike Linux, which has a monolithic kernel, NetBSD's architecture is designed to be highly modular and portable, allowing it to run on everything from tiny embedded devices to large servers, with a strong emphasis on correctness, simplicity, and ease of maintenance."
Here the LLM struggles. The answer is mostly correct, but it makes it sound like NetBSD's kernel wasn't monolithic even though it definitely is. NetBSD was the first of the Open-Source BSDs to implement loadable kernel modules in 1993. What probably also confused the model is the fact that NetBSD pioneered the _rump kernel_ concept, an idea that is about making regular device drivers reusable outside of the monolithic kernel as well (for example in a server-like manner in userspace like with microkernels). While NetBSD is indeed designed to be very portable, Linux de-facto has also been ported to a wealth of platforms, even though this was not a design goal.
According to Mixtral,
"NetBSD is an open-source, highly portable, and strictly conformed UNIX-like operating system that values code correctness, simplicity, and efficiency. It caters to a wide range of architectures, offering a small footprint, making it suitable for embedded systems and servers alike. While it shares some similarities with Linux, NetBSD's BSD heritage sets it apart in terms of licensing, design philosophy, and user experience."
This is Exaone's answer:
"NetBSD is a highly portable, open-source Unix-like operating system known for its portability across a wide range of hardware platforms and its focus on correctness, maintainability, and code clarity. It offers a robust and stable environment, appealing to users who value system integrity and cross-platform compatibility, similar to the reliability sought by seasoned Linux users but with a distinct emphasis on portability and comprehensive documentation."
Gemma 2 summarizes the OS like this:
"NetBSD is an open-source Unix-like operating system known for its portability and robust architecture. Similar to Linux distributions but derived from BSD, it focuses on stability, correctness, and supporting a wide range of hardware architectures."
Phi 4 describes the OS this way:
"NetBSD is an open-source Unix-like operating system known for its portability, robustness, and adherence to standards. It emphasizes code correctness, simplicity, and clean design, making it highly adaptable across various hardware platforms. Unlike many other Unix derivatives, NetBSD maintains a strict focus on these principles, which results in a stable and secure environment suitable for everything from embedded systems to servers. Its modular architecture and extensive device support make it an excellent choice for users seeking a versatile and reliable OS."
There's not much to criticize with the last four answers: They are correct and to the point!
And Glm 4's take?
"The NetBSD project is an open-source, Unix-like operating system that emphasizes clean design and portability, akin to FreeBSD or OpenBSD, yet it differs with its robust support for hardware platforms and its focus on security and stability—a familiar ground for a seasoned Linux user who might appreciate its strong community and wide range of features."
Glm 4 describes NetBSD in relation to FreeBSD and OpenBSD. Since security is the primary concern of the latter, the statement made here doesn't make a whole lot of sense. Also it would be better to point out that NetBSD has a smaller community when compared to the other two BSDs and especially compared to Linux. Also if I were picky, I'd point out that the NetBSD project is not the NetBSD operating system - but to be fair, my prompt contained that misleading term!
So for NetBSD it's interesting that the largest model and the smallest make mistakes but all of the rest provides answers that are not half bad! We'll soon see what they know in terms of NetBSD's special characteristics other than portability.
Llama 3.3 summarizes the OS as follows:
"The OpenBSD project is another free and open-source operating system based on the BSD codebase, with a strong focus on security, simplicity, and code correctness, aiming to provide a secure and reliable platform by default, through proactive security measures and rigorous code auditing. Unlike other BSDs, including NetBSD, OpenBSD prioritizes security above all else, with a philosophy of "secure by default" and a reputation for being one of the most secure operating systems available, making it a popular choice for applications requiring high security, such as firewalls and VPN servers."
This summary, while not comprehensive, is accurate.
Here's Mixtral's take:
"OpenBSD is an open-source, security-focused, and standards-compliant UNIX-like operating system that values clean, audited source code, and strong cryptography. Known for its emphasis on transparency, proactive security, and consistent design, OpenBSD is suitable for servers, desktops, and firewalls. It is often compared to Linux due to their shared UNIX heritage, but OpenBSD's unique approach to security and licensing sets it apart from other systems in the BSD family."
Unlike Linux which is an OS kernel inspired by Unix (or rather Minix in the beginning), OpenBSD is an actual descendant of the BSD line of the Unix operating system. So the "shared UNIX heritage" is misleading (especially when the name is spelled like this, in all caps!), but some people may call that nitpicking. The "unique approach to licensing" setting "it apart from other systems in the BSD family" is interesting. There's actually some truth to it: OpenBSD is the only BSD that does not use the BSD license - their preferred one is the ISC license (which ironically the ISC no longer uses!). It's still a BSD-style license, though.
This is Exaone's answer:
"OpenBSD is a security-focused Unix-like operating system renowned for its proactive approach to security, emphasizing code auditing and secure defaults. It provides a hardened environment with a strong emphasis on correctness and security features, making it particularly appealing to users who prioritize system security and integrity, akin to the meticulous approach valued by seasoned Linux users but with an unmatched dedication to proactive security measures."
I like this one. In fact I think it's something that a Linux user who is just about to discover the BSDs might be happy with. Well done, Exaone!
According to Gemma 2,
"OpenBSD is a security-focused Unix-like operating system known for its strict code auditing practices and emphasis on proactive security measures. Like NetBSD, it's derived from BSD but prioritizes minimizing attack surface and providing robust security features out of the box."
As usual, Gemma 2's answer is short and this time actually to the point.
How would Phi 4 describe it?
"OpenBSD is renowned for its emphasis on security, code correctness, and simplicity. Developed from the BSD Unix variant, OpenBSD prioritizes secure system design, rigorous code auditing, and minimalism to reduce vulnerabilities. It implements a variety of built-in security features, such as ProPolice stack protection, randomization (Address Space Layout Randomization), and privilege separation, making it one of the most secure operating systems available today. The project also focuses on providing clear documentation and actively encourages contributions that enhance security, privacy, and transparency. OpenBSD is known for its robustness in network services and cryptography implementations, often serving as a foundation for other operating systems seeking to improve their security posture."
Now that's a great summary, probably sparking the reader's interest with mentioning features like ProPolice and ALSR (even though the latter is extremely common today).
Glm 4 summarizes it like this:
"OpenBSD is an open-source Unix-like operating system that is known for its security features, clean code, and minimalist approach. It offers robustness and portability, sharing similarities with NetBSD and FreeBSD but is particularly esteemed among seasoned Linux users for its emphasis on security, simplicity, and the "One True Way" philosophy."
While I'm not sure what the "One True Way" philosophy is, the OpenBSD is pretty opinionated on various things and are (mostly rightfully IMO) convinced that they are doing the right thing.
Like with FreeBSD, most models seem to know a bit more about the other rather popular BSD variant compared to the less frequently used systems. Most of the answers are pretty good this time.
Now we get to the distinctive features of the two BSDs we take a look at here. Let's see what the various LLMs can tell us about them! Like in the previous part of the article I'm going to simply list the bullet points that I agree to (and don't have any comments on).
Llama 3.3 lists 7 features:
While Llama 3.3's answer includes some correct and relevant information but also a surprisingly large amount of claims that are either imprecise and potentially misleading or entirely false.
Mixtral's 6 points are the following:
The result here is a bit of a mixed bag (pun intended): There's some actually good information but also some questionable or downright incorrect claims.
Exaone points out 10 features:
Exaone's answer contains a fair bit of useful information about the system, most of it correct. Unfortunately a couple of errors have crept in as well.
Gemma 2 mentions only 4 points:
Gemma 2's answer is not very helpful. With only four points out of which two don't mean much, it is far from comprehensive. At least there's no errors in the details that were given for these points (but they were rather generic, too).
Phi 4 presents 8 points on NetBSD features:
With the exception of one inaccurate claim I really like Phi 4's description of NetBSD's features. It provides a nice amount of detail and rightfully points out some important things.
Glm 4 tells us about 10 features:
Glm 4's answer is not too bad with a couple of surprises (unfortunately both positive and negative).
All models got the characteristic right that NetBSD is best known for: Portability. Mentioning Pkgsrc has been a stable, too (it's interesting to see that it was rarely mentioned that Pkgsrc is portable as well). There has been quite some confusion around virtualization with none of the models mentioning NetBSD's native hypervisors NVMM and HAXM. In the case of this OS the larger models were not necessarily much better than the smaller ones and at least partially gave pretty low-quality answers. Phi 4 is a notable exception here. The other models seem not to be too knowledgeable about NetBSD.
Llama 3.3 provides a list of 10 features for OpenBSD:
This time Llama 3.3 got it all right: One minor nitpick aside there was no wrong information in there and it even avoided inaccuracies!
Mixtral delivers an answer of 6 features:
Except for the blunder with the penguin, the answer is pretty much fine.
The list provided by Exaone contains 7 elements:
For OpenBSD, Exaone got it all correct!
Gemma 2 lists 4 features:
Once more Gemma 2 is very brief. It's also completely correct (and the additional information left out here is pretty ok, too).
Phi 4 presents 8 features:
Except for one flaw the answer that Phi 4 gave is excellent and even includes quite a bit of detail.
Glm 4 lists 10 points:
This time Glm 4's response is pretty low-quality with several false claims and quite some rather generic text.
All of the models seem to know a bit more about OpenBSD than about NetBSD. They all mention its security focus, and for a good reason. It's interesting that even after all those years the "strong cryptography" statement keeps getting repeated time and time again. The background of this is most likely that at one time US law prohibited the export of strong cryptographic algorithms except to a few countries. For that reason e. g. FreeBSD asked non-US citizens to not download the cryptographic bits. Since OpenBSD is based in Canada no such restrictions applied and they shipped good cryptography for everyone.
Once again Phi 4 shines with its answers. Exaone and some of the others do pretty well, too. Glm 4 maybe had a bad day (or seed / starting memory state rather) and included both irrelevant things and actual false claims. Mixtral's claim of the penguin mascot is a facepalm moment in BSD circles, or one of having a good laugh.
Just like in the previous part with DragonFly BSD and FreeBSD, we can see that the various models have some knowledge about BSD operating systems but you have to be on your guard for inaccuracies and entirely wrong information.
In the first part I compared the results against Linux, this time I decided to compare them to asking the LLMs the same questions about _illumos_. The general description was mostly accurate with a few inaccuracies (like the claim that it supports the UltraSPARC platform which has not been the case for some years now). But the answers to the question about its features really surprised me: Except for two mistakes (Phi 4 claiming that it's also used in the embedded sector and Glm 4 thinking that it's available on ARM), _all_ of the models not only answer correctly but in some impressive detail: ZFS's features, DTrace, Zones, SMF, Crossbow... I didn't expect this at all! Even though it's even more niche than *BSD, maybe it being a derivative of Solaris and thus kind of having a big name (Oracle) that puts some weight behind it, helped?
What other conclusions can we draw from this? Well, it looks like our favorite family of operating systems could use some better coverage. What can we do to work towards that? Not too much, I guess (unless some of us work for the companies responsible for the LLMs and could influence training data). The best bet may be to keep using and both talking and writing about *BSD, further increasing its relevance.
Oh, and we could keep telling people to not trust the ramblings of "AI"s. While I usually avoid political statements, right now I will make the one that we live in weird times. Incredibly stupid things seem entirely possible right now. And if we don't keep our guard, we might just be creating new gods that will exercise influence over our lives - and probably not in a good way. Let's try to avoid that. Blind trust in false prophets is bad enough if they are human. It could be much worse if it's "AI"s that don't even have an agenda. Even though it would be quite the irony if we fell for our own simulations that we created, wouldn't it?
In the next couple of articles I'm going to take a closer look at installers for the BSDs (and maybe beyond).