EPISODE 01

Omarchy! Omarchy! Omarchy!

Sunday, September 13, 2026, 3pm EST95 min7+ speakers130 tuned inrecording 1:11:47
Listen to the Space The post on X

Summary

as posted on X by @NixFred

Omarchy! Omarchy! Omarchy! X Spaces Episode 1:

Listen here: https://x.com/i/spaces/1nxeLMVNppbJX

Recorded Sunday, September 13, 2026, on X Spaces. 95 minutes, 7+ speakers. First of many. See you next Sunday!

What came out of it was better than a show-and-tell. Underneath the demos there's a set of hard problems every person on that call is hitting from a different angle, and some are solved some not. The second half of this is about those. The first half is what got said.

WHO WAS IN THE ROOM

@ncfrontiersman Wes built OmaStorm, the NEXRAD radar plugin, and is in the mix on Omarchy Mac. Me, co-host: Blip & Infomarchy. @thisisgm who wrote Flea. @gh_TechLuddite, 13 years at an MSP, found the security holes in Blip, building Opinionated Omarchy and Omarchy Fleet. @thetundeo in London, runs ZeroRAMP, hosting the first Omarchy London meetup on September 24. Joel, Rails developer, built http://theomarchyguild.com. @MuchmoreIT, a 25yr friend, Slackware 1.0 vintage, AMD only, $1B PB data centers.. @1280MHz, who found us through OmaStorm. @Wappy_x, who wants to know when Apple Silicon M4 lands. A chemical and electrical engineer running Omarchy on an old Windows laptop who called himself "a hardcore vibe coder." Plus a room full of awesome folks who just listened.

PART ONE: WHAT WE TALKED ABOUT

The infrastructure guy's advantage

Wes and GM were on the point that writing code faster doesn't make verification any less important, and if you're not careful, verification is where the money goes. @thisisgm does a lot of adversarial review, throwing diffs into separate code reviewers so one session doesn't "loop around in useless workflows." Everybody in that room has watched an agent chase its own tail.

I pushed back on one thing. Wes had said it might take a real programmer just as long to write something themselves as to have AI do it. I agree, for programmers. I'm not one. I do enterprise storage for a living. Yawn, I know. I know what good looks like when I see it, but I couldn't sit down in an IDE and start writing Java. For people like me, AI didn't make coding a little faster. It made it possible. Wes's take: infrastructure and DevOps people have more agency here than programmers, because we're used to making things work at the edge. That part of the job didn't get compressed. The typing did.

GM said it for all of us: "Contrary to popular belief, I am an infrastructure guy. This would not have happened if it's not for AI."

OmaStorm: a radar app built over a long weekend

Labor Day weekend. A new model drop and Fable 5.1 landed right as Omarchy was taking off. Wes stood Omarchy up in a VM, fell for how configurable it was, and asked himself what he uses every day. Weather radar. He's been hooked on weather since Hurricane Andrew.

First question was data, because "if you don't have the data, it's just a toy." He assumed a commercial license was coming. It wasn't. In the US, every NEXRAD radar reading gets published to a public S3 bucket. Free, no license. So he pointed Fable at the bucket, asked how to render it fast, and asked for it to look 8-bit and pixelated instead of the smoothed-over radar every phone app shows.

The architecture decision is the interesting part. Omarchy plugins aren't compiled. Whatever's on main gets pulled down and run, interpreted, on install. Great for iteration, terrible for a rendering engine. So Wes split it: the plugin is the "dumb terminal," and a headless daemon does the work and talks to it over a socket. First try in Python, too slow, torn down. Second try in Rust, a language he didn't know, picked because the problem was bounded (same data in, PNG tiles out) and memory management mattered. First time it rendered he couldn't believe it worked.

He tweeted it. It's around 100 stars now with a steady stream of contributors, several of them on the call. His line on the whole thing: "It reminds you of what's good in the world. Building cool software is something we can all hopefully agree on."

One thing he's too polite to say. Last week Larry chased a stall in OmaStorm's live poller at volume rollovers all the way upstream into the NEXRAD parsing library it depends on. It wasn't OmaStorm's bug. That's what "swarm to lift the maintainers" looks like in practice.

@gh_TechLuddite : three years fighting AI, five months building with it

He's worked since he was 12. He's 42. Thirteen years at an MSP, with a CTO he calls the best leader he's ever had. For the last three years his job, in his words, was to be "a meat proxy for safety and data governance." You can't let a tool like this run wild among people who don't understand the blast radius. Then his CTO gave him a reality check: it's time. He's been learning in earnest since May.

His warning comes from somebody who does vendor qualification for a living. He's watching companies switch vendors because the vendor's AI implementation doesn't meet basic security standards. He brought up the McDonald's data leak as a system that didn't need to be engineered the way it was. And he said, flat out, that when these LLMs land inside MSP and enterprise management tools, "the IT AI apocalypse is inbound."

His first learning project was building an AI to play an AI in Universal Paperclips. Not a joke. He wanted fast discernment in a system with no API, then action in another gated system, which is a real shape of problem.

I asked how he found the holes in Blip. He wanted to use it, forked it, and ran his agents across the fork for safety, "because I'm freaking paranoid." Then he couldn't help sharing what he found. And he added the thing that stuck with me: I was the first person who responded to him at all and welcomed his additions. That's why he's here. That's why any of us are here.

@thetundeo found Omarchy by accident, watching a review of a 5K monitor he was shopping for. The reviewer was running Omarchy. He didn't connect it to @DHH, whose products he'd been using for a year and a half, until later. He handed his MacBook Pro to his daughter for uni, bought a Dell XPS 14, and installed Omarchy. His CTO and two developers are on XPS 16s. Only the designer is still on a Mac, and if 37signals ships a creative suite, the designer converts too.

DHH asked for meetup volunteers. Tundy put his hand up without thinking. First Omarchy London meetup is September 24 and it has 70 signups. He also had an obscure streaming webcam with no config software on Linux. Built the config tool in two days. Once it's tidy, it goes public.

@joelga from the Omarchy Guild, because Discord is a mess

@joelga was already building http://railsguild.com, a verified portfolio site for Rails devs. When the Omarchy wave hit, he cloned the idea and built http://theomarchyguild.com in a couple of days. Sign in with GitHub, it pulls your repos in. There's a leaderboard. The motivation was blunt and everybody agreed: the Omarchy Discord is huge, vast, and impossible to find anything in. Wes: "Maybe we're getting old."

The Guild repo is private right now. Joel will open it if there's enough interest, with the honest caveat that opening a repo creates work.

What do we call ourselves?

Joel told me anybody who's seen my output would say I'm a programmer, whatever I call myself. I said what I always say: I give Larry an idea and he comes back close to done. It's my ideas, not my coding.

Tech Luddite had been wrestling with exactly this. He doesn't feel like a coder. Where he landed, the night before the Space: what he actually had to learn in the last five months wasn't AI, it was software development. So he's a software developer now. Not a coder, not a programmer.

Joel offered a second frame: mid-level manager. "Hey agent, go build this," then making sure your little sibling did what you asked. Tech Luddite had tried on orchestrator and coordinator and rejected both, because they leave out that it was your idea and you drove it the whole way. And it's more than vibe coding when you're dedicated to it, fighting confabulation, insisting on accuracy.

Wes's answer covers all of it: the definition's changing, and what it comes down to is, can you verify it, and do you have boundaries.

Amalgamafi thesis: small local models are the new app store

For Omarchy to win for real, we need small local models trained on everything the community is learning by using Omarchy. Most of your token spend right now is the agent rediscovering your machine. That should be local, in a model you download like any app and plug into your main agent. And the bigger idea: this is the new app store, and it's open source. A sub-gigabyte model that already knows plumbing, on a hundred-dollar computer, already wired into the Omarchy environment.

His proof of concept: he told an agent to grab Chromium, rip it apart, and run models in Wasm inside the browser. Fed it a video stream and got an hour of transcription back in about twenty minutes, locally.

My answer was two numbers and a claim. Right now about 8% of my prompts go to my local machine. I think that's 50% by this time next year, and the gains come from quantization and open source software, not hardware. And the people you'd hope are thinking about a model trained directly on Omarchy: I can confirm they are. That might be one of the big announcements. I said "maybe." I meant it.

Wes summed it up: "The cavalry ain't coming. We're in."

I commented...."the people who put $15 million into Omarchy didn't put it in to get where we are now. We're 10% into this. In a year we'll say "I can't remember before Omarchy" the way we say "I can't remember before AI."

@gh_TechLuddite 's counter: Opinionated Omarchy

He hears "train a model" a lot and isn't a fan. His project is Opinionated Omarchy. The idea: AI fixes for Omarchy should be consistent and cheap, for every user, everywhere, without a cloud. Can a model do that? Yes. But training one is expensive to build, expensive to iterate, and hard to peer into when it's wrong.

His experience with quantized local models in production roles is that they break things. He's had a local model running shell commands go sideways and had to bring in a frontier model to repair the session. So nothing local runs in a production role for him.

What he's betting on instead: skills and a tight troubleshooting corpus. Skills cost more input tokens, which are cheap, and produce fewer and better output tokens, which are expensive. "What's the point of saving input tokens if it gets you output you can't use?"

Opinionated Omarchy today is a troubleshooting corpus with a GitHub Pages front end. Next stage: a skill plus a SQLite database so it can be searched to shape the input before the model ever sees it. MIT, run it locally. His own agents called the repo "the most documented empty folder in all of GitHub." He asked, in plain words, for help.

He's also got Omarchy Fleet, built on the Omarchy Kids work. "Business by repo": a pizza shop is a repo, combine it with Fleet, and you stand up their machines from it.

Apple Silicon, AMD, and big iron

@Wappy_x asked when M4 support lands. Wes: M1 and M2 are the current focus, M3 and up come next, and it won't be in the upcoming 4.5 release. Omarchy Mac lives in a fork today, and every bug fixed in one place has to be ported to the other. Getting it back into the main repo is "stop the bleeding first," then things accelerate. Help is welcome.

@MuchmoreIT came in from the other end. AMD only, testing Omarchy across Radeon boxes because that's what he builds at scale: EPYC, NVL72 racks. He installs Google TPU systems on-prem at tier0 companies, the BIG ONES. His contribution is going to be stretching Omarchy onto bigger and bigger systems, the way CIQ did with Rocky Linux. Omarchy linux? @dhh y'all need to connect. Known Heath for over 20yrs and I'll say what he won't, he's the real deal.

Heath also wants to collaborate with Wes on NOAA and he wrote a hurricane model against the same NEXRAD data OmaStorm reads.

Building in the open, and building together

Wes made the case for building in the open. It takes vulnerability. But people learn, you learn, and the community gets lifted. His ask: if anybody's drowning in PRs, let's swarm and help.

@gh_TechLuddite 's reply was the most quoted thing on the call. He started building in the open five months ago and it got lonely, more lonely than building privately. What changed it was finding a community to build with. "Build in the open, but find ways to build together." He also gave the maintainer's side: when you send someone a PR, you're asking them to support your code. He's got things he wants to throw at Infomarchy that he won't, because he doesn't want me carrying that burden. That kind of restraint is rare.

I said the obvious thing people avoid: you don't have to accept every PR. It's your baby. Be firm and nice. "I appreciate what you did, but I'm going a different direction." Somebody sent Blip a full Apple Contacts integration, and the right answer for Contacts is http://icloud.com/contact installed as a web app. I felt terrible saying no. I still had to.

@ncfrontiersman went further. Code generation is the easy part now; the approach is the hard part. Open an issue first, agree on the approach, then write the PR. Otherwise somebody burns their tokens building German radar for OmaStorm and the data license doesn't allow it. He's thinking about gating PRs behind approved issues.

@joelga proposed the thing I think happens next: a shared contributor instruction set any of us can drop into a project. When a contributor's agent reads the repo, it's primed to help the way that maintainer wants. Tech Luddite added: PRs have a draft status. Use it for discussion. And like I said on air, it's a pull request. Request is implied.

Blip, Claude on Mac, and what I've learned about memory

Before Omarchy, I had Larry on an M4 Mac and asked it to go look at what it could see inside the Apple ecosystem. Turns out the iMessage database isn't encrypted. It's wide open. That became a repo called Claude on Mac. It gives Claude Code or Codex the ability to read and write Apple Mail, Notes, Photos, Messages, all of it. First thing I did was have Larry read every message my brother and I ever sent each other and summarize the relationship. It came back with "you send 58% of the text, he responds, you respond faster." It can send on your behalf. I locked that down to ask every single time, and it's never burned me.

When I moved to Omarchy, the first idea was to proxy to the Mac over SSH. Then: why not a plugin? Version 0.0 was "show me my iMessages, quick and dirty." Then make it look like Messages. Then I got greedy: photos, contacts, send and receive. The setup is a Mac on your desk with SSH open (Amphetamine keeps it awake), Tailscale so you can be anywhere (if you don't use Tailscale, shame on you, go get it), and the plugin pulling from the Mac directly.

Twelve contributors now. Folks like @gh_TechLuddite got in there and showed me I had security problems. They were right. How I run it today: Larry Prime on my Omarchy box keeps up with the PRs, merges the easy ones, and for anything that needs a decision it prompts me with a pick-list and a recommendation. If you use Claude Code, bang its head into the ground until it always uses AskUserQuestion with recommendations. Nobody wants "here are the twelve things I need you to answer."

I started building Larry about 3 days after Claude Code came out, out of frustration at re-explaining everything every session. Each piece I built showed up in Claude Code native about a month later. Not because anybody saw my code. I'm nobody. But it told me I was on the right track. And here's what I've learned from pointing Larry at every new memory system that drops and telling it to cherry-pick whatever makes it better: every memory system saves everything. Memory isn't the problem. Recall of the right thing at the right moment is the problem, and nobody's solved it.

PART TWO: THE QUESTIONS STILL UNANSWERED

Every one of these came up on the call. None got resolved. Some of them nobody even owns yet.

Q1. Who owns the contribution protocol?

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?

Q2. Local models: train, or retrieve?

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?

Q3. Recall, not memory.

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.

Q4. Install and distribution.

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.

Q5. Security and blast radius.

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?

Q6. Discovery and duplication.

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?

Q7. Desktop or platform?

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.

Q8. What do we call ourselves?

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.

Q9. Data rights and regional reach.

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?

Q10. Maintainer loneliness and burnout.

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?

Q11. The platform we're meeting on.

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?

Q12. Not your grandfather's Linux.

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?

ASKS ON THE TABLE

Tech Luddite wants help on Opinionated Omarchy: skills, SQLite-backed corpus search, anything past skills. MIT, run it locally, "I would love anyone to even just talk." JoelGA will open the Omarchy Guild repo if enough people say they want to contribute. Say it. I want a real next-next-next installer for Blip. Point your agent at the repo, then come find me. Wes and the Omarchy Mac effort want help getting the Mac fork back into the main repo, then M3 and up. Tundy has 70 people coming to London on September 24. Heath and Wes are talking NEXRAD and hurricane models offline. If you've got weather data chops, that thread wants you. Everybody: follow everybody who was in the room. We're small. We've got the same likes.

NEXT EPISODE: Weekly, probably. Same format: raise your hand, tell us what you're building, tell us where you're stuck.

Don'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.

No politics, no ego. This is about Omarchy and nothing else. If you were in the room, you've got the same likes as the rest of us. Follow everybody.

"The cavalry ain't coming. We're it" - @ncfrontiersman

SEE YOU NEXT SUNDAY!

Transcript

machine generated · speaker labels are the tool's, not names
Speaker 1

I'm thinking about, like, yeah, we can write code faster, but that verification process is still so important because you're not... you're never really going to know if it's going to work like you think until you're actually exercising the binaries, the executables, you know?

Speaker 2

Yeah, and if you're not careful, it can get extremely expensive.

Speaker 1

Oh, totally.

Speaker 2

Like, I run testing on subscriptions, and I ideally like as much manual work as you can do you try to do, but at the end of the day you're tackling a bunch of different issues at the same time, and it's like, the machine is going to do a better job regardless. You, as the human, have to, like, actually make sure it's doing the right things and verify as much as you can.

Speaker 1

Yep, totally. Totally. Well, let's pause for a sec.

Nix, did you get connected?

Fred Nix

Hey, man. I think I'm on.

Speaker 1

All right. Awesome. So, uh, GM was just telling us... I mean, I don't know what you heard, but kind of what it's been like with Flea and how it kind of just exploded overnight and...

Fred Nix

Well deserved.

Speaker 1

He's trying... yeah.

Fred Nix

Hey, I want to comment on, uh, you know, earlier you said you think it might take just as long to code it yourself, and I agree with that. Um, however, I'm an infrastructure guy by trade. I do enterprise class, uh, storage systems.

Uh, yawn. I can feel everybody yawning from here.

Um, but what... AI might take you guys that are programmers just as long, but for me, I could code, but it might take me 10 times longer than somebody that knows what the hell they're doing. So for me, it opens up an ability to do something that was out of reach for me.

I know what good looks like when I see it, but I don't... I couldn't just, like, start out and write Java code, you know, inside an IDE. Um, so, you know, for some of us that are technical... I'd call myself technical, but certainly not programmers... you know, it reaches down and allows us to create things, you know, as long as we can explain it well and we know what good looks like.

Speaker 1

Yeah. 100%. Yeah, it's, uh, it... yeah, I mean, and I think, if anything, infrastructure folks, DevOps folks, you guys have a lot more agency in some ways than programmers because you're used to making the shit work, like wiring it up, you know?

Fred Nix

Exactly.

Speaker 1

At the edge, dealing with all the issues, and, uh, you know, for us that are, you know, we're programmers or whatever you want to say, like, that kind of has been compressed now.

Fred Nix

Yeah.

Speaker 1

And it's like, now, you know, we're solving that stuff too, but now it's like we also, you know, we're having to make sure it works, so.

Fred Nix

I also might say that it helps infrastructure people. It's given us a bigger boost than programmers.

Speaker 1

Yeah. Yeah, I mean, some, for sure, yeah. And there's all kinds of programmers, you know?

There's some people that just write the code, there's some people that, uh, are, you know, all about crafting the perfect, you know, clean functions and all that, and no shame to that, no shade to that, but.

Fred Nix

Yeah, and, you know, the... not all the PRs I've received, you know, we wouldn't be... I wouldn't be where I am or Blip or any of the rest of them. Um, if anybody that's listening wants to speak, please raise your hand and let's, um, let's get folks involved. I'm the co-host, I don't want to... excuse me... step on you ass, but.

Speaker 1

No, no, sounds good. Yeah, uh, real quick, though, GM, anything else you wanted to share? Um, I appreciate you coming in.

I thought you said you had work today or something, so I appreciate you jumping in.

Speaker 2

Yeah, I mean, contrary to popular belief, I am an infrastructure guy.

Speaker 1

Awesome.

Speaker 2

So I'm not as skilled when it comes to the UI and front-end stuff, so AI helps a lot, for sure. Like, this would not have happened if it's not for AI. And, yeah, uh, I have a bunch of processes that I do a lot of adversarial reviews with AI.

Like, I just throw the diff into code reviewers, and that does wonders for my workflow because sessions can get caught up in context if you're not careful, and it's going to loop around in useless workflows.

Speaker 1

Yeah.

Speaker 2

And, yeah, I mean, like, to summarize, yeah, Flea is a work in progress. I am grateful, uh, beyond words, for all the support it's been getting from all of you guys around the world, and, yeah, I'm going to keep working on it and hopefully things happen.

Speaker 1

Wonderful.

Fred Nix

Just for, uh, housekeeping, I would like to let everybody know this is being recorded. There will be summaries put back out later. Um, if you don't want to be recorded, don't speak.

If you'd like to speak, raise your hand.

Speaker 1

All right. Well, um, awesome. So let's see.

Anybody else, um, have something cool they're building you want to talk about?

Let's see. Um, trying to see who all's in here.

We've got some folks. Um, I mean, I can... I can take a moment if no one's got anything right off the top of their head.

I can talk a little bit about Oma Storm and how that...

Fred Nix

I'd love to hear about that. Oh, I'd love to.

Speaker 1

Yeah.

Fred Nix

How did it... you know, let me ask you a few questions. So how did the idea come to you? When was it that you went, "You know what, I'm going to build it"?

Did you kind of, like, start making something just for yourself and then go, "You know what, I think other people would want this"? And I've seen your work, Wes, and, you know, basically in my mind, if you're interested in it, I think a lot of folks are going to be interested in it.

So, you know, tell us the origin story and all the places you went with it and...

Speaker 1

Yeah. Uh, it was interesting. So Labor Day weekend, you get a long weekend, right?

New model comes out, Astra comes out, um, you know, we had Fable 5.1 out, and I... and at the same time, Omachi is just really taking off, and I'm like, "All right, I don't want to... " so I got it installed on a VM, uh, to begin with, and started playing with it, and I'm like, "I love this. I love I can configure things.

I love how easy it is to, uh, spin up stuff," and I'm like, "I think I'm going to build a plugin. I wonder, you know," and so I started thinking, um, what would I like to see?

What do I use the most? And it's... I've loved Wether.

I've loved Wether since Hurricane Andrew years ago. Uh, been following it, um, and, you know, something that I do pretty much every day is open up the radar app on my phone, and I'm like, "All right, where are the storms coming from?

Where are we tracking?" I'm like, "Man, I bet I could make a sick radar app on Omachi that... " and I'm going to say Omachi, so y'all just have to be okay with that, but, uh, a sick radar app, and, uh, it's like, I want it to, you know, I gave... I started out with Fable, and I'm like, "All right, I want it to look like it was built in, you know, just like the other plugins: network, um, the agents plugin, like, just native to how it is,"

um, and I said, "You know, let's talk about where we can get this data from. We can start from the US," because, you know, it always starts with, "You could build cool, you know, apps and plugins, but if you don't have the good data, you know, the access to it, then it's just a toy," and so my first thought was like, "Well, dang, I bet I have to pay for some, you know, commercial license and all that," so I did some research, and it turns out, at

least for the US, they put up all of the records, the readings from these radars that are, you know, they put up all of the records, the readings from these radars that are, you know, uh, reading this information up on an Amazon, like, S3 bucket that's publicly available for free to download and anybody can use, and so I was like, "All right, Fable, go take a look at that data.

Think about how we could, in the most performant way, have a plugin that shows the radar, uh, for a given location," um, and I want it to render it, and it should look like 8-bit, pixelized, you know, not the traditional, like, smooth over radar that you see in some apps because it's just like, "Hey, that'd be cool," right? And so I started doing some mockups with it, um, not, like, just HTML mockups, and I got to design a light, and then I said, "All right, let's

think about this. Let's split it up," and, um, because of the way the Omachi plugins work, it's kind of like whatever you push to main, um, when somebody, like, up on GitHub, when somebody goes to install it, it pulls main down and runs it, and it's not a compiled thing. It's not like you're compiling, uh, the Qt code.

Like, it just runs, like, interpreted, and so I'm like, "That might be a little slow for, like, a rendering engine," so, you know, what can we do to have the plugin be the view, the dumb terminal, you know, like we used to say, and then the backend kind of run like it's this headless daimon, or demon, or however you say it, that, um, could then talk to it over a socket connection? And so I started riffing with, uh, with Fable, and it's like, "All right, I can

do something," and it started with Python, and I'm like, "No, that's not going to cut it. That's going to be too slow," so I had to tear that down, and I said, "All right, I don't really know Rust. I know, you know, I've got experience with others, like Golang and all the front-end, you know, JavaScript, TypeScript, I know Ruby, but as far as, like, you know, really fast, you know, I've heard the memory management in Rust is great, and, uh, you know, slow to compile, the

syntax is a little hard to follow," so I'm like, "All right, this sounds like a good use case. Uh, for something that's bounded, you know, it's going to always take this type of data in, and I want it to generate, you know, an output of PNG images that then the plugin can just lay on top of a map, a control," and so it got to the point where it started rendering, and that first time I saw it, I'm like, "Oh my God, I can't believe this works.

Like, this is so cool," and so I think I shared, like, a tweet of, like, you know, "This is something I'm building. What do you guys think?"

and, you know, everybody was like, "Wow, this looks incredible," and I was actually kind of surprised how, like, much people liked it. You know, I didn't know a lot of people like Wether or not.

Maybe that's a little bit of a niche thing, but, um, yeah, so I started getting out, and then it was like, "How can I, you know, release this to people so they can test it? What's involved with that?"

and, uh, pushed it up, and, uh, people started downloading it, and, you know, I think Omach has 100 GitHub stars now.

Fred Nix

Wow.

Speaker 1

I've had so many people contribute, um, people on this call, I'm sure, and, uh, it's just been fun. I did open source years ago, took a break for a while, and it's just been so good to be back.

Fred Nix

Awesome.

Speaker 1

Doing it again, um, because it's like this... it's like when we build as a community, um, it reminds you of what's good in the world, you know? We all get caught up in politics and everything else going on, but, you know, building cool software, you know, that's something that we can all hopefully agree on, so, um, yeah.

Fred Nix

That's awesome, man. Where do you see it going? Like, what's on your roadmap?

Speaker 1

All right. I don't know if we lost Nix or not.

Fred Nix

It muted me when you started talking. So what on your roadmap can you share?

Speaker 1

Cool. Well, that's enough of me talking for a while. Let's see.

Anybody else, uh, have something they want to chat about? Any questions or?

Fred Nix

Can you not hear me?

Speaker 4

Yeah, check, check. Can you hear me all right?

Speaker 1

Yeah. Yeah, man.

Speaker 4

Yeah, awesome. I, uh, I found you through Oma Storm, and I installed that right away when I saw it. The vibe is real.

And Wether, I don't think it is a niche thing. I think that's a universal thing, right?

Speaker 1

Yeah.

Speaker 4

It doesn't matter where you are in the world. It touches you, and it's something that belongs in the operating system, and we've seen how that's happened with other operating systems. I mean, Windows tried that stuff, but Oma Storm's awesome, and I think that that captures a lot of.

Speaker 1

I think... did it cut out?

Fred Nix

Can you unmute me, maybe?

Speaker 1

Let's see.

Speaker 4

Maybe he was hit by a storm.

Speaker 1

Yeah, maybe so. All good. All good.

Um, we'll catch up with him in a minute when he comes back. Uh, let's see.

Anybody else building anything cool?

Doesn't have to be a big deal. It could be small.

It could be a theme, you know, figure out how to set up something.

Speaker 4

Just, um, yeah, prevent us from doing that stuff. There's great hardware that is locked up, and to be able to break those walls and access that hardware with open-source software is the coolest thing to me, so I'd love to hear Nix talk about that a little bit and hear about some other things people are working on in that vein because that's what really gets me going.

Fred Nix

Hey, I'm back. I just restarted. Something muted me, and the button wouldn't unmute.

Sorry about that. I believe your question is a bridge away from, like, Microsoft and Apple?

Is that what you want to know about?

Speaker 4

Yeah, I mean, that's just what gets me fired up, is breaking the walled gardens and allowing open-source software like, like, Omachi.

Fred Nix

Yeah.

Speaker 4

And Linux and whatever to run on Apple hardware and whatever hardware. Phones even. It's coming.

Fred Nix

Yeah. If you look at what DHH is really trying to do, and you upclick the highest level, um, Omachi, it really... we've got to get to the point where it's not for developers anymore, you know, it's not your grandfather's Linux, if you're old enough to know what that means. Um, so when I started Omachi, uh, you know, I struggled with the keyboard as well, but I knew that there was something different about it, and so I think I alluded to Wes as like, "How did you get

started building it? Was it just for you?" and Blip, turned out, for me, was one of the earlier things that I started tinkering with, and, you know, again, I think I've said it enough.

I'm not a programmer, so, you know, I just started talking to Fable and saying, "Hey, here's what I'd like to do," and my first instance, you know, I was shooting low, and I just said, "Is there any way that you could just make, you know, if I get a message, you know, from, um, from my message, is there any way we can bring that on?"

and, um, and I'll mention things here, and please don't think I have an ego about it. Um, I don't like talking about myself, uh, but the story includes it.

So I've been using and building what I call Larry on, uh, Claude Code since about three days after it started, and I didn't understand exactly what it was, and I had frustrations, and I started, you know, couldn't remember crap, and so I started down, you know, what thousands of people did was, you know, "How can I make this thing, you know, how do I get to the point where when I start a session, I don't have to tell it everything fucking again?" Um, and over time, the

interesting thing was that each of the pieces that I built for Larry, um, Anthropic started, and it wasn't... certainly wasn't because they saw my code. I'm a nobody. Anthropic started building these pieces into it, and so each time I'd make something custom for Larry, about a month later, it would show up in Claude Code, um, which, again, they didn't see my stuff, but what it did do is tell me that I was on the right track, um, and there's lots of folks that have better memory systems

out there, but Larry, for me, is so customized now, um, when I do see a good memory system, and I have no shame in this, uh, even just last night, I saw a new, good memory system, and I wish I could repeat the name of it, but I don't remember. Um, I believe I retweeted it, and what I do is I point Larry at the repo and, um, unabashedly just say, "Hey, go check this repo out and see if there's anything we can cherry-pick to make

your memory better," and what I've found, and it's really started to hit home in the past three months, is that everybody's memory system that they do, and I'm assuming everybody knows what I mean about memory, is just, you know, not having to explain it over and over and, you know, telling it to take the fucking em dashes out, you know, they drive me crazy, um, and over time, Larry, you know, started understanding what I wanted and what I was trying to do, so new memory systems come

out, I point those people's repo and say, you know, um, without any sense of shame, "Go pull out the pieces that'll make yourself better," and what I've learned in the last three or four months is that pretty much anybody's and everybody's memory systems that they're using for codecs, for, um, uh, for Claude Code, and the one that I really built mine off of was a guy by the name of Daniel Miesler, and I'm probably killing his last name, but you should look him up. He has a

thing called LifeOS, I believe it's ourlifeOS.com, and I've long since left his code behind, but he was kind of the guy that got me started. Um, I'll leave comments for him on the table. Um, good dude.

Um, anyway, what I learned in the last three months is that everybody's memory system pretty much has 100% memory save rate. All the memory systems save everything.

Memory is not the problem. Recall of the right data at the right time and the right situation is the hard part.

So fast forward, and let me know if I'm getting too long. I'm happy to shut up.

So I got onto Omachi. I'll back up a little bit.

So when I was coding for Larry, it was on a Mac, just a little M4 Mac, and I started tinkering around and asked Larry to go look into the, uh, Apple ecosystem and what it could, you know, where it could see into, and come to find out that iMessage database is not encrypted. It's pretty wide open, um, and so over time, I've got a repo out there called Claude on Mac, um, and you can go grab that, and what that did... this was before Omachi... and what

it literally does, and you can install it, it's fairly well tested and used by a couple of folks, maybe a dozen. What it does is it unlocks your Claude or your codecs to reach into your Apple Mail, your Apple Notes, your Apple Photos, your Apple everything. Um, so it has the ability to make modifications and do those things, and so as I tightened that up, iMessage became one of them.

Now, I didn't... I stayed inside the Claude Code CLI interface, so for me, Larry was just able to, um, go read my iMessages and, um, you know, what I did early was just say, "Hey, go read my 1,500 messages," because I never deleted an iMessage, "Go read all my messages back from my brother on iMessage and, you know, summarize our relationship," and, like, it would go down and do details like you send 58% of the text, he responds, you respond quicker, you know, give me all kinds

of data. Um, it can send on your behalf, um, but I locked it the hell down.

Hey, Sir Jill, thank you for joining. Uh, X just dropped us.

We're hoping to get everybody back in.

Give me a thumbs up if you can hear me, sir. It'd be helpful.

Oh, man, we had 56 folks on and it died.

Hey, thank you, Sir Jill. Appreciate that.

Hey, we're going to get started here in just a few minutes. I'm waiting on Wes to rejoin.

Uh, X just dropped the session, sess the, uh, X system, so sorry about that. Hopefully, we can rebuild some momentum.

Hey, Wes, go ahead and request being a host.

Hey, can you hear me, guys?

Speaker 2

Yeah. I, I, uh, actually shared another one. I think people are maybe over in that one.

Fred Nix

All righty. Well, where's the... where's the net?

Speaker 2

Where should we go?

Fred Nix

We'll join.

Speaker 2

Uh, let's stay here. This is fine.

Fred Nix

You want to tell them to rejoin? Nick's read?

Speaker 2

Yeah.

Fred Nix

I, uh, actually shared another one. I think people are maybe.

Speaker 2

Oh, weird. That was me.

Speaker 4

Like my dad.

Fred Nix

Just end it, Wes, and or tell people to come over here and end it.

Speaker 2

Okay. One second.

Fred Nix

Yep.

Hey, 1280. Um, I know we had plans to do some other stuff, and I know I got to this before I got to you, and I want to publicly let you know that, uh, I meant no disrespect.

Uh, we will get to it.

Speaker 4

Thanks, brother. Don't feel bad. Uh, part of it's my schedule.

I got a lot going on. I got two kids and running a small company, you know, it's.

Fred Nix

Yeah, man. I understand.

Speaker 4

It's all right. We'll link up. You know, there's always so much going on.

Fred Nix

Right.

Speaker 4

That I think it's kind of fine. It doesn't even matter when we talk, because whenever we talk, there's going to be some groundbreaking stuff going on. It's going to be a crazy week, I'm sure.

Fred Nix

Absolutely. Well, you know, I study human behavior, and the most powerful negative human emotion is being left out, and I don't want you to feel like that, and I did not want you to think I was skipping over you or for some reason. Anyway, hey, um, Tech Luddite, I'm so glad you're here.

Uh, would you request to speak? I'd love to chat with you and hear more about you, sir.

Here, I'll just throw you... if you want to. Um, if you don't want to speak, just deny the invite.

It's all good. Um, where were we?

Um, should we keep going or?

Speaker 2

Yeah. I was trying to cancel the other one. There's no way to do it.

Like, I don't.

Fred Nix

Just end it at the top, right?

Speaker 2

Uh, well, I have to go out of this to do that, so I'm trying to do it on my laptop.

Fred Nix

Um, you know what I would do? What would help me is just close your browser and just rejoin us. It'll actually put you back as speaker automatically.

Speaker 2

Okay. All right. One minute.

Fred Nix

Be back. Hey, 180. I appreciate... or 1280.

I appreciate you, uh, talking to my other buddies, um, that I introduced you guys to. Uh, that's really cool.

Tundy. My man, Tundy joined.

Oh, he disappeared. Hey, everybody, I really would encourage everybody to follow around each other.

Um, we're a pretty small community. Um, you know, we don't do politics.

We don't do ego. So, you know, just hover over everybody and follow everybody, and if you join this session, um, I'm following you for sure, and if I miss anybody, please poke me and let me know.

Um, I don't give a shit about ratios and all that. There's Tech Luddite.

Hey, man. I really appreciate all the work that you've done on the PRs and helping me out.

I'd love to hear your story and, uh, how did you find all my bugs that I... they're like, how did you do it, man? You're amazing.

Oh, we can't hear you, Tech.

Speaker 2

Yeah, I was just reaching for that, uh, mute button, hoping my Omachi hotkeys would hold up. Uh, here I am. Am I coming through okay now?

Fred Nix

You're good, man. You sound great.

Speaker 2

Excellent. I think this might be the first time I've spoken to anyone through this laptop. Uh, my, uh, my wife got it for me, uh, for a birthday present just before that, uh, vacation, which led me to find your software.

To answer your question, I was, uh, just going wild with this laptop on vacation.

Fred Nix

Nice.

Speaker 2

Like, uh, like a typical bad tech dad.

Fred Nix

Yeah.

Speaker 2

Uh, but I did minimize it, and, uh, and your software really helped me through that, because trying to put a little bit of time in a lot of things.

Fred Nix

Our software.

Speaker 2

Yeah. Uh, I'm at a loss for words, honestly, and to respond to that here with others is, uh, is a lot, because you're the first person that I think responded to me at all, and you welcomed my additions, and it's hard for me to even articulate how much it means to me, so thank you.

Fred Nix

You bet, man.

Speaker 2

And, uh, yeah, so, uh, you asked my story. I have a tendency to burn time, and I know we don't have a lot of it, so I'm going to condense it to, like, four sentences, if I can even do that, but, you know, if I had more time, I'd write a shorter letter.

Speaker 4

Wait. Let me stop you. We got all the time in the world.

You know what I mean? Like.

Speaker 2

Not this kind of time, uh, but, uh, yeah.

Speaker 4

You're saying.

Speaker 2

We all have long stories too, so I would more than as much as I love to rant, uh, and let me tell you, my coworkers, um, suffer it, um, I would love to hear a little bit about a lot of you, and so I'll try to make mine short, and if anyone gets on a roll, I'd love to hear it all too, but please, uh, I take no offense to me being stopped, and especially no offense to being corrected. Like.

Fred Nix

I'd rather be corrected than wrong.

Speaker 2

You know, I'm giving people classes left and right and then having multiple AI, uh, systems audit my own speech and bring it back to me so we can reiterate on it.

Fred Nix

Right.

Speaker 2

Because there's a lot of people need to learn so much, and it needs to be taught in a, in the right way.

Fred Nix

Hey, if anybody out there knows how to open up the chat space, can somebody type something into the chat space to get the chats running so we can drop links in here? I can't find it. And also, we have a very special guest from the UK, somebody I really want to talk to.

Uh, I may murder his name, and I'm not even going to try the last name. Tundy, you know it's you.

Would you request to talk? And, uh, if you don't mind, if you don't want to talk, I get it, um, but I'd love to speak with you, man.

Anybody else that wants to speak? Hold on.

I'll get it. I'll invite you.

There you go, Tundy.

Um, let me finish up my rant, and I'll keep it a lot shorter on how Blipp got there and get my crap out of the way before a bunch of people join so we can talk about more important stuff. The bottom line is, um, I don't know if you heard the last, but Claude, I built Larry.

Larry, uh, before Omachi came out, I, uh, I really wanted to integrate the Apple, um, infrastructure with notes and photos. I didn't have a UI.

I just gave Larry the ability to go look at all my Apple information. He could write Apple notes.

He could send iMessages, but I locked him down where he has to request every single time to do it. Now, I know that's just best effort.

He never burned me and sent iMessages. I don't really have, uh, my business stuff is elsewhere, so I don't... there's no real danger of sending the wrong person.

So I had this, uh, Claude on Mac software that's a repo. You can go find it out there if you want.

And it gives your either Codex or your Claude code the ability to really, um, opens up the Apple infrastructure. And so when I got over to Omachi, and I kind of got a week into it, it occurred to me to go ahead and, uh, give the Larry... and I use Larry everywhere.

It's either my MSI laptop that I got a couple days ago is my Larry Prime. Uh, there's probably 15 Larrys out there, which does cause problems, but, um, at any rate, I thought, why don't I install, um, Claude on Mac on my Omachi box and then just, um, have it proxy over SSH with an SSH key over to my Mac and get the stuff that I need?

And then I started thinking, well, shit, why don't I just make a plug-in? So, you know, version 0.0 was, all I want you to do is just show me quick and dirty, uh, any iMessages.

And then I thought, you know, make it look more like an iMessage machine. And then I kind of got greedy.

Like, I want the photos, and then I said, um, I want to be able to, uh, send and receive, um, contacts. And I started exploring it, and the way it works is you have to have a Macintosh, uh, with SSH.

I recommend, I mean, a software that's free called Amphetamine that keeps it open without shutting down. Uh, and that's important.

Um, there are projects and there are plug-ins that will do it directly to your phone through Bluetooth. There's trade-offs there, and I'm not going to go into those.

The main reason that I like having, uh, not using Bluetooth on my phone is because of Tailscale, and if you don't use Tailscale, shame on you, go get it. Um, tailscale.com.

And so my Mac sits on my desk at my house. I can be anywhere because of SSH keeps the SSH open to not to Larry, just directly to the Mac, and it delivers, um, my messages.

And then I think it's... I think we're up to 12 contributors now, and folks like, uh, Tech Luddite and some others really got in there and showed me, like, Fred, bro, like, you got security problems, and this is completely fucked up, and, you know, and what it's become, you know, basically now I just have Larry, uh, Prime on my Omachi box, um, you know, keep up with the PRs. Um, and I tell him, you know, if it's an easy PR and it makes sense, just go ahead

and do it and commit push. If it's something I need to think about, um, you know, prompt me with ask user questions. If you're a Claude code user, bang its head in the ground to always use ask user questions with recommendations.

That's the menu that allows you to pick. Uh, one of my pet peeves is when it goes, "I need some questions from you.

Here's the 12 things I need you to answer." I'm like, "I'm not typing all that."

So at the end of the day, and I'm going to bring it home and let somebody else speak, and thanks for the platform. Um, I couldn't do it without all the contributors.

Um, it's really solid. Uh, it's pretty easy to move around.

If anybody's interested in a tighter installation, uh, method where you literally do next, next, next and install, I'd really appreciate it, and I'd work on it with somebody. Right now, the best way to install it is to point your AI at my repo and say, uh, you know, "Help me get this installed," because there's a lot of things that are missing on the install side.

And now I'll shut up. Tundy, my man.

I want to know all about you, man. We started talking a couple times, and you have a really interesting story, man.

The floor is yours. Unmute.

Tundy.

All right. Well, Tundy, uh, works on getting unmuted.

If anybody could help one of the speakers that's muted, could you direct him to where the unmute is? Um, what?

Speaker 2

Yeah. Lower left on mine, and I'm in Firefox. Uh, lower left-hand corner.

If you've gone back to X trying to get into the chat, uh, the bottom, like, footer should relink you back into the live session.

Fred Nix

Yeah. And one of the anybody that's a speaker, uh, jump in here. I've been talking too long, and this is not the Fred show.

Speaker 2

Well, I, uh, I didn't actually deliver those three sentences before that I was thinking of. I can try to cram that in while I do the, uh, the mute, uh, situation to resolve. So, uh, as some of you may have seen, if you've been watching my post, I'm trying to make it apparent that, uh, that I spent the last three years fighting, uh, AI at work, not embracing.

Fred Nix

Yeah, you mentioned that.

Speaker 2

Yeah.

Fred Nix

You're an MSP guy. I hope that's not too much to share.

Speaker 2

Uh, I do work for an MSP, and.

Fred Nix

Sorry if I outed you.

Speaker 2

No, that's all right. Um, none of my personal work, uh, has, like, a full-on direct relationship to them, although I do give them credit and shout out in some of my projects. I've worked for them for 13 years, and my CTO is the best leader I've ever had, and I've had a few.

Um, I've worked since I was, like, 12, and I'm 42, so, yeah, I've seen all types, and, uh, so we're presently, uh, a private company and not doing open source, but, uh, without saying too much, it's on the table, at least, and we're discussing it, and we're doing a, you know, AI is changing everything, so.

Fred Nix

It's a big step.

Speaker 2

Hopefully, there'll be a story later to share there. Go sorry, go ahead.

Fred Nix

What are your concerns?

Speaker 2

Like, there's got to be a lot of people out there in your shoes that are thinking, "I'm not sure about this." Like, what's your thought process, and what will be the decision maker for you? Yeah.

Yeah, I guess. Well, let me actually take a stab at those three sentences, and if it doesn't become apparent, I'll wrap up with it.

And it's basically that, uh, you know, when there was this big explosion of capability that, uh, came along with, uh, ChatGPT's user explosion, uh, what, three years ago, three and a half years ago. Um, and at that time, uh, I wasn't going into it right away.

I'm intrigued by machine learning and anything computer science, although I'm, you know, not, like, a capable, uh, researcher or anything like that. I'm no computer science master, right?

But, uh, I find it intriguing. Like, it's computer poetry is bad, and I love it, right?

It's just I'm into it. And, uh, when I saw this all coming about and I did some early testing, I mean, let me tell you, I have corrected so much of this output is unbelievable.

Uh, and not until about five months ago did I really start focusing on, um, fixing output a little bit upstream.

Fred Nix

Right.

Speaker 2

Right? And I think the term, uh, and it feels weird in this position now, living effectively living this way for three years, but I was a meat proxy for safety and data governance purposes and, you know, all things MSP. You cannot let a tool like this run wild with people that don't have a deep understanding of it or the blast radius that it's about to touch.

And so, you know, I helped us take benefit, uh, in a safe but really ineffective way. And my leader, this is why I mentioned him earlier, um, gave me a reality check that it's time.

Like, I can't even I'm doing having to redo vendor qualifications and dig into how companies, uh, are implementing AI and, like, switching vendors because they're not adhering to good security standards with how they're doing it. You know, like how McDonald's lost everybody's info.

They did not need to engineer their system in the way that led to that and others. And there will be many.

Um, and in some tools that these LLMs are coming into, like MSP tools and enterprise management suite tools, when people start making mistakes with these, I mean, I see the IT AI apocalypse as inbound. Uh, but I'm also hopeful now, too, that I've spent that my leader, uh, really gave me that reality check and put me in a different mindset, and I've since May, um, I've been trying to learn it.

And my first project, uh, shout out to Frank Lance, I believe his name is. He made an autoclicker way back in 2017 called Universal Paperclips.

It's a wonderful, uh, web masterpiece about an older thought experiment related to machine learning. Uh, please take a look.

But my first, my first, uh, project to learn what's what now was, uh, using AI to make AI to play as an AI, that's you, in Universal Paperclips. Uh, and, um, I just thought it had some good crossover for things that we need, like fast discernment in systems that don't have any API.

Fast discernment and then action in another system based on info, uh, that's gated. And, uh, I had a lot of fun doing it, and I have not.

Fred Nix

That's really cool. That's really awesome. And I bet a lot of people appreciate it, um, you know, because there's insight, and it's something that everybody's dealing with.

Wes, what you got? Oh, and Wes, go ahead and speak.

Tundy, you ready, man?

Speaker 4

Yeah. Yeah, I want to hear what's been going on with Tundy. How's it going?

Fred Nix

Am I saying it right?

Speaker 4

Sorry, man. Hi, can you hear me?

Fred Nix

I can. Am I saying Tundy right?

Speaker 4

Yes, it is correct. I'm so sorry if the, uh, I couldn't get the microphone.

Fred Nix

No, man, it's all good. It's all good. Spaces are new.

We're going to get this thing tightened up, though, for and probably do it quite at least weekly. So tell us about yourself and how did you get to Omachi and what do you like about it?

What are you doing with it?

Speaker 4

Okay. Um, I'm Tundy. Um, I run a small software company here in London called ZeroRAMP.

Fred Nix

Can you give us the URL or put it in the chat?

Speaker 4

Uh, yeah, I'll do that. It's zeroramp.com, Z-E-R-O-R-A-M-P dot com. Um, how I discovered Omachi was by accident.

I was shopping for a 5K monitor, and I watched this gentleman's review. I think everybody knows Linux, by the way.

Fred Nix

Yeah.

Speaker 4

A popular YouTuber, yeah. And, uh, he reviewed a particular, uh, monitor, the one I ended up buying, and the OS he used to review that monitor was Omachi. And I thought, "Hmm, what's this?"

So I've been following it since then, and I didn't connect that it was a project run by DHH, whose products I've been using for a year and a half now. So once Quatro came out, it was around the time I needed to swap my Mac Pro, my MacBook Pro.

I'm giving it to my daughter. She's going off to uni, so I bought the XPS 14, and I installed it there.

I've been using it since. My CTO uses the XPS 16.

My two other developers also use the XPS 16. Only our designer still uses a MacBook Pro, so maybe if, uh, No Signals releases its creative suite, I'll convert our designer as well.

Um, what have I been doing, uh, with it? Uh, one, once David offered or mentioned anybody who wants to run meetups, should put up their hand.

So I offered to run the London meetup. So I'll be hosting the first Omachi London meetup on the 24th.

Fred Nix

That's amazing, brother. Good leadership there.

Speaker 4

Yeah, thank you. Uh, we just did it. It was one of the first things that I just did without thinking.

And, um, all of a sudden, we have 70 signups.

Fred Nix

Wow.

Speaker 4

All of a sudden, I have to yeah, I have to figure out what I'm going to do, how I'm going to manage it, etc. So, but it's fun. It's fun.

And, uh, by way of dev work, um, there's some stuff that I was relying on, on the Mac, that I've just started implementing in Omachi. So I use a very rather unpopular streaming webcam, uh, that didn't have the configuration software available on Omachi.

So I just built that in two days using Grok Build. Um, once it's tidy, I'll release it, and people can freely contribute.

So I'll tidy that up. I'll do that for another webcam.

And I have another I have a whole laundry list of stuff that I need to fix, but I'm enjoying it so far. Thank you.

Thank you very much for having me here.

Fred Nix

You bet, Tundy. And I hope to become ex-friends with you even deeper, man. You're a really cool guy.

Speaker 4

Thank you so much, sir.

Fred Nix

Yeah, man. Uh, Wes, let's turn it over to you, man. What you got?

Speaker 2

Yeah. Um, I mean, does everybody else feel the energy with Omachi? Like, it feels like we all kind of hit at the same time when this, like, just strapped ourselves to a rocket ship, and all of a sudden, you know, everybody is just like, you know, "I don't know how to handle all these contributions."

You know, there's, like, the meetup you're talking about that you're doing, Tundy. Like, it's just like, "Wow, what is going on?

This is cool." It's almost like, uh, it's almost like a renaissance in some ways.

So, yeah, I'm just super stoked to be a part of it, and I'm really glad that all of you are here, too. And so we'll keep having these conversations.

Um, you know, I did I see Joel, you're here. You know, if you want to speak and tell us a little bit about the Omachi Guild, that sounds cool.

I mean, I love that you're building that. Um, you want to give us a little info on that?

Speaker 5

Hey, everybody, can you hear me?

Speaker 2

Yeah.

Speaker 5

Cool. Hey, thanks, Wes, first of all, for organizing this. I think this is so cool.

Like you said just a second ago, we're all kind of on this rocket ship trying to figure out where all the controls are and, heck, build some controls as we go. And it's I've it's a really cool feeling.

Um, and having this community that, uh, that is happening right here today is really neat. So thanks for pulling this together.

I, uh, I appreciate it. Uh, and before I talk about the Guild for a sec, I want to react to something, Fred.

Uh, Nix, you said a little while ago where you said you're not a programmer or a developer, which I get in the traditional sense, but I think anybody on this call here who has seen your work would beg to differ that you are a programmer, man. Um, I don't know how many little tidbits and plugins and things you've built in the last couple of months, or maybe it's been longer, but you were building some really cool stuff.

And, uh, I haven't had a chance to dig much into it, but, um, there's so many cool things, especially for Omachi, that I want to dig into. So thanks to you for pushing out all that, man.

You were just making lots of cool stuff. So thanks for that.

Fred Nix

I built nixfred.com and was just putting all kinds of things up there and got my tooling and my, uh, scaffolding and Larry all tightened up. And Omachi came around just as I kind of had Larry just humming. So literally, I use Herder.

Um, I've got a bunch of machines out there on Tailscale, and Larry just I give Larry an idea, um, and he comes back pretty close to the end. So it's really more my ideas, not really my coding.

Um, and I'll be the first to say that.

Speaker 5

Well, I think I think most of us on this call would probably say that's what we're all doing these days. Um, maybe we're looking at the code a little bit to make sure it's dialed and it's what it needs to be doing and it's safe. But, you know, I'm having real-world conversations, real voice, natural voice conversations with Claude and saying, "Build this, do that, tweak that," which I think most of us are doing.

Um, so, you know, no matter what your background has been traditionally, that's where we're most of us are at these days is using these models to build for us as we guide them, so to speak. So, um, so theomachiguild.com.

Speaker 2

Joel, I'm sorry. Do you mind, before the Omachi Guild, could I hop in on just the.

Speaker 5

Yeah, absolutely.

Speaker 2

That topic we were just on because I've really been struggling with this.

Speaker 5

Hit it. Yeah, let's talk more about it.

Speaker 2

Saying, "Oh, I've built this," because I don't feel that way, and I don't feel that I'm a coder. But I've spent years even I mean, I have big dev shops that have software, I'm sure, that has touched most of your lives, actually, as customers, and I help them with technical things. But I, uh, I don't feel like I'm a coder or a programmer.

I've worked in and around helping people with various languages over the years, but I think where I finally landed last night on how I'll position or propose myself in this, and I think it makes sense. It's really what I had to learn more of than AI the last five months, and it's software development.

So I think I'm a software developer now, not a coder or a programmer.

Speaker 5

That's a good way to put it.

Speaker 2

I'm an expert already.

Speaker 5

Well, into your point, uh, is that I agree that we're not doing as much of the coding or the writing. I mean, heck, we're doing almost none of it. But I've kind of looked at myself as kind of like a mid-level manager, you know, managing the agents, saying, "Hey, agent, go do this.

Please build that," and then making sure that my little sibling, my agent, is doing what I had asked or hoped that it would do. So I agree.

I'm not coding, but I'm managing something that is, and it happens to be a computer or a network of computers. So that's a really good way to say it.

You're a developer, not a programmer or a coder. I could get behind that for sure.

Speaker 2

Yeah. So I was starting initially to land in those areas, like orchestrator, coordinator, that kind of thing, because that's what I was doing, as you said. You know, I was coordinating or orchestrating the creation of the software.

But I felt like it left out that it's your idea and that you drove it the whole way and that some people are going to feel like, "Well, that doesn't really capture it." So I think I'll be pushing for that for us all of us.

And it's more than vibe coding, too, when you're really dedicated to it. I think.

Speaker 5

Oh, totally.

Speaker 2

Things to fight confabulation and ensure, you know, accuracy and truth. So, yeah, I think software developers may be what we should start saying about ourselves, us non-coders that are doing so much of this. That's it for me.

I want to hear about the Guild. I'm so sorry.

I'm sure interruption. Here we go.

That's my thing. Five minutes later.

Thank you.

Speaker 5

This is organic, and we'll keep it flowing and going. I love it.

Speaker 2

Yeah. Let's keep it.

Speaker 5

All right. Well, the Guild, um, uh, thanks, Wes, for letting me chat on this for a hot second. Um, the Omachi Guild, Omachi Guild, say it as you all do.

I say it five different ways every day. Um, came from an idea of, uh, something I was I had started building a couple weeks prior for Rails Developer.

Um, I'm a Rails developer and, um, wanted to create a place where, um, myself as a developer could show off my Rails work, um, kind of like a portfolio, if you will, a showcase for all Rails developers to be able to verify their projects that they have created and have live, and it has verification and all this kind of stuff that folks who want to, you know, see what your work is can see your verified work. So I was building that, and I'm just about to release

that, actually. It's called railsguild.com. Um, and Omachi, the wave was so big and strong, I said, "I got to build something for Omachi in a similar vein."

Um, and so that's where the idea came from, kind of a spin-off of the Rails Guild, and as a way to centralize all of, um, the showcase of work that we're doing in one place because the Discord is so huge and vast, and it's a-traditional in terms of it's not like you can navigate it easily to find things. It's such a mess.

Speaker 2

You know what? Maybe we're getting old, man. I don't know.

But it's wild. I'm with you on that.

I want to like Discord, but it's just so much going on. So I appreciate that.

Speaker 5

It's such a mess. Yeah. And as a I've drunk the 37 Signals Kool-Aid for two decades now, and base camp type stuff is the way to go.

I'm done with that thing there. So I wanted a place that where we could kind of have all of our work just show up, and you could see it, and it would be easy.

So again, like all of us, I started talking with I think it was Fable, and he said, "Hey, let's take this Rails Guild idea and clone it and tweak it." And over a couple of days, I got to where it was, and I've been making tweaks on it every day.

But, um, if you want to visit it and add yourself to the Omachi Guild, just create an account and add your GitHub authorization, and it pulls all your stuff in. Super simple.

You got to manually do the plugin stuff because that's a little trickier, but just drop a few links in, and you're up and going. And then you've got a directory.

Um, and you can kind of I've gamified it a little bit, too. As some of you know, um, there's a leaderboard on there where you can see who's moving up and down in the rankings, and, uh, it kind of makes it fun in that regard.

So, um, that's the two or three-minute version of what theomachiguild.com is all about there. So thanks for thanks for the opportunity to share that, y'all.

Speaker 2

Yeah. Thanks for building it. Um, and I hope everybody hops on there.

Let's make it fun.

Fred Nix

Yeah.

Speaker 5

I'll pop a link in the chat, and y'all can check it out.

Speaker 2

I'm joining right now.

Fred Nix

Please do. Please do.

Speaker 2

Sweet.

Fred Nix

It's so cool that we're all building things, and there is some overlap. If you go on plugins and if you look around, there's variations of different software. And if you're building something and you see something that somebody's already built, you really have two options.

You can kind of abandon it and go, "Let's just use what's already been made." Or what I like to do is check their license, make sure that they're cool with you poking their code, and fork it and make it your own.

And, you know, I do that a lot. There's other things about system updates that become a problem, but that's for another discussion.

But you can fork it and make it your own. And then, you know, like Tech Lite and some others have done, you know, you can send them PRs to fix holes and things, and that's how open source grows, and that's why we're going to win this.

Speaker 2

Yeah, 100%. Um, yeah, build it. It takes a little bit of vulnerability, I guess you'd say, to build, right?

Build in the open. Um, because, I mean, number one, I mean, nowadays, it's not as, like, a big a deal.

I get maybe it still is because of, you know, the advent of AI and all. But, like, used to, it's like you put your code up, at least speaking from programmer experience.

It's like a reflection of your intelligence and how well you're spoken. And also, it's like, I'm putting this out there for other people to, like, comment on it.

You know, what's the deal with that? So if you can get past that, I think that, you know, building in the open, it's just such a great thing because people learn, you learn, you're helping the community out.

Um, and yes, there can be times when you get the occasional bad actor that's nasty about something. Why isn't this working?

And, you know, it's like, well, it's just my side project, you know, or whatever. But, like, in general, I think it should be something where, you know, we're lifting each other up.

So I would love to see us continue to do that. If all of us together, you know, are sharing the projects we're working on, maybe it's the Omachi Guild website or whatever, you know, if anybody is having issues getting inundated with PRs coming in or issues coming in, let's try to help.

Let's try to swarm and kind of lift the whole Omachi plugin community up. I think that would be awesome.

Speaker 5

So that's a really good point, Wes. And I'll just jump in if I can real quick in that I've currently got the repo for the Omachi Guild website private. And I half anticipated and still would consider opening it up so it's public and accepting PRs and stuff.

But when you do that, it creates more work in many ways. You know, you've got folks coming in and wanting to add this feature or do that, which would definitely make it cooler and easier to add cool stuff to it, but it also adds a layer of more work.

So if there's enough if there's interest in folks like, "Oh, I've got cool ideas. Let's do this," I can easily take them privately and just talk to my agent and have them done.

But I guess what I'm saying is I would be open if there was enough interest to open up publicly, too, because there's nothing secret in there, so to speak.

Speaker 2

Yeah. But you make a good point. And I think you can address some of that in making you know, even if it's just having a conversation with your agent and say, "Hey, I'm getting ready to make this public.

What can we do to make it obvious to other agents that are working in this codebase, like the patterns that should be followed, you know, the check for open issues first or check for other open?" Because, you know, we get this whole, like, five PRs that do the same thing five different ways.

It's like, how do you make so I think doing that's good. I think having a good contributing doc somewhere that just kind of explains, you know, here's how we do things, and then maybe labeling, you know, labels to tackle, like, needs author, needs review.

It just gets a little bit crazy, you know, sometimes. But I think if you get all that kind of set up ahead of time, it can mostly, you know, be okay.

And, you know, yeah, I don't know. I'm just I'm rambling at this point.

Speaker 5

That's good feedback. And I've never been the head, if you will, of an open repo that's got anything more than one or two folks poking around on it. So those are all really good things to consider.

And I definitely have to lean heavily on Claude to help me. I've got the repo mostly prepped, and there's nothing secret or tokens or any of that stuff in the wild.

But all those things, the contribution guidelines, the tags, all that kind of stuff is stuff that can definitely make the flow go easier once it's open.

Speaker 2

Yeah, for sure.

Fred Nix

Tech Lite? That's like raising your hand?

Speaker 2

Yeah, yeah. Thank you. I appreciate it.

Oh, little feedback. Thank you.

I wanted to comment on two of the things that were said just recently. One was Wes mentioning building in the open.

And absolutely, it has benefits. But I started building in the open five months ago, and it got really lonely, more lonely than building privately, if that makes sense.

And what really changed it for me, has ignited it for me, is finding a community to build in and build with. And that's what Omachi opened up for me.

So, yeah, build in the open, but find ways to build together. And on that, coming straight off that, Wes, I'm sorry.

I think you're priming a response. I don't want to lose the track.

Something that Joel, I think, hinted to a little bit as a mature developer, knowing I don't want to say burden necessarily, but it is a good word for it, of pulling in other people's code. And I think us that are new to this and have a lot of ideas and want to throw them everywhere and, "Okay, let's build together," I think it's important to understand, too, when you submit pull requests, even if you're doing all the right things, like Wes was saying, and maybe one of us

should make a skill or something for doing PRs on other people's repositories. I made a couple of attempts, but nothing published yet. But the thing to remember is when you give someone else a PR, you're asking them and their, you know, fellow maintainers on that repo to support your code.

Whatever you're throwing, I have things I wanted to throw at InfoMarchie that I just won't because I don't want Nyx to have to take on that burden of these, like, insane things that I'm thinking. And I started with forking, and Nyx, you asked, I never answered.

I found those holes initially by having a desire to use your software, forking it, running my agents across it for safety because I'm freaking paranoid.

Fred Nix

Sure.

Speaker 2

And that's how it got going. And I just couldn't help but share where I landed. And, yeah, I wanted with these two things that were just said about building in the open and also building together, those were lessons, I think, that were important for me to learn early.

Fred Nix

Yeah. And you don't have an obligation to accept every PR either. Like, it's your baby, right?

And so.

Speaker 2

Yes.

Fred Nix

And, you know, some people don't like confrontation, but you could be firm, assertive, but nice and be like, "Hey, I appreciate what you did, but I think I'm going to go a different direction." And most of the time, people are fine with that. And if they're not, maybe you didn't want to work with them anyways.

Yeah. For example, on Blip, I got I'm not sure who it was.

I got a PR. They wanted to add an entire contact, Apple Contacts, into it.

And just I think having Apple Contacts, the easiest way to do that is to do iCloud.com/icloud.com/contacts and then install it as a web app. And the same thing for photos.

So, you know, my apologies. And I do feel like I feel so bad when I tell somebody, "No, I don't know if somebody mentioned that," but you have to.

Speaker 2

Yeah. Yeah. I mean, if you do your due diligence ahead of time to have good contributing docs to be like, you know, I don't know how to you know, we'll have to come up with some way to say, basically, don't send me something thinking it's going to get merged.

Like, sometimes the best way to handle it is to say and, you know, I've done some open source in the past is before any PR is open, let's start a discussion or let's open an issue up and talk about it there first and agree on the approach. And then you can go off and do your PR.

Because nowadays, code generation is pretty straightforward. It's coming up with the approach, like, how do you want to solve it?

Because, like, with Omastorm, somebody could just put up a PR to add support for Germany radar, and it may not be legal. Like, you know what I mean?

Like, I have to deal with licensing and all that kind of stuff. And it may not be something I would even want to do.

And then they've done all this work and spent their tokens. And so, you know, something I'm thinking about is, like, maybe turn off PRs or I think there's some automation we could do to be like, "Hey, unless you've, you know, created an issue and we've, like, approved the issue, then don't just open up a PR."

So just something to think about.

Speaker 5

What if we could come together on a good contributor-type instruction set? Maybe not one that we're directing people to, but some of us start keeping in our projects where we have open for PRs because then we've defined the shape. And if you know we want something different for that project, that's defined, too.

So right at the top of the README, if you're welcoming PRs, you can reference that community framework that we've built customized to where you want help maybe on that project. And then the person that's contributing gets their agent primed for helping the moment they start looking at your project.

Perhaps that is a good way forward for us to work together.

Speaker 2

Yeah. I think it's a great idea. Yeah.

I mean, as anything, you know, we'd maybe try it out in the micro first, see how it goes, and then can expand. But I think it's a wonderful idea.

Speaker 5

Well, thank you. And it's always no pressure, too. I appreciate the warm response, too.

There's always just extremely consciousness that never stops. And as I said before, never any offense taken.

Fred Nix

Hey, I'm.

Speaker 5

Where are we going next? Sorry.

Fred Nix

Okay. One housekeeping thing. I'm following everybody I can.

If I don't follow you after this, please poke me, at me to make sure I do. And I'd really like everybody follow everybody here.

We're such a small group that if you're here, we've got the same likes. You know, we may differ in some other things, but this is about Omachi, not anything else.

So just go through, take five minutes, and follow everybody.

Speaker 2

Yep. Absolutely.

Fred Nix

Thank you. Who's up, Wes?

Speaker 2

Oh, let's see. Michael, you got anything you want to talk about, brother?

Michael Y, I think it is.

Speaker 4

Hey. Sorry. One other thing I forgot to mention before Michael hops in.

Forgive me, Michael. PRs have a draft status heads up.

And I've been trying to use that. I have a PR right on Omachi for changing up how screensavers are done a little bit.

And I want discussion there. And that seemed like the app and existing way to have discussion, but get my idea in there.

Speaker 2

Yeah. Yeah. I think that's valid.

I mean, you make it clear, "Hey, I've just put this up as a draft. You know, what do you think?"

I mean, because honestly, it's a pull request, right? You're putting it up saying, "Hey, look at this.

What do you think about it?" kind of deal.

So I mean, it's sort of implied, but the draft definitely helps with that.

Fred Nix

And request is implied.

Speaker 2

Yeah. That's right. That's right.

Cool. Yeah.

We had some more people jump in. Let's see.

Lennox, Ninja, Andre. But, yeah, I don't know.

Michael, did you have anything? You've been hanging out for a while here, so I wanted to give you a chance.

If you don't want to, that's cool, too.

Fred Nix

I think Sir Joel wanted to speak up. Sir, what you got?

Speaker 2

Yeah.

Speaker 6

Hello. Can you guys hear me?

Fred Nix

Oh, yeah.

Speaker 2

I can. Yep.

Speaker 6

Perfect. So you guys talked about the pull request stuff, the makeup of the Omachi lock screen.

Fred Nix

Can you speak up just a little bit?

Speaker 6

Yeah, yeah, yeah. Wait. Can I turn this up?

Wait. I will come back.

I will come back.

Fred Nix

All right.

Speaker 6

Just continue talking. Yeah. I will come back later.

Fred Nix

Okay. Well, please do. I'm not trying to push you off.

We do want to hear from you. Who is next in line?

This thing should be in order.

F4M?

Speaker 2

Amogama.

Fred Nix

There you go, brother. Thanks for joining.

Speaker 7

Thank you, guys. Okay. I have like, I have a stream of thoughts about, like, key things that I think are going to happen that I can share.

And maybe that could open up discussion for.

Fred Nix

Let's hear it, man.

Speaker 7

One of the key things for this to win for real is that we need local models that are small, that really know, that are just trained with everything that we're learning from using Omachi. So it's like, okay, you got to fix your Bluetooth. Amazing.

Or whatever I/O setup that you have. There should be a model that is very small that has the capacity at this stage that we know there's a couple of papers about this to do it.

So one of my questions, this is the thing that I wrote a message to the agent or a reply to the agent, but I'm trying to, like, figure out something that I think is critical is that if someone is working on local models for Omachi because your token spend a lot of the time is just factoring your machine setup. Like, it's not really about spending the tokens on really going into your computer and looking at all your setup.

Like, that should all be local. And you should have just a very small model that you can download as you download any app and just connect it with your main agent.

And if someone wants to take it to the next level, then you set up, like, a bigger rig or your local agent setup. You know, but I think this is, like, the moment for this to happen.

I think this is, like, I think the system has matured enough, or Omachi and the whole community has matured enough, but I don't know if this exists or not or someone is working on it.

Fred Nix

So let me help you out here. I don't have a whole lot of details. I do know that the right people are looking at exactly what you're talking about.

There's a lot of independent discussion. I was a part of one with Mark yesterday.

I'm looking for the post, and I'll put it up here. Lots of discussion about local.

If I look at my InfoMarchie right here right now real quick, I can see that on my machine I'm sorry, Bitburner. My machine is offloading 8% of all my prompts to my local machine.

I believe by this time next year, we'll be as high as 50%. And that's going to come from quantization, not from better hardware.

We'll see gains in hardware, but software and open source is where it's going to be. Your thoughts about running a local LLM that has, you know, more than just the /omachi AI skill to where the LLM has already been trained directly on Omachi.

The smart guys that we would hope that would be thinking like this, I can confirm those people are on it. It's one of the big announcements, maybe.

I'm saying maybe that DHH is going to announce. I'm not trying to sound like.

Speaker 7

I just feel it. I'm a very intuitive person. I like to follow the older people.

But I just feel that that is obviously the case. And I'm not trying to be a big shot saying that I know that.

Please don't take it that way, but I can confirm you're right. I mean, that's the feeling that I have.

And my point here is, even though that's the case, this is the new app store. Like, local small models that are targeted for example, if you're in an industry that's requiring heavy workload on AI and you have this whatever, I don't even know, to put an example, pipeline setup, you know, it's like, why isn't a or my question, this is the thing because now I've been very on top of Omachi since the first week, but this is the first time I'm actually, like, now engaging with the

community at any level. My point is, is there someone, apart from this thinking that I just expressed, is there someone seeing that this is the new app store and it's open source? And it's like, okay, you want to have the best small model that's going to help your plumbing thing on your computer that you can just buy a computer for 100 bucks that already knows everything about plumbing.

It's a file that is less than, you know, 1 gigabyte. And it's going to do an amazing job because it is already enabled with Omachi environment for this stuff.

And plumbing experiences.

Fred Nix

Right. So.

Speaker 7

And, like, yeah, yeah, yeah. You go ahead. I'm just like.

Speaker 2

Yeah. Well, yeah, go ahead, Fred.

Fred Nix

So I'll wrap it up real fast. The people that have put $15 million in did not put $15 million in to get where we are now. Again, not trying to sound cool, but we are 10% into this, fellas.

And, like, sorry. We are 10%.

I mean, this thing is just 2026. We'll look back and go, wow, you remember before Omachi, just like before AI.

Take it over, Wes.

Speaker 2

Oh, I mean, I'm just going to say also, the cavalry ain't coming. We're in.

Fred Nix

That's right.

Speaker 2

So if you've got an idea to build a local model, let's talk. You get started on it.

Speaker 7

Yeah. That's the thing. That's my point.

It's like, I think it's not like, let's wait for them. Like, this is the moment where it's obvious.

At least for me. And then I'm like, okay, I need to type on my computer to see if there's someone in the same wavelength.

And that's what I want to find out, you know?

Speaker 2

Yeah. Yeah, totally. I mean, I've thought about it outside of the context of Omachi.

You know, my day job, some of the things I build or help the teams build, it's always the same types of apps and different things. And it's like, if I had if I trained a model to my taste and how I build things, and it's more than skills and rules, like, that just doesn't really cut it.

And it's like, I don't need a model that can tell me how to do open-heart surgery, right? I need one that can tell me how to apply this change set on this platform in a predictable way, right?

And some of the answer to that is, like, actually not using AI, using a deterministic system. But also, I do think there is room for having these models that are tuned specifically to certain things that are small and lightweight because they don't have to do everything.

They don't have to do what Fable does or whatever.

Speaker 7

So this is my also as a concept, it's like, why some exist? I literally did, like, small tests. This is just some I'm letting everything out now.

I've been in a container for, like, three months, I feel. I don't know.

And it's like, okay, I did a small project of grabbing a browser. You know, Omachi, no, I didn't write anything.

This is just all verbal prompts. I told it, hey, grab Chromium, rip everything apart, just run Wasm.

And I want you to just download models. And, like, for example, you know YouTube?

There's not all videos have transcriptions. So I'm like, okay, the best way to prove my point that this is the thing that's going to happen for me, like, 120%, is that I build a browser harness that just acted as an app of receiving the streaming of any video, processing it within the browser through Wasm hosting the AI, quote-unquote.

And it gave me an hour of video transcription in four times, like, in 20 minutes, locally.

Speaker 2

Yeah. It's wild, man. It's wild.

And so let's pivot a little bit and let's talk about what it means to have Omachi is a like, why is it different than any other Linux or whatever? Like, what is it that makes it special?

And I have some ideas, but I want to hear what you guys think. Yeah.

Yep. Yep.

Yeah. And I think.

Fred Nix

Because everybody accuses it of that.

Speaker 2

Yeah. Well, I mean, to an extent, though, it's really well-curated. Like, it's like somebody being DHH and others on the team, but it's like they went through and said, all right, I'm going to go through and make sure everything works.

I'm going to take it all the way through so that my workflow can be done on it. And then they said, all right, we're going to do a snapshot of that, right?

We're going to curate that and then build it as something that people can then take and customize themselves. So they came up with the quick shell and the plugin system.

It's so easy to build plugins. And I think part of that is because they put some thought into it, to what that primitive should look like.

And so, yeah, I'm curious what other people think.

Speaker 4

If I could plug a personal project, it's a dream of Omachi being a self-fixing OS. I have some thoughts on this whole model training because I hear that often, and I'm not a fan of that idea for what may be my own hallucinations on the current state of things. But if I could share that project and ask for help.

I've been asking for help here and there, but not campaigning for it hard. But I can tell you it's got an incredible amount of tokens put into it.

My own agents have described it as I don't know if it's true, but and I've said it in a few posts now as well because it was a hard thing to hear my own agent say to me. But it's the most documented empty folder in all of GitHub.

And the repo itself is quite brimming at this point, I would say, although the corpus is intentionally tight. I plan to expand it to so much more specific to Omachi and building and all that.

But may I share the project?

Speaker 2

Yeah. Drop it down in the chat if you want to. That works.

Speaker 4

Okay. So I've thank you. Thank you, Wes.

And Wappy's had their hand up for a bit too.

Speaker 2

Yeah. Yeah. Let's hear what Wappy's got going on as well.

Speaker 4

It's called Opinionated Omachi. It's just a stub name. The idea there is everyone was throwing that out as, like, a joke or something.

But my opinion on it is that the fixes from the AI should be consistent and cheap. And so my Opinionated Omachi project is an effort to achieve that for everyone, for every Omachi user everywhere without a cloud.

Can a model do that? Absolutely.

But the problem that I see, at least from my kind of early perspective, is that's pretty expensive, not just initially, but to iterate on, even small ones. It's hard to poke and peer in there too and really fix things.

That's like a big challenge right now. And a big it's like a different project almost to take on even.

Because if you can master that, you've fixed everything else. And I think a lot of people are working on that quite heavily.

Quantization, yes, that's letting us get work done locally. But from my experience, it's led to problems.

And I've had local models break things, like when it's running shell commands and had to bring in a frontier, okay, check the session, fix it. Skills may feel a little shallow, but they're quite effective in retooling or tuning the output.

They cost you more input tokens, but those are cheaper when you're using the cloud. And they usually will result in fewer output tokens and better tokens you want, right?

Because what's the point of spending fewer input tokens to save when it gets you output tokens that you can't use?

Speaker 2

Right. Right.

Speaker 4

So then you just have to run it again and fix it, and it's just it's screwed. You know, I do not like doing it. And that's why I don't have it local models running production roles anywhere.

But from inspiration I've gotten from Nix, I'm going to make another attempt at that maybe tonight if I get time.

Speaker 2

All right.

Speaker 4

Opinionated Omachi is currently it's a troubleshooting corpus, and it has a GitHub pages arm of its release. So you can go there, and it works like a web page to help you with Arch and Omachi troubleshooting. I don't even there's a security thing that came out of it.

Never mind.

Speaker 2

Yeah. No, that's good. I appreciate it.

We're coming up on the end of the half hour here, so I do want to give anybody else a chance to pipe in or answer anything. But anything you want to close with, Tech Ludite?

Speaker 4

Yes. The ask really was, here's are open. And I really could use help.

It's going to be a skill and probably a SQLite database, so we can get searching with it too to improve how the input is built first to improve the output. But I'm desperate for help on this idea of making Omachi self-fixing locally.

And if we try to do that with quantization now without something supplementing it, I don't know. If it's not a skill, what is it?

Sinks or context vectors? There's more.

That's going to be my next stage. And if people know about these things and are already doing stuff with it, I could use the help.

And there's a lot there to build off of, but it's also MIT, you know? So just it's meant to be downloaded to your machine, the whole dang thing, and you can just run it locally.

And I would love anyone to even just talk.

Fred Nix

I see you he's much more. Speak up. What have you got next, Wes?

Speaker 2

Yeah. Oh, I just saw it looks like Zach hopped in. I wanted to just say to Zach, see if you wanted to talk about anything.

Have you tried out Omachi or?

Speaker 4

Wes, sorry. If we've been waiting for Zach, I didn't want to forget or have Wappy be forgotten. His arm is tired.

Speaker 2

Oh, yeah, yeah. That's right. Mine's pretty quick.

How far off are you guys from having, like, M4 and M4 Pro support? And, like, I guess, is it stable enough for somebody to install it and then throw tokens at it?

Or is that not really how we tackle Apple silicon? Well, so I know we're working on M1 and M2 right now, specifically.

There's some differences with M3 and up that'll come soon, but I don't think it'll be part of Quatro 4.5, which I think is coming out pretty soon. So it's coming, but for now, I think your path is something like the Trio Omachi, where you can install it as an app and use it or put it on a VM somewhere.

You know, I've had success with that as well. But yeah, I don't know if I'll have to see.

I'm just getting in there talking to the guys to get an idea of what's going on. But I know for the release coming up, like, right now, Omachi Mac is in another repo, and it's forked.

And so our goal right now is to get it back into the main Omachi repo. So we're not dealing with two sets of code and, you know, a bug gets fixed in one, you got to port it to the other.

So that's kind of like stop the bleeding first. And then once we get that in place, we're going to start accelerating pretty fast.

So I'm sure we could use some help with that.

Speaker 4

Cool. Is Discord the best place if I'm willing to, like, I don't know, mess around with my own higher I don't have any M1 or M2. So, like, I'd be very much, like, willing to put the computer at risk and just format it.

So is Discord the best place to connect with people there?

Speaker 2

Yeah. There's an Omachi Mac channel in there, I think. So that'd be a good place to start.

Or we can just message on X about it.

Fred Nix

Mr. Much More.

Speaker 4

Hey, Fred. How are you, brother?

Fred Nix

Just fair notice. I've known this guy for 25 years. How long have you been using Omachi?

Speaker 2

Not long. I started to, I don't know, probably let's go crazy and say a week or two before you did and started posting about it, right? Because I've been following it for a while because I'm a long-time Linux person.

I mean, I started back on Slackware, like 1.0. So Unix and Linux has been just part of my life.

So, you know, my thing is that I've been going through the iterations because I'm a former AMDer. And so, you know, how smooth and fast it's been on my AMD systems that I have because I don't I mean, I'm just going to say it.

I'm biased. I only run AMD.

I don't run anything Nvidia. So that's what I've been mainly testing out is making sure that it kind of goes through the iterations on Radeon as well as anything because, again, a lot of what I do is supercomputing and AI types of things.

So the Epic as well as all the big stuff. So I'm actually looking to see how this is going to play out for me on the bigger scale because of the size of systems that I do and just how efficient it is.

So I'm just getting into it. But I got a lot of, like, side projects because I saw somebody had a weather project.

I got a lot.

Fred Nix

Yeah. I did the you talking about the radar app, Omastorm?

Speaker 2

Yeah. That's me. Yeah.

I would love to collaborate with you because I wrote a Noah is one of my customers. And so I wrote a something to go against them because my hurricane model is more efficient than theirs.

Fred Nix

Oh, yeah. Okay.

Speaker 2

Well, hey. Yeah. I mean, it points to Next RAD data that's on, you know, just S3 bucket.

We can chat offline. I'll be happy to talk about it.

Fred Nix

Guys, you definitely follow each other. And I'll say it because Heath won't. Heath deals in billion-dollar data centers with petabytes of data.

He's the real deal.

Speaker 2

Wow. Incredible.

Speaker 5

Yeah. I just won a really big deal with the FBI last week. So I'm getting ready to put in, as far as I know, we're the only company on the planet who's authorized to install Google TPU systems on-premise outside of Google.

So we're going to install a bunch of those.

Fred Nix

Smart guy.

Speaker 5

Yeah. So we're going to do that. Nvidia NVL 72.

So my point is, the thing I'm going to try to contribute to the platform is to see if I can install it on bigger and bigger and bigger systems and see how far I can stretch its boundaries versus it just being a desktop kind of thing.

Speaker 2

Okay. Yeah. I mean, I know there's been talk about an Omachi server too.

So something to think about.

I mean, I've heard, you know, they're building their own kernel, right? So who knows?

Speaker 5

Yeah. Because I'm going to try and piggyback off of CIQ's Rocky Linux kind of premise and just see if we can make Omachi in that same fashion.

Fred Nix

Hey, I'll connect you up with a few people.

Speaker 5

Roger that, man. Thanks.

Speaker 4

I'm sorry to throw another plug out there. I'll make this one much more brief, but I have another repo called Omachi Fleet that leans on the Omachi kids work because I would love to. I've replaced countless, countless machines the last two years, and I'd love to have a really great Linux option.

And I think kids laid some of the groundwork for me to have an option that I feel comfortable or could start considering rolling out for businesses. And the idea is too that we could design businesses by a repo.

And then with Omachi Fleet, you could have a repo that's a pizza shop or whatever your friends, families, customer, business needs. You could just build it in a repo and then combine it with Omachi Fleet or perhaps something that gets pulled in natively.

Speaker 2

Cool. All right. All right.

I don't want to go on too much longer. I want to try to keep these this one's been awesome because it's been the first one.

We've been talking over, feels like an hour and a half now. We'll definitely do this again.

Maybe it'll be a week from now. We'll see.

Fred and I will talk about it, see what's going on. But I appreciate everybody.

Speaker 4

Hey, Wes.

Speaker 2

Yeah.

Speaker 4

Hey, I just wanted to introduce myself to you.

Speaker 2

Yeah, man.

Speaker 4

Just say hi and say, you know, your projects, you know, what you did with Omachi Storm was awesome. I just started using it. I talked to Nix or commented on one of Nix's or Fred's posts and put it on one of my older Windows laptops.

I'm a chemical engineer slash electrical engineer, so much less on the software development side and more trying to apply these to process simulations and stuff. But getting into I have a Linux box right now that I run, and then I'm mainly on Windows just because I use AutoCAD so much.

But just trying to kind of dive into it and see what it's all about. And good to hear from, you know, people that are much deeper into it than I am.

So just trying to learn.

Speaker 2

Yeah. We're just getting started. Like, it sounds like we've been in it a while, but honestly, maybe a month.

Like, I don't know.

Speaker 4

Yeah. I guess I mean more on the development side. I'm learning.

You know, I took a deep dive into, you know, AI development, but I would consider myself a hardcore vibe coder, not necessarily a software developer.

Speaker 2

Right. Well, the definition's changing, right?

Speaker 4

Yeah. You guys have much more experience on the architecture side. I'm just trying to apply, you know, general engineering principles.

Speaker 2

Hey, that's good, man. That's what it comes down to at the end of the day. Can you verify it?

Do you have boundaries? You know, that kind of thing.

So awesome. Well, it's good to meet you.

Good to talk to everybody. Like Fred said, let's try to follow everybody, and we'll keep the conversation going.

I appreciate it. I hope you all have a good Sunday, whatever time it is.

It's Sunday afternoon here. I'm probably going to go see if my bucket here is lost or not.

I'm guessing that they did because they were getting killed by the bangles. So we'll see.

But yeah, thank you all, Fred. You got anything you want to say?

Fred Nix

Appreciate Fred being here. Again, follow everybody. Keep contributing.

Don't be scared. If you don't know what you're doing, ask somebody or ask your agent.

We will do this again. Appreciate everybody.

Follow everyone. Have a great Sunday.

Thank you very much. Have a good evening.

Speaker 2

All right. See you all.

↑ back to the summary