I noticed this hasn’t exactly been announced, so I figured I’d post about it here to increase visibility: the first Fedora Hummingbird community meeting is scheduled for 2026-05-28T12:00:00Z on Jitsi. I may not be there as it’s ruinously early in the morning for me, but if you can bear the pain of a meeting without my company, feel free to show up 
2 Likes
Update: there appears to be some confusion about what time this meeting is happening, as @jspaleta had it down as 8am Eastern (which I believe would be noon UTC) - that’s 2026-05-28T12:00:00Z, and that’s also written on the Wiki page. Jef will try to clarify the situation.
Update 2: It appears that 2026-05-28T12:00:00Z is the correct date and time. The Fedora Calendar entry has been changed. I’ve changed the first post and topic to reflect this.
I’ve posted that to devel as well: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/NWYHOKS3WANRLHNCNJEYHJJEFC642CLJ/
I couldn’t make it to this meeting. Are there any meeting notes or a recording?
1 Like
I’m going to tag @fatherlinux in here in case there is a recording 
However, I can share some brief notes from memory. We started out talking about what people were interested in using Fedora Hummingbird Linux for. After people started sharing, it was clear that, while there is a lot of interest, there were a lot of underlying questions on how exactly this will work and what the experience will be like for both contributors and end users. So, we went through a presentation talking about the what and how, with a lot of breaks for discussion and questions. From a contributor’s standpoint based on what I heard, there’s opportunity to engage with the factory (tl;dr: think artifact repo plus CI/CD pipeline plus security scanner plus SBOM generator. That’s Konflux) through helping to write the AI skills and agent documentation. There’s also opportunity to help review the patches to approve builds, write the documentation, and build up community knowledge overall on how this works.
We had a lot of deeper technical discussion, including discussion about the infrastructure and where it lives, but I want someone else who was there to weigh in as I didn’t take explicit notes this round, and I don’t want to get my technical details wrong and get everyone confused. However, if we don’t have a recording available, I can check my memory with folks and post a follow-up here with those technical details once I’ve confirmed them.
Thank you for tagging me in Laura! We had a bit of an error with the recording part way through. I just checked, and the webm file for at least 35 minutes of it worked. I’ll see if I can “fix” the other chunk, and stitch them together. I’ll also try to find the right place to publish this (and transcripts, etc). I’ll need a little while, have some travel next week, but I’m hoping to get it out in days, not weeks.
Kind Regards
Scott M
Hey Scott I have some follow on info for you.
I think part of the issue is this:
I am guessing the typical Fedora user has no idea what that means. I get what you mean, but not a lot of others do. So let’s go over it for folks, here’s my version of your pitch:
New high-speed, frictionless Linux distribution is purpose-built for AI agents and the people who build them, with services and support options planned through a Red Hat subscription
“Whatever this is, Red Hat is making it, so of course they want it in Fedora first. That sounds like cloud AI bs, but I guess the free one in Fedora will be fine, win for me.”
Then you do the research and you look, and what you’re getting is basically free hardened images. Most of the industry sells these for a lot of money and you want to give them away? Like the old days? And even ones with kernels in them?
The people asking “who asked for these?” I would say organizations with the pressure to pass all sorts of compliance issues but can’t afford all the security goodies enterprises get. That is a ton of organizations. Including mine I plan on making heavy use of these in CI, heh.
Much of the background maintenance and feature integration is performed by AI agents with human-in-the-loop oversight.
We have both independently reached the same conclusion. Here’s our prototype:
We have issues where humans triage whatever, add a certain label, agents pick it up, send PR, other agents review, human does the final ack. We’ve fixed actual, real bugs. And we’re feeding the system → bluefin clients can ujust report issues, which then can go in a queue. And then I as a maintainer can send my agent to do the gruntwork of reading all that gunk and tell us where the problem is.
Then it goes back into the queue. Humans control the decisions. This is all built on tech called hive, which is one of those orchestrator framework things that the Hummingbird article wants people to build. He’s been helping us out because Kubestellar is a CNCF project and has been making good use of this and you all should check it out.
So here’s a riff with the Fedora, repurpose/slice as you see fit.
“Fedora Factory, powered by Hummingbird”
Ever want to contribute to Linux but you’re technical but just not technical enough to fix the things you want fixed? Here you go:
- Download ISO, install on homelab
- Your own Forgejo with out of the box Fedora workflows, tailored to upstream contributions, start on the right rails every time.
- Also acts as your personal issue tracker.
- But also set up so you can set up your own projects with prefilled templates
- Server goodies
- Built in Fedora QA testing with automagic VMs, help Fedora directly by automatically testing for issues, test branches, etc.
- Pick whatever container web front end thing you want
- Silverblue for the Home
- Full konflux pipeline, build your own OS in minutes, fully signed, reproducible
- Podman Desktop on your clients discover this server so all your developer stuff can just run there for efficiency.
… and so on.
OK now everyone go BACK and read the Red Hat article again. NOW that makes way more sense to me.
Give every Fedora contributor an automated way to do testing, that is setup like the real thing. QA budgets are not awesome right now we need to automate our way out of this. And what human resources we have should be used to coordinated automated work. That makes sense to me. We have an issue to add CoreOS support to knuckle, which is effectively a “home server installer” for these style systems.
Fedora CoreOS for the OS, Fedora Hummingbird for the containers. And that scales out from single node podman to k8s so that covers everybody.
Also … “Ever want Universal Blue but without that one guy? We hired all the approvers!” would have gotten you a landslide victory.
4 Likes