Twitxy Saved

Gregory M. Kurtzer

Verified account This account is verified. Learn more

@gmkurtzer

CEO - @CtrlIQ | investor | AI researcher | biochemist | software infrastructure | open source founder: @CentOS, @Rocky_Linux, @OpenWALDO (open source AI), Ginux
Reno, NV gmkurtzer.github.io Joined January 2021
337 Following 8K Followers 3.8K Posts

Latest posts 5

Pinned post

Gregory M. Kurtzer Verified account @gmkurtzer
Seeing @herrold again inspired me to share the foundation story about the creation of @CentOS...

Back in like 2001-2002'ish it all started with this wacky idea I had to create a community distribution of Linux, kinda like @debian, but RPM based.

I started talking to a bunch of people about it locally around Berkeley as I was architect of the HPC team @LBNL and joint appointment to @UCBerkeley at the time. We coordinated on IRC, and several of us met weekly at @JupiterBerkeley to have beers and discuss how we would do something like this.

I started calling it Community Assembled Operating System: cAos. I put the capitalization in the wrong place, because I was young and stupid (I'm still stupid, just not as young now).

On the IRC side, @herrold was the main coordinator and advocate. @herrold was very well connected in Red Hat, RPM, and various open source communities already, and I assumed he loved the project because he was one of the biggest developer advocates. He brought in Seth Vidal (creator of YUM, predecessor to DNF), Jeff Johnson (one of the founders and main developers of RPM), among a bunch of other amazing people.

The project started getting more attention, and when I asked who should lead this, @herrold was one of the people who pointed at me.

So Caos Linux was born (and I refuse to do the weird capitalization thing now). Also, I created a 501(c)3 non-profit for the Caos Foundation to host the project and infrastructure.

The operating system was based on a core foundation of Red Hat Linux (before RHEL), which was freely available. This allowed us to have a very compatible foundation as we were hoping to have enough API and system similarity with Red Hat Linux that many apps, tooling, and experience translates directly over.

The foundation of the operating system was designed to be completely API/ABI stable, locked, and self hosting (e.g. the core can rebuild itself, install those RPMs into a new chroot, and rebuild itself again, and constantly do this on every change). I was the developer and maintainer of the "Caos Core".

On top of the Caos Core was the extended operating system. These were all of the packages that were community maintained. I developed a portal for developers to sign up and join to either contribute packages or adopt packages leveraging CVS under the hood.

I'm not sure if @esrtweet remembers, but he actually joined up to maintain some of his code like Fetchmail. @jsquyres (project lead and scientist of LAM and OpenMPI) joined up to maintain MPI. Rémi Cohen Scali, who was a Gnome maintainer at the time, joined to contribute and maintain Gnome, and many other amazing people...

The project grew wonderfully!

Red Hat then introduced Red Hat Enterprise Linux (RHEL) version 2 (there was no public version 1), and shortly after that version 3, and shortly after that (IIRC), announced the end of life of Red Hat Linux, the freely available operating system that most of the world was using at the time.

This affected Caos Linux, because some of our core depended on it.

One of our project contributors, Rocky McGaugh mentioned that he's been doing RHL rebuilds for his HPC company, and we can base off of that. The Caos "tech lead" Michael Jennings, also maintained a rebuild of RHL at the time called Vermillion, which many people may remember comes from VA Linux back in the day. So we had expertise in rebuilding RHL already, and they offered to start rebuilding RHEL internally so we could use it as needed.

Shortly after that, there was an announcement on Slashdot of a new Linux distribution called White Box Enterprise Linux (WBEL) built by John Morris. I immediately reached out to John congratulating him on the project and offering him resources within the Caos Foundation if he wanted any help. He mentioned to me that he is beyond overwhelmed at the attention he is getting and didn't intend to maintain it long term.

Upon sharing that with the members of the Caos Foundation, it was decided that we should release our rebuild work under the name "Caos EL" (Enterprise Linux), Rocky would lead development of the RHEL 3 rebuild, and John Newbigin would lead development of the RHEL 2 rebuild, and my role was to lead the project itself.

Another Caos member was Lance Davis. Lance was busy with his own company out of the UK (uklinux IIRC), and he reached out to me proposing the name "CentOS".

My main response to it was that I didn't like the "OS" being capitalized, because that puts emphasis on the "Cent" part of the name and this project is designed to always be free, so I said I liked "Centos" (prounced without the "ohh-ess" at the end), and thus the project had a proper name.

Funnily, many people in the community seemed to immediately adopt the "cent-ohh-ess" pronunciation, and is probably why there was always three different ways to pronounce it: "cent-ohh-ess", "cent-ahh-s", and "cent-ohhs".

Sorry, probably my fault. LOL

Rocky announced the name change from Caos EL to CentOS on the email list and that he is almost all the way through v3, when he went AWOL. Not to disclose too much out of professional curtesy, but he passed away before he ever got to know how much of an impact his work would have on the world (Rocky Linux is a H/T to him, thank you Rocky!).

Lance took over the final bits of CentOS-3 and drove it to release, with John having CentOS-2 done shortly after.

From there, Caos Linux usage subsided and CentOS took off as one of the most widely used operating systems ever.

And now you know, the rest of the story...
Gregory M. Kurtzer Verified account @gmkurtzer Oct 28
I love it when I randomly see someone I used to know in my follower list!

It's great to see you here @herrold!

And to those who don't know this "Orc", he is one of the other early/initial @CentOS co-founders.
View quoted post on X
7 replies 9 reposts 50 likes 5.6K views View on X
Gregory M. Kurtzer Verified account @gmkurtzer
"When (AI) is 'thinking' it is deaf and blind and can not 'perceive' anything else about what is going on ..." -- Mira Murati, CEO of Thinking Machines Lab (loosely quoted).

But, that is changing! Interaction models will be much more real time and immersive in how it communicates and interacts with people and the world.

We are getting to a point where we will not be able to distinguish communicating with other humans Vs. communicating with machines over text and/or audio.

She continues with... "And by contrast, our interactions with each other are very rich, there are a lot of information in our interactions..."

She is absolutely correct about communication being very rich. For example, non-verbal communication is an important factor in communication, and we will see AIs also on the screen in front of us via video and audio, mimicking how we communicate, tone, body language, facial queues, etc. Soon we won't be able to distinguish real humans over AI in any digital medium.

Today's AI/LLM I really like. It is very powerful, and useful as a natural language processing engine.

But I'm curious, the concerns about truly multi-modal interaction models seem obvious, but what about the benefits?
2:08
Gregory M. Kurtzer Verified account @gmkurtzer
What if....

America continues to face backlash from the community against AI, data centers aren't supported, AI becomes super regulated so there is very low US competition, tier-1 AI labs don't have successful IPOs and are unable to continue at the pace we need...

What happens?

Companies globally must remain globally competitive and thus will become dependent on the only option that will exist, which will be the Chinese open-weight models.

Is that bad?

Open-weight models are actually not open-source, and this matters considerably because even though they are licensed with open source licenses (Apache, MIT, etc), we don't know what is inside of them. We don't know how they were trained. We don't know what their alignment is or will be.

This is critically important because they are black boxes that will have zero accountability.

It's become the freeware of the 2020s, but worse, because people are enabling them with harnesses without appropriate controls and no US based company to be held accountable to.

Considering how dependent AI has already become to the success of much science, medicine, research, energy, cyber-security, and software development (among others), the US loosing control and leadership of AI is actually very dangerous.

Is there a fix?

I believe there is, let's go back to talking about open source.

Historically open source has been the fix to transparency, ethics, massive growth, collaboration, and safety.

OpenWALDO is the first (and only) truly open source community for AI where anybody can join, collaborate, audit, and contribute to the entire AI corpus, tooling, and compose definitions for models.

openwaldo.org.

I've seen what a community of amazing people can do when they come together. It is what pivoted my own career over 25 years ago from biochemistry to Linux. And it is exactly what AI needs now.

Come and join the open source community. Learn about AI training, inference, infrastructure, and (not to sound cliche, but...) be part of the future!

@OpenWALDO
openwaldo.org OpenWALDO — The First Open-Source AI Project and Community Community, training data, tools, models, governance, and provenance: everything people expect from open source, brought together for AI.
13 replies 17 reposts 53 likes 2K views View on X
Gregory M. Kurtzer Verified account @gmkurtzer
This is where open source has historically taken over, when everyone collaboratively joins and works together to change the world.

OpenWALDO.org, you should join, just sayin’…
openwaldo.org OpenWALDO — The First Open-Source AI Project and Community Community, training data, tools, models, governance, and provenance: everything people expect from open source, brought together for AI.
Mark Kretschmann Verified account @mark_k 2d
We must build superintelligence faster than the commies can ban it.
View quoted post on X
6 replies 8 reposts 36 likes 1.3K views View on X
Gregory M. Kurtzer Verified account @gmkurtzer
Status update on Ginux...

I pivoted the full bootable kernel to be user-space instead. This means it can now run as a program on any Linux distro, executes as a command, and without using kernel namespaces it runs a container instance in a hardened virtual environment as an unprivileged user without virtualization.

This idea came from "domains" in the bootable Ginux kernel, which allowed Ginux to treat container instances as separate virtual environments, with enhanced separation over existing containers and with much greater fine-grained control over the kernel to user space ABI.

It is still very early, but as of last night, I have Rocky Linux 10 doing a full systemd boot, with very fine tuned permissions of what is allowed and what isn't.

Once a binary is executed, the ELF itself is passed to the host kernel and runs as an unprivileged user but Ginux captures and evaluates all system calls, ioctls, permissions, and file IO, and can figure out what each process is allowed to do or not based on the runtime configuration of that instance.

There are already a number of pending contributions and reported issues, but I'm going to hold off on getting to them until I complete the base architecture and defined gates.

If you are interested in the project, check out the GitHub repo here:

github.com/gmkurtzer/ginux