Every ask anyone put on the table, and every question the room left open. All of it is Fred's text from the episode summaries, newest first. If you can help with one, come up on Sunday and say so.
Try Uber TV at ubertv.dev. Plug it in and tell DevSec what you think. Bug reports and new feature ideas are welcome.
EP 06Want in the X group chat? Message Wes, me or Called To Construct and we will add you. That is how we stay in touch all week.
EP 06If you know somebody at a phone or hardware company, point them at Wes. He is serious about the Omarchy phone.
EP 05Hammer test Flea and Atmos. Click around, try to break them, and tell GM and Chris what you find.
EP 05@StudioXRadio 's Omarchy M4 alpha is headed for the wild. It is experimental, so read his disclaimer before you wipe anything.
EP 05@m_tolhuijs wants people to try OmarKit and reach out. He is building it as we speak, aimed squarely at the submission pain.
EP 02@cfaulkingham wants collaborators on the local fine tune. He has a LoRA adapter, a model that runs on modest hardware, and a clear view of why it is hard.
EP 02@ncfrontiersman wants to see a proper plugin dev loop with hot reload, and offered to contribute his version to OmarKit.
EP 02@hancore_linux wants patience while the new marketplace gets built, and recommends OmarKit and the security skill in the meantime.
EP 02@MuchmoreIT is building toward federal ready Omarchy and would welcome anyone who knows that world.
EP 02NEXT EPISODE: Weekly, probably. Same format: raise your hand, tell us what you're building, tell us where you're stuck.
EP 01Don't be scared. If you don't know what you're doing, ask somebody, or ask your AI agent. Ping me! I want to help.
EP 01Lee is one person who found this on his own and rebuilt his life around it. He is not going to be the only one. Nobody is looking for these people, and there is no obvious place for them to land when they arrive.
hancore is building eye-controlled mouse input. Lee is building cognitive scaffolding. Neither of them started out as an accessibility project and neither of them is coordinated with the other. Is that a thing the community should organize, or does it work better as people scratching their own itch in public?
Lee's problem generalizes to everyone with a memory system: it stores what you believed, and has no idea when you stop believing it. His answer is a scheduled audit that asks you questions and appends the correction without destroying the history. Nobody else in the room had solved it. Most of us have not noticed it yet.
Three people ran into the same thing from different directions. It does more than you tell it to, by design. Tech Luddite moved his to a dedicated folder because he was not comfortable with the default. Porsche runs hers on a private server. Nobody stated a norm. What is the safe default install for something that acts on its own?
Colin wants it to go further and become the core of Omarchy. Everybody agreed it is the killer feature. Nobody knows what "further" looks like or who decides.
Kozha's machine broke mid-update over bad Wi-Fi and the only path back was a full reinstall. Tech Luddite raised checksums after download and said he has not verified whether that happens. That is worth someone actually checking.
Victorio said out loud that the hard part is a sponsor and a venue. Tundy's answer two weeks ago was just do it and ask a friend for a room. The pattern is repeatable but nobody has written it down.
This came up three separate times from three people who had never spoken to each other. @NoPowerPoints asked it directly. @m_tolhuijs built OmarKit to solve it. @cfaulkingham found jankeesvw's security skill and OmaVM. @ncfrontiersman points his agent at the marketplace source to work out the checks ahead of time. @hancore_linux says a new submission system is coming but cannot say what it will require yet. Three partial answers, no official one. Does a preflight skill ship with the new marketplace, or does the community keep owning this?
@NoPowerPoints said the quiet part: having the agent run the initial submission is great, and also a little scary, because it is using your credentials. Nobody in the room had a policy for that. What should an agent be allowed to do with your GitHub credentials, and what should always be a human pressing the button?
@ZryMiller asked and we gave him tactics, keep the feature set small, but not an answer. He is waiting for the day he uses it without hitting constant bugs and suspects it never comes. What is the actual signal that a thing is ready to put in front of people?
@cfaulkingham did the work and reported back: overfits easily, has to run on a potato, automatable but weak without serious human curated training data. Competing on general capability is a losing fight. So what is the narrow job where a tiny local model beats sending the prompt to the cloud, and who builds the training set?
Everybody now has the habit of asking the agent if somebody already built it. That is a workaround, not a system. Discovery across the marketplace, GitHub and X is still manual.
@ncfrontiersman thinks a lot of the Arch crowd would change their minds if we got to talk to them. Nobody tested that this week. Is there a real move here, or do the walls just relocate?
I raised it and we moved on. I run my own plugin site partly as disaster recovery, and we put import and export into Atmos as a piece of it. The Omarchy crew will probably beat me to the full answer, but right now there is not one.
OmaStorm is going into the next version of Omarchy, the first time that has happened to somebody in this room. What changes for the maintainer, for the people sending PRs, and for the people who forked it? We did not ask @ncfrontiersman directly and we should have.
A NIST tested crypto validation suite is the better part of a year for one busy person. Omarchy Fort Knox is a real opportunity and right now it is one guy with a day job at the end of fiscal season.
Every maintainer is getting five PRs that do the same thing five different ways, PRs for features they don't want, PRs written by agents whose owners never read them. Wes wants issue-first gating. Joel wants a shared CONTRIBUTING framework that primes the contributor's agent. Tech Luddite wants a skill for making PRs on other people's repos. I want the "no" to be easier to say. These are all the same document, and nobody's written it. What's the minimum a repo needs so an agent reading it behaves? Who maintains the shared version? Can the gate be automated without killing the drive-by fix that actually helps?
Amogama wants small models trained on Omarchy, distributed like apps. Tech Luddite has watched local models break machines and is betting on a curated corpus plus skills plus a SQLite index. I said 8% of my prompts are local today and 50% in a year. Nobody knows which approach produces consistent, cheap, correct fixes on a stranger's machine, because nobody's measured it. Who curates the corpus? Where's the line between what a local model may do (read, diagnose, propose) and what needs a frontier model or a human (run the shell command)? If DHH's team ships an official Omarchy model, do community efforts complement it, compete, or fold in? How do you test a local fix without a fleet of VMs?
Every agent memory system saves everything. None reliably surface the right thing at the right moment. Every person on that call re-explained something to their agent this week that it already knew. Is this a per-person tooling problem or a community one? Could the Opinionated Omarchy corpus be the shared recall layer for Omarchy problems, with personal memory on top? "Did the agent already know this" is measurable. Nobody's measuring it.
Plugins run whatever's on main, interpreted. That's why they're easy to build. OmaStorm needed a Rust daemon for performance. How does a compiled backend get built, shipped, and updated through a system that only pulls QML from main? Blip's best install path is "point your AI at the repo." That's not an installer. Forking is the community's default move and system updates break forks. There's no versioning, pinning, or rollback story most users understand.
Who reviews plugins before people install them? Right now the answer is "Tech Luddite forks it and runs his agents across it." That doesn't scale. What's the norm for agents with send or write capability? Ask every time is best effort. Allowlists are better. Nothing is standard. When agents with shell access run local models that break things, who's accountable? And Tech Luddite's MSP view: vendors are already failing security qualification on AI integrations. When LLMs land in enterprise management tools, the blast radius is every managed endpoint. Should Omarchy, as a community, be saying something about that now?
Discord is unusable for finding things. The Guild is one answer. Plugin catalogs are another. Neither stops five people building the same plugin and never finding each other. Fork or abandon? Forking grows the ecosystem and fragments it. Who says "this already exists, go contribute there" before somebody burns a weekend?
Wappy wants it on an M4. Heath wants it on TPU racks. Tech Luddite wants to roll it out to a pizza shop from a repo. Nobody publishes a supported hardware matrix. Is there an Omarchy server, or is that rumor? Business fleets are a different support burden from enthusiast laptops, and Omarchy Kids and Omarchy Fleet are each one person.
Developer, coder, programmer, orchestrator, mid-level manager, vibe coder. It sounds like vocabulary. It isn't. It decides who feels allowed to say "I built this" and who gets taken seriously when they file a bug. Tech Luddite's answer, "software developer, because that's what I actually had to learn," is the best one I've heard. Nobody's adopted it.
OmaStorm works because US NEXRAD data is free. Somebody will send a PR for German radar and Wes may not be legally allowed to merge it. Blip works because the Mac's iMessage database is readable. Apple could close that tomorrow. Which of our plugins depend on access that could disappear or doesn't exist outside the US?
Building in the open got lonely for Tech Luddite. Wes can't keep up. Joel isn't sure he wants the work of going public. I've got twelve contributors on Blip and an agent doing triage because I can't. "Swarm to help" got proposed. Nobody defined how you ask for the swarm or who answers. Who backs up a maintainer who goes quiet for a month?
X dropped a room of 56 people. The chat couldn't be opened. Speakers couldn't unmute. This is our community infrastructure right now. Is X Spaces the venue, or the discovery layer for a venue? Where do the recordings and summaries live so they're searchable?
Zoom all the way out on what DHH is doing and the goal is that Omarchy stops being for developers. Every single person on that call was a developer of some kind. What's the first non-developer use case that actually ships, and who's building it? Meetups are the first thing pulling in people who don't live on GitHub. What do we do with them when they show up?
Machine readable: asks.json.