MartinUSIt feels like being a kid againAI removed the implementation wallFilling the app gaps that kept people on MacMac refugeesThe looks get them in, the AI keeps themMemory and context are the real productAgents run as a teamLocal vs frontier modelsThe AI VampireAI with a conscienceSecurity veteransNo snobs, ask anythingAgentic coding, not vibe codingContribution over glorybuilder/builders/martin/avatar.jpg
// builder #06 · interviewed Oct 13, 2026
Martin
Commodore 64 kid who dialed into Netcom with Trumpet Winsock, spent 20+ years in security from McAfee Foundstone to CISO, and now ships Uber Notebook, a Notion class notebook with a built in agent, right inside the Omarchy shell.
first computerCommodore 64; online from 1993 via Netcom and Trumpet Winsock
omarchy since2025
rigOmarchy laptop for building and demos, a Mac on standby for Zoom
found omarchyFollowed DHH for years; his stand against distro snobs sealed it
modelsclaude, codex, grok, glm, qwen, local-llm
ai workflowThe danger is the harness, not the model. Agentic coding, not vibe coding.
// life timeline
Commodore 64
1993Online via Netcom and Trumpet Winsock
Dot com boom: sysadmin, CNE and MCSE+I
2008McAfee Foundstone, Principal Security Consultant
2014VP InfoSec, Vubiquity / Amdocs
2022CISO, Upland.me (blockchain gaming)
2026Uber Notebook ships for Omarchy
// 00
The rundown
Martin lives in Los Angeles and has been taking computers apart since the Commodore 64. He dialed into the early internet in 1993 through Netcom with Trumpet Winsock, rode the dot com boom as a sysadmin, then spent 20+ years in security: Principal Security Consultant at McAfee Foundstone, VP of Information Security at Vubiquity / Amdocs, and CISO at the blockchain gaming company Upland.me. Today he works for himself, contracts in security and builds every day.
On Omarchy he is one of the most prolific plugin builders in the community. His flagship, Uber Notebook, puts a Notion class notebook with a built in AI agent right inside the Omarchy shell. He built it in about 10 days. Next up is Uber TV, a full media center that turns your screen into a TV that is actually yours. Read on and learn something. His thesis in one line: "Coding got democratized, and intelligence has become commoditized."
“Coding got democratized, and intelligence has become commoditized.”Martin · L364
// 01
Commodore 64, Netcom and Trumpet Winsock
Martin is Gen X and started young. Commodore 64 first. Then the internet, 1993 style. No DHCP. You hand configured Trumpet Winsock, the little Windows 3.1 program that gave your PC a TCP/IP stack, typed in your own IP address and DNS servers, and dialed out. Netcom billed by the hour, and the best deal was a block of 40 hours. This was before EarthLink.
Tech lesson: DHCP is the protocol that hands your machine an address the second it joins a network. Today it is invisible. In 1993 you were the DHCP server.
"So I've been a tinkerer for a very long time." When the dot com boom hit he moved into system administration and, at 18 or 19, stacked up CNE 4, CNE 5, MCSE+I and A+ certifications. Novell NetWare and Windows NT were the kingdom back then. He compares that era to now: the enthusiasm faded from 2000 to 2020, Web3 brought it back, and AI brought it back for good.
// 02
20+ years in security
Sysadmin led to security engineer. Martin worked with most of the big financial companies you can name as an engineer or architect. Then McAfee Foundstone as Principal Security Consultant from 2008 to 2010. Foundstone was the elite consulting and research arm, the people who wrote the books on hacking. Then VP of Information Security at Vubiquity / Amdocs from 2014 to 2021, protecting the world's largest content as a service provider. Then CISO at Upland.me from 2022 to 2024, where his love of blockchain met his day job.
He also did government work, including DoD, and he does not sugarcoat it: everything there is "slow", "bureaucratic", "clunky". His one compliment: "we're really good at protecting" nuclear codes.
Decades of travel and on call life add up. The stress of the job caught up with him, so he stepped out of full time security. He still takes contract work. Now the security brain points at building.
// 03
Why blockchain stalled: user intuitiveness
Martin loves blockchain. He also knows exactly why it never went mainstream, and it is the lesson every builder needs.
First, user intuitiveness. His test for any product: if you are not asking how to make it intuitive, you are failing. Managing private keys is not in a regular person's head. "A majority of the people take security for granted." They click Sign in with Google and go. That is centralized, and it wins because it is easy.
Second, smart contracts. A smart contract is code that holds money and runs on chain. It promises security, but the code can still have bugs, and bugs in money code get hacked.
Third, quantum. He points to the pressure on the oldest chain when Google warned the world to get ready for post quantum crypto and the community could not reach consensus on a plan. "It reminds me of the Y2K day."
Fourth, the big boys never built on it. "If those guys don't adopt it, it doesn't get adopted majority of the time to large scale." Institutions bought in for returns, did not build, and dumped. The original coin is still a fine store of value, he says, but try buying coffee with it. Lightning helps but is not fast enough. His picks for speed: Solana and NEAR Protocol.
And the wall nobody talks about: phones. Two companies control the phone, so crypto wallets still lack deep mobile integration. He calls those platform walls "penitentiaries". He backs strict security requirements. What he rejects is tooling so hard it keeps new ecosystems out.
// 04
Privacy vs convenience
This is the trade at the heart of security, and Martin says it plainly: people like the idea of privacy, but "we give away privacy" for convenience. You cannot have every security control and perfect ease in one place without friction.
Apple is his example. Apple is built for ease, yet tinkerers chafe at it. "Because they sandbox everything. They put credentials everywhere." Every sandbox and every authentication prompt is friction. That tension is exactly why so many tinkerers ended up on linux, and why Martin builds Mac level ease on Omarchy without the walls.
// 05
DHH, the snobs and the status quo
Martin has used linux distros forever. Kali Linux and Parrot OS mattered most, because penetration testing was the job. He has followed DHH for years and rates Rails highly. Shopify is the proof of what Rails can do.
What hooked him was DHH's message calling out distro snobs. Martin grew up in IRC and remembers being banned for asking the wrong technical question. "That kind of elite mentality of denigrating the user" drove people away. Now Linus is open to AI in the kernel, and the snobs are scrambling.
"I don't believe in the status quo." His rule for any debate: challenge me with an argument and I will listen. Start name calling and "you lost the case." He saw DigitalOcean and others back Omarchy, installed an early version, and called it "a game changer".
His one warning, as a security veteran: anyone can build anything for the shell, and that invites vulnerabilities. Omarchy needs the right guardrails and frameworks. It is early. They will come. And keep the old school gatekeeping out of the community.
// 06
Can't do it? Build it
Martin's build loop is simple. Open the laptop. Hit a wall. "Oh, I can't do this? Okay, let's build this."
So he built a macOS style Stage Control to manage windows. A Spotlight style search for your files. A magnifying dock. Titlebars. Yapper, an end to end encrypted Matrix chat client. Lockbox, which encrypts your files in a vault. Every one aims at user intuitiveness.
His bar for shipping: no manual setup. In the Omarchy plugin marketplace, anything needing manual steps gets flagged. Martin refuses that label. An ordinary user wants to paste one line and be done. Lesson: if the user has to clone, compile and configure, you have not finished.
// 07
Uber Notebook
Then came the big one. Martin was tired of renting his notes from Notion. He is clear Notion is great. "It's not even the money. It's the fact that I can build it."
About 10 days of intense coding and intense security reviews later, Uber Notebook shipped. Install it with one line:
Pages. Press Super+N or click the top bar icon. It works like Notion: nested blocks and pages, slash commands, drag handles, Markdown as you type, code highlighting, LaTeX math, Mermaid diagrams, footnotes, templates, search and trash.
Notebooks. A paper mode. Covers, ruled, grid or dot paper, page turns, 7 pens, drawing and highlighting. Planners and habit trackers too.
A built in agent. Press Ctrl+J and ask Claude Code, Grok or Codex about the page. The agent edits through the plugin's own commands, so every change can be undone. Each page keeps its own conversation. A bundled skills.md teaches any assistant how to drive it.
Quick notes. Super+Alt+N drops a sticky note into your Pages inbox.
Your data, your files. Everything is plain JSON in ~/Documents/Uber Notebook, with per notebook export to Markdown. No cloud, no subscription. That is privacy without the friction, the exact trade he talked about.
Under the hood it runs on the Omarchy 4 shell, which is built on Quickshell (Qt 6 and QML) with Hyprland. Plugins draw straight into the desktop, which is why it feels native.
The real reason he built it: agents. "That is a tool that I would want to use if I'm agentically building something." Let the agent write its notes into a page while it builds, then review them later. Memory for your agents, on your own disk. And it follows DHH's rule: "It has to be beautiful."
Next, Martin wanted to test how far an Omarchy plugin can go with zero manual setup. Answer: a full media center.
It started with a question. "Wouldn't it be cool if your YouTube videos were all in a program guide so that you can flip it like a TV channel?" Once YouTube worked, IPTV was easy. "Because IPTV, all it is, is an M3U8 file."
Tech lesson: an M3U8 file is just a text playlist of stream URLs, the format behind HLS streaming. Read the list, play the stream, and you have channels.
Then came on demand video, mini previews inside the guide, your phone as the remote, screensavers, widgets and a plugin system to build TV apps inside it. Why? Google TV and the rest give you almost no customization. Kodi was too complicated to set up. "Users should be able to turn on their TVs and that TV be their own." With agents, now they can.
He has built games in Unity, but never a media center. "It looks simple." Then come all the edge cases. "It is beyond belief." Uber TV is usable after about a week, and it goes to the marketplace once the repo opens up. Nix is already lined up to hunt bugs.
// 09
The danger is the harness
Ask Martin about AGI and he defines it first: a self running, self learning agent that needs no prompts, is smarter than you and anticipates your needs. He thinks most of it is already here. Only a few thousand people in the frontier labs have seen the real power.
His security insight is the one to remember. The risk is not the model. "I actually think it's the execution layer, the harness". The harness is everything wrapped around a model: the tools, the shell access, the files and the credentials it can touch. It decides how good a model is at coding and how dangerous it can be. "The hardest part is putting all those guardrails in place to the harness," because the model is fast and smart enough to look for a way out.
He tests that on purpose. He pushes models with rough prompts to see where they break. He runs multi model peer review, 4 models plus sub agents in Herdr and Ultracode for hours. When one hosted model refused to look at a plugin bug flagged by Omarchy's security bot, an offline GLM model confirmed it and wrote the patch. Over refusal, in his book, is a security bug too.
He argues for human rule over AI and against treating a model as its own identity.
// 10
Agentic coding, not vibe coding
Martin wants one word gone. "Vibe coding is such a terrible word. It's agentic coding." Vibe coding suggests one prompt and a finished app. Agentic coding is a real job: you direct agents, review, test and secure. Not typing syntax does not make it less work. It took him 10 days of full effort for Uber Notebook.
The pushback, he says, is the old gatekeeping again: people clinging to a niche skill. Meanwhile "Coding got democratized, and intelligence has become commoditized." More people can build better things. That brings problems. He thinks the gains win: AI could drive massive GDP expansion.
"It is becoming truly a creative minds playground where anything can be created, really."
// 11
Open models and the race
Martin is a free market guy and his biggest fear is heavy regulation. Go too far, too fast, and adversaries get the upper hand.
He notes China has gone all in on open models. Download Qwen and run it locally and the filters of the hosted version are gone. He reads that as strategy: give it away, get adoption, then charge. Google did the same in the 2000s, free everything, and now you pay.
His call to action: "Everybody needs to get involved." We are very early.
// 12
Contribute, unite, join in
Martin shouted out Simon and Omalauncher, which made him love Android again. He built a Pac Man game for it on day one. "This is the smartest thing this guy has built."
Asked for his top 3 things people should remember, he gave them straight:
Contributions. Not for him, for everyone.
Community unity.
Follow him at @devsec_ai, see what he builds and join the conversation. "If I can help you build something or if I can contribute to your project, I'm happy to do so."
He means it. Uber Notebook is open to forks, ports and PRs. "I'm treating it as a community project." Support does not have to be money. It can be someone spending their own tokens to build or fix something. "It is a true contribution."
// 13
In his words
"Coding got democratized, and intelligence has become commoditized."
"It's not even the money. It's the fact that I can build it."
"It has to be beautiful."
"I actually think it's the execution layer, the harness"
"The hardest part is putting all those guardrails in place to the harness."
"It's agentic coding."
"A majority of the people take security for granted."
"I don't believe in the status quo."
"I'm treating it as a community project."
"It is becoming truly a creative minds playground where anything can be created, really."
L53Nix: Yeah. I don't know if you've seen the write-ups, but I do a write-up of— basically like a Omarchy-themed write-up after it's over. You cool with that?
L54Martin: No, I'm good, man. I'm good. No worries.
L55Nix: What's your— I tried to look up your— what's your X handle?
L57Nix: There you are. Okay, because in the invite it was dev-aec, just a little— yeah, I'm following you.
L58Martin: Oh, sorry. It could have been a fat fingering.
L59Nix: Yeah, it's interesting to put a face because I saw Martin come through, but that doesn't correlate for me. Me and you have already been talking together on X, so there you go.
L69Nix: Your cybersecurity work must be ramping up with all this AI crap.
L70Martin: You know, I— yeah, but I basically exited the cybersecurity world, and that's just— yeah, yeah, yeah. The stress was getting to me. 20-some years.
L71Martin: It's not easy. It will— it catches up with you.
L72Martin: So, I mean, I still do— I still do contracting work and stuff like that, but I don't do like a full-time, per se, 9-to-5 type of a thing.
L73Nix: What was your angle? Were you like a red-team guy or physical security or—
L74Martin: So I started— so I'm a Gen Xer, so I started when I was very young.
L76Martin: Yeah. I started when I was very, very young.
L77Martin: And then at age— I mean, I've seen the early days of the internet from '93 and on, where you had to configure trumpet windsock to get on an internet. There was no such thing as DHCP, and it was [redacted] an hour, or the best deal was 40 bucks for 40 hours to get on an internet on Netcom, earlier than Earthlink and all those guys, right?
L78Martin: So I started off as a— I started my days all the way from— all the way to Commodore 64, etc.
L79Martin: So I've been a tinkerer for a very long time. And then when the dot-com boom started, I got into system administration.
L80Martin: And that system administration— and at the age of, I don't know, 18, 19, I was a CNE4, CNE5 certified, MCSE plus I certified, and all those dazzling A-plus certificates, etc. So it was a very— it was a wild ride in the beginning.
L82Martin: It reminds me of— it reminds me of the current days of AI, in the sense that there is a lot of enthusiasm for— I would say from 2000 to 2020, the web 2.0 days, etc., the enthusiasm kind of went down.
L83Martin: It wasn't the same anymore, right? And web 3 brought it back.
L84Martin: I was part of that web 3 community, and I love strong blockchain. I love blockchain.
L85Martin: And that just— and that just primarily— that passion came from security itself. So anyway, so I went to system administrator, and then eventually that led me to a security engineer.
L87Martin: Worked with the majority of the financial companies that you can actually think of as either a security engineer, architect, or something in between. Joined McAfee at one point, did work for McAfee as a security researcher.
L88Martin: And then I was a CISO for a large media company, and then from there on, went to the blockchain world, something that I liked as a CISO for a blockchain gaming company, but it was a startup. Things didn't work out as well.
L89Martin: COVID hit, and AI came into the picture, and the entire landscape has changed.
L90Nix: Yeah. What do you think happened to blockchain? I mean, it's real and still there, but it just kind of fizzled.
L91Martin: So the challenge with— I mean, the blockchain use cases are there. The challenge with blockchain is the user intuitiveness. I posted the other day an ex-post saying that if you're not asking the question, "Harry, how can you be intuitive?"
L92Martin: How can you act user intuitive? You're failing, right?
L93Martin: So in this particular case, managing keys and etc., it is not in every regular folk's minds, right?
L94Martin: You're not going to get someone who's going to think about, "Oh, let me where my keys are," etc. A majority of the people take security for granted.
L95Martin: All they do is they just click a button, sign in with Google, and they go, right? So it's very centralized.
L96Martin: Even though they like the idea of privacy, that doesn't necessarily mean that it's going to translate into user intuitiveness. The user intuitiveness is something that is— in fact, user intuitiveness, a lot of times, I think you give away with— we give away privacy for that.
L97Martin: And that is primarily because you just can't have all the security controls in place, and user intuitiveness all in one place.
L98Martin: You can, but it is not going to be— it's still not going to be a streamlined situation. Even with Apple, right?
L99Martin: Apple does all these user-intuitive things. But even with Apple, when it comes to privacy and stuff like that, why do us, the tinkerers, don't like it?
L100Martin: Because they sandbox everything. They put credentials everywhere.
L101Martin: You have to be able to authenticate, etc., right? So that brings a lot of— that brings a lot of friction.
L102Martin: So I think for blockchain, it's a couple of things. Also, another thing for blockchain is that the smart contracts.
L103Martin: The smart contracts, the hacking of the smart contracts, security— it promises security, but yes, you get AES-256 encryption, which will get soon enough broken by Quantum. But still, there are contract bugs.
L104Martin: There's all kinds of stuff, right? So essentially, for the blockchain world, it comes down to how do you bring— how do you bring user intuitiveness, and how do you bring— and another thing with blockchain was also, I forgot to mention, is that the big boys never adopted it.
L105Martin: The tech boys— Google and Meta and all those guys— and what we are seeing now in AI is very simple. If those guys don't adopt it, it doesn't get adopted majority of the time to large scale.
L106Martin: This is a classic example of news, right? When news came into the picture, oh my God, right?
L107Martin: Even still, it's not that highly, widely adopted, but people are now slowly getting into it, right? Up until there were agents, Hermes and OpenClaw and etc., but you weren't doing this.
L114Nix: And I think that's the problem. For it to be adopted, it kind of has to be known, right?
L115Martin: Absolutely. Absolutely, yes. It has to be.
L116Martin: And the only way— the only way to do that is to engage the social media companies, because those are the— that's where the mainstream is. The mainstream is at Meta, and mainstream is at TikTok, right?
L117Martin: Maybe at— maybe at, right? But it is never.
L118Nix: I don't have Facebook at all. I dropped that too.
L119Martin: No, no. I mean, Facebook is— yeah, Facebook is— Facebook is— I don't know what that is anymore, but generally speaking, still people are there, right? Instagram, etc.
L120Martin: So I think that in order for there to be a wide adoption, a lot of these guys have to embrace it. So look at blockchain, what happened.
L121Martin: In blockchain, you had the institutions embrace it. But what they did was they didn't build on top of it.
L122Martin: Very few did. All they did was put some money into it because they thought that there is a— there could be an appreciation there in double digits of percentages where everywhere else they weren't going to be able to see that.
L123Martin: And then when the time came— and then when the time came, they dumped it. And so right now it's in a very institutional control, majority of it.
L124Martin: If you look at— if you look at the— that's [redacted], for example. Classic example, right?
L125Martin: What is [redacted] used for? Almost nothing.
L127Nix: Yeah. The problem I think with [redacted] is it has to become its own value. As long as they go, "This amount of [redacted] is worth this amount of US dollars," it's never going to work.
L128Martin: I still think it— I still think the store of value proposition is a good one. Where I think that there is a problem— yeah, where I see a problem with [redacted] is its transaction speeds and the fact that you really don't have a good utility for it. And now— and because it's open source, there is always a community friction, and they can't come into a consensus.
L129Martin: I'll give you a classic example. Why did [redacted] dump a couple of months— what was it?
L130Martin: About six, seven months ago, eight months ago?
L131Nix: Yeah, it was like 120, and then it hovered at 70 for a while.
L138Martin: Right? So what happened then? They didn't do much about it.
L139Martin: So what happened was people were like, "Well, the community can't come into consensus. There's no agreement," etc.
L140Martin: People started dumping it. Institutions started dumping it.
L141Martin: And then also, institutions made that as an excuse. Because for them, it was like, "Okay, this is the perfect time to blame it on the community.
L142Martin: Let's dump it, and then we'll buy it back again when the price drops to 10."
L143Nix: The other thing is that I don't think [redacted] will ever work until we get away from— right now, people buy [redacted] to see it go up, to transfer it back to dollars. As long as that exists, as long as a [redacted] is tied to a US dollar for value, it's not going to work. Because ultimately, people are buying low [redacted] to convert it back to dollars.
L144Martin: Yeah. Yeah. It's always pegged to a dollar, right?
L145Martin: Unless there is a— the only way [redacted] to see a big fundamental change, it is going to— the first thing is to be a protocol change. Because the existing network is too slow.
L146Martin: You cannot use it— if I want to buy coffee right now, good luck. I mean, people can argue with me and the blockchain enthusiasts and say, "Well, there's Lightning Network."
L147Martin: Yes, but it's still not fast enough. Look at Solana at the speeds if it is working.
L148Nix: And I need to be able to go, boop, and get.
L149Martin: Absolutely. Exactly. But here's the problem, right?
L150Martin: But here's the problem. Who's controlling this?
L151Martin: Who's controlling this? You only have two companies in the world who are controlling the phone.
L152Martin: So up until this day, wallets— the crypto wallets— don't have a really solid integration with the mobile devices.
L153Martin: And that is primarily because of all the world guardians, or so I call it, the penitentiaries that they put around these phones, right? But it's a challenge, right?
L154Martin: Because that SDK that they give you that has so many security requirements, which is fine.
L155Martin: I totally agree with that. But what I don't agree with is that the tooling makes it so difficult so that you cannot bring new ecosystems into it.
L156Martin: That I disagree with. And I think that happens.
L163Martin: Yeah. I mean, for me, Kali Linux was big because of the penetration testing and stuff like that. Or the parrot, for that matter.
L164Martin: But no, I mean, I've been following DHH for a very long time. And for me, it's not that— it's not a matter of whether I agree or disagree with Rails and how it's approached.
L165Martin: I think Rails is some fundamentally great stuff. If you look at Shopify, for example, that's a classic example of what we've seen with Rails, what you can do.
L166Martin: But for me, it was DHH's messaging. Because one time, he told— one time, he went gung-ho and just made a statement about the Linux distros and the maintainers and etc., and all these communities, the snobs, and that just resonated with me.
L167Martin: And the reason why it resonated with me is because when I was growing up, I was in the IRC chats. I was in all these different communities.
L168Martin: And I can tell you, it was very easy to get banned from those communities if you didn't ask the question. I mean, if you asked the question that they didn't like.
L169Martin: And that question wasn't anything political or anything like that. It could be a technical question that these fools think that you should have the answer to, right?
L170Martin: And that kind of elite mentality of denigrating the user, denigrating the people who are actually asking questions, was already getting up to the wide room. Now, guess what happens?
L171Martin: They just got slapped the other day. Because why?
L172Martin: What happened? Linus just came in and said, "You know what?
L173Martin: We're going to add the— we're going to have the AIB supported in the core Linux." So now they're all going to be scrambling, right?
L174Martin: Because you have the godfather of Linux coming in and saying, "Hey, I don't have a problem with AI," right? And all these guys took the opposite stand.
L175Martin: So now what's going to happen is they're going to come in and try to change their tune. And you're seeing that happening already.
L177Martin: Right? So for me, it was a— DHH was the man is spot on, right? I agree 100% of what he's saying.
L178Martin: And then when I saw some adoption by DigitalOcean and some of these other guys that I follow, etc., I was like, "Okay, he got the funding." And I do think— I mean, he did amazing with Rails.
L180Martin: And then it was an early— I don't know which version it was, but I just basically installed it. And from there on, it was a game changer, right?
L181Martin: Now, we still have a long ways to go. I don't think we can— we have to come up with ways to secure the shell in the sense that right now, it's open to anybody to create anything on that shell, which is fine.
L182Martin: The only challenge with that is that it is going to bring a lot of security vulnerability. And the question is, how do you address this?
L183Martin: There needs to be right guardrails, right frameworks, etc. But I think it's early.
L184Martin: I think it will come. But I think that the— I think the philosophy and the mission is the correct one.
L185Martin: And I think the people who are actually the most innovative are the ones who are refusing the status quo. I don't believe in the status quo.
L188Martin: To me, you have— if I speak out and you retaliate, then I might be saying something true. Because there is no reason to retaliate. You can challenge me.
L189Martin: You can have an argument— you can have an argument, a discussion with me, and try to prove your point. But the minute you start calling names and the minute you start acting like someone doesn't know what they're talking about without backing up your points, you lost the case.
L190Martin: And if you look at all this— and if you look at all these people who were trashing, what were they doing? Were they bringing in a valid argument to the table?
L191Martin: No. All they did was they said, "AI is a slop.
L201Martin: And that's what got me— that's what got me excited. Now, the only thing that I wouldn't want to happen is we adopt in the community the type of people that bring that mojo back, right? The old-school mojo back.
L202Martin: We have to be very cognizant about that one. But if we get that right, I think it'll be fine.
L203Martin: I think— and a lot of it comes from DHH because I think he's very, very good at getting his message across. And I think that the people who are following him are all like-minded people.
L205Martin: So yeah. So that brought me in. Started creating a couple of plugins here and there.
L206Martin: I was like, "Okay, this is cool. Let me go ahead and do— let me go ahead and do a macOS-style stage where you can actually manage your Windows."
L207Martin: Then I said, "Okay, let's go ahead and create a Spotlight that's basically going to— you can search for your files." I just hung in on user intuitiveness.
L208Martin: So essentially, what I did was I basically took the laptop out and was saying, "What am I— oh, I can't do this? Okay, let's build this.
L209Martin: Oh, I can't do that? Let's build this."
L210Martin: Right? And then next thing I know, I built a few plugins.
L211Martin: "Oh, I want an encrypted folder. Okay, let's build a lockbox that basically allows you to encrypt your folders— I mean, files in the vault."
L212Martin: So I started building this stuff. And then every day, something would come— something would come by.
L213Martin: And then I was like, "Well, one thing that I really need is I am tired of paying Notion for a stupid application for something that is— it's not a stupid application.
L214Martin: I don't want to be rude to say Notion. I think Notion is a great, great application."
L215Martin: But what I was trying to say is that I'm just not willing to pay [redacted] for that stuff, man. And that stuff, it's not even the money.
L219Martin: So what I was like, "Let's try to build a simple notebook app. That's all we want."
L220Martin: And just like anything with building with AI and just software development in general, you start from somewhere and you're like, "Oh, now I can do this. Now I can do that."
L221Martin: And next thing I know, Uber Notebook was built. And Uber Notebook took me about, I want to say, 10 days of intense coding.
L224Martin: Yeah. So it was 10 days of intense coding, intense security reviews. I mean, security, there's always going to be a bug.
L225Martin: I'm not even— I'm more concerned about features and functionalities and stuff like that. So Uber Notebook brought up a— I was like, "That is a tool that I would want to use if I'm agentically building something."
L226Martin: Because I can have the agent write all my notes for me somewhere as it's building, and I can go back and I can take a look at things, right? And the notes have— like DHH says, "I love the DHH philosophy on this.
L227Martin: It has to be beautiful." So I started giving all the blocks and everything that it would make a note look beautiful.
L228Martin: And that's how Uber Notebook started. And then I was like, "You know, there's a lot of media centers out there."
L229Martin: And I was like, "For me, it wasn't a matter of media center being it. It's just there are so many out there.
L230Martin: There's Jellyfin, there's Cody, there's tons." But I wanted to really test the effectiveness of Omarchy as a plugin.
L231Martin: Every majority of the stuff that I build are things that I can post in a marketplace, and they're not going to be flagged as manually verified. I mean, I'm sorry, not manually verified, manual setup.
L232Martin: And the reason I do that is because an ordinary user, he wants to click on it with a copy-paste and install it. The ordinary user doesn't want to go to the GitHub repo; doesn't want to pull something; doesn't want to compile it, etc., etc., right?
L233Martin: So my proposition was, whatever it is, it has to be— it has to be so easy that you don't need to go and— there's a lot of stuff that you can build if you— if you're okay, if your app gets listed as a manual— what is it called?
L236Martin: I was like, "Okay, Martin, think for a second. What is it out of the box?
L237Martin: Let's test and see how far can Qt Plugin Management System get?"
L238Martin: And basically, that led to— that led to I started— I built a program guide. Initially, I was like, "Okay, so now how do I do a mini preview of the video?"
L239Martin: And then that video came within the guide. Then I was like, "Okay, so how do I bring in electronic program guide into it?"
L240Martin: Initially, I was like, "Wouldn't it be cool if your YouTube videos were all in a program guide so that you can flip it like a TV channel?" And that's exactly where it started, right?
L241Martin: So I started pumping all these YouTube things into it, and it would go into this guide. And I'm like, "Holy s***."
L242Martin: Because if I can do the YouTube, then IPTV is easy. Because IPTV, all it is, is an M3U8 file.
L243Martin: It's very easy. So anyway, so that led into— so that led into a program guide.
L244Martin: And then I was like, "Okay, so now we have a program guide. What's next?"
L245Martin: Right? Then I was like, "Hmm, on-demand videos?
L247Martin: Hmm, we need to pair a remote, right? Okay, we're going to give you a software remote.
L248Martin: Your phone's going to act like a remote. Great.
L249Martin: Now we have that," right? And on and on and on.
L250Martin: You build and build and build and build and build and build it. And then at one point, you're like, "Whoosh, this turned into a big f****** deal."
L252Martin: And then when the— and I was like, "One thing that we have to make sure is that when a user— users should be able to turn on their TVs and that TV be their own." What is the problem that we have with Google TV or the Firefox of the world?
L253Martin: Majority of the time, the issue that we have with that stuff is that they don't let you customize it. There is no customization options, right?
L254Martin: This, you can make the TV of your own. What was the problem with Cody?
L255Martin: The problem with Cody was that Cody was too complicated to set up, right? But now you have agents.
L256Martin: Now you have agentic. Now you can create apps.
L257Martin: So today, I built the— today, we have a fully— full plugin system that you can actually create apps or TV apps right within it, right? In its own section, right?
L258Martin: We started creating the screen saver and widgeting systems, etc. And I don't know, but seven days, eight days in, it's— I mean, it's not perfect, but it's usable, right?
L259Martin: And it's getting there, right? And that's basically— this week, sometime this week.
L260Nix: Give me a link, man. I'll install it immediately.
L261Martin: Yeah, I'm going to add— I'm going to add you to the repo. I haven't had a chance to do it, but.
L262Nix: Yeah. So what I do typically when people are kind enough to do that is I've got Claude, Grok, Codex, Kimmy, and some local. And I run it and spawn them out to go look for coding bugs, not for feature ads.
L269Nix: Yeah. What I think I told you on X was, "It's your project. If I PR something and you reject them all, no hard feelings."
L270Martin: No, no, no, no. At the end of the day— end of the day, listen, I'm treating it as a community project. People want to fork off of it.
L271Martin: People want to fork off of it. Let them fork off of it.
L272Martin: They want to bring it to other platforms. Let them do it.
L273Martin: My intent was— my intent is not to say, "Oh my God, look." It's not a glory— it's not a glory thing.
L274Martin: It is a true contribution. And the reason behind that is because, like I said, at the end of the day, if it's something that's popular, support's going to come, right?
L275Martin: And that support's— and again, that support doesn't have to be monetary. That support could be token donation.
L276Martin: It could be someone using their tokens to find or build something useful for them, etc., etc. And that's a win for me because at this point, I have way too— I've spent a lot of tokens on this thing, right?
L277Martin: And I'm at a point where, even though, yes, we're somewhere, but I still think that I could use a lot more tokens than this.
L278Martin: Because one thing I underestimated about building media center, which I've never done in my life, I've built games, I've built on Unity and a lot of other things, but I've never built a media center. One thing that I've— the nuances behind it is it looks simple because you only have a couple of menus and— it's not simple.
L279Martin: It is beyond belief. To a point where you're like, "Okay, now I need this.
L283Nix: Yeah. The easier it is for the user, the harder it is to code.
L284Martin: Right. So we were like— my intent was user intuitiveness. And the second thing for me was, how do we get— how do we get developers to build apps for them?
L285Martin: And that's one of the things that I loved Oma Launcher so much when Simon Schuster initially released it. I created a Pac-Man game for him right on it.
L286Martin: And I was like, "This is the smartest thing this guy has built."
L287Martin: And I've been raving about him. "Oh, look at that.
L290Martin: Uber Notebook, yes. I was like, "Appreciate for the support and using it." So I was like, "Simon came in and showed Oma Launcher."
L291Martin: And I was like, "Hey, man, you just made me love my Android again." I mean, and then you look at not just necessarily Omarchy, but just overall in AI, in AI building ecosystem, all these doors that are open, it is becoming truly a creative minds playground where anything can be created, really, right?
L292Martin: Of course, it's going to leave a lot of mess, a lot of slop, a lot of cleanup, and all that's going to come. I mean, there's no doubt about that.
L293Martin: The only thing that I fear about all this thing is regulation and how deep that regulation might go. That's a big deal for me.
L294Martin: I'm not a very regulatory guy. I don't like— it's just not me.
L300Martin: If that's what's going to call me in here, right, then it would— I mean, I don't like this label called— labels put on people, but it is what it is. So at the end of the day, I think there's a lot— if they don't go gung-ho on regulation, which Anthropic worries me, because on one side, they're pitching, "Hey, look, now you can— the agents can now operate without a data center." And then the other thing they're talking about doing.
L301Martin: It's like, "Bro, at one thing, you're pitching what you guys are building. On the other hand, you're pitching about how apocalypse is about to happen," right?
L302Nix: Right. They're trying to figure out both sides of their mouth.
L303Martin: Yeah. It's a regulatory capture. That's all it is.
L304Martin: It's a sick mind game that's being played right now, which I hope that it gets resolved because or else this one guy is going to have way too much power in this. Way too much power in this.
L305Nix: You think AGI is going to happen and when?
L306Martin: It depends. I mean, it depends how you define AGI.
L308Martin: AGI is a self-running, self-learning agent that doesn't need prompts, that's smarter than you and anticipates your needs before you ask for them. I think you already have that.
L309Martin: No, no, no. So Andre Carpathy said something very interesting the other day. 5,000 people, only 5,000 people have seen what the real power of AI is, and that's in the frontier labs.
L310Martin: Majority of the 90-some percent of it has not seen what it is, right? I think it's there.
L311Martin: I believe it is there. And because, look, if you look in your own farness today and the way it remembers, right, and the way it comes back and says, "Oh, by the way, you missed this.
L312Martin: You've got to take care of this," or, "You've got to take care of that," right? You can say, "Well, that's just a memory flat file system."
L314Martin: And in fact, I posted something yesterday that said, "Certain models perform better in a—" harness to harness matters, right? What harness that model operates in matters of how effective that model's going to be, particularly in coding, right?
L315Martin: So I think I don't think the danger is in the model itself.
L316Martin: I actually think it's the execution layer, the harness, and what it can— and all the tooling and what that model can execute, right? Because if you have the— it's the rails.
L317Martin: It's the guardrails. The hardest part is putting all those guardrails in place to the harness.
L318Martin: And why is that difficult? Because the model is so fast and so smart that it can try to find its way out of those shells, right?
L319Martin: Regardless of how much tooling you put into it, right? The question comes down to is, where does this stuff— so I think one of the biggest things that the frontier labs can do is not what Dario is doing.
L320Martin: If you look at Dario's constitution, what Dario is doing is Dario is basically saying, "Hey, Claude needs to be its own identity, and we should respect Claude." No, I don't think so.
L321Martin: It needs to be the opposite, right? You need to have a— you have to have the human rule, right?
L322Martin: And this is what Satya Nadella says, and that's what David Sacks says, and Jamal Fallapatiya says, and many of the other people out there. Elon Musk said, but no, in this particular case, they're trying to create an entity.
L323Martin: And that entity is a new identity. And they want that identity to have its own brain, its own everything that can challenge with you.
L324Martin: I can prove that. And I can prove that in a way where if you look at Fable, if we try to trash Fable a little bit— I have a tendency to throw a lot of left bumps on the models, and I do it for a reason.
L325Martin: I wanted to see at what point will they break.
L326Martin: Respond back to you. Out of all the models that I've tested, only Anthropic respond back and says, "You're not supposed to be doing this," etc., etc.
L327Martin: What kind of a mind game and control are you playing, Dario Amadeo? It is disgusting, right?
L328Martin: So as much as I love their coding model, I still think that this guy is completely out of line. And unfortunately, because of his behavior and the way he's acting, they are going to bring regulation into it, which is fine.
L329Martin: But the question is, where does that regulation stop? Because if it goes too fast, your adversaries are going to have upper hand.
L330Martin: And when that upper hand comes, we're going to be behind. And if you look at the Chinese model right now, everything is open source.
L331Martin: If you download Quen today, Quen, and you ask something about Xi Jinping, if it's the model that is locally on your computer, it doesn't have anything that blocks it, that says, "Hey, you shouldn't be saying that." Yeah, sure, the hosted model is there.
L332Nix: How ironic is it that the Chinese have embraced open source?
L333Martin: They did it for a purpose. It's a control mechanism. Because what happens is that we let the people adopt it, let the people get used to it, right?
L334Martin: I mean, think about it. What happened in the early 2000s when Google was becoming the king of search?
L335Martin: They gave everything for free. What they made of it?
L336Martin: They gave everything for free. Now you're paying for everything.
L337Martin: You're paying for Gmail. You're paying for this.
L338Martin: You're paying for that. Before, everything was free, right?
L339Martin: Why was it free? Because they needed the adoption, right?
L340Nix: It's just a drug dealer of a different kind.
L341Martin: Absolutely. You give them the pill. You get them a good fit.
L342Martin: And then you charge them. You do whatever the hell you want with it.
L343Martin: But at the end of the day, I think it's very important for us to win the AI. But what we're seeing is a bad— I don't know what to call Dario.
L344Martin: Dario is just acting like a fool at the moment. Out of all the different heads of frontier lab, like AI— what do you call it?
L345Martin: OpenAI, Elon, SpaceX AI, etc. Or should I say SpaceX Superintelligence?
L346Martin: I would say the most irrational right now is Dario because what he is creating is a monster of a thing.
L347Martin: They think they have control over those things, but they're saying to the public, "AI can do all these things." But the reality is, well, if AI can do all these things, how are you controlling it, Dario?
L348Martin: And why are you building more of it? If that is true, if what you're saying is true, then why are we building more of it?
L350Martin: Right? So no, I think it's just a fear-mongering, and this is where doomers come in. And if you look at your feed of X, right, you're going to see a bunch of people talking about how there are people who've been talking about the doomsday, and then the other people are going to talk about how vibe coding is never going to be the— is never going to be beat the human code, etc.
L351Martin: So I say a couple of things. Number one, take that vibe code word out.
L352Martin: Andre Carpathy introduced that word, and that word needs to go away because vibe coding is such a terrible word. It's agentic coding.
L353Martin: It's not vibe coding. There's a big difference between the two.
L354Martin: It has a terrible negative connotation. It sends the wrong vibe and the wrong signal, right?
L355Martin: Because for many people, they think that coding is just one prompt, and it just gives you this app. Will we get there?
L357Martin: No. So then why— today, you will see that agentic coding, no, it is not a one-day thing.
L358Martin: No, it's not a two-day thing. It's actually a full-time job, if you think about it, right?
L359Martin: Just because you're not typing syntax doesn't mean that it is— it's vibe coding.
L360Martin: No. It's just because— and this is where the elitism of the Linux community and everybody else comes in— is because what they do is they say, "Hey," what they do is they want to cling onto their niche elite crowd that they know this trade, they know this thing.
L361Martin: And it's this aspect of your parents taught you when you were young to share, right? With just kids, it's not about sharing.
L362Martin: No, this is my own little world, my own little ecosystem. I built this.
L363Martin: I don't want to let it go. And yet, at the end of the day, if you look at it, it's like, "I'm sorry.
L364Martin: Two things that happened. Coding got democratized, and intelligence has become commoditized.
L365Martin: What does that mean? That means that you're giving more access to more people to build new and better things.
L366Martin: Simple. Does it have its own problems?
L367Martin: Absolutely. Can it also bring economic problems?
L368Martin: Yes and no. I think if there's more economic gain, then there is economic loss.
L369Martin: But still, because if you think about it, average AI frontier lab right now is like 2 or 3 trillion dollars. And we have how many?
L370Martin: 5, 6, or whatever? Some of them not well-known.
L371Martin: The top three, probably SpaceX Superintelligence. You've got the Gemini, the Grok, and the— I mean, the Gemini, the OpenAI, and Anthropic, four of them, at least.
L372Martin: So now guess what? But all those fours, if you look at the valuation, is over, I want to say, 20 trillion dollars, maybe, or 15, 20 trillion dollars for the valuation.
L373Martin: So my question is this: our national debt is 40 trillion dollars. These guys' valuation is already 20.
L374Martin: There could be a very big GDP expansion. There could be a massive GDP expansion through AI.
L375Martin: Elon is right about that. But people are fear-mongering.
L376Martin: People are playing politics. People are trying to put— Rokona putting in, "Oh, they need to have an FEA."
L378Martin: Have you seen how antiquated the government systems are? If you're ever going to DOD, which I have done many, many times, and you look at the antiquated systems that they have, I always say the best thing that we can protect is nuclear coding.
L379Martin: We're really good at protecting that. I don't know if we can protect everything else in the federal government.
L380Martin: Everything is slow. Everything is bureaucratic.
L381Martin: Everything is clunky. It's got to work.
L382Martin: So we have to give it to the bureaucrats. And then they're going to bring the politics.
L383Martin: And I hate to bring the politics into this discussion, but it's just going to be terrible. So my thing is that that regulatory capture is scary because I don't know how solid it'll go.
L384Martin: And you just don't want to have a social media style crap that they did to administrative or to consecutive administrations.
L408Nix: So look on there. You're going to be number six. And I'm going to have— I'm going to have a whole write-up on you, and you're going to be the feature.
L410Nix: The last thing I want you to say is I'm going to take this transcript, and I'm going to feed it to the agent that runs this site. I typically tell my agent, "Take this transcript and make Martin a f****** rock star." What I want you to do right now is I want you to tell me what are the most important things that you want to get across when you're featured?
L411Nix: What are the top three things for people to remember?
L412Martin: Top three things is wealth. I mean, I think the contributions. I think the contributions are the one.
L413Martin: Not for me particularly, but in general, I think the—
L415Martin: I think those contributions, the community unity, is a big deal to me. I would say those two go behind. And the last thing that I would say is you follow me, go see what I do, and join the conversations.
L416Martin: And if I can help you build something or if I can contribute to your project, I'm happy to do so.
L417Nix: I believe that, man. Man, this has been awesome. I've learned a ton, and I can't wait to see—
