Start9 CEO Matt Hill joins Jordan Bravo and Stephen DeLorme to discuss the evolution of sovereign computing and the upcoming StartOS 0.4.0 release. Learn about Start9's mission to democratize server administration, their revolutionary new networking capabilities, plans for an open-source router, and innovative community programs for scaling support and development. Plus: why dignity matters as much as privacy and security in computing.

Chapters

• 00:00 Opening Quote: Sovereign Computing Definition
• 00:33 Introduction and Show Sponsorship
• 01:51 Welcome and Sovereign Computing Origins
• 03:21 What is Sovereign Computing?
• 04:54 Privacy vs Confidentiality and the Dignity Factor
• 06:15 The Undignified Reality of Modern Computing
• 08:37 Making Server Administration Accessible
• 10:00 Democratizing Advanced Computing Skills
• 11:42 Familiar User Experience Design Philosophy
• 14:09 Learning from Mobile OS Evolution
• 15:45 Bringing Linux Admin Experience "Above the Hood"
• 18:05 StartOS Evolution: From 0.0.1 to 0.3.5.1
• 19:28 Dependency Management and Configuration Breakthroughs
• 20:17 Moving from Docker to Custom Container Runtime
• 21:33 StartOS 0.4.0: Complete Architecture Rewrite
• 24:44 Two Years of Development Hell and Breakthrough
• 26:30 Custom LXC Container Runtime Development
• 27:35 Advanced Networking Capabilities in 0.4.0
• 28:29 Granular Interface Control Example
• 31:49 Private Domains and DNS Management
• 35:06 The "Digital IKEA" Philosophy
• 36:34 VPN Tunneling and Network Abstractions
• 37:57 Start9 vs Tailscale Comparison
• 41:13 Router Hardware Prototypes and Development
• 46:47 Router as Standalone vs Integrated Product
• 52:00 StartOS and Router Integration Benefits
• 53:26 Router Release Timeline: Not Before Mid-2026
• 54:44 Community Tech Program: Scaling Support
• 01:00:18 Community Developer Program Announcement
• 01:05:02 Package Development and Crowdfunding Model
• 01:11:56 AI in Development: Limited but Useful
• 01:17:01 Self-Hosting AI Challenges and Hardware Limitations
• 01:24:37 TabConf 2025: Workshop and Package Development
• 01:26:55 Conclusion

Links

• Jordan Bravo
• Stephen DeLorme
• Boost in on Fountain.FM
• Start9 Website
• TABConf 2025 Conference
• Matt Hill on Twitter

Transcript

SOV022 - Matt Hill

Matt Hill: [00:00:00] Sovereign computing effectively means that you, your family or your business, whoever you happen to be, whatever entity you are, can have control over your computing devices and the data they collect and the communication interactions that you have with other devices that you are in control.

We view sovereign computing as control over your computing resources such that you can have privacy, security, and dignity with assurances without trust.

Jordan Bravo: Welcome to the Sovereign Computing Show, presented by ATL BitLab. I'm Jordan Bravo, and this is a podcast where we teach you how to take back control of your devices. Sovereign Computing means you own your technology, not the other way around.

Stephen DeLorme: This episode is sponsored by ATL BitLab. ATL BitLab is Atlanta's freedom tech hacker space. We have co working desks, conference [00:01:00] rooms, event space, maker tools, and tons of coffee. There is a very active community here in the lab. Every Wednesday night is Bitcoin night here in Atlanta. We also have meetups for cyber security, artificial intelligence, decentralized identity, product design, and more.

We offer day passes and nomad passes for people who need to use the lab only occasionally, as well as memberships for people who plan to use the lab more regularly, such as myself. One of the best things about having a BitLab membership isn't the amenities, it's the people. Surrounding yourself with a community helps you learn faster and helps you build better.

Your creativity becomes amplified when you work in this space, that's what I think at least. If you're interested in becoming a member or supporting this space, please visit us at atlbitlab. com. That's A T L B I T L A B dot com. Alright, on to our show.

Jordan Bravo: Welcome to the Sovereign Computing Show. I'm Jordan Bravo, and I'm here as usual with my co-host Steven Delos up. [00:02:00] And today we have a special guest. we have Matt Hill from Start nine. And before I kick it over to you, I wanted to let you know that the, the name of this show, sovereign Computing Show, I came up with the name because I, I'm pretty sure I first heard it from you guys at Start nine. And the reason I named this show, the Sovereign Computing Show, is because even though we talk about things like Bitcoin, Linux, open source software, that kind of stuff, privacy, self-hosting, these are all think things that we cover, topics that we get into, but it's not a show about any of those things.

Those are the tools. Those are the means. To get to the end. And the end is really encapsulated, well by the phrase, sovereign computing, right? We want to have, devices that serve us and not someone else or some other organization. I've already given my spiel in the, the first episode of Sovereign Computing Show.

I explained what [00:03:00] sovereign computing means and why I think it's important and why we should all care and be striving for that. But, I'd love to hear you talk about it because I think you do a great job of conveying that message. And so maybe we could start off by, you could introduce yourself and you could answer the question, what is sovereign computing and why do we care?

And then of course, feel free to, segue into what Start nine is and how that's related.

Matt Hill: Yeah. Great. Uh, and thanks, thanks for having me. Early on in Start Nine's history, this is late 2019, wanted to get t-shirts made and we wanted to put something on the shirt besides our name. And so the, that's where the debate started. And we landed on sovereign computing, and this was for the UN Confiscated conference in February of 2020.

And so we had our shirts made with our logo on the front, and then sovereign computing on the back. And, um, while it was the outcome of that conversation about what we should put on the shirt. [00:04:00] The months that followed and the years that followed, we have never been able to think of a better combination of words that succinctly describes what it is we are trying to do and trying to enable in the world.

So sovereign computing, effectively means that you, your family or your business, whoever you happen to be, whatever entity you are, can, have control over your computing devices and the data, they collect and the communication, interactions that you have with other devices that you are in control.

what's interesting about that is that privacy is actually a derivative of that. a lot of people talk about privacy, that's like a really big popular word in the broader computing world, even with companies like Apple. Side note, they don't actually mean privacy, they mean confidentiality, right?

You're not private from Apple, you're just trusting them to keep your information, confidential. [00:05:00] But, we actually saw sovereignty as a broader category, that included both privacy and security. And one that most people don't ever talk about dignity. So we view sovereign computing as control over your computing resources such that you can have privacy, security, and dignity, with assurances without trust.

So.

Jordan Bravo: I, I've heard you before when you talked about that dignity, it really got me fired up in a, in a good way. Of you, you gave this analogy of when you are using a permissioned device that is reaching out to a company's servers to access your own data, you're, every time you're trying to access your photo of your family member, you're effectively saying, Hey, Mr.

Apple or Google, may I please see the picture of my family member? And they, you know, auto, this is happening automatically on an API level. They're saying, yes, we're allowing you this particular time to [00:06:00] look at that data. But, that's not always the case, right? Like, they can deny you from your data that you think is yours, but in reality with the terms of service, like they have full access to that and they can cut you off from it at any time.

Matt Hill: Yeah, and I would put, I would classify that as censorship resistance, which is arguably a subcategory of security, keeping things secure, and available to you, the person who owns them. but yeah, it's, it's, you know, totally undignified to have a conversation where 10 different employees, belonging to 10 different companies are effectively sitting in the room, taking notes.

even if you have nothing to hide, even if they're not, you know, uh, using those notes to hurt you, know, immediately at least, uh, it's still just weird. It's still just undignified. It is not how humans, have interacted forever. It's a very new phenomena where, conversations, interactions, [00:07:00] activities that people expect to be private.

Are in fact not, that's a new thing. in the past that was considered, you know, like eavesdropping, surveillance, like, you know, tapping somebody's wire required a, you know, bench warrant. we live in the era where, even your private interactions are, absolutely being, being listened in on by many parties and you granted them permission, right?

They, they're not doing this secretly. it's right out in the open. most people just don't understand how computers work, how the internet works. Nobody reads the terms and conditions. but it's not a secret. It's just. Unspoken. Nobody says it, but more and more people are right. This is becoming a thing.

More and more people are waking up to the reality of cloud computing, software as a service and modern mass surveillance practices, and hopefully are beginning to reject them and look for alternatives. And that's where we come in. We [00:08:00] believe that we are building the building blocks and foundation of an alternative computing paradigm called sovereign computing, where it is possible for normal people to be able to, exist, on public networks, private networks, private computers, and in such a way that there are no third parties, who have access, who can control access, can charge for access.

Or otherwise. So yeah, we just think it's a, a proper and dignified foundation for, the future of computing.

Jordan Bravo: I really like the, ethos and the mission of start nine. And, I am a software engineer and I have a lot of cis admin and infrastructure skills, but not everybody does. Right? This is my family is not able to do this. My extended family, my mom and dad, et cetera. Like, they are not gonna be installing a Linux [00:09:00] server and maintaining it and, administering it and installing apps and keeping their security posture tight and spinning up a, reverse proxy so they can access it from outside their network.

And, and on and on and on. these, the ability to be sovereign with our data and still have most modern conveniences, I would say exists at the current moment, but the skillset required to do so is just beyond 99% of the population has. And so I think it's really cool to see that you guys are making something that, that the average person can use the the server operating system.

Matt Hill: Yeah, we absolutely are not inventing, um, a new concept. Here. We are inventing a new tool, that makes a previously powerful capability that was accessible only to people with advanced skill sets, uh, available to anyone, right? We're trying to democratize access to sovereign computing. We are not inventing [00:10:00] sovereign computing.

In fact, you could argue that computers as a default, uh, were sovereign. And it was only through decades of painstaking efforts by governments and big tech corporations did it become what it has become. And, we're just trying to get back to the, to the roots. and. To align with the rest of human history where, you know, property and communications were inherently, defensible and private.

Jordan Bravo: I see sort of two major reactions when people realize the extent to which modern tech living is surveilled and censored. One of them is I'm gonna go become a Luddite live in a cabin in the woods, in a shoe all technology. other, which I think is the preferred route, is to focus on sovereign computing. and so for people who might be seeing people who might not have heard of start nine before and who have never [00:11:00] used start os that's your operating system. they realize the problems with modern surveilled cloud computing and they wanna try getting more sovereign with their computing, and they've heard us here on the show talk about often how self-hosting is a superpower, right?

It gives you all of these modern conveniences and tools useful, productive tools without the downsides of some third party handling all that. when somebody downloads Start os or buys a device from you guys that's preloaded with it, what is the sort of. Steps they have to go through, you know, how technical or non-technical can they be?

And, and what is the basic, what does that look like for them? The experience.

Matt Hill: so as a, higher level, objective, our goal was to make administering a server as familiar as possible, right? We want, we don't want people to feel like. They have to acquire an entirely new [00:12:00] skillset to run a server because they already have the skillset to run a laptop, a tablet, and a cell phone.

So desktop and mobile devices, they already have those skillset. and we believe we, we hypothesized that it would be possible to create abstractions in the Linux CIS admin experience that would sufficiently overlap, right, with the client administration experience as to not terrify people, as to not make it look like some totally foreign alien technology.

luckily for us, big tech, uh, companies have spent billions of dollars over the past two decades. Doing user experience r and d. like we know at this point we humanity, we know what [00:13:00] an average person is expecting when they pick up a computer and try to do something because that user UX research and product evolution has taken place for so long.

and so we did our best to try to make this feel familiar, to make our operating system start os feel like Windows, Mac, os, iOS, Android, while at the same time diverging from those user interfaces and user experiences only to the extent that is necessary to. Do the different things that somebody will do on, server as opposed to what they would do on a client or mobile device.

Right. On a laptop or mobile device. Just like when Mobile OSS came out, iOS and Android, they did their best to try to match the abstractions of the previous computing paradigm, which was the, the desktop. But they had to diverge from them a bit [00:14:00] because mobile is not desktop. They are different, devices and they are used for different things and therefore the operating systems could not be one-to-one.

In fact, Microsoft hypothesized that they could be one-to-one, and in the original, I think it was Windows 10, actually tried to create, they, they tried to do a single os, right windows that you could install onto their Windows phone or onto a, pc and it flopped

Jordan Bravo: UI at the time. Yeah,

Matt Hill: it like.

Stephen DeLorme: windows eight. But yeah, that's the one phone

Matt Hill: Yeah, it like

Stephen DeLorme: computer.

Matt Hill: totally flopped because they were trying to unify laptops with mobile devices in the form of like the Surface computer and the Windows phone. And ultimately Windows phone just died. Right? And it wasn't just because Apple and Android were killing it, it was because they were horrible. I had a Windows phone when they first came out with one, 'cause I like them, thought that it might be amazing.[00:15:00]

This was, again, way back before I learned how awful Windows was in general and all the, I forget, let's not go there. So, like those who came before us, right? There's a new computer now that we are trying to make available to the masses, right? There's always been servers. And there's always been people who could administer those servers, but all that administration was done via the command line.

That is how you administer a server. You install some flavor of Linux or you know, Mac os. Windows can do server stuff too. But regardless of which operating system you select to administer your server, you have to pop the hood and go to the command line, right? Operating systems have two layers. They have what's above the hood, and they have what's under the hood.

And all server administration takes place under the hood, AKA, the command line. And so our goal with Start Os was to bring the Linux CIS admin experience above the hood, was to create a point and click graphical user [00:16:00] interface, through which a ordinary person with normal computer literacy, right, meaning they know how to administer a laptop, they know how to administer a cell phone, would understand.

But would be able to do all of the advanced techniques that a, an experienced linuxy admin could do from the command line. So our job was to gify the Linux command line for the sake of running self-hosted open source software. who understand what that means, what it means to be a linuxy admin and have it, and networking skills and, you know, background, would argue that what we're trying to do is inherently kind of undoable that.

Asymptotically, you can only get so close that you're never gonna be able to total Totally. Just like dumb it down, so to speak. I think some of that rhetoric comes from a place of, um, pride, right? Like people who have these it [00:17:00] linuxy admin networking skills and years and years of experience, like anyone who possesses a a, a a craft that is difficult to acquire are going to immediately recoil at the idea that a normal person could do what they do, right?

It's this whole like, oh, you can't automate me. You know, like, I'm amazing. I host things, I know how to spin up tour hidden services and blah, blah, blah. And we get that. But at the same time, we felt that that skillset could be not automated, but partially automated. And the parts that could not be automated could be, diluted down into, comprehensible buttons on a screen for a normal person to click.

And, start OS 1 0 0 0.1 was our first, stab at this, and it came out in February of 2020. And, um, it was a, you know, it was neat. It, it was about as good ASEL is today, right? It was like a, it was like this really kind of [00:18:00] basic naive approach of just taking Docker containers and like representing them as buttons on a screen.

And then you click it and it installs it and then they run and you assign it, you know, a Torah address. And it was, it was like very basic stuff. It only took us like four months to write the O Point one version of Start Os but it worked really well and people liked it. And we sold a bunch of devices and we gained a bit of a community.

You know, we didn't think it needed to be perfect to go to market. and it only had four services on it, uh, when we first, you know, came out with it. but then we recognized all the mistakes we made, right? Not just mistakes, but shortcomings, all the things that couldn't do, right. All the things that we as the linuxy admins wanted to do with ROS, but we're like, crap.

To do that, we have to, we have to go back to the command line. Okay, how can we bring that to the gui? How can we automate it? How can we make it easier? How can we make it less time consuming? And we, we came up with some new abstractions and we rewrote it from scratch and we came out with 0.2 of [00:19:00] start os I think it was late 2020 or early 2021.

But don't quote me exactly on this. The get history is there for anyone who wants to take a look. oh 0.2 was a big improvement. It had dependencies. Now, so we could do things like L and DCLN and they could depend on Bitcoin and we could manage those dependencies and make it such that, you know, if Bitcoin wasn't installed, the system didn't just shit the bed, it told the user, Hey, there's a dependency that's missing.

Click this button if you'd like us to satisfy it. Furthermore, we could do dependency configuration. So l and D could now say, Hey, I want Bitcoin to be, you know, I want the ZMQ interface turned on, which is an option in bitcoin.com and l and D could make this declaration. And if the user did not have the ZMQ interface enabled, then START OS could detect this and tell the user that l and D is unhappy because this thing is not enabled.

Would you like us to enable it for you? You click okay and it does it, and everything starts working again. So that was the big breakthrough of oh [00:20:00] 0.2. oh 0.2, ran into all sorts of. challenges related to networking, but also just kind of our abil reliability and performance issues. Because we were really leaning on Docker and Docker compose, and those are just kind of piles of shit, right?

Like they're extremely, opinionated and buggy. And you need to often like, go in and like, kind of, kind of like kickstart 'em again, like restart the docker demon. And we just weren't getting the performance and reliability that we wanted out of Docker. likewise the backend was split into both. Should I be doing this right now?

Do you guys want this like,

ATL BitLab: Yes,

Matt Hill: history? I think I'm like, I kind of went off the rails of your original question, but, somehow I got here so I'll keep going. the backend at the time was split into two code bases because. Two of the founders, Keegan McClelland and Aiden McClelland Brothers. Aiden was all about Rust and Kegan was all about Haskell.

And so they had sort of divided up the back end of start os from a, you know, production [00:21:00] standpoint. And Keegan wrote all his stuff in Haskell and Aiden wrote all his in Rust. And we made the decision to move everything to Rust. And so, 0.3, not only gets rid of the Haskell, code base, but it introduced, all new abstractions.

It introduced Pod man instead of Docker, gave us a little bit more control and reliability. but uh, yeah, I mean the change log to oh 0.3 is just ridiculous. If you look at the oh three oh release, um, it was basically a new start os right? Totally new architecture. We redesigned the front end.

Um, and, uh, and it's held well, right? Oh 3 0 3 oh came out. What's the date on that guy?

Stephen DeLorme: March 1st, 2022.

Matt Hill: I do. Yeah. So we are still in the oh three series. so it's been, you know, three and a half years. we've gone through, you know, we're at oh 3 5 1 right now. And I would argue that the changes that were made between oh three oh and oh 3 5 1 we're probably bigger than [00:22:00] all of the changes made prior to O three Oh.

Combined. But none of them were, um, breaking. None of them were non backwards compatible. Right. We were literally just building upon the O three oh paradigm, and when we got to about O three three, we realized that we got it wrong again. Right. We were getting really good, like Start OS was starting to get really good in the O three series, and we just realized that to do what we wanted to do.

Meaning, I mean, we can get into the, the potentiality of this technology here, but ultimately, you know, connecting iot devices and that are all talk to each other on private networks such that I can go to the other side of the world and control, you know, my entire home security system from my phone in such a way that nobody on earth even knows the security system exists.

But I can watch live video footage from the other side of the world. Like we want, like we want the whole shebang, we want all of computing to be able to be done in a private, self-contained, sovereign way. And we [00:23:00] realize that start os O three O and you know, the O three series, O three X, had the abstractions wrong that we weren't going, we were gonna reach a ceiling at some point, where we wouldn't be able to do everything.

Uh, our functionality would eventually be limited. And so, um, we set out to. Do it one last time, and I swear we're not doing it again if, if oh four oh is not right. So, one last chance to get the fundamental architecture right. Right. One last hard fork and um, so we set out to build oh four oh in mid 2023.

So it's been just over two years. In the meantime, we were releasing oh three series, right? Like we, well no, that's oh three four. Okay. I was gonna like, no, we definitely came out with oh 3 5 1, like, sometime in 2024. Pretty sure. Uh,

Stephen DeLorme: uh, for, for those just listening, I was, uh, we're we're looking at the change log page on GitHub right now [00:24:00] as, uh, Matt's dogging. But

Matt Hill: so,

ATL BitLab: of oh three releases here.

Matt Hill: yeah. Yeah. So while we were building oh four, oh, the totally new start os by the way, this is not like some code change. We rewrote it from scratch, everything. Not just the backend, not just the front end, but also the packaging. SDK for like packaging services for startups. Everything got rebooted with all the very, very, very hard lessons learned from the years.

Uh, that came before. while we were building O four oh, we were also iterating O three X, based on bugs and, and, you know, features that we deemed necessary. But we are now on the eve of oh four OH'S launch. It's been the longest, most unpredictable, most stressful product development cycle of my life.

Uh, I have served as the, you know, the, I'm the CEO of start nine. I also am the effectively the product manager, if you want to think of it that way. For start os I also write a large chunk of the front end code base, and I [00:25:00] designed, from a product and some code perspective, the SDK that developers use to package services for START os.

And I am just. I, I, I, I simultaneously feel almost embarrassed at how long this took us because we advertised and genuinely thought that it was gonna be ready about a year ago, and just didn't understand the magnitude of what we were taking on and how difficult it was going to be. because we were like really figuring things out as we went, things that people have never tried before.

but at the same time, I'm running oh four, oh Alpha nine, which we're looking at on the screen right now, on my production server for my personal life. In fact, I'm running Alpha 10, which will be published later today. Uh, we just made the last commit to Alpha 10 about 30 minutes ago. so Alpha 10 is coming.

Alpha 10 is a dramatic and incredible improvement on Alpha nine. Uh, and it's probably maybe the last alpha of start os oh four oh before we go to public beta. [00:26:00] And, I am happy, to say now after years of effort and all sorts of big promises that this is better than I ever thought it was going to be.

It's better than anyone thought it was going to be. We nailed it. The architecture is correct. The features are incredible. it's reliable, it's performant. Fuck Docker, fuck pod man. We rolled our own LXC container runtime. We just invented it. We, Aiden, Aiden did all of this. Aiden invented a new container runtime based on LXC.

You can think of it kind of like LXC compose, which is not a thing. We sort of invented LXC compose, right? and it's just, I mean, the, the performance metrics are off the chart. The UI has been rewritten from scratch using this new component library, uh, called TGA ui, which is an open source library. It's wonderful.

It's a component library for angular. And, um, Alex who goes by water plea online is actually the, developer of that, of TGA ui, the [00:27:00] component library. And he also wrote the new front end for Start Os uh, with a ton of contributions from me as well. But, it's, it's beautiful for the first time ever, right?

We've always just thrown up like stuff that Debs would expect, you know, like pretty, pretty basic kinda dark mode. HTML looking stuff. the New Start OS is beautiful and um, the networking capabilities especially are where we really broke through here. You can now do networking things on Start os that were previously impossible and are impossible in any other system other than hand rolled setups, which takes forever, right?

Networking is like the hardest thing. And on Start OS oh four oh, you can now install. I'll give you like a quick little example. I could install uh, Bitcoin, right? Since we're all Bitcoiners here, I can install Bitcoin, on Start os and I can granularly control its various interfaces so I can control.

Its RPC interface, its peer-to-peer interface, it's ZMQ interface. We might even throw a [00:28:00] user interface on it at some point so that you can kind of visualize your node and its connections and stuff. There's no UI on it right now. There doesn't need to be, but it has these interfaces. I could go into the RPC interface and say, okay.

I want my Bitcoin RPC to be reachable at, this to address, uh, screw that to address. I want my RPC interface to be reachable at this vanity Tor address instead. And I can generate a vanity to address and, you know, mine for the public key. And, uh, eventually, you know, put a vanity Tor address on my Bitcoin.

RPCI could go. Okay, let's do another Tor address. 'cause I want two for some reason. One random one and one vanity one. Let's do that. So I add another tour address. I could now say I want to add a public, domain. Right. RPC dot Matt hill.dev. I want that to be my Bitcoin rrp C. So I can add that public domain.

Now I can attach that public domain to any of my gateways. Gateways are something you can now manage in Start Os. So your router, obviously by default, is going to be your sole gateway for most people. And that is a private gateway, meaning it only creates a, A lan. [00:29:00] And to access that LAN from the outside would require port forwarding and some firewall rules and all the rest.

But you have this gateway. Let's say I want a different gateway, right? I wanna spin up a wire guard reverse tunnel. On A VPS. We invented a thing called Stark Tunnel that we're gonna be announcing soon, which is a totally open source. Wrapper for wire guard that you can install on any VPS with a single line command.

The result is a very powerful but also very, very simple piece of software that is set up to automatically reverse tunnel traffic to a set IP address, right? So I would go into my VPSI would install start tunnel. I would say start Tunnel connect server. I would put in the IP address of my home, my home IP address, and hit enter.

The result of that is a wire guard config file. I copy that wire guard config file. I go over to start os I click add New gateway. I click that it is a, you know, a private gateway and I upload the wire guard config file and I hit save. And now start OS views. My wire [00:30:00] guard VPS. As a virtual router, right?

It now has two different ways of accessing the internet and two different ways of being accessed from the internet. And it views them both as effectively routers, right? This is just a skinny router in the sky. So now I can go add a public domain to, to my Bitcoin RPC interface. So I do RPC dot matt hill.dev, and I click save.

Oh, and I attach it to the wire guard gateway because I don't wanna expose my home IP address to anyone accessing that Bitcoin RPC interface. So I, you know, I add the domain, I select the wire guard interface, I give it a name, and I hit save Start Os now tells me, okay, you need to go create a DNS record to point your, you know, either star Matt hill.dev, or rpc.

Do Matt Hill do dev to this IP address, the IP address of the wire guard gateway, the V VPs that I created, the new gateway on. So I go do that 1D NS entry Start [00:31:00] os takes care of the rest, right? Oh, I also have to forward a port, right? Because Start Os is gonna tell me, to access this address, you need to do two things.

You need to point your DNS to that public IP and you need to go to your wire guard, VPS reverse tunnel and forward a port and start tunnel has port forwarding capabilities. It's a single line. It's start tunnel, port forward, create, and then you just add the, the port and then the internal IP port of the, container you want to hit.

But again, start os tells you all of this in like three easy steps. So you add the domain three steps later, your Bitcoin R RPC is now up on RPC dot, Matt Hill Dev for example. I could then add a private domain. To, let's use a different service. Now I have next cloud. I mean, I could do it for Bitcoin too, but there's no reason to have a public and a private domain on the same, interface.

So I could go to Next Cloud, for example, and you know, I install Next Cloud and it's immediately available on a unique, local address and [00:32:00] port on my lan. It's available on my lan, IP and port. It's available to Torah address. But I'm sitting here thinking to myself, well, I want my friends and family to be able to use my next cloud with me, right?

I want us to be able, I want to have them to all have accounts, but I don't wanna expose it to the public internet. I don't wanna put next cloud on next cloud dot Matt hill.dev because it now has this attack surface where it could be denial of service attacked. Someone could try to brute force passwords.

I just don't want it on the public internet. I don't want anyone to even know that it exists. But my family's not gonna use Tor and my family also is kind of weirded out by IP addresses and ports. Okay? So sure I could go install. VPNs on all of their phones and laptops, right? And turn the VPN on my router on or the V or use Start Tunnel and say, okay everyone, if you want to use Next Cloud, you gotta go to 1 9 2 1 6 8, 3 3 and at this port.

And they're gonna be like, I'm not putting that in my browser. That's scary. What I could [00:33:00] do instead is I could create a private domain for Next Cloud and I would call it Matt Next Cloud, or Matt dot, fuck you next cloud. It doesn't matter. I can make up any domain I want because it doesn't have to exist, right?

It's just, I could use facebook.com, could be my next cloud instance if I wanted it to be okay. So I add facebook.com to my next cloud user interface. and start os now tells me, okay, if you want this to work, you gotta do a couple of things. So one, you need to go to your router and tell your router to no longer use, you know, Google for DNS, right? You need to tell your router to use Start os for DNS. So go put this single rule, this IP address into your router for DNS start os when I added the private domain, already created the DNS entry to route, you know, facebook.com to my next cloud internal IP port. So once I point my router to my server for DNS, now I can go put VPN clients on all of [00:34:00] my family's phones and laptops and tell 'em to go to turn on the VPN client and go to facebook.com and it just pops up as our next cloud instance, right?

But obviously we wouldn't name it facebook.com 'cause they might actually want to go there too. Hopefully not, but maybe. And so they can now visit my next cloud instance via VPN on a clearnet domain with a root ca right, signed by my servers root ca. So I'm also gonna have to, you know, put my root ca on their phones and laptops.

But what this allows me to do, I'm talking about really advanced shit right now, right? that not a lot of people are going to do, but the amount of work that it takes to do the things that I'm describing on Start Os is a fraction of a fraction of what it would take to do on Ubuntu, right? And a different sort of setup.

Like Start os has these primitives abstracted such that you can push a few buttons in a ui, maybe go to your router and run a single command and you're done. Right? It's like if we, we think of [00:35:00] ourselves now, like the new Start OS has so much more power than the old ones, but you have to follow instructions.

And so the way that I've been talking about this to people and the way that I'm going to market it once this is out in the wild, is that if you want to know what we do and what we are and how to use it, we are the digital equivalent of IKEA furniture. That is sort of how you should think about start os is it's like anything that you want to build, anything you want to do, we have an instruction set to do it and you can do it.

I promise you the instruction set is not too complicated, but it's the middle ground between do it yourself, figure it out and do it yourself and push a button. We do everything for you, but you have to trust us. And by the way, everything goes through our servers and we can spy on you. It's the middle ground of that, which is you can do it, we can help.

That's the Home Depot slogan. But I'm thinking about taking it, it's like exactly, exactly what we want. You know, you can do it. We can help. And we being start os and our, technical support team is [00:36:00] there to hold your hand every step of the way. Okay. Super long-winded, but I am, could not be more excited about oh four oh and, um, I, I just think it's gonna be a hit.

Jordan Bravo: Exciting. Yeah, I'm really excited about that. I've been looking forward to it. And, um, following the, the progress, the, the issue as you kind of alluded to with Tor, well, Tor has a great upside, which is that there's no configuration required for a whole punching through your gnat. So, you know, for people who are familiar with that terminology, it means you just use it and it's good to go.

But the problem of course, is there's a couple problems. The trade offs are. One is your URL is this long, complicated looking string that's hard to remember. another problem is that tor is often slow and unreliable. So the alternative that I've done for myself, of course, and like many other people who are more technical, I've set up a VPS with a wire guard and you know, I've [00:37:00] also used tail scale to, to make that easier.

And it's nice and ergonomic, but like you said, if I had to set that up for my, like there's just no way that my family or anybody less technical is gonna do it. I mean, for, for me, even it was a many day like head banging against the wall challenge. and that's, it's just really cool to hear that it's gonna be coming to start os for anybody to be able to do it.

Even just following some basic instructions because the user experience, in my opinion, the, the, the benefit that you see when you have these wire guard tunnels, it's, it's amazing. It's, it's like having your own. the same experience with speed and reliability that you get when you have these third party servers with giant engineering teams and ops teams running 24 7 to keep, give you this great experience, but now you get to have it with your own self-hosted data.

So really cool and I'm, I'm excited to hear that's coming out.

Matt Hill: Yeah. Thank you. you mentioned tail scale, [00:38:00] right? We, we get asked a lot about tail scale. and, you know, we looked into it and we looked what? Kind of how it works and what the requirements are. And we we're not fans. Uh, I mean we're, we think it's a cool project. We just, it sort of is, is a, a across the line from what we like in terms of sovereignty and control, user control, right?

Like tail scale requires a tail scale account. You have to be running their software on the device you're trying to reach. and none of that is necessary for what it's doing, which is why we built Start Tunnel. So Start Tunnel, like I said, is just a, it's a wrapper for wire guard. It's totally open source when release it under MIT, we hope people build on it.

and it is just a, it's, you can think of it as like an alternative to tail scale, but that you self-host on a VPS, and it's just simpler. There's, it's just very, very simple, right? You provision a VPS with any. Debian based Linux distro. So you can throw up Debian, run a single line command to install start tunnel, run a single line to connect it to your server and [00:39:00] you're, and, and then copy paste the resulting config file to your server.

And you're done. You now have this like permanent reverse tunnel, going to your server in such a way that start os views it as another lan, as another, it gets a lan ip, it gets a wan ip, it gets, ports it, and it just start os almost can't tell the difference between this thing and your router. It just thinks of it as another router.

and so it's, it's your own personal hand rolled tail scale for the price of, of VPS, whatever that costs you. And in most cases, people will be able to get the cheapest one because the resources required by Start Tunnel are. Effectively negligible, right from compute, memory and disc. You don't need anything.

The only thing that really matters is transfer, right? How much bandwidth the VPS has. But most people, don't have high upload speeds from their homes anyway, and it's upload that matters when you're running a server, not download, like when you're running a client. So if your home upload speeds are like 30 megabits per second, [00:40:00] the biggest VPS transfer speeds you would want is 30 megabits per second.

That's the upper bound of speeds you're gonna be able to get when accessing your server via the reverse tunnel anyway. So for most people. The cheapest VPS they can find is going to be perfectly sufficient for this. people who have like fiber gigabit internet obviously get gigabit upload speeds. Not that they need that, right?

Unless you're running a website from your home that gets millions and millions of clicks and or you're trying to do, you know, stream video to the world and compete with Netflix illegally or something. unless you're doing something like that, again, bare bones VPSs are gonna be just fine.

Jordan Bravo: It's a really powerful concept and for people who have never rented a VPS before there, you can get the lower end ones for like $4, $5 a month. probably even less if you, if you're doing it for like a package of 12 months or something like that. So really cool. I think it's accessible to everybody.

Matt Hill: There's some that take.[00:41:00]

Jordan Bravo: Yes. In fact, we have an entire episode on private VPS hosting and how you can do that with Bitcoin and anonymity and not, not KYC and yourself and all that stuff.

Yeah. That's great.

you mentioned routers and the last time that I was at the Bitcoin conference in Nashville in 2024, last year, I saw the start nine booth and you guys had these router prototypes. can you tell us more about that? What's the update on that and what, what do those do?

Matt Hill: Yeah. boy. Oh boy. Yeah, we, we are so excited about the router. we were a little premature at Nashville because we didn't realize the workload of START OS oh four Oh, right. Like, we are a tiny team. We, we don't have like a team working on the router and a team working on Start os it's just me and Aiden, you know, and Alex.

It's like, it's just a few guys trying to build all this. so, uh, the router has made progress. We have, command line prototypes in [00:42:00] place for a lot of the features that the router needs. we are, for those who don't know. Start nine has previously announced that we are building a fully open source, MIT licensed, router operating system, whose base it'll be based on open WRT, which is a, a Linux router os um, that is almost infinitely powerful and configurable, but notoriously impossible to use because it's so many options and it's command line driven and they've built some gooey for it, but it's unintelligible to any human being.

And so, we set out to use this as a, as a base because it's again, really API driven open WRT and we weren't planning to use the GUI anyway, so we forked open WRT, we're calling it start WRT, cause we are not that creative. And we, built a custom GUI for it. So a user interface that in just in line with Start os tries to take really advanced.[00:43:00]

Linuxy admin stuff and represent it in ways that a normal person can understand. Router gooeys are notoriously terrible, right? They're from 1995. I know there's some newer ones that that try. Uh, the best router gooeys out there are actually the ones that come stock with like your ISP router. You know, if Comcast comes and drops a router at your house and they're like, Hey, you can administer it through this gui.

Well, if you ever notice, the GUI is@likecomcast.com, log into your account. And so it's like you're administering your home device through a Comcast server and you're like changing your wifi password. I just, the insanity of how horribly insecure and in private that is, is just, it's just crazy. So the best router, gooeys, the ones that most people are capable of actually using.

Are, tend to be tied to cloud-based ISP owned, [00:44:00] software, that it has remote access to your LAN and can change your wifi passwords on their servers. So, we wanted to create a GUI that was not only as easy to use as those gooeys, but afforded many more, uh, powers, you know, many more features, advanced stuff.

and we realized that open, WRT, open w is a, a plugin based operating system. So the, the, the basis of open WRT is fairly skinny and then you can write these plugins and there are many, many plugins out there already that give open WRT kind of these specific features and, and powers. and we realized that what we wanted from our router ui, because that's where we started, we actually just designed the UI and the experience that we wanted to bring to somebody and then worked backwards and said, okay.

What existing operating system can satisfy this the best and with, you know, not only from a performance, security, openness standpoint, but from a level of effort on [00:45:00] our side. Like, we don't want to build something that already exists. So that's where Open WRT came from. It checked all the boxes and we realized that we didn't have to build that much.

When I say that, I mean it's not like building star os right? It's much, much, much smaller than what we had to do for start. Os Star os was based on Debbie and it's almost not worth saying that anymore because it's, it's in another galaxy. But, but like, you know, I hope that that doesn't happen with open WRT.

I don't think we need to do a whole lot more. 'cause open WRT almost does everything already, but it didn't do what we wanted. It was missing. A couple features and so we're writing custom plugins for open WRT. We'll license them, uh, MIT and then ultimately start WRT then is just open WRT with a few custom plugins and a custom GUI and start WRT.

And uh, we will sell routers physical hardware with start WRT pre-installed just like we [00:46:00] do our servers today. And we will warranty them. We will provide excellent on demand customer support for them, but you don't have to do business with us. Start WRT will be fully open source and any router hardware that you already possess that was capable of running open WRT, will be able to run start WRT.

It'll be compatible with any hardware that open WRT is compatible with. so we expect many DIY people will just do it and we will encourage that and, and yeah. I could talk a little bit about some of the unique features that the router offers, if you want to get into it.

Jordan Bravo: Yeah. Quick

Matt Hill: or we can just wait for the announcement.

Jordan Bravo: If you have op, if you have the Start nine router and you don't have Start Os is, are they completely standalone products that you can use? One without the other.

Matt Hill: Absolutely. Yeah. Yeah. We, we expect that the router, will be, will by far exceed the, the market for servers, right? So we, [00:47:00] right now, when we sell star OS servers with Star os pre-installed, we are selling to a very tiny corner of the world, right? Like we're, we're selling to. A market defined as, anarchists and libertarians who have had enough and are aware of Bitcoin, but also don't want their data and messaging in the hands of Google and Apple and care enough to actually do something about it, but not too much because they're also lazy, like ev you know, in their busy with their lives.

And so it's like a really, really niche market. and even with that tiny little niche market, like we're proving out the business model, um, we're doing well from a business standpoint and it's like a, a tiny test market and we hope that that market grows, right? We very much want the market for self-hosting and personal servers to grow by orders of magnitude over the next decade.

And I think that it will, we definitely want the number of people who wanna run Bitcoin nodes to grow over the next decade. And I think it will. And we continue to, we will continue to try to be the go-to best way to do Bitcoin nodes and all [00:48:00] Bitcoin self-hosting related activities. But we're not just banking on that.

And it's not just a business decision either. We are genuinely trying to invent sovereign computing, right? Not invent. We are trying to bring sovereign computing to ordinary people, and the router is necessary for that. You can't be a, you cannot have a sovereign computing paradigm with a Comcast router.

You just, you're, you're done. It's, it's the elephant in the room. You, none of your traffic is safe. None of your passwords are not safe. Your LAN is not safe. They can see every device on your lan, even if they can't control it, which they might be able to. It is just, you're using closed source software on closed source hardware on a server you don't control.

And it's just like, come on. It's, we, we, there needs to be a usable, powerful, open source router, with the proper abstractions for self-hosting. So one way that we're gonna market this router. Is it's, we're gonna market it as like the self hosters router, right? This is the router you get if you want [00:49:00] to, run servers from your home.

Right? If you don't, you can still use this router. It's gonna have incredible VPN capabilities, incredible security, uh, abstractions for control over the land. Again, like we, we came up with, I'm so proud of the designs for the routers. It's not coded up yet, right? Not in total, at least, but, it's just everything I've ever wanted from a router and everything that everyone we talked to ever wanted from a router.

And, um, so, and it's fully open. Okay? And I don't just mean the software. We are trying to, and we're working on something right now that I cannot announce yet because it's still uncertain. But we think we might deliver fully open hardware, like completely, fully open hardware. Like not a single closed source, firmware blob, not a single closed source package.

not even.

Jordan Bravo: risk? Five chips.

Matt Hill: It certainly would. Yes, we are, we're working on something really cool right now. and it's still a little indeterminate, but we, we absolutely might bring the first [00:50:00] ever, risk five fully open from, from chip to, you know, plugin open W RT, plugin router into the world. But it also just happens to be the most powerful, router because even if our GUI doesn't enable you to do something, you can always, open the stock open WRT gui and you can always go to the command line.

So there's no limitation to what you can do with this router. We've just taken all the things that we think most people will want to do. We'll call it the first two standard deviations of users that we imagine all those features are in the gui. If you're a third deviation user, if you're doing something really wild with your network, you're probably gonna have to pop the LUCI GUI or go on the command line.

But that's fine because you know what you're doing if you're in the third deviation of users. So. We think that our gui, the start, WRT graphical interface captures what 99% of people are gonna ever want to do with a router in a highly intuitive click button, highly [00:51:00] performant way.

Stephen DeLorme: I think, from a product perspective, you could imagine that start OS four is launched and that, you people run into these scenarios where the os tells them as you described, you know, to do this you need to go to your router and you need to turn on port forwarding. And it, it could almost fail at that point in the user journey if the UI to their home router sucks or is difficult to use and you don't really have a lot of 'cause because of like privacy and stuff, you don't have visibility into what router operating system. know, they have what kind of router? there's not really like an easy way from like a, I guess a customer support perspective to tell them exactly where to forward that port on that janky router os. So it, it seems like a wise product decision to and, you know, build something better for the router as well.[00:52:00]

Matt Hill: Very natural extension. Um, and one of the other reasons we chose open WRT is because that it is, kind of designed to be utilized via API with authorized access. So guess what you're gonna be able to do, right? Yes. The router is a standalone product that you can use as a router with not care about start nine or servers or any of that.

The server is something that you can use and not give a crap about the router, but if you have both right, the, some of the, uh, the hole is greater than some of the parts. Now you can take your start os public key, your SSH public key, which is displayed in the about menu of Start Os. And go to your router and authorize your server's public key to administer the router.

And now when you add a public domain to start os and the instructions pop up and say you need to go, you know, forward this port in your router. It can also [00:53:00] say, would you like me to do it for you? And you click yes, and then it just forwards that port in your router.

Stephen DeLorme: Amazing.

Matt Hill: when, when you combine them, it it, it gets really cool.

Yeah.

Jordan Bravo: Well, I'm looking forward to this. This sounds like something I absolutely want to get my hands on. And so let me ask you the question with the caveat that. Software takes time and hardware as well. when is this gonna be available or estimated to be available.

Matt Hill: We are gonna, we're shooting for 2026, so, um, definitely, I, I can, uh, let me tell you when it's not gonna be available. It will not be available in the first half of 2026.

Jordan Bravo: Okay.

Matt Hill: That's all I can say. That's all, that's all I know. I know it will not be available by July 1st, 2026. I, we are going to try to make it available by the holidays of 2026, but if this shit doesn't come out till the holidays of 2027, don't, don't shoot me.

We have no idea. We, we thought we knew, we thought we knew it was start os and we were [00:54:00] so wrong that I'm just not repeating the mistake. I'm not gonna do it. So, I'll tell people when it's not gonna be available. I would not tell 'em when it is going to be available. All I can say is that we are busting our ass and we're super excited and we work very long hours.

But we are a tiny team. We were a bootstrap startup even six years in and we like it that way. You know, scale comes with productivity gains, but it also comes with a bunch of costs, and we're not ready. We're not, we don't wanna do it. We have parts of the company that we can scale like our technical support later this week, I guess it's getting to the end of the week here.

probably tomorrow or Friday I will be announcing a new program that we are doing. Uh, are you guys familiar with our community tech program?

ATL BitLab: Yes.

Matt Hill: Okay.

Jordan Bravo: and, and, and some listeners might not be. What is

Matt Hill: Uh, so there's a, a neat little writeup about it@blog.startnine.com. Um, in fact, if you wanna pull it up on the screen here, we could, we could do it. I will talk through [00:55:00] it as well.

But I figured, you know, uh.

Stephen DeLorme: This one

Matt Hill: So go to community tech program now hiring, with hiring in parentheses. so the community tech program was our novel approach to, uh, kind of employment and scaling a support, department. And so rather than hiring dedicated community support and technical support people and having to train them and then having the, uh, the risk of them quitting or leaving or not being available, going on vacation, et cetera, and constantly dealing with the thing that employers deal with all the time, which is how do you get the people you need when you need them at the price you need them, and all the rest.

We, because of my previous life where I was developing applications that helped with labor market efficiencies, I came up with this idea of, well, why don't we just do group training? Why don't we just. Call for applicants, they'll need to be some kind of minimal competency or experience, you know, with IT or Linux stuff.

they'll need to [00:56:00] be good communicators, et cetera. So we'll do a minor vetting and then invite all the people that got through that vetting process to a giant group training program, right? So it'll be a weekend long intensive. Here's what start nine is, here's what start OS is, et cetera, et cetera. And at the end of that program, everyone who gets through it is effectively now certified.

Okay? We, we'll just use that word loosely since it's not like an official you know, thing. It's just basically me saying somebody's good enough. But you're now certified. You're a start nine certified technician. With that certification, you gain only rights, not obligations. You never have to work, right?

You get that certification. It's like Uber, right? You get qualified to drive for Uber, but you never have to get in the car and drive. You now have your certification. Having that cert allows you to, uh, but not requires you to pick up shifts and monitor start nine's, community [00:57:00] channels. So like our private matrix server, our community, discourse, which is a self-hosted community management software, email, right?

Like our support channels, you are now qualified, right? We have certified you to monitor those channels. And if you want to grab a shift, we publish an open calendar. We just publish a calendar and people who are certified kind of put their name into the calendar and it fills up. and then we get the coverage that we need.

These people are all over the world, so we have good time zone coverage. And, um, if anyone ever can't work, like they claimed a shift, but then they realize that they can't work, they use, automated shift trading software, which I wrote back in 2016, , to basically offer their shift to everyone else who is certified and the first person to claim that shift gets it, and then it automatically shuts down for everyone else.

The manager is notified that the trade took place and we can get our payroll right at the end of the month. So basically, people work when they want, if they want, however much they want, so long as they maintain their [00:58:00] certification, which again, is a very informal process at this point. And at the end of the month we tally up the hours and I send everybody a Bitcoin and, and that's it.

And everyone makes the same amount. In the future, I could assume that there might be some stratification where we'll have like senior tax and mid-level tax and, you know, junior tax and they make different amounts and this, it, it could get more complicated. But because we do support in the way that we do, meaning we're not answering phone calls one-on-one for everybody, we do what I call group therapy, right?

Which is this idea of a giant telegram channel. Someone comes in and asks a question and they get a answer either from a volunteer in the community to somebody who is lurking around and could help them, or from the person who's on staff at that time, the certified person. And even if the community is helping the community, the certified person is still there monitoring the chats to make sure that nobody, you know, is wrong or lying or scamming, et cetera.

scamming doesn't happen on our support server because it's private and gated and paid. so we don't have that issue and that's why we did that. But, [00:59:00] um, it, it works really well. It scales really well. And eventually we can use all of these support interactions these years and years of like one-on-one expert support interactions that are in a, a chat thread categorized by topic by the way.

'cause we have different rooms for different topics to eventually train, you know, a really competent tier one support AI and the community techs would then be there to observe the interactions between the AI and the customer and for the purposes of escalation. So we think we sort of cracked the code on how to scale a, support for technology, and.

All the support that we do today is knowingly being sort of cataloged for future training for an ai. And so start nine support will always be incredible. It'll always be backed by humans who are on staff and fully qualified. But, again, so this is our community tech program. We implemented this a couple years ago and, um, it's been just fantastic.

It, it has been a, a home run, hit Start nine offers incredible [01:00:00] support to all of our customers, all of our users. and uh, that's that. This week we're announcing a new program, which is similar in concept, uh, sort of, you know. Expands upon the success of this program and, uh, we're calling it our community Developer program.

And ultimately the goal of this program is to get more and more software packaged for START Os. This is to expand the number of service offerings on Start Os. And uh, to get a piece of software onto START os requires it to be packaged for Start Os right Start OS is not debon, it has a unique package format.

You can't just take an existing Deion package and install it on Start Os for the same reason that you can't install a Mac application on Windows. It's a different operating system. And so software has to be packaged for Start Os and we in oh four oh have now made that process very, very easy. Okay, [01:01:00] we, we've dramatically simplified.

Uh, what it takes to package something for Start os while at the same time dramatically expanding the number of features package developers can utilize in Start Os. The Start OS package. APIs have expanded by probably five x what they were in oh 3 5 1. Like you can do almost anything. We actually have yet.

This is yet to, in oh four oh packaging run into a circumstance where we said, man, wouldn't it be great if the Lightning Network demon, if the l and d package could do blah and we weren't able to do it. We have not run into that circumstance yet. It is, you can do anything. If you can think of it, you can do it.

That is the, that is the, the grand claim of Start OS oh four oh SDK in service packaging is, if you can think of it, you can do it. And the essence of [01:02:00] packaging for Start OS is that you, the package developer, you are the IT professional. Linux sis, admin, network engineer dev, okay? You're the guy who know who does this.

You're the guy who does this on Ubuntu, right? You're the guy who did automated your home using open source tools and spent did it over the course of months and months, so you know what you're doing. What if you could encode yourself into a package, a piece of software such that everyone else could on earth could benefit from the capital that you built up?

Right? So, so many home labber spent hours and hours and hours setting up their personal home labs, but it's like. Why, why can't others benefit from the work that you did? Like, you, you wrote some scripts, right? You wrote some scripts to monitor the health of your next cloud server. And if it, if one of those pings ever fails, then you, you issue a notification and it sends to you via SMTP, but it also automatically shuts down next cloud until you go change a [01:03:00] config variable in the underlying PHP config file, right?

It's like, wow, you're incredible. Why didn't you do that for start os like go write that exact same code for the next cloud package for START os Now when somebody installs next cloud on start os they get you too. You encode yourself into these upstream packages, right? So star packaging for Start OS is an incredibly one simple experience because going bare bones like getting to Hello world is actually really easy.

Now you just take, let's say a package has already been dockerized, there's a docker container for your package, getting it to run on start os is like. Cloning a repo, a template repository and swapping out the URL that it grabs the Docker container from Docker Hub done and maybe changing the name and the id, right?

And boom, it's running on Start Os. But now you get to sit down and go like this and be like, what else can I do with this? What other advanced crazy stuff do I know how to [01:04:00] do that? I think mom and pops all over the world could benefit from, I'm gonna encode myself into this package now, and I'm telling you, it's one thing for me to say it right now, but the people who are learning how to do this and who are weekly attending, we have biweekly calls for office hours where I get on the call and I literally pair program with people and show 'em how to use the APIs.

We do screen sharing, we talk about how to achieve things, we review other packages and how they did things. it is like this endless and very fun and rewarding experience of like, creativity. It's not just let's put next cloud on start os it's like, let's put next cloud with all this cool stuff around it on Start os.

and again, anything you can think of you can do. So just make something up right now and I'll tell you if it's possible if you come up with something while, while we're on this call later. Okay. I could give you like some examples if you want, but I might be for a different call. Anyway, we're announcing this program, okay, the community developer program, and the way that the program [01:05:00] is going to work is similar to the community tech program.

I am going to create some webinars like 1 0 1, like Introduction to Packaging for Start os kind of, you know, screen share webinars. And I will introduce what Start OS is. I'll introduce what a Start OS package is. We'll go through the hello world template, we'll review the SDK and all the different APIs that you can use.

And if you love it, show up for the 1 0 2 call. It's in, it's in two weeks and these will be going forever. This will be a recurring just every month we'll have a one oh one call every month. We'll have a 1 0 2 call. And as you start going through and learning, then you might be like, all right, well I'm gonna go package up cal.com.

I like using that. So let's go package cal.com and. You know, you give, you throw, you throw your hat into the ring, go try to package it. You're struggling. You don't, you have a question. Show up to office hours. We're gonna do office hours twice a week for two hours a day, right? Four hours a week, you're gonna get me or another senior level package developer who can just sit down and work with you directly on whatever you're working on [01:06:00] while other people watch.

Right? So it's like everyone is benefiting from everyone's work. And, um, so you show up, you get help and eventually you've done a package. Well, eventually, eventually, me, or again, one of a couple other people at this point will be able to be like, Hey, you know what? I think he can do it. Like this is somebody that has been on the calls, he's packaged a couple of things, he's sort of proven that he can do this.

Well, again, very informal. Just like the community tech program, you are now certified by us. Okay? This is just a stupid credentialist thing. It's like start nine has said you are a start nine approved or start nine certified package developer. Why would you want that? Here's why. Once you have your very unofficial start os start nine certification package, developer badge, we will now broker you to anyone on earth who wants things packaged and we get hit up constantly, constantly by companies that are like, I write this software and [01:07:00] we, our customers are asking us to put it on Start Os and we don't know how to do that and we don't wanna know how to do that and we don't wanna learn.

Eventually they will. I think this will be a thing at companies and they'll just package for Start OS in the same way that they do Mac Os and Windows. But we're not that, we don't have that kinda leverage yet. So they come to us and they're like. Put our software on your marketplace and I'm like, I don't have, I, I don't have time.

I'm sorry, but we're busy. But now that I got this pool of like 10 certified package developers, I can just throw a bounty out and be like, Hey everyone, there's somebody asking for this thing to be packaged. They're willing to pay X amount of money for it. Start nine would, as a company, would love to see this too.

And we would benefit because now all of our existing users get more superpowers. So we'll add to the bounty as well. We'll throw some money in too. Then we go on social media and we say, Hey everyone on earth, so and so, the company wants their stuff packaged for Start Os Start nine wants it packaged for Start Os if you too want it packaged for Start os throw some money into this BTC pay server crowdfund that we spun up and you know, [01:08:00] Bob over there is gonna package it up.

He's, he's, you know, committed. He said that he will be the one to package it up and we know he can do it because we trained and certified him. So it's like everyone wins. Bob makes money. Start nine, gets more packages on its marketplace. The upstream development team gets their distribution for their open source software.

The whole thing is a marketing event online with the crowdfund and just everyone wins. And customers now have more services available to them. I shouldn't even say customers, just anyone using Start OS now has more services available to them, and perpetuity deals could get cut. So, for example, start nine has committed publicly to maintaining about 25 to 30 s.

Critical services, right? Software that we're like Start OS just isn't start OS without Next cloud. I'm sorry it's not. Next cloud is the killer app for self-hosting and it must be there. Well guess what? The next cloud team don't give a shit about us. They don't know who we are. We tried to [01:09:00] reach out to 'em, they're never gonna respond to us.

They're too good for us. They don't care about START os fine. We're still gonna put next cloud on Start os, but we have to do it. We have to take responsibility for it. We have to make sure it's up to date. We have to make sure it's bug free. And so what I wanna be able to do, rather than doing it myself, 'cause I'm busy, is I wanna be able to turn to one of my certified package devs on the other side of the world who doesn't have a job, by the way, is totally voluntary and just be like, Hey, do you wanna maintain next cloud?

I'll give you X amount of money per month payable in Bitcoin at the first of every month. And what you need to do is basically just keep it up to date. Like make sure that if next Cloud comes out with a new version, the new version is reflected on Start Os. Make sure that if. there's a way to improve the next cloud package using Start OS APIs.

Please improve it, make it as good as it can be, make it bug free. And, if Start os comes out with a new version or a new feature, make sure you keep it updated to that. Ultimately what this means is that every package is a different level of complexity. Some packages are gonna be, you know, [01:10:00] basically nothing.

They come out with a new version every year and it's just a bug patch from the previous version. And so, you know, it's not gonna be worth a lot. But other things that are changing constantly and requiring lots of changes are much harder to maintain and they'll be worth more. And so each package maintenance kind of bounty deal will be cut on a case by case basis and somebody who thinks it's worthwhile to maintain it and earn passive, not passive, but earn, you know, good value for hours put in income.

Right? Because again. Even if the package is heavy. As somebody who's been maintaining these packages for years, I can tell you that even the heaviest ones sometimes have months where nothing is required. Most months there's like a minor version bump and you literally have to change a single string in the package and hit deploy and you're done.

And some months it's like, oh boy, we swapped out the APIs on you. We broke your shit. And you know, you need to put in 10 hours of work to fix it. So it really is kind of variable and we're just gonna try to do our best to, to come up with a, a fair value for [01:11:00] maintaining those packages. And if someone claims it, great.

If not, we'll up the number until someone does claim it. And that's how we intend to scale, not only, support for the services that we are dedicated to maintaining on the Start nine registry, but also just expand in general, the number of services available on Start Os. we just want to be a, a, a glue, a liaison that connects.

Demand for new services on Start Os with supply, the people who are capable of packaging services for Start Os we just wanna connect those people as best we can, such that the ecosystem grows and we want to do so in a cost effective way by not literally hiring developers, uh, because there's just no way that would be efficient.

Right. The Bounty program is more efficient. It's a, it's, it's modular, it's piecemeal. Yeah.

Stephen DeLorme: It's really cool. I never realized that, uh, start at nine, like [01:12:00] was involved in all of this kind of, don't know, like supply and demand kind of matching stuff. Um,

Matt Hill: We're getting there. Yeah.

Stephen DeLorme: if, if, I can segue a little bit, I kind similar related topic, like just the idea of a startup with, you know, so few employees and, and dedicated engineers. Well, very dedicated engineers, but, you know, so few full-time engineers is what I mean. I, a lot of, I know a lot of people who work for small startups and I feel a, like, you know, in the one to three person startup range. is like using ai, to, to code with. And I was, somehow though, I can't imagine, um, I can't imagine someone such as yourself like spinning up a bunch of instances of clawed code and like, u using or, or GPT five or something like that.

So I was kind of curious how you [01:13:00] approach, AI or coding with it, if at all. For start, start nine.

Matt Hill: I think I'm representative of the other developers at start nine. again, there's not many of us, but um, I don't think any of us start with ai. We don't, we don't, we don't go to it like vibe, coding style, prompt and be like, build this app or build this module. Right? We, we, we, we imagine, we architect and when, then we start building.

And I, where I think AI shines in the programming world at the moment at least, is not on the architecture, uh, side of things. It, it is on the scripting side of things. It's really, really good at like, You know, syntax and kind of, you know, pure functions. Like write me a a bash script that does this thing.

Or, you know, come up with a regular expression that satisfies, you know, these requirements or write a type [01:14:00] script function that takes in, you know, uh, an, parameter of this type and then puts out, you know, this result. And it can do these like kind of these, these limited scope, pure function, or, or kind of u utility function stuff, you know, like UU utils.

and so you can we architect and design and then sure you can like plug in these discrete things on a, uh, as needed basis if you try to get AI to like code for you to do the actual kind of hard work that developers do, which is really around. creating abstractions, uh, you know, architecting things in ways that are, um, coherent and non-redundant and dry.

Just AI isn't there yet in, in our opinion, and maybe I'm just not using it right, but the, the sort of testing that we've done around [01:15:00] it, it always kind of puts something out that. appears okay, but might not hold water in when it's actually spun up. Like it, it only does in a way what you tell it to still, it's still kind of a mongrel, you know, like it's, I'm actually blown away at what its capabilities are, just like everyone else on earth.

But there are certain things where it's just like, what? No, like you, you have to be such a good prompter, basically. You have to be the, the best prompter in history to get it to produce functional code. Not functional as in functional programming, but like code that works, you know? you just have to be incredible at prompting.

And when you get to that level of prompting, why not just write the code? Right, like, like I don't wanna spend an hour writing prompts to get code that I could write in 30 minutes. It's just like, it's actually not a time saver at that point. But however, like I said, there are absolutely times when I'm just like, I don't wanna sit there and, and, and [01:16:00] like come up with the, the very particular, like, I know AI could just pump it out in two seconds and so I, I'll go to it, right?

It's, so, it's on a case by case as needed basis, not as a, not as a sort of fundamental strategy.

Stephen DeLorme: Makes sense and, and like if, when you do have to turn to it, like, I mean, I, and I agree with you about the like architectural approach. That's kind of what I've found too, is that like if I want to build something that I know is gonna like last, that I'm gonna want to still maintain in six months, then I need to have pretty strong opinions about the architecture of it.

And I need to, you know, really, really, um. force the AI to, stick to my architecture and build it the way I want to. but yeah. Yeah. So I mean, I, I, I agree with like your overall assessment, like the architecture and the hard work is, is, that's not replaceable yet, but like in those few moments where you do have to turn to ai, but you just like going to cloud code, or do you, do you have like a self-hosted [01:17:00] setup or something?

Matt Hill: No, unfortunately we're not with the, we're not to the self-hosted yet. So Aiden really likes perplexity. Um. I've used Claude, I've used perplexity, I've used chat. GPT. Ultimately it doesn't really matter 'cause they're all about equivalent at the things that I use it for, right? I think they might differ in certain ways, but I've never run into any differences and I've never, had that like, nagging feeling of why am I using someone else's server for this?

Because what I'm putting up there is like, write up peer function that takes these inputs and these outputs and it's like, I don't feel like I'm sharing anything personal. I don't think it's, it's all open source code anyway. There's nothing proprietary. But that said, I am always striving to do, cloud things in a self-hosted way.

but oh four oh is really important for that, right? So, one of the things that we're doing with oh four oh is enabling hardware acceleration. So, you know, the AI stuff is, I, I can't be bothered. I can't be waiting 30 seconds [01:18:00] for response from a. Small model, you know, like give me power, like I, I, the, the, the power and accessibility of it is more important than who's hosting it, believe it or not, right?

But all things being equal or even close to equal, the self-hosting starts to get really important. So we have not yet been able to using our own operating system and hardware, do self-hosted AI in a satisfactory way, right? So, for example, we, uh, currently use Shopify for our e-commerce solution, right?

Shopify is a, is an amazing, incredible, massive, endlessly powerful e-commerce solution that is 100%, third party trusted, SaaS product. Super expensive. And it just kills me every single day that we are using that. I know that we don't have to, there are self-hosted solutions, but I have to make [01:19:00] these tough choices just like I do in my personal life about trade-offs and costs.

Like what is the cost of not using Shopify and what does that mean for the longevity and success of the business as a whole? I have to operate at that scope, not just my personal desires. And continually the equation keeps resulting in stick with Shopify because it does things that, you know, are not easy to do in a self-hosted, sovereign way, all the way from networking to specific features.

And so, what we do is I allow that pain to exist. I use it to inform the product development at start nine. This is how we do product developments. How we know what to build next is that I as an individual, refuse to self-host anything that's not running on Start os. So for me it's like, oh, I really want a security camera.

You know, but like, I won't do it [01:20:00] until I can do it using my own product. And so the more I want something and the more start nine as a company wants something that the product does not yet support is actually what informs what we're gonna build next. So I like it. I like that my personal and my personal life and my company, have it almost as a principle that we will only do it if we can do it using our own product and thereby, push the product forward.

That becomes the motive force of product development. And AI just isn't there yet. I was just saying the AI stuff just wasn't there yet. And it's not that it wasn't there yet from a software perspective, because we have, O Lama and Open Web UI available for start os it's easy on start os to install a chat.

GPT, like identical experience running an open source model that you got from hugging face via oh Lama, you know, and it's like easy and it works and it's great, but it's [01:21:00] slow and it's stupid, you know, and that mostly comes down to hardware. That's a hardware limitation. And, um, part of that is something we can't control, right?

Because GPU acceleration, like inference, just is very expensive, right? If you want something that can do, you know, enough tokens to satisfy your, your urgency, it starts getting very, very expensive. Like many thousands of dollars for these, you know, powerful NVIDIA GPUs and you're introducing closed source, hardware and software into the equation.

So what we are really excited about is we are talking with many different, you know, companies and keeping an eye on the, the development of, uh, consumer grade, yet powerful and cheap, acceleration because it's coming, it's coming fast, right? This is an initiative that companies all around the world because of the AI boom.

And so we think that in the coming, you know, five years there is going to be, uh, a rapid commoditization [01:22:00] of consumer grade, inference. And that we're going to see, GPUs that are not Nvidia, that are potentially maybe even risk five based and fully open source GPUs that, can do like really powerful models, you know, even upward of like 70 billion parameter models, and get you immediate great responses, for like.

Cheap for a few hundred bucks, you know, and like that's our hope. We're not doing that. We don't know how to build those, right? We're not operating at the chip level. We're not like that kind of computing company yet, maybe someday. but we are talking with those companies and trying to be, to get that hardware available for our products as soon as it becomes available upstream.

So that's number one from a hardware problem is that the components, the GPUs are just too expensive. But number two is that, start os needed to be able to intelligently outsource [01:23:00] these tasks to the GPU, right? That we needed to have the, um, the APIs available in start os for package developers to be able to tell start os that they want to use certain, certain hardware components.

And we wanted to build it in an abstract way. not a like fragile bespoke way just for like AI acceleration. We wanted package developers to be able to say, Hey, I have a hardware requirement that there's a second wifi chip on this device. Right? Like we wanted package developers to be able to inspect the hardware on which start os is running to see if it satisfies their needs and message to the user if it does not, and why it does not, so that the user could potentially go get that hardware, right?

That's the degree to which we've thought through the, the kind of package developer and user intersection is we want package developers to be that, that expert person in your life that just like knows everything about self-hosting and what [01:24:00] hardware is needed. And we wanted them to be able to encode it without limitation into the package.

And that includes hardware acceleration, which just isn't done yet. So it is going into oh four oh, but it's not an Alpha 10. It's one, it's actually like the. The last, dare I say, kind of feature wishlist for oh four oh. We might even leave it off if we feel like it we're, we will see.

Jordan Bravo: This is beautiful. It sounds like, uh, with Start Nine's help, that self-hosting is finally getting the love and attention that it really deserves.

Matt Hill: That's our goal.

Jordan Bravo: I I just have one more question for you, and that is. Let me, let me say for people who don't know what Tab Conf is, it's an annual focused, technical small high signal conference here in Atlanta, and I've gone the last several years and it's been a great experience.

I've gotten to meet and hang [01:25:00] out and chat with the start nine guys, including you, Matt, in person. And my question is, are you going to be attending tab conf this year in 2025?

Matt Hill: Uh, yes, we'll be there and we'll have a pretty strong presence actually. We're going to, um, be there for both builder days. We'll have a, a station, uh, like a workshop, a perpetual workshop for those two days where people can come over and do two things. one they can come and build their server, and get start os flash to it.

like bring your own hardware type of thing. Uh, we will also help you get start os set up on a VPS. If you don't have any hardware, you can come over and we will help you provision A VPS and, you know, flash start os to it using the, the iso. And uh, yeah, so we'll help you get set up with a server. . I don't know if we're gonna bring servers.

Maybe we'll bring a couple to sell, but that's not really the point of our attending tab Conf, people might want 'em though, so we'll [01:26:00] bring a couple. but also we will be doing package development. So anyone who has any interest whatsoever and packaging anything for start os Bitcoin Lightning or unrelated entirely can come over and learn how to do some package development.

We will pair a program with you on, packaging, whatever it is that you're working on. You could just watch me code up some new package, right? Like, we'll pick something out in the world and we'll be like, all right, let's package that for Start os and we will just in real time, like code it up and get it packaged and release it onto the marketplace.

Like, so basically I'm just gonna be sitting there coding for a couple of days, and building servers and, uh, so anyone's welcome to come over and participate.

Jordan Bravo: Great. Yeah, I look forward to it. I'll definitely see you there. is there anything else before we conclude that you wanted to cover that we might not have gotten to yet?

Matt Hill: No, that was, that was a lot. I, yeah, I leaked a [01:27:00] bunch of crap. I'm not supposed to, but Okay.

Jordan Bravo: Well,

Matt Hill: It's fine

Jordan Bravo: This was a, a great conversation and we appreciate you taking the time to join us here.

Matt Hill: guys. Thank you so much. Really appreciate it. Thanks.

Jordan Bravo: take care. Bye-bye.

Matt Hill: Okay, bye.

Stephen DeLorme: Hey, thanks for listening. I hope you enjoyed this episode. If you want to learn more about anything that we discussed, you can look for links in the show notes that should be in your podcast player, or you can go to atlbitlab. com slash podcast. On a final note, if you found this information useful and you want to help support us, you can always send us a tip in Bitcoin.

Your support really helps us so that we can keep bringing you content like this. All right. Catch you [01:28:00] later.