5BSD - a verifiable FreeBSD derivate

cracauer@

cracauer@

Developer
[no affiliation, I just spotted this]

5BSD — Private Compute With Verifiable Integrity

5BSD provides secure execution infrastructure for workloads where data exposure, tampering, and unverifiable state are unacceptable. Open source, permissively licensed.
www.5bsd.org www.5bsd.org

Webpage is a little pretty and light on technical detail. Better info is on Linkedin in the author's posts:
www.linkedin.com

Kory Heard - Dallas-Fort Worth Metroplex | Professional Profile | LinkedIn

I build 5BSD the base operating system for security sensitive appliances. Other… · Location: Dallas-Fort Worth Metroplex. 500+ connections on LinkedIn. View Kory Heard’s profile on LinkedIn, a professional community of 1 billion members.
www.linkedin.com www.linkedin.com

The source code is forked from FreeBSD. https://github.com/5BSD/5BSD


5bsd.jpeg
 
Should it be explicitly stated that this project has been clearly vibe coded? To spot that you don't even need to look at the commit log (which I did). The website is 100% similar to all the slop we see everywhere nowadays: huge amounts of rambling text without much sense, but of course full of buzzwords with no real technical details that would tell us at least something concrete about the project and/or its developers.
 
sualbee said:
Should it be explicitly stated that this project has been clearly vibe coded? To spot that you don't even need to look at the commit log (which I did). The website is 100% similar to all the slop we see everywhere nowadays: huge amounts of rambling text without much sense, but of course full of buzzwords with no real technical details that would tell us at least something concrete about the project and/or its developers.
Click to expand...
Not convinced of that. Look up the core figures and public crtitics. I might be superstitious but it has the smell of a digital ID / age veficitation platform trying to become lawful "OS provider" next year with a special security infrastructure.
What has Oracle to do with it?
 
Is someone managed to try this thing?

I don't understand anything. I don't even see where to begin.
I feel that as just words and schemes void of any sense.
 
MG said:
Not convinced of that. Look up the core figures and public crtitics. [...]
What has Oracle to do with it?
Click to expand...
Where should I look that up? All I could find about the author is countless very abstract, if not meaningless, posts on Twitter and LinkedIn, plus only one (still pending) small FreeBSD contribution. Please share your findings, if you have more.

But what I also see:
- 100% LLM-generated website with no attempts to hide it,
- the project has only one author (excl. Claude),
- the author makes basic git workflow-related mistakes,
- 1.3 million LOC were added by 577 commits since May 15,
- these commits touch many different FreeBSD subsystems,
- some commits I checked clearly lack human validation and contain obvious mistakes even for me (who spent like 10 minutes on it).

I could try to build it just to be sure, but I'm already pretty sceptical about this verifiable/isolated/auditable/tamper-evident/etc BSD as it is. And, by the way, "Oracle" is just a name for its service manager, which has nothing to do with the company. Just another buzzword.
 
sualbee said:
Where should I look that up? All I could find about the author is countless very abstract, if not meaningless, posts on Twitter and LinkedIn, plus only one (still pending) small FreeBSD contribution. Please share your findings, if you have more.
Click to expand...

Google for the developer and critics. I just downloaded that github archive. What's going on? It contains apparent binary /bin/sh/sh but it can't be accessed. I can't determine the binary format. It feels like a borked archive. The sh manpage ssems recent FreeBSD.
 
sualbee said:
And, by the way, "Oracle" is just a name for its service manager, which has nothing to do with the company. Just another buzzword.
Click to expand...
Thanks I was trying to figure out who would run Oracle and bluetooth....
Neo would have been my choice.
Verifiable Integrity.
BBB stamp of approval? UL listed?
 
5BSD Kory here. Glad I’m getting some eyes on the project. To address the awesome feedback, yes I use LLMs to assist with coding. If that’s a no go for you I completely understand. I’ll note that many current FreeBSD devs and folks in the community also use them so you may want to check if this is the place for you.

The website needs a lot of work and will be updated before the V1 releases in early November.

The three technologies I’ve developed eCapsicum, mac_Capability, and Capsian are largely compatible with FreeBSD but you all need additional MACF hooks, a couple of file descriptors, multi label ZFS, and new hooks around sending descriptors, and transforming them during fork/exec. A language model could generate that code for you but they couldn’t have invented those ideas. That’s all grass feed Texas engineering 😁.

Here’s some images from my upcoming talk the Design and Implementation of the 5BSD Operating System.

IMG_2806.jpeg
IMG_2811.jpeg


As always it’s BSD licensed. Fork it, sell it, just don’t sue me and don’t advertise with my name.
 
sualbee said:
Should it be explicitly stated that this project has been clearly vibe coded? To spot that you don't even need to look at the commit log (which I did). The website is 100% similar to all the slop we see everywhere nowadays: huge amounts of rambling text without much sense, but of course full of buzzwords with no real technical details that would tell us at least something concrete about the project and/or its developers.
Click to expand...

Vibe coded abandonware takes time. I’ll give it a few months tops.
 
XNUKory said:
I’ll note that many current FreeBSD devs and folks in the community also use them so you may want to check if this is the place for you.
Click to expand...
Well, I'm aware of the current policy, but let's also look at the numbers:
Code: 
$ git -C /usr/src log --oneline --shortstat --since 1.year -i --grep 'co-authored-by:.*noreply@anthropic.com'
3800525cd2cc Fix aarch64 build failure by removing earlyclobber (#18532)
 1 file changed, 9 insertions(+), 9 deletions(-)
ddf19dcbe166 initramfs: fix incorrect variable rename
 1 file changed, 4 insertions(+), 4 deletions(-)
c214a3ae9f03 initramfs: fix use of renamed variables
 1 file changed, 2 insertions(+), 2 deletions(-)
6495dafd58b9 range_tree: use zfs_panic_recover() for partial-overlap remove
 1 file changed, 1 insertion(+), 1 deletion(-)
If I get this right (I didn't, see EDIT), we're talking about 4 very small and fairly trivial commits that had to be reviewed by humans, which is incomparable to more than 1 million lines of (I suppose) mostly blindly generated code.

XNUKory said:
A language model could generate that code for you but they couldn’t have invented those ideas. That’s all grass feed Texas engineering 😁.
Click to expand...
This sounds familiar... Aha!
Thread enabled-the-nvidia-gpu-passthrough-inside-a-linux-virtual-machine-via-qemu-accelerated-with-bhyve.102901/

EDIT: Sadly, I got it wrong. There is also an 'Assisted-by:' trailer that I forgot about, which reveals 39 additional commits. I'm not sure whether it means that such code was generated entirely by an LLM or simply that LLMs were used at all in the process. From my POV, there is no difference.
 
You must log in or register to reply here.
Back
Top