Blog draft: Bluefin Server/Dakota/Utah #4905
Replies: 8 comments 14 replies
|
no fear, waiting dakotaraptor being shaped |
|
//nice tattoo, will you mirror it on your left hand? :) |
|
Desktop agnostic here - Firefox and a terminal emulator are where I spend most of my day. I don't really even need the Zoom / Discord / Slack apps - they're installed because Zoom / Discord / Slack want me to install them and use them. Speaking of hummingbirds, my login picture is the hummingbird because it's the only dinosaur I could find in the menu on Dakota. 👍 |
|
you should also include mastodon's 'the wolf is loose' in the next PL :) |
|
If I'm reading this right, one of the things you're saying is that all the heat is in Dakota, and we should (if we can) be running that. FWIW, Rich Bowen was really helpful and welcoming to me during my time with CentOS. |
How I Talk to Agents: Natural-Language Prompts from Building BluefinA collection of real prompts I used while working on Project Bluefin (July–August 2026). Kicking off work
Just add the thing
Directing the fleet
When it goes wrong
Vibes and judgement calls
The lessonShort, imperative, context-loaded. The agent has the repo, the skills, and the docs — |
|
I think the comments on Fedora (Silverblue and more broadly) are generally valid; there's a big mix of structural issues (a lot of fragmentation, some natural and some really avoidable) and silly technical issues. I am definitely a fan of what Hummingbird has done and definitely looking forward to infusing a lot of that DNA into this! Some of this probably needs some overdue real-time discussion with a few focused people in the room. Maybe Devconf.us? |
|
My Rush friends! Seeing the right people respond to the videos really told me something about people. It's no secret we're "Mom and Dad Linux", so the idea of making a heavy metal thing with the aspirations of Hemispheres, starring our git logs of YOU people? There was gnashing of teeth on reddit but it was fine, check this out. I think this last video we made about taking advantage of LF Training will be a watershed moment for us, along side Utahraptor. And I think now we can offer a new thing. Those bazzite numbers attract the right eyeballs, and we're all gamers at work, of course. I want you to imagine a world where we write that content. The pioneers will be those of you who will take that Kubernetes course. That is an expensive course, and in today's competitive environment I think it's awesome. I didn't even know the LF did this for this level of course. @cgwalters @clubanderson - I've come to believe that this proves that it's about the operator's use of the tool, with an eye towards teamwork, understanding that AI is here and wants to dig in. But also, clankers. Because this does suck if you do it wrong. I am currently doing it wrong lol. This new cohort, ahmedadan, hanthor, joshyorko, and the rest (stand up and be counted!) Once people started to mess with AI it became clear that there's enough of you out there to start to catapult hive to way more success, and thanks to, the oddest of allies. Go look at the other hives, it's a bunch of software like ours from IBM. LOL. It's not revolutionary it just needs to be solid and good and the leading edge. But now this also means that a bunch of smart people and our people are now in their own discord and that thing is going to be it's own thing. And it can all be done locally and the people doing it now are doing cool stuff. Now that the entire toolchain is pure cloud native we can simplify everything. If you're that expert who wants to do ONE magic thing a month, now's the time to stand up. We're also looking for people with strong opinions - but we need people who understand that less is more, and there's still some fat to trim, and this Fall-ish. In case it's not obvious, I too feel the pull to slurp in the gnome extensions we need that would be automagically ported with gnome +1 at all times. I think we should keep these human. I love plopping some cash to extension developers, that part is UX and touches your finger or mouse. That needs the love of a human. We can probably so a bundle pack easily though, something in userspace. We do have an idea for bluefin "thicc mode" that brings in some of the more geeky stuff that you just need. Like the earbuds things. I think someone should drive this and figure out something tasteful. Forking Just Perfection is off the table out of respect for the original maintainer who is in a warzone. Agents can trivially do this, GNOME's problem is curation now. And Sri's got a plan for that (surprise! GNOME ED is a Bluefin Contributor LOL). Since we're basically the ops people anyway a lot of the important apps are in flathub so the stock experience can be lighter. DX, GDX, gaming mode, etc. can go full bird too. We have to maintain the images anyway let's stretch Dakota's legs. And lastly I took a week off, and now it's ahmedadan's turn, now is the time to learn buildstream and RPMs. I took some time to build an RPM Packaging Abomination. My parting gift to this accursed land. And also it catches utahraptor up to dakotaraptor, James is gonna do the whole hog, v3 baseline, shooting for FIPS(?) because why not. That means we have to build a bunch of RPMs. But this is k8s native bridge the RPM birds needed, there's no reason why Utahraptor and Dakotaraptor cannot go head to head. I think this will be awesome. Utahraptor's game mode will be easy because it'll just reuse the Bazzite RPMs. And the custom image part is a set of github actions, so your "custom image" is basically another module in bluebuild or whatever you use. And we can build it from source all the way to the end, giving S-Tier SBOM support! I'll update this post over time. But TLDR is the more experienced need to go get this academy off the ground because this thing is way easier than people think it is. James runs his own Tuna-Os thing that has everything. This isn't a wheel and spoke. This is a mesh network. Especially custom image people, just start fixing stuff and sending PRs and being loud. Now would be the time for someone to step up to a cosmic too if they wanna take a stab at it! |

Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
We're failing LTS users, which sucks. There's a new person rocking Bluefin LTS maintainership in the new org. How do we get to this point? We didn't arrive here by choice:
https://www.youtube.com/watch?v=1l9r4apMJzs
They're the dad's in the story. Here's who we are:
This was going to be a surprise, but like Bluefin herself, this is not the final form. But it shows my commitment. Because in a few weeks after this is healed, we're going to fill it in with The Blue.
But we're in production, ublue-os/bluefin has had a cosmetic issue. When I ran into it I realized it had been reported. I didn't get to it. I could have fixed it, but many of you won't understand how to prioritize. I believe that cosmetic issues aren't important at this phase in our life. And many of you know how much of an annoying perfectionist I am. Same thing with the motd. etc. If I am working on something that makes that not my priority, then it's important. But we can't break people's extensions like that. And dakota has that built in.
My coworkers depend on it - I told them my options are, solve the problem forever or deal with a cosmetic issue. The choice is clear.
[ reassure ublue-os people that we're not bailing ] - we are purposely keeping you change-averse for now.
Rebase status:
ublue-os/bluefin-lts -> projectbluefin/bluefin-lts -> opt in, needs finishing
ublue-os/bluefin -> projectbluefin/bluefin -> opt in, needs finishing
ublue-os/bluefin -> projectbluefin/dakota | People have done it needs fixing |
ublue-os/bluefin -> projectbluefin/utah | May or may not work |
ublue-os/bluefin-lts -> projectbluefin/dakota | Who knows? |
ublue-os/bluefin-lts -> projectbluefin/utah | Who knows? |
We can't do it alone, Apache Buildstream is the tool we were missing. And when combined with CNCF tech like bootc and Kubernetes it is unstoppable:
[ New video goes here ]
Utahraptor is the newest murder chicken. It is the obvious Bluefin we've been building in front of everyone. But just like Kubernetes needed openshift, we need a competitor. And the only one that shows up is something like Hummingbird. This is the natural fit for those of you that come from RPM land. But I hate all of that stuff, and if you stay with me, you will be underserved and you will be mad.
Project Hummingbird is a result of Red Hat's acquisition of a few Universal Blue core team members. In many ways when I say "ublue's mission is complete" this is what I mean. It will offer what Dakota never will.
The Bluefin Factory will be a DDI image based on freedesktop components. It is the farthest away from "Linux distro" as possible. All the parts are interchangeable. I have one node that is Fedora, one that is Flatcar, and one that is Bluefin Server.
Distros don't matter, they're all kubernetes. You can run any combination and it doesn't matter. Every factory can publish anything you want. If you're a RHEL kind of person you'll be building on our tech. Your list of images will be different. It doesn't matter we all work on the same thing.
This factory will also feed your WSL images, your mac images. All of it. Because linux distributions and operating systems don't matter, you have to learn the same thing everywhere. (Link to Mac VM and the dx, ai, gdx, and other brew bundles as supported on mac and windows)
Ideal spitball roadmap (this is an aspirational proposal)
What does this mean for Fedora?
Here is the FPL presenting our efforts to Fedora:
And then something like (how do we get those folks to work with us more)?
Now. I consider this chart proof of our success. But unfortuntately I don't think the world's largest RPM factory would consider this chart awesome. But we all know humans aren't going to be doing any of this crap. (p5 did 3000 RPMs already or whatever crazy number he said last hummingbird meeting)
It would be madness to not include Fedora, since they are literally part of the team now.
What does this mean for other RPM based bootc images?
I have no clue. All I know is that you can read what's happening here and read what Fedora says it's about.
I will NOT put this project at risk when leadership of a project "committed to innovation." are asking new people to understand what a mailing list is. Projects like Hummingbird break them out of this chain. I can't help them, but we can help Hummingbird. [Join the party here], [looking for RPM people here], [convince sherman to rebase ucore onto hummingbird], etc.
To answer Jef's question, "how do we get these folks to participate" I'll follow up with some info:
I will not send our best and brightest to figure out mailing lists and technology that will actively harm their careers. We will send them to Kubernetes. Which is where Red Hat is investing already and getting a FANTASTIC RETURN in their investment. Evidence: Project Hummingbird's talent. Happy 100k live bootc devices!
[Offer live supported rebase from any RPM bootc image to this one, maintainers commit to working with Hummingbird - we still keep the old stuff around but it will continue to be left behind.]
All reactions