Parker BrownCASpreading the word by shippingAttention is the corrective for slopAgents run as a teamThe looks get them in, the AI keeps themIt feels like being a kid againThe AI VampireLocal vs frontier modelsContribution over gloryThe docs still matterFilling the app gaps that kept people on Macbuilder/builders/parker-brown/avatar.jpg
// builder #09 · interviewed Oct 14, 2026
Parker Brown
The self proclaimed Omarchy Propagandist. A Kelowna builder who went all in on linux in 2006, built Terminal Delight, a headless and fully themable work surface for running agents, writes music with agents from scratch, and preaches one rule: attention is the corrective for slop.
fromBorn in High Level, Alberta; grew up in a southern Alberta prairie town, a decade in Edmonton
first computerThe family Windows 95 PC; then a PC he built himself for university, PCI vs PCI Express and all
linux since2006
rigOmarchy with Terminal Delight as the work surface
found omarchyNot stated on the call
modelsclaude, codex
ai workflowAttention is the corrective for slop
// life timeline
Doom 2 at a booth in the local hockey rink; Wolfenstein 3D and Doom at home
Ancient Domains of Mystery, the ASCII game that taught his hands the keyboard
Builds his own PC for university
2006First linux install: Ubuntu
2016Moves to Kelowna
2025Leaves the IDE for agents in the terminal
2026HTML decision briefs; starts Terminal Delight in February
2026Builder #9 on omarchy.builders
// 00
The rundown
Parker Brown lives in Kelowna, British Columbia, and he has a title nobody gave him. He took it. On his X profile, right at the top, it says Omarchy Propagandist. When he joined the call, Nix greeted him the only way that made sense: "The original Omarchy Propagandist is here."
Parker laughed it off. He says the title is a work in progress, and that for now "we're just living in the future, holding on for dear life." (line 10) But watch what he ships and you will see why the title fits. He runs linux since 2006. He built Terminal Delight, a headless, fully themable work surface for running a whole crew of agents. He makes short story clips scored with music his agents compose from scratch. He reads documentation for days on purpose. And he has a one line philosophy that every builder on this site should tape to their monitor.
"And so attention is the corrective for slop." (line 257)
Read on. He teaches as he talks, and this story is built to teach you too.
“And so attention is the corrective for slop.”Parker Brown · L257
// 01
The Omarchy Propagandist
Propaganda has a bad name. Parker does it the honest way. He builds things, posts them, and shows people what is possible on Omarchy. His X bio says it plainly: "Build the best looking agentic development environment in the world: Terminal Delight." That is the pitch. Not a slogan about linux. A working tool you can install today.
Where did he get the nerve to just start? From the community. He took his lead from what he calls "some real beasts" in the community, and their message was simple: "you can just go build it. Just go do the thing." (line 15) That line is the whole Omarchy ethos in 9 words. Nobody asks permission. You build, you share, you help the next person.
His GitHub reads like a field journal of that ethos. There is his awesome omarchy list, a curated set of Omarchy themes. There is his fix everything observatory, a scrubbable replay of every issue and PR in the Omarchy repair swarm. There is Terminal Delight itself. When Nix asked if he could get a repo, Parker did not hesitate. "Terminal Delight's fully open source." (line 445)
And he sees the scale of what is coming. "there's 200,000 Omarchy users." (line 655) If even 1% of them pitch in, he says, the project has its hands full for a long while. That is the propagandist math. Every person you reach is a person who might ship the next fix.
// 02
Doom 2 at the hockey rink
Parker was born in High Level, Alberta, way up north. He grew up in a small prairie town in southern Alberta, about 8,000 people, cold winters, wild summers. He describes it as a strange place that is "technically a desert because it gets no rain, but it's also technically a wetland because the water doesn't go anywhere." (line 91) Oil country. Salt of the earth people.
The first computer in the house ran Windows 95, and it came with a printer. Suddenly there were printouts everywhere. Spreadsheets. Phone lists. Parker connects that era straight to today: a new machine arrives, you think you will crush the whole backlog, and then you discover the backlog is infinite. That, he says, is exactly what agents feel like right now.
Then came the moment that hooked him. The local hockey rink cleared the ice for an expo, booths everywhere. He found the booth for Doom 2 and basically moved in. "And just, yeah, like, minds blown." (line 128) The home computer could not run it. So the family PC went down the Wolfenstein 3D and Doom 1 road instead, and Parker became a proper gamer. Diablo. Final Fantasy. Hours on the keyboard.
// 03
Faster than the computer
Here is the memory that explains everything Parker builds today. As a kid he played Ancient Domains of Mystery, known as ADOM, a roguelike where every wall, monster and item on screen is an ASCII character. It is driven entirely by keyboard commands and shortcuts. You can play it as fast as your machine can draw it.
Parker got so good that the computer could not keep up. The machine was still loading and rendering, and his hands already knew the next 5 moves. He compares it to what DHH talks about with the keyboard: "you're not thinking anymore." (line 196) And then: "your hands are just doing the stuff and you're kind of watching it happen." (line 197)
Tech lesson: this is why Omarchy is keyboard first. Every trip to the mouse is a tiny context switch. A tiling window manager like Hyprland puts every window on a keystroke, so your hands never stop and your brain never leaves the work. Parker felt that flow state as a kid in an ASCII dungeon. He has been chasing it ever since.
// 04
It's just got to be linux
University meant his first real computer of his own. A friend was building PCs and put him on the path: order the parts, put it together. "And my goodness, was it fun." (line 178) It was also brutal. Which motherboard socket? Which bus for the video card? PCI Express was brand new, so was this card PCI or PCI Express? Then first boot, and a blue screen.
He spent about a year in that loop of hardware and operating system trouble, walking the halls of his residence asking the older tech guys for help. It worked while they were around. Then he was on his own, and that is when the lesson landed. If he was going to fix his own machine, there was only one answer. "it's just got to be linux." (line 186)
In 2006 he installed Ubuntu. He wanted it simple and clean. "I was floored." (line 188) He opened a terminal for the first time and felt both lost and free. "Then the first experience on the operating system of, like, total freedom." (line 198) He could change the background. He could make something and just put it there. He ran ls, then learned that ls with the a flag shows the hidden files too. "Oh, I'm a programmer now." (line 200)
Tech lesson: on linux, any file or folder whose name starts with a dot is hidden by default. That is where your configs live: .bashrc, .config, the whole personality of your system. Plain ls hides them. Adding the a flag shows all of them. That one flag is the doorway from user to tinkerer, and on Omarchy almost everything you will ever want to tweak lives under the .config folder.
He took engineering at university in Edmonton, where they taught some C and some assembly to move a robot around a screen. It never quite landed in class. He left engineering for his own path, moved to Kelowna in 2016, and never looked back. That was 20 years ago for linux. The terminal never left him.
// 05
The death of the IDE
For years Parker lived in the IDE. JetBrains first. "I was on JetBrains, so it was a little bit heavy." (line 286) Then VS Code. "whoa, so fast." (line 289) He calls the editor what it was for a whole generation: "That was our workbench for decades." (line 291)
He went agentic about 3 years ago. He watched agents learn to write functions, then good functions, then tests. But he still kept them inside the IDE. The break came around October 2025. "oh, you can run these things in a terminal." (line 297)
At first the terminal felt like a very advanced chatbot. What changed his work was the format of the conversation. First markdown, which he found useful but heavy. "But markdowns are just so much reading." (line 299)
// 06
HTML is the new prompt box
Then, around February or March of 2026, Parker hit on the idea that powers his whole workflow. Have the agent speak in HTML. He found HTML "was actually a really nice way" (line 299) for agents to convey dense information, because it is so simple and so visual.
Here is how his loop works. The agent does not dump a wall of text into the terminal. It escalates a decision brief to him as an HTML page with graphics and structure, built to hold his attention. Then the trick. He tells the agent to give every element a commentable box. "And so you can now comment on everything, right?" (line 302) Then a button feeds the whole HTML map of his comments right back into the right agent. No retyping. No copy and paste, unless "you're a Luddite living in, you know, 2025." (lines 412 to 415)
He demoed it live. Each brief flags inflection points. You stamp a concur on the ones you agree with. Or you hit concur all and let the agent fly. "Yolo." (line 403) And because it is contextual, save and send goes to the exact agent that owns that file. "And so the prompt is auto done." (line 419)
Nix nailed the point: Parker is cutting every step between his decision and the machine. Parker put it in 4 words. "Just closer and closer and closer." (line 426)
Tech lesson: try this yourself. Ask your agent to write its next plan as a single HTML file. Ask for a comment field on every section and a button that exports your comments as structured data. Your feedback stops being a vague paragraph and becomes a precise map of the doc. The agent knows exactly which part you mean.
// 07
Terminal Delight
All of that lives inside the tool he built. Parker wanted one place to run many agents, keep himself organized, and make it beautiful. So in February he started Terminal Delight. When Nix saw it on screen he said it looked like Herdr, the agent terminal from another builder. Parker made it completely on his own, and he gives the other teams plenty of credit, naming Herdr and Human Layer as people he respects.
What makes it his is the style. "So it could be 70s." (line 364) You can switch CRT mode on and off. You can retheme it whenever the contrast or colors are not quite right. Nix was sold on the spot. "I'm going from Herder to this immediately." (line 368)
Nix runs Omarchy every day, and still he wanted Terminal Delight that badly. That is what happens when looks meet low friction. The looks pull you in. The workflow keeps you.
// 08
Going headless
Here is the big unlock, and it is real engineering. Every terminal user knows the pain. Control C copies in some apps and kills your prompt in others. One fat finger and your half written prompt is gone. Parker gives Omarchy credit here. Super C copies, so you never wipe the prompt by accident. "such a simple thing" (line 310), he says.
But he wanted more. He noticed the strongest agent terminals had gone headless, so he did it too: he separated the terminal process from the terminal rendering. He closed a window on screen, opened it again, and everything was still going. "prompt is still there, right? Because the process is still running." (line 380)
Tech lesson: a terminal window is just a view. The shell or agent running inside it is a separate process. Tools like tmux have long let you detach from a session and reattach later. Terminal Delight builds that idea right into an agent work surface. Close the view, the work lives on, and you can grab it back when you remember you were not done. The tradeoff, as Parker points out, is that the agent might still be running. So he added a trash bin for finished threads. "I think it's really important to have hygiene with our agents and our terminals." (line 562)
Open source raised a harder question. Should an agent in a public project be able to grab a terminal? On his own machine his answer is loud. "I'll 100,000% just like, yeah, agents get terminals." (line 545) In an open source project anyone can install, he got much more careful. He called it a rats' nest for security, and he put guardrails in so that if the project takes off, it stays manageable. That is the call a grown up maintainer makes.
// 09
Agents can write music
Then Parker showed the part that made Nix's jaw drop. The Cinnamon Delight studio. Parker saw a gap: nobody was making short, musically powerful content that tells a story. A 30 second clip. A 1 minute clip. Today in emoji. This month in emoji. "also like the year of linux is 2026." (line 466)
And the music is not pulled from a library. "It's like agents can write music." (line 472) Not mixing. Not remixing. Scratch tracks, built "MIDI style the way that a composer would do it." (line 473) He directs it with taste, the way a director talks to a composer. The French horn is too tinny, change the tone. His voice to text drops the note into the HTML brief, and the agent gets to work.
Tech lesson: MIDI is not audio. It is a list of instructions: which note, how hard, how long, on which instrument. That makes it text, and text is what language models are great at writing. A composer agent writes the notes, a synth renders the sound, and you give the feedback. Parker turned agents into an orchestra.
// 10
Attention is the corrective for slop
Nix asked Parker where he is pushing his limits now that a year of struggle is a 3 minute prompt. Parker's answer is the heart of this story.
The limit is no longer what you can do. "when you can do anything, um, doing everything becomes counterproductive." (line 245) You need guardrails beside you, not in front of you. "you can do all of these things, but you can't do every one of the things." (line 251)
So what is left? "It really comes down to attention." (line 255) Then he went further. "The world is created by attention." (line 256) When you pay attention to your agents, the work gets unbelievably good, unbelievably fast. When you stop, it slides into a high entropy decay. Slop. "And so attention is the corrective for slop." (line 257)
That is why every piece of Terminal Delight exists. The decision briefs grab your attention. The headless sessions protect your flow. He described what he is going for as staying locked in: "staying in, staying in, staying in is really what I'm going for" (line 394). And when you leave, you leave on purpose, for X or a meeting, not because a window crashed.
He even sees it in public work. When you build in the open, people can see exactly where your attention went. Where you cared, it got good. Where you did not, it did not.
// 11
Always read your docs
In a world racing to go faster, Parker did something radical. He stopped. When a new decision engineering tool he calls Jev came out, he pulled out of his usual just keep building mode and read its documentation for about 5 days straight, about 15 hours of it.
His insight: deciding well is ancient. The Romans wrote everything down, ran it through the courts, put it through its paces. "now we can do it in a microsecond." (line 278) Speed changed. Judgment did not. "for all of this advancement, the documentation still matters." (line 282) "To read the manual and to understand the problem." (line 282)
He brought it to life in his demo with a line every builder should keep. "keep reading your docs, always read your bloody docs." (line 406)
His rule is to know when to go fast and when to slow down. That is how you stay the engineer and not just the passenger.
// 12
Token debt and Burn Bar
Near the end, Nix showed Parker Burn Bar, his plugin that tracks your agent subscriptions against the weekly reset and tells you how much time you have banked. Parker laughed. He lives on the other side of that bar. Way in debt on every agent, burning hard, then stepping away from the computer when the tank is empty. "I'm massively on token debt mode." (line 679)
Then the most builder moment of the whole call happened. While they were still talking, Parker's agent was already building Burn Bar into his setup. "my guy's doing your burn bar." (line 670) A few minutes later it was live on his toolbar. That is the speed he works at. An idea on a call becomes a running feature before the call ends.
Tech lesson: agent subscriptions run on rolling limits, often a short session window and a longer weekly window, and they differ by provider. If you track both, you can pace your burn so every subscription lands near empty right at its reset. Use it all, waste none. That is cost aware routing in practice.
// 13
Simple and hard
Parker closed with the reason he shows up for any of this. Under all the tech, the flash and the outrage cycle on social media, "really we're looking for, for just human connection, and it is that simple." (line 518) He did not pretend it is easy. "so simple and hard is the way" (line 519).
That is the propagandist's real job. Not hype. Connection. Showing up, sharing the code, giving away the tool, and pulling the next person into the most fun corner of linux right now.
Parker Brown. Omarchy Propagandist. Builder #9. Go install Terminal Delight, and pay attention.
// 14
In his words
"And so attention is the corrective for slop." (line 257)
"The world is created by attention." (line 256)
"it's just got to be linux." (line 186)
"Terminal Delight's fully open source." (line 445)
"It's like agents can write music." (line 472)
"keep reading your docs, always read your bloody docs." (line 406)
"you can just go build it. Just go do the thing." (line 15)
L15Parker: Yeah, I mean, uh, I took my—took my lead off of, uh, oh jeez, we've got some real beasts in the—in the community just saying, you know, uh, you can just go build it. Just go do the thing.
L23Nix: Um, so let me just get some stuff out of the way. I am recording this, and I will publish your story on Omarchy Builders and promote you as a f****** rock star that you
L25Nix: are. I just want to get a verbal okay that you're okay with me putting your stuff out there.
L26Nix: So say whatever you want at the very end; I'll ask you, is there anything that you want to exclude? So I want you to feel comfortable that, you know, I'll take anything out that you want.
L27Nix: And let's go from there. You're the 8th person, so let me tell you how this started.
L28[redacted: private, lines 28 to 29]
L30Nix: me to a guy in Singapore that I'd Zoomed with called James. And they're both up on the Omarchy.
L31Nix: builder site. The website's actually Omarchy Builders, with no dot com, which is pretty, pretty cool.
L79Nix: So where exactly are you right now, and where did you grow up, and where were you born, and all that kind of stuff?
L80Parker: Yeah, yeah. So, uh, born in high-level Alberta.
L81Parker: So pretty way up north. Um, and moved around a lot as a kid.
L82[redacted: private family detail, lines 82 to 87]
L88Parker: And, uh, and that's sort of how we live our lives. But yeah, Edmonton, I spent a decade, um, grew up in, uh, a southern Alberta town, kind of a—a nice prairie little,
L90Parker: little house on the prairie kind of a town, uh, what, 8,000 people there. And so we, uh, yeah, we—we had cold winters and, uh, and—and really wild summers.
L91Parker: It was, um, a mix of—it's technically a desert because it gets no rain, but it's also technically a wetland because the water doesn't go anywhere. So it's this weird, weird place.
L92Parker: But, uh, yeah, it's, uh, a really nice place to grow up. Lots of—it was an oil community, so lots of good money in there, lots of, uh, lots of, um, yeah, just—just salt of the earth kind of people.
L93Parker: And so—so that was growing up. And went to university for a few years in Edmonton, uh, decided that the—the engineering wasn't for me on the—on that track.
L94Parker: So we, uh, yeah, so then moved to Kelowna back in—in 2016. And yeah, never looked back.
L108Nix: Yeah, life's tough for everybody, I suppose. So what was the— what was the first computer you ever used, and then transition that all the way, if you can, to Omarchy?
L109Nix: And just so you know, there are no time limits. I'm here with you until you're ready to quit.
L116Parker: Um, so we've got, uh, what did we get? Our first computer in the home was—it was, uh, the Windows—it must have been Windows 95.
L117[redacted: private family detail]
L118Parker: And she had—she was going to use this computer to—it's just, like, agents now, right? Like, it's—you've got your infinite backlog, and you're like, "Yeah, I'm just going to crush through everything.
L119Parker: " And then you realize that it's an infinite backlog, and now you're going so fast, and it's—you're just hanging on for dear life. So, so that's kind of the—the thing that happened.
L120Parker: Got a computer, got a printer back in—when it must have been—maybe it was, like, I don't know, Windows 3 or something was the—was the name of it, but it—it would have been about 1990. Yeah, it must have been Windows 95.
L121Parker: And, uh, and so we got the computer and the printer, and so all of a sudden there's, you
L123Parker: know, that blue tack with stickies all over with printed out spreadsheets and—and phone lists. Uh, you're old enough to remember phone lists where we had phone numbers once upon a time.
L124[redacted: private family detail]
L125Parker: And so we, uh, yeah, we—mom had the computer. Until one day we went to, uh, what do they call it?
L126Parker: It's like, uh, it's like an expo, but like the—the local arena, the local hockey rink, uh, cleared off the ice and—and there was just a bunch of booths there. And so we—my brother and I found, uh, the booth for—for Doom 2.
L127Parker: And so we, yeah, we, uh, we just—we just hung out there. And, like, kind of scabbed it and kind of, you know, were just proper mall rats with all the other 13-year-old kids, all the other 8-year-old kids, whatever it was.
L128Parker: And just, yeah, like, minds blown. And we realized that
L130Parker: our computer couldn't run Doom 2 because it was, like—it was a nice computer maybe 2 years ago or whatever, but then also Doom 2 is, like, expensive. Like, it's a—it's a video game.
L131Parker: It doesn't, you know, it's not a book. It
L174Parker: So, uh, so yeah, we got, uh, lots of video games. And that first PC, the, you know, the first Nintendo Entertainment System for Christmas back in whatever it must have been, '89.
L175[redacted: private family detail]
L176Parker: I was like, "I need a—I need a good computer. I'm going to be a proper student."
L177Parker: And so, uh, I got, uh, my friend was getting into building his own computers. And so he got me on the track of, like, "Oh, yeah, you just go here and order the parts and whatever.
L178Parker: " And my goodness, was it fun. But you just—you're over your head and you're, you're like, like, "Whoa, what kind of a motherboard socket layout is this?
L179Parker: What kind of is it, uh, you know, back in the day? Like, what's the bus for the video card?
L180Parker: " And PCI Express was brand new. And it's like, "Is this PCI or is it PCI Express?
L181Parker: " And—and all of that hardware stuff that goes along with it. And then you fired up for the first time and you're like, "Blue screen!
L182Parker: " And, like, "Do I have the right version of, uh, at the time it was Windows for me?" Because that's what I was familiar with.
L183Parker: And I went through kind of a year of that, uh, cycle on the—the hardware software on the, uh, operating system, just trying to figure out, like, going a different guise in my residence. It was an all guys, uh, residence.
L184Parker: And so going to different guys and saying, like, "Hey, like, we need to figure this out." And there was a couple of tech geeks in there that were able to help.
L185[redacted: private]
L186Parker: But as soon as I had that computer on my own, I realized that without all those other dudes there, it was just hard to—hard to fix. And so I realized, like, if I'm going to be self-sufficient, like, Linux, it's just got to be Linux.
L187Parker: So in 2006, I, uh, put my first install of Linux together. And it was just—it was Ubuntu, right?
L188Parker: Because I wanted it to be simple and clean and—and whatever. And I was just—yeah, I was floored.
L189Parker: Like, the—the feeling there was, like, this is opening all of the doors and so I opened the terminal and it's like, "Oh, I don't—I don't know." Like, I—I used to use, what is it, uh, DOS, where you—you enter your—your command to run
L191Parker: your game. And as a kid, you know how to do it, but you don't know what you're doing.
L192Parker: So as a kid, like, we—we had a video game called, uh, Ancient Domains of Mystery, which was all ASCII characters as the graphics. And so, you kind of walk your little dude around this ASCII map and—and it's this—it's crazy labyrinth of—of commands and—and shortcuts and—and keyboard sort of joy.
L193Parker: And you're just—you're just flying and you're going through this game and you can basically play it as fast as your computer can run. And so our computer, what we got to, is that we could play the game faster and faster.
L194Parker: And our computer couldn't keep up with us. And so, like, the—the time it took just
L196Parker: to load and render and all of that kind of stuff, even the ASCII characters, like, we just knew the patterns of what to do. And it's, you know, exactly like—like DHH talks about with—with the—the keyboard kind of thing, that feeling of just, like, the—the workflow efficiencies where, like, you just—you're not thinking anymore.
L197Parker: You're just—you're just in with the computer and you're not—your hands are just doing the stuff and you're kind of watching it happen. And so, so that was the—the first experience was, uh, was—was that game on the—on the keyboard.
L198Parker: Then the first experience on the operating system of, like, total freedom. Like, this is mine now, was installing Linux and—and Ubuntu and seeing, like, "Oh, I can just go and make a background.
L199Parker: I can go make something and—and just put it there." And so that first sort of LS and then learning that, like, LS-A does a different thing than LS does.
L200Parker: And—and that first sort of, like, "Oh, I'm a programmer now." Because you—you learn in
L202Parker: engineering first year, you learn some—you learn some C and you learn some—some, like, maybe assembly, just to, like, move a robot around the screen or whatever. But it never quite lands because you're in a class and there's goals and whatever.
L203Parker: But, uh, but yeah, so—so that's that. And, uh, and then now we're 20 years later.
L204Parker: So, so—so that computer went, uh, yeah, go ahead.
L205Nix: Right. So you— you made a couple of— you made a couple of points I want to kind of follow up on that are really interesting.
L206Nix: So you talked about struggling for a year to get things working.
L208Nix: And you also said that you're— and I don't want to paraphrase, but you kind of said you're not afraid to put— you intimated that you're not afraid to put yourself out there, kind of thing.
L214Nix: how it really works in the age of there not really having value in knowing how it works because you learned all those things in that year of struggling without the internet, talking
L218Nix: and all those things that were really difficult without being able to Google it and without being able to do those things. And you learned what good looks like.
L219Nix: You're not afraid to put yourself out there. You learned all that stuff.
L220Nix: And now you're in the age of, like, that year of struggle where today would be a three-minute prompt.
L224Nix: I see you as a person that pushes yourself harder than you believe that you can do it. And I see that already in you.
L225Nix: Where are you pushing the limits now that the things that you struggled with in the past were a problem?
L226Parker: Yeah, that's a good question. So—so two.
L227Parker: Address the—the minimization of problems over time. It's like, people used to wash their clothes in a river.
L228Parker: You know, I go hunting. I know what it's like to go out there; you take your bow, you have a stick, and you need to get this stick through that animal so that you can take it home with you.
L229Parker: It does not want you to do that. And so—and not only that, but, like, nature's trying to kill you.
L230Parker: So it's cold and windy or raining and maybe it's—it's snowing. Like, you're up on the mountain.
L231Parker: Don't break your leg. You've got 160 pounds of meat to pack out now.
L232Parker: And so, like, how do you do that? There's bears.
L233Parker: It's—it's—it's proper engagement and there is no room for—I mean, there's room for you to mess up, but the—if you do mess up in the wrong way, it's—it's pretty catastrophic. And so, honestly, the—the limit for me really goes back down to the—to the basic stuff like that.
L234Parker: And when it gets into computer land, it almost goes the other way of, like, I can't tell when—when I've done something that is outside of the bounds anymore so much. Um, because you build something and it's like,.
L235Parker: You solve a problem, you build something, what am I looking at? I'm looking at the—the guys on the Mac.
L236Parker: I think the—the Omarchy M, um, and the Omarchy
L238Parker: mobile, Omarchy, I should say. Uh, to be proper about it, Omarchy.
L239Parker: Put my—put my—put my pinky up. I—I—I got my—my voice prompt, uh, whatever, whatever voice prompt, uh, preload phrases to, like, really get it dialed so whenever I say Omarchy, it knows.
L240Parker: Um, but it was, yeah, fun thing. So yeah, like you say, the—the problems aren't what they used to be, but it's—I go back to the infinite backlog example.
L241Parker: It's not as though the problems disappear, but when you realize that the—the conveyor belt or the highway of problems that you're—you're addressing is becomes never-ending, then it becomes a sort of choose-your-own-adventure book where you have really important decisions to make. Um, and the world isn't up with it yet.
L242Parker: That's—I think one of the biggest things to—to friction with is that, um, I was talking to some guys yesterday and—and you know, had the—the AI chat, like, "Oh, Jev, this," or, "Oh, LLM that." And—and understandably, when you're mooted in with a business, you can't just claude.
L243Parker: And—and dump all your—your, you know, your proprietary code or your database info or customer stuff into Claude. Like, that's just not an option.
L244Parker: And so it's like, well, then you're looking at, like, uh, Metal AI and stuff like that. And so—so what I'm finding is that, uh, the sort of the—the constraints, the boundaries that—that limit aren't in front of me anymore.
L245Parker: I need to get them beside me so that, uh, the goals that you're aiming for, uh, make better sense. Because when you can do anything—when you can do anything, um, doing everything becomes counterproductive.
L246Parker: And so—so it's like, the limits of what I'm capable of on a computer, it's like, I really have the attitude—it's like, yeah, we can fix everything. We can build the perfect machine.
L247Parker: Should you, like, not interested in should right now, I don't think that I'm—I'm the should guy. I think I'm—I'm not changing things in that way.
L248Parker: So the things that I am changing are, like, well, should I, you know, change this theme? And it's like, well, yeah, of course.
L249Parker: But should I, you know, um, push a company and—and put pressure on them to—to really say, like, AI is the future and the sooner that you onboard, the deeper into the culture it will be, uh, when it becomes a non-option. Because right now it feels like an option, but it's going to be a non-option pretty right around the corner, sooner than later.
L250Parker: And so pressuring people, it's like, should I pressure people? And it's like, well, that's a—that's a harder question to answer.
L251Parker: But for myself, technically, um, yeah, the guardrails have to come up with, like, you can do all of these things, but you can't do every one of the things. And so having prioritization on what is really important, uh, becomes the key to—to—to
L253Parker: do that. I've kind of lost our question a little bit.
L254Parker: I'm trying to—trying to answer the—the question that you—you seemed that I remember you asking as kind of like in a limitless world, what—what do you find your limitations to—to be and how do they present themselves? And so—so yeah, really just the—the limitations on—on attention.
L255Parker: It really comes down to attention. Is that the world is—is made up of attention.
L256Parker: The world is created by attention. And so what you notice with these agents, when you pay attention to them, the thing gets unbelievably good, unbelievably fast.
L257Parker: When you don't pay attention to them, it kind of has this high entropy decay sort of state, like the slop state. And so attention is the corrective for slop.
L258Parker: And so what you pay attention to, uh, what we pay attention to when we're doing this, uh, futuristic kind of stuff, um, is—is—is managing outcomes, right? We're trying to get the thing.
L259Parker: We have a vision. We're trying to get there.
L260Parker: And so—so So what? Where there was—there was a point that I just had that dropped away.
L261Parker: It was, um, yeah, pushing—pushing the attention and—and really staying in that—that attentional groove. I think that what better smarter more sophisticated engineers than myself, um, are onto is the kinds of things like, uh, graph engineering and loop engineering and the kinds of stuff to manage like agent sets as systems.
L262Parker: And I think the work there is really, really cool. And that's like something that I aspire to and drive towards, but it's just not what—uh,
L271Parker: first of all, that's stupid and inefficient. Like, you're—you're definitely doing it wrong, but that kind of idea of—of being able to with more resources, um, produce more, produce more bad products, um, that—that's something that I'm really interested in is like the harness engineering side and the—and the—and the what?
L273Parker: Yeah, one thing—one thing that I has—has anchored it for me to shift gears slightly is with Jev coming out. That was pretty groundbreaking stuff.
L274Parker: And system one type engineering, like decision engineering isn't new. We've been doing it for thousands of years.
L275Parker: Like—like what—how do we make this decision? You go to the Romans and they have a whole way of documenting, writing everything down, putting it through the courts, putting it through the paces, et cetera.
L276Parker: And it's—it's this—it's just not a new thing. And
L278Parker: so Jev really illustrated like it's not a new thing, but now we can do it in a microsecond. And so—so a slow down a lot from the sort of day-to-day, uh, just keep building, just keep building.
L279Parker: Mode. And literally just read Jev documentation for like five days, like—like 15 hours, of just reading documentation.
L282Parker: for all of this advancement, the documentation still matters. To read the manual and to understand the problem.
L283Parker: And this goes back to your original question is what do we do with the practice of gaining understanding when that seems not to matter because the end of the—the machine will understand it better?
L284Parker: Is that it's—it's incumbent on us now, I think, more than ever as—as old men, let's say, to—to know when to go fast, but to know when to slow down and to know when to read the documentation, to know when to like go, line by line through, not code maybe, but through the—the ideas to where it lands for me in my usual workflow is—is so I have a—an IDE like sort of story, uh, used to use an IDE, loved it, loved the speed
L289Parker: Um, I switched to VS Code for a while and it's like, whoa, so fast. And—and you know, we all have our story, right?
L290Parker: Because that's what we worked on. That was our work surface.
L291Parker: That was our workbench for decades. And so—so the IDE, when it finally died for me, because I was on agentic about three years ago, um, and—and I saw the light with it, but it was like I was just kind of waiting, just like waiting.
L292Parker: It's like this thing can write functions now. This can write good functions now.
L293Parker: This can write—uh, you know, tests to—to establish cases. And—and so we've been on for a ride, right?
L294Parker: And, um, and so I knew that agentic was a thing, but it didn't occur to me that the IDE wasn't the correct, uh, work surface for this until—until, yeah, basically, November, maybe October. Yeah, October I think 2025.
L297Parker: Today, when it's like, oh, you can run these things in a terminal. You just fired up in a terminal and—and get direct interaction.
L298Parker: And to me, that didn't feel any different than texting back and forth with it as a chatbot kind of, as a very advanced chatbot. What changed was when we started recognizing markdowns as very useful.
L299Parker: But markdowns are just so much reading. And so then kind of maybe February, March of this year, um, realizing that HTML was actually a really nice way because it's so simple and—and straightforward, a really nice way for agents to, uh, convey like dense information and to essentially put documentation up that, uh, that is useful.
L300Parker: And so I realized to interact with your agent it's like typing into the prompt box is this sort of, uh, transient way of doing it. But if you get them to escalate a decision brief up to you, that has—you know, cool graphics and—and interesting
L302Parker: things to look at to keep your attention, again, it's with attention, that, um, that you tell it, well, give me—give every element a commentable, uh, box. And so you can now comment on everything, right?
L303Parker: And then realizing that you can tell it, okay, now make a button to just like feed this back into the prompt box. And so it feeds the—the HTML map of your comments back in.
L304Parker: And the agent has the file, the agent has the map, and the agent just goes back to it. And so you're not prompting through the terminal anymore, but the terminal is sort of
L306Parker: that, uh, focal point where everything comes together. And then because I—I built this terminal work surface, um, to do like multi-agent stuff and keep myself organized and made it all colorful and themable.
L307Parker: So—so when—when all of that was happening, it's like basically if you had a crash or if—if you accidentally wiped your prompt, like Control C is copy but Control C is also wipe your prompt in—in a lot of terminals. So it's like well, shoot, like how do we manage that?
L308Parker: And Omarchy's done it actually really well, uh, with—with Super C. I—that's Super
L310Parker: C is—is killer. Um, such a simple thing, but it's like now you don't, you know, wipe your prompt out.
L311Parker: So—so little stuff, but, uh, what I realized was when I looked at other people who are bigger and better than me, like Herder and like, um, Human Layer and, uh, who else is doing it? Anyway, the—the other—other terminal players for building these agentic work surfaces, uh, in—in a way that's nice, is that they had gone headless.
L312Parker: And so you could kind of open and close terminals, blah, blah, blah. And, uh, and run them, you know, non-locally and all that cool stuff.
L313Parker: And so the other code that got cracked with the whole HTML markdown interactivity was getting the.
L314Parker: Getting the—the terminal process separated from the terminal rendering. So kind of going headless.
L315Parker: So the terminal process is—it ends up being this—I'll just flip a—a screen share because it might be easier to show than it is to tell.
L341Parker: like whose face belongs to what thing and nobody really wants credit. Like that's not what we're doing here.
L342Parker: Um, but—but it's like it's good to know like, uh, like—like, uh, Tom Ballard with his, uh, with his familiar plugin. It's like that's—that's pretty killer work.
L343Parker: That's like—that's actually a really big deal. And, uh, and—and this meeting on the Omarchy thing, it's like, yeah, that's going to—that's going to sync for Omarchy users quite a lot of, uh, I don't want to say like revenue or anything, but quite a lot of, uh, clout, I'll say, uh, knowing that—that, um, here we go.
L371Parker: But, uh, changing sort of like—oh, the text isn't quite right or the colors aren't quite right. I need a little more contrast or—or whatever the thing is.
L374Parker: The—the classic IDE problem of like trying to zoom in like so the control zoom and—and so—so the big unlock here was, okay, so you've got all these terminal sessions running and look, you just close it down.
L378Parker: Everything's still going because the processes are still going. And to the point of like, because it's a terminal process, it's like you type in your prompt, you're half done, close
L386Parker: So we'll work flow, right? Like when you're really in the flow, you really want to—to stay on.
L387Parker: Like—like maybe you don't run, maybe—I mean, I see you've got some—some arrows and a bird in the back that looks like—like, uh, a bird hunting implication, but you're not going to catch
L390Parker: a hummingbird with those arrows. I'll tell you what.
L391Parker: I've seen. I've seen those guys move and you're not catching a hummingbird with that—that kind of a—a big—big beefy stick.
L392Parker: Anyway, um, but when you're—when you're really—really on the path, when you're really plotting along, when you feel like you've got everything rolling and that all that momentum, you—if you context switch, even to like
L394Parker: open your HTML, you swing over here and, uh, I don't—yeah, you kind of—you have to come out of it, reestablish, then you got to, you know, look at some files and it just—it just changes your mind a little bit. And so staying in, staying in, staying in is really what I'm going for for that—that lock-in.
L395Parker: And then when you disengage, uh, you—you do it intentionally. You—you intentionally, you're disengaging, you're going over
L397Parker: to something else. So you're going to a—a Twitter X.
L398Parker: You're going to, you know, a meeting. You're going somewhere else.
L399Parker: And so really being there is—is what the goal is with this, uh, attentional surface. Um, but—but yeah, so we popped open an HTML, um, I don't know if I told my guys to turn this into.
L400Parker: Yeah, we got some comments. So we've got one frame from each beat.
L401Parker: Uh, so you can throw your comments in there. Yes, because just yes, everything.
L406Parker: And then you come, come down, keep reading your docs, always read your bloody docs. Uh, it'll give you your decisions because it's a decision brief.
L412Parker: thing is this, uh, you probably can't see it and I'm not going to be able to zoom, uh, copy map. Uh, lets you copy and paste if you're a Luddite
L422Nix: You're looking at minimizing steps in workflow that slow you down, and you're parsing things together that need to keep moving without having to cut your flow state.
L423Parker: That's right. And so the decision is—is made very quickly and—and so you can establish, you know, when you reopen that HTML, you can establish what you're thinking.
L424Parker: You just, you know, voice to text, you can say, oh, this part of the song, uh, the French horn is really tinny, we need to change the tone. And so it'll dump that in there.
L425Parker: Uh, probably I've got my mouse over here, so emoji is going to get confused about where we are, but anyway, um, so you—you talk to text it, dump it in there. Uh, and—and yeah, like you say, get that—that decision closer to—closer—get the layers between you making the decision and the agent seeing the decision.
L434Parker: It's saying, you know, Cloudflare R2, blah, blah, blah.
L435Parker: It's giving you all of this stuff and you need to, uh, flag it as useful, not useful, or—or impactful and requires decision. Uh, and—and just go on and on and—and you just go through these things and it's—it's a—a way to slow down and to get your attention kind of right here right now.
L436Parker: Like why did you open this document? If you're not going to read
L454Parker: So it's, uh, yeah, so—so back to that question of—of what's the limit? And it's like you need—you need guardrails.
L455Parker: Some people, I think, are like—are really quite focused and they can work on a thing and get it done. And that's super cool, but, uh, but for guys like me, it's, um,
L459Parker: self-destructive unlock that lets you just spiral out of control and do a bunch of mistakes a lot faster.
L460Parker: And so it's, uh, I mean, the cool thing with this emoji, uh, the—the way that the path is kind of taken me along was, uh, yeah, it's like you realize there's a gap. It's no one's making
L465Parker: like a little 30-second or a one-minute clip telling a cool story. And, uh, and yeah, so this one, uh, [redacted].
L466Parker: And so that's what kind of inspired this whole idea. Is to like, well, you know, today and emoji, but also like this month and emoji and also like the year of Linux is 2026.
L467Parker: And so like how big do the ideas get? And so that's where the music anchoring it down really—really comes in.
L471Nix: Do you use Suno or something like that, or where do you get your music from?
L472Parker: No, it's all—it's all totally self-produced. It's like agents can write music.
L473Parker: Like I don't mean—I don't mean like mix music or remix music. I mean like this is all scratch done tracks like basically done MIDI style the way that a composer would do it.
L474Parker: And then you go in there with your like profoundly good musical taste you'd basically tell it like, oh, you know, the—the tone here or the voices here, whatever musical terms, if you're musical, uh, you'll kind of know what I'm getting at. But.
L516Parker: thirsting for it with, um, with a lot of this technology in the socials. Is, uh, we're just looking for connection, we're just looking for other people, and— and we'll get lost on it and be like, well, isn't that flashy, and— and that makes me feel nice, or that makes me feel angry,
L518Parker: whatever the, the, you know, emotion de jure is. Uh, but— but really we're looking for, for just human connection, and it is that simple.
L519Parker: And even though it's simple, the simple things are rarely easy, so simple and hard is the way, and, uh, you know, taking the time to do these meetings, it's, uh, it's simple to do, but, you know, it's much appreciated, because I know it's not easy.
L528Parker: on there so it doesn't— if it does take off a little bit, it should be manageable and whatever, but, uh, yeah, totally open to ideas and stuff like that. Um,
L531Nix: folks, I spent— so I have Claude Codex, Kimmy, Grok, all of them. And so when IPR for folks, I specifically say, "You're not trying to add any features.
L532Nix: You're not trying to change the design. Just look for things
L534Nix: that are could be problematic in the code itself." And so if you get a PR for me, that's what it'll be like.
L535Nix: That's what it will be— I don't think that you need any help at all with the design. And I'm not even suggesting that there's errors in the code, but if I find some, I will PR them.
L536Parker: Cool, yeah, yeah. Appreciate the tokens.
L539Parker: terminal, there's been some, like, very significant decisions of, like, okay, so do you want an agent with an open-source terminal project to just be able to grab a terminal?
L545Parker: I'll 100,000% just like, yeah, agents get terminals. What do you— what do we do in here?
L546Parker: But, um, it's— it's a computer, and the agent's driving, so it's like, you'll, you know, tell it where to go or whatever, but it's— it's, yeah, it's going. But, uh, but to do that into an open-source side of it, and it's like, oh, no, like, that's a lot more careful.
L547Parker: And so when I first did that, I— I kept stumbling over that decision of, like, oh, yeah, like, of course we'll do this thing, because it's exciting and new and— and— and big, but, uh, but yeah, putting the— the constraint on there of, like, wait a second, like, that is a security and—
L549Parker: and open-source kind of a, a— I don't want to say a nightmare, but a rats' nest for sure. And so— so what you might find is some, like, little sort of ragged, loose ends
L551Parker: of ideas that happened and then got tamped down. So it'll be really interesting to see, you know, what your— what your guys uncover.
L552Parker: Um, but yeah, to finish off the idea with terminal delight, uh, it went headless, um, even to the point of, like, if you— if you close down the terminal,
L556Parker: It's like, oh, shoot, I wasn't quite done with that, actually. Um, that kind of stuff.
L557Parker: It comes at the cost. There's always trade-offs, because then that terminal process is still running, that agent might still be running.
L558Parker: Um, but at least if you accidentally— you know, hot key and fat finger yourself out of a— a process that was running, it's not, you know, go find an agent, tell it to search for that session that you had going, that you need to recover, and— and it's all this friction.
L568Parker: and, uh, and it's a little bit of fun, so. And then, yeah, like, tons.
L569Parker: Of— tons of features. It's— it's been everything that, like, this is it, is, like, we can fix everything and so you're like, shoot, like, I need this thing here, and so agent usage.
L594Nix: processes it, puts it up on Omarchy.builders, writes the exposed, and then puts it on my—
L595Nix: Puts it on my clipboard, and then I post it. So my last thing I want you to talk about is, is there anything in here that you don't want to be public that you have said?
L606Nix: this arrow represents my subscription week, and it's all on all of them. And so if I were to burn the exact amount of tokens that it takes to burn
L612Nix: So that means that I've got 40% banked. That means if I burn hard, I'm still under pace.
L613Parker: Oh, I see what you're saying. So— so based on the different session limits, because codecs and claude have a different 7-day versus 5-day session limit relationship.
L616Nix: So go to my GitHub, Nick's Nix, and look for Burn Bar, and then grab the code out of that to use it for yours. Because banked for me is the most important thing.
L618Nix: And if I get closer and I can't find one because I haven't been using it, it'll say on pace, and on pace means I'm using it at the right level that I'll— that I won't run out.
L622Nix: Right. And so it takes it to the next level that when I kick off— I've got all these subscriptions.
L623Nix: When I kick off a terminal, if I allow it to, it will call out and use agents, and the goal is to efficiently use them all so that every one of them, when they get to their reset, is at 1% left.
L629Parker: yeah, I— I'm going to be honest, I don't have that on my plate or bandwidth. That's— that's— I would— if you already have that plugin, you should be able to basically just, like, give an agent terminal delight and say, like, make this plugin for terminal delight, uh, put it in the subscription panel as a separate tab, and you're off to the races.
L632Nix: I already have my agents working on PRs for you. Again, if you don't like it, throw it away.
L633Nix: I don't take it personally. It's not that level.
L634Nix: But go grab Burn Bar, and if you want, you can use that code. I would— you can just install Burn Bar and use it on your system to just see— you don't have to add it to your system.
L644Nix: Go to Nick's Nix on GitHub and then search for Burn Bar.
L645Nix: And I'm not special, but I see everybody telling me what the percent of it is. And it occurred to me one day, I want to know where I am in the process.
L648Nix: the reason I'm so far behind is because I've had so many conversations for hours now that I'm not actually coding, which is kind of cool.
L649Parker: I mean, yeah, I appreciate it. The— that attention really matters, right?
L650Parker: That attention really shows, um, I think the— the really cool thing, honestly, with agent— uh, agentic and— and where we and people are at, is where your attention go is going becomes more obvious.
L654Parker: to this or that, it got good. And when you didn't, it didn't.
L655Parker: And so— you know, I see— I see these little— little— little now, like, you just started whatever last week, these little meetings, like, this is a— this is a little thing, but there's 200,000 Omarchy users. And if even 1% of them are contributing in even a small way, you're— you've got your hands full for a while.
L667Parker: doom and glooming it. There's a lot of people that are hyper-optimistic, like, frightfully optimistic.
L668Parker: Um, and— and it's like, for— for us normally down— down here in the trenches, it's kind of like, well, I mean, it's really cool, but, like, yeah, we— even the guys at the top, I don't think really see this thing. And so it's— it's a little bit scary.
L669Parker: It's a little bit too fast. It's a little bit— uh, you know, a little bit of everything.
L670Parker: And, you know, is it going to— going to— oh, I think my— my guy's doing your burn bar.
L687Parker: Yeah, I mean, the— the dashboard itself, though, is, like, a really slick— I love the— I love the style. And I see that you're doing lots of stuff, and I don't know what to make because there's so much stuff.
L688Parker: And, um, and I really love Omarchy of the box, so I— for most of the stuff, and I know that you're not going to take offense to it,
L695Nix: Thank you so much for spending this much time with me.
L696Nix: Like I said, it'll kick off, put it on Omarchy, dot builders, I will post a little lead-up and mention you please retweet for your own sake. And hope to get to talk to you another time sometime soon.