Is AI Freedom Tech? Part II

Welcome back to Freedom Tech Friday, a live and interactive show taking place every Friday at 9AM eastern, 2PM UK time across the ungovernable network. Each week, we dig into the latest freedom tech, including Bitcoin, Monero, encrypted messengers, and privacy tools.

If there's any news item, tool, or topic that can help you take back some control in today's digital panopticon, then we wanna talk about it. This show and the topics we cover are powered by freedom.tech, a daily news desk that uses AI to monitor hundreds of sources so we can continue to bring you the signal every week.

I'm Max, the CCE of the Ungovernable Network. I'm joined by Seth, who is the COO at Cakewallet. This show is live and interactive, so you can help steer it by commenting live, asking questions, boosting the show, or just sharing it with your friends.

Top support for last week with Jordan comes from Kaz Peland who sent 1,042 sats and said, can you please tune down a lot of the GIFs? He said GIFs, but I think GIFs. Moving pictures. My fountain on a Pixel four can't handle and starts stuttering.

I can only listen offline. We had a bit of back and forth on that, and I think, basically, this is a fountain issue, but, I've suggested just download it. And, I think the streaming stats won't work, but you can still have a listen.

Thank you for your support, Kaz. And like I said, don't worry about streaming the stats. You supported us a lot over the years. Just just tune in, enjoy, and message when you can. So without further ado, let's dive into the show.

We don't have q and a with us. He is on holiday. I can't remember if he told everyone where, but he is away sipping sangria, I think, from memory. And, yeah, we'll we'll catch up with him next week. So, Seth, how are you, mate?

I am doing well. I'm excited for this. Yeah. Been away last week. I know y'all y'all missed me dearly as I was not sipping sangrias. I don't I don't know. Queue queue with these these constant vacations, but some of us only take time off from work to work more. So, you know, you know, the faithful view.

Of He's away sipping sangrias. We're just about to get into your San Francisco visit. What were you shitting in the street and smoking crack, getting in with the locals, or or how was it? Not this time. I was in the I was in the the fancy part of San Fran, not the,

the normal part that you see all the all the expos about. So in this area, you get nice French food. You walk around a park. You watch the Golden Gate Bridge. You pretend that you're obscenely wealthy. It was it was nice. It was nice. It's very very different. It was in the Presidio, which is like this absolutely gorgeous

park area. It used to be an army army, barracks and hospitals. Got all these cool older buildings that they've restored, which is where the Presidio Bitcoin is, the coworking space and podcasting space and all the other good things that have going on there. So,

yeah, it was a ton of fun. It was very different than the San Fran that I had seen previously, so it was nice to get a taste of kind of some of the good side of San Fran. Get some good steak frites while I was there as well, you know, gonna get some good food. So yeah. Very nice. Very nice. Yeah.

Well, as q is away this week and you've just been away, I thought it'd be a good time to touch on some of this AI stuff, which you are deep into the weeds with. Obviously, being the open source AI summit, I thought, like, a perfect time to to do a bit of a recap of where are we with open source AI and

your thoughts on on the whole sort of space and where it's going. So don't really have I've got some bits in the plan that we can go through. We we don't have to really stick to it too much. We can go a bit loose. But, really, like, what was your impressions from being there? And, like, the was there anything that stood

out to you that you weren't expecting or sort of went beyond above and beyond what you were expecting? Yeah. It was it was a really, really good event. I'm glad that I that I made the made the journey over. It was, obviously, was a chance for me to get to go to the Presidio Bitcoin, which

which was awesome. I love Bitcoin Park. I've been there many times and, love this new concept of, like, Bitcoin and Freedom Tech centric coworking spaces, and they they lend themselves to building communities around them, which is what they've they've done out there. So this was a kind of offshoot out of the Presidio Bitcoin,

the open source AI summit, as you said. And it was really fascinating, I think, because I mean, like, we're just we're so steeped in the Bitcoin ecosystem.

Like, that's the vast majority of the events I go to. If it's anything outside of that, it's like a Monero event or maybe an Ethereum event. Went to one of those this year for the first time. So I'm I'm almost always only running in these, like,

insider Bitcoin circles, which are great. I love them, but they're also you see the same people at all these things over and over again. Insidious. Yeah. It's just yeah. It's there's a there's a small core of people relatively that are out there.

So That wasn't the right word. I was supposed to say incestuous. Insidious too, man. Think I felt that that bad about the the Bitcoin cabal.

It's more incestuous than it is insidious, but it's it's a little bit of both. I I haven't seen any of that going on Yeah. At these events, but, you know, maybe maybe behind closed doors. You never know what the devs get up to.

Little hackathons, quote unquote. So, yeah, never been to one of those, but you never know. You never know. But it was it was a it was really nice to get outside of that scene and circle and get to hear from a lot of speakers that are not like, a lot of them are Bitcoiners, but they're not like, Bitcoin is not their identity Mhmm.

Which is kinda refreshing to be able to to to chat with a lot of people, hear from a lot of people who are similarly focused on Freedom and Freedom Tech. And we can dive into it a little bit more, but also heard a a new term that I really loved that is aligned with Freedom Tech. And

that was really, really refreshing to be able to hear. And and just to see too that a lot of these people are are also Bitcoiners. Like, they are focusing on local AI, or they're focusing on open source or open weight models, or they're focusing on harnesses around using AI more effectively,

like Goose, like, all of these things that are tangential, but are all kind of fitting into the the puzzle of what does a Freedom Tech stack look like.

So, yep, that that was a ton of fun getting to to chat with and hear from a lot of people that I really respect in the AI space that I'd I'd never seen before again because they don't they don't run-in these these Bitcoin only circles.

And then lastly, just the the norm of getting to have all the side conversations was a ton of fun. With a lot of Bitcoiners, I mean, there was I would say it was probably, like, 50% normal Bitcoiner type of person and, like, 50% AI centric people from a bunch of different AI companies, AI organizations,

open source start ups, stuff like that. So, yeah, it was it was it was really, really good. I I definitely enjoyed it in the end. It seems to me it's the sort of event that I would be excited to go to and without, like, slacking off most of the Bitcoin events. I I I couldn't care less. I'd rather just, like, be at home.

But these sort of things, it feels like there's a real excitement and, like, openness is is my read of it in terms of trialing new things, not exactly knowing where everything is going, but knowing that you kind of you have to be involved if you're a Bitcoiner even if you don't want to. If you're running a Bitcoin business

of any kind, really, you kind of you can't just go, I'm not really interested in AI. I'm just gonna ignore it because you're gonna get attacked if you don't. And so well, you're not gonna keep up if you don't. So whether you want to or not, you're kind of dragged into this AI world, and it seems to be extremely fast paced. And

every week, I'm seeing changes and new innovations and new unlocks.

And that's kind of a refreshing thing compared to Bitcoin when you've been around for a long time. Don't me get me wrong. Like, it's it's fucking great. Like, it works. You can send payments around the world of any size to anyone and, like, especially with the amount of issues that I have in banking

over the years. That is an incredible superpower, but it's not fast paced, it's not changing all the time. It's like, it just works, and there there are tweaks and there are things. And, you know, we have the odd crazy moment, like a hack here or there or whatever, but,

it it feels a little bit with AI. Like, you could be like a a kid in a candy shop again. Yeah. It really is. And, I think, like

I don't know. Like, Bitcoin is such a powerful tool for freedom and is essential, but it's also boring. Yeah. Like, I'd like, I say that as someone who has spent the last, I guess, four years now working full time in mostly Bitcoin.

Like, I love it, but it's also it's it's very simple. It's very straightforward. Not a lot changes. There's not a lot of cutting edge things happening, And those things that are happening are very slow in terms of adoption. Takes a long time to get uptake, things like silent payments, things like page one v two.

And that's not again, that's not necessarily a bad thing, but it's also, like, I don't know, we're we're clearly enthusiasts. I'm definitely like, I have always been a tech enthusiast, and I love getting to to try out and play with new things. And that that's part of what's been so exciting is this space is moving

so insanely fast. Like, I was you know, we might talk about this a little bit, but I was working on slides for an event that I'm I'm gonna be speaking at about local AI. And just digging back through since I first started getting into, really, AI, but more specifically, local AI.

Like, there were seven massive new open weight model releases in six weeks, all with different pros and cons, all of which I tried to run or did run-in some capacity.

And that, like, that's on top of all of the other stuff that's happening and all of the ways that I'm figuring out how to use it at work, how to use it on for personal things. Like Mhmm. It's just a space that is so fascinating and moving so quickly and that I used to believe was just kinda like, yeah. It's gonna do its thing.

It's fine. I'll do my thing. But I really do think it is it is shaking up a lot of the space. I mean, we've seen in Bitcoin how it can have positive and negative impacts at the same time. So it's something Oh, yeah. We have to stay on top of.

But, yeah, there's there's a lot of excitement and a lot of the I feel like the early excitement around Bitcoin of, like, what can this disrupt is happening again in the especially the local AI scene. Yeah. It's like, there there is a an insidious group trying to control our access to intelligence,

very similar to how they're trying to control our access to money or the tools around money. Mhmm. And we actually have tools to fight back now. And so there's just so much excitement and interest in how do we make these things as good as possible, as simple to run as possible,

what are the pros and cons, what is all the licensing craziness? Because it was named the Open Source AI Summit, but a lot of the models we talked about are technically open source as well. So there's, like, there's all of those caveats that go with it.

And we've seen that in the Bitcoin space too with source verifiable versus open source and having to figure out, like, where does that matter, where does it not. Spoiler matters a lot. Yeah.

But, yeah, it's that that excitement has definitely been been a lot of fun. Yes. I think it's that ground that can be made in in this phase. You can make so much ground, whereas, like, with Bitcoin and cryptocurrency stuff, it's kind of like you you kind of are where you are, and then you're butting up either against regulators

or just, like, limits of what can be done. Mhmm. And so you you kind of not at a standstill, but you it's like you're much further ahead, and it is where it is. And then this stuff is like there's loads of ground to be made, and it is so fast. And

like you say, like, it's now I assume you're gonna be starting to butt up against, like, the controls, and it's still in that phase of working out all the different ways. It's it's like excitement, isn't it? It's just just like huge excitement. I'm I'm excited for less so the stuff that you and Q do because it's just beyond me.

But the the the art side and the creative side and, like, for someone like me being able to get to a point where I can use CAD properly and three d modeling and, like, more of that kind of side of things, especially when it's open,

it kinda blows my mind because it it feels to me like we're getting into a stage now where we all sort of if we have a little bit of money behind us to buy some hardware, we can, in our own way, be a little Tony Stark and, like, building

really fucking useful and cool stuff. Yeah. Yeah. For sure. And it's it's kind of the it's the AI equivalent of of self custody, really. Yeah. Like, it's it's different. Obviously, it's it's more expensive too. Like, a good hardware wallet's a couple 100, $300.

But there's a lot of time that you spend on figuring out what's the right stack for me. Do I do a multisig? K. Maybe I'm looking at more like a thousand dollars in hardware wallets at that point.

How do I figure out how do I make this useful for my family as well, not just me? Like, you get into the Uncle Jim stuff with Bitcoin as we we all have over the years with running nodes for other people, with managing Lightning nodes for other people,

with BTCPay servers that you you help onboard merchant to merchants to so that they don't have to do it themselves. Like Mhmm. We we see all that as well on the the local AI side where it's it's this stuff is powerful. It's changing the way a lot of things work.

And at the same time, it's also it's changing the way a lot of us interact with and build around Bitcoin too. I mean, like, just even I know it's not technically Bitcoin, but just even all of the amazing stuff that Q's been doing with the website, with y'all's podcasting workflow,

with the speed at which you can you can make clips and publish them, like, all of that, that actually has a tangible impact on Bitcoin adoption in a way that I think is

honestly more impactful than a lot of the technical stuff, because the technical side of Bitcoin is just pretty slow right now. Like, a lot of the core technical problems have been

solved enough for a lot of use cases. Obviously, there are tons of improvements we could still have, but without something like covenants, there's a lot of limits to what we can do. And so a lot of the gap, like you said, is butting up against something. Is it regulators, which obviously is always a thing?

Is it, running into just a a lack of awareness, or is it people they want to start using Bitcoin in a more, self sovereign way, but they don't know how? Like, the education piece is so huge for it.

And Mhmm. Like everything else, it's it's accelerated the pace at which y'all can ship and make good content, and it enhances y'all's creative abilities and y'all's technical know how. And that that's part of, like, why it's so exciting because it just like having sovereign money empowers and

serves as a foundation for all of the other things you want to do to free yourself and your family, I really do think that owning your own intelligence does the same thing, where it starts to to to fund and drive all of the other things that you're working on and give your family a superpower

that they wouldn't otherwise have and one that can't be cut off or limited or restricted and that just, honestly also has all of the privacy assurances that are so critical to being able to to use it freely and not have to worry about who's reading what or who's gonna steal your your mathematical discoveries,

all of the the things that that could happen. I worry about that every day, mate. I think he's gonna be sitting a new theorem. He's he's seen the type. He's seen the type. Yeah. Yeah. Exactly. It's like it's a double edged sword, isn't it? It it can be so useful

just using the, you know, like, anthropic or one of those sort of things for the normal person. But at the same time, you're giving it all your data. The same thing with with Bitcoin. Like, you you can go to an exchange and give all your details and

you know, to your bank and Revolut, and then they leak everything, and you put yourself in danger. So there's there's that side which is bad, and then there's the run your own node, you know, by peer to peer, like, proper freedom sovereign side.

And and you can always see the tempting side of the ease of the other way. But as Jordan would say KYC is the dragon that man must slay. And on top of that, I think that these closed source Snoopy type things are also the dragon that must be slayed, and we we we have to start moving towards

this open source and self hosted stuff. So for anyone who maybe hasn't listened to the previous ones who will be joining us either live or in the in the pod once it goes on the feed, I think it'd be useful to do a bit of a run through of what you're personally running now. I think last time it was

last time we spoke, I think it was, like, eight sparks or something like that. So just a rundown of, like, how how you how you made the decision to to run those. Are there any plans to adjust what you're running, add to what you're running?

What does that achieve for you and for cake? And then maybe a little bit of a a thought, which is something I've been thinking about a lot, which is it doesn't seem like the hardware prices are gonna come down anytime soon. Mhmm.

But was there anything that when you were away in San Francisco that made you any more confident that the people who don't have 5,000 to blow or is probably even more than that per GTX Spark, that they might be able to run things, that these things might get more efficient

to run on on smaller hardware. So you don't technically get a saving on Seth. Oh, you wanna share screen? I yeah. Sorry. Jumping on a little bit. But yeah. Yeah. I figured I can it's a good way for me to show the the options here.

Yeah. Do you see what I'm trying to say? It's just that, like, don't want people to listen to this and be like, oh, yeah. Great. You wanna be Tony Stark and be able to do all this kind of stuff, but, you know, you need $50 to to throw at it and these prices won't come down.

It would be it would be good to know, like, if you think these models would be able to run on smaller hardware and people aren't priced out. Yeah. For sure. For sure. Yeah. So I'll I'll use a sneak peek from a a presentation I'm going to give in a in a few weeks, but was working on this week.

All made obviously, I knew exactly what I wanted to talk about and what I wanted to cover, but all made with local AI running on two Sparks specifically.

Okay. And using this open slide tool that I I just talked about on Twitter yesterday as well. But it's a good it's a good way to show kind of, like, what to expect on the different tiers of hardware, because I think this is a really common common question. And before I dive into this, I will caveat, like, yes,

we are running eight Sparks. Yes. They're running in my house. No one sane would do that for, like, personal use. Even for most business use, that would not be necessary. I would say even for, like, our use case, we could we could get away with four with the core of what we absolutely need, at Cake.

But it it frees us up to use them for a lot more by having eight because we run three separate models, across the eight.

Well, and you're exploring as well. Yeah. You know, it's like it's like having having a a toolbox. You have, you know, more tools than you possibly need, but you might be able to experiment and build some extra stuff that you you weren't even thinking about. So I think having all that extra power at your disposal is no

bad thing. Yeah. For sure. For sure. And it lets us very easily compare when a new model comes out, how does it perform on our specific workloads compared to the current one without tearing down the current one to test that and and stopping all of our workflows at the same time.

So, yeah, definitely a lot of a lot of benefits there, but, is overkill for the the vast majority of people. But to get an idea, I think a lot of people think you need way more hardware than you actually need. Mhmm. So I'll cover, like, some of the tiers here. Obviously, all of this equipment is relatively expensive.

Yeah. It's not cheap, but it's not all, like we have 40 k sparks that no one no one needs that. Before you do this bit, I just wanted to make a point that I I was thinking about. I saw a tweet with someone saying that

because, obviously, you're just about sure of MacBook Pro m m five chip or whatever. Like Yeah. I don't even know what the price is, but it's thousands and thousands. So I saw someone make a point that Apple do, like, a finance option Mhmm. Where

I think you then you know, a bit like a trade in and and trade up a bit like you have PCP finance on cars or bikes and stuff. And it'd be like, you you put a small deposit down, then you pay us very small monthly fee, and then you give it back at the end or or trade up. And that their their hope is that you trade up.

And so they were making the point that actually with a a lot of this hardware, you can kind of get it for similar money as you would for, like, a Claude Max subscription anyway. So I just wanted to make that point that it's not always like, oh, I need to find $10 down the back of the sofa.

You you might actually be able to get more than you were hoping for using finance even though yeah. Okay. Maybe not people don't wanna use finance and blah blah blah, but it it can be done.

Yeah. For sure. For sure. And the the other piece of that too is, like, a lot of this, you're already going to get something like this. Like, for instance, this MacBook Pro, a lot of people, if you're doing any sort of technical work, even if you're not, you're probably gonna get a MacBook Pro anyways, quite honestly. It's really normal.

So this is more the difference between getting a base level MacBook Pro and bumping up the specs so that you can also use it for local AI or getting the base level let me get the right one. The base level Mac mini or Mac Studio and bumping it up so that you have something that can run larger models. So it's also not like a 0 to 100 spend.

Yep. But, yeah, it's definitely it it is more than you would normally spend. But when you pair it against that, like, I think the the new Mac Studio Ultras with $2.56 gigs, which are, like, $2.50 256 gigs for the long term, I think, a good target.

Okay. That I think I saw that if you do the, like, the lease to own, it's, like, $230 a month or something like that, which ClockMax is $200. So it's Yeah. Exactly. Yeah. And it's not I'm not gonna guarantee I'm not gonna promise you you're going to get the same level of performance and,

power today as you would get out of ClockMax, But something different that you get and something that I think a lot of people just forget when it comes to this stuff is you get something you can run twenty four seven

and not have to care about limits, not a five hour limit. You don't have to do this stupid bursty work where you do a little burst work. You hit your five hour limit in one hour, and then you have to wait because everything that you're doing depends on that. Like, you don't run into that.

You can just run it overnight and not care about token counts. You can run six different things at once and not care about your limits. And I think a a little bit of the shift that has to happen is people are used to this fast, bursty AI workload.

Yeah. And something that shifts when you start to use local AI is I can just I I should can and should just have stuff working all the time. While you sleep, it's just churning through the bay I mean, that's what Q does for the website. It's like Mhmm.

While he sleeps, it's churning away. It's making clips. It's it's doing show notes. It's doing, like, the tasks that it repeatedly has to do every day or every couple of days, and Mhmm. It just does it while he sleeps, which is which is yeah. It's it's amazing. And like you say, you don't need all that power

right at that moment. It can just cruise through the night. Yeah. Yeah. You don't you don't need instant frontier level performance for the vast majority of things. And that's something that I've been finding too. Like, in for the majority of this week, I've been primarily using a smaller, faster model, called Quinn 3.8 Flash Next,

And it's not nearly as intelligent as Opus, technically. It's not nearly as intelligent as DeepSeaQV four dot one Flash, technically. But I found just for the vast majority stuff I actually need, it's it's good enough. It's way faster,

and I don't have to care about anything while I use it. So I think there is some some some perspective shift that comes with it.

But, yeah, I thought it I think it's interesting, like, when you look at all the different tiers of hardware, like, another one that's really common is a lot of people have gaming PCs. Like, if you're a tech enthusiast, you've probably built a gaming PC at some point.

Maybe you already have something like a thirty ninety. Maybe you don't, but you have a rig where you can just swap out the GPU and get something that would be more performant, and you can get really good local AI with that. I think the I think, like, a 39 to use is, like, three k right now.

You can get multiple and pair them together in certain gaming PCs as well. Mhmm. And that that can be something where either you already have the hardware or you have the vast majority of the hardware, and then you have something that works really, really well,

especially for just just yourself or maybe a couple more users at the same time. That sort of feeds into Bond's question he just asked. Would you run that as your main computer, though, or would it be an AI box almost strictly thinking about security?

That's something that is very much the way I was thinking as well. I haven't pulled the trigger and bought any hardware, but I was kind of thinking, like, that would just do like, I'd have a box, and it would do everything to do with ungovernable, everything that I need, image gen, this and that. And then I'd probably have another one

that is I might be retarded. It's just how I think. And another one that would be, like, my normal life one for everything to do with my finances, health, family, you know, all all the stuff that I have to think about and do manually Mhmm. Have that and never the two shall meet or touch.

So I get his question here. Is that are we being, like, too and, like, security focused about this or not? Yeah. Yeah. I think so. I mean, not I I understand, and I think it's it's a good idea to think that way and to be careful. I But I think it's kind of the same thing as, like,

people treated their Bitcoin node as, like, I can't run my Bitcoin node on the same hardware that runs anything else. Like, I need to keep it separate and secure. And it's just realistically, it's not really a problem.

Also, just if you're talking, okay. Now you I want dedicated hardware that's just for this on top of it, then, yeah, obviously, you're talking more cost. Because now if I if I needed a MacBook anyways, now I need to get a second MacBook to run this. Or

if I have a gaming PC, but that's also my daily driver for everything else, now I have to get a second one and get GPUs and get a CPU and RAM and all that stuff. So I I think, like, if you have the privilege to be able to do that, like I do, where I have separate sparks that can run everything

Yeah. Obviously, yes. That's fantastic. Like, I think that that would be good so that you have some some separation. But I think what people also forget is the thing that's running the model is different than the place that you're

prompting it and running things. Like, yes, the sparks are running the model and answering the queries, but, like, if I'm doing anything with AI, it's happening on my laptop or on my desktop. And any local commands, any work in Git, all of that is happening on that machine.

So even if I had like, let's say I have a separate box that I'm using to run the model from, all of the sensitive stuff is still happening, all the potentially dangerous stuff is still happening on my machine that I'm actually prompting it from.

So it's not the same as, like, a Bitcoin node where everything is happening on that box and you're just making requests. Like, if if I'm running sensitive operations or generating an image or working in a Git repo. That's all happening locally, and then I'm using the model when it needs it.

So the security more matters on the host that you're actually prompting from rather than what's running the model itself. That makes that makes sense. I I think I also think about this in my very specific case, which is I'm a NIM that has a public profile, and I have, like,

two lives I walk. I walk a life of just, like, normal work and normie land and Fiat work and Fiat friends and everything. Mhmm. That's just normal. And then I have my ungovernable stuff, and they're just two completely separate lives that cause or would cause enormous problems if they cross. Mhmm. And and even if something's

being done and I think this could be the case for a lot of NIMs anyway, but, like,

a lot of NIMs aren't maybe quite so publicly facing or putting out so much information and stuff. So mine is like, oh, suddenly, like, doing all this stuff for uncovernable, and then that ends up in a tax return or like that. Do know what I mean? Like, that ends up in a and so I think that's maybe what Bond's

getting out there. But what you're saying is that it's done locally, and, yes, you're connected to the Internet and everything, but you'd have to manually push something. It's not being stored on someone else's server. You're not putting this into ChatGPT.

It's gonna go and grasp you. That cross And your your model, like like, I am doing some personal stuff on the Sparks in their downtime. The the model and the device serving the model isn't going to cross those streams. That's not how it works. Like, that's not something that

is going to happen, and it's not storing any of that information locally anyways. So if I do something personal and then I do something for work, it doesn't have access to that somehow magically through the Spark. That's it's not not the way that it functions. To keep them separate, the best way, like

even if you wanted, like, a complete separation, then you could, like, have a let's say you you have a mini PC that you only use for ungovernable ungovernables. Mhmm. And you you're running all of your AI workload from there,

and then it's connecting to whatever is running the models, whether that's your main PC or not. Mhmm. And then everything sensitive is happening on that device. All the history is stored on that device.

You can use something like Herder to connect to all the different devices that you want to to run prompts from. And it can have its own siloed sandboxed. This box only knows about ungovernables.

And then on your personal laptop or whatever, don't do ungovernable stuff and and only do the personal stuff. Something like that. It could be also I mean, it could be user profiles on a Mac.

Like, there are lot of ways to do it Yeah. And how separated you wanna go. But it more matters where you're prompting from and where you're working with all of that data locally more than actually what's serving, the model. Obviously, they're like, yes, someone could hack the device

where you're serving the model and then log prompts and data. But if that happens, you have bigger problems to worry about, honestly. Yeah. So I I think a lot of this stuff comes down to, and I don't think I'm alone here. Just the the how this stuff pieces together similar to the early days of Bitcoin. Like Mhmm. How

like, you just said something there where you're like, oh, it's not a filter. You said, oh, you said a word of, like, oh, that's gonna connect that to that, and I can't remember what it was now.

But there's there's information that people who are new to AI Mhmm. Just lack in terms of this is a big puzzle piece.

And you sort of know that there are x amount of puzzle pieces, but you don't know that, oh, if this one goes near that one, you're fucked. Or, like, this one's really important, and it saves you having to do this. And I think I I don't think I'm alone in that. And I think that's maybe why people people are thinking, like, separation

of the actual hardware when actually it's it's like permissions or sandboxing or, like, more technical ways to do it rather than just a simple dumb approach of, like, one box for that, one box for that.

Yeah. Yeah. And that's that's why we're here too. So we can Yeah. We can talk through it and answer questions and and help people understand the differences of of how these things work, which is yeah. It's fun. It's fun. Yeah. It it we've got a lot of questions in the chat. I don't wanna derail us too much. But Very good.

Ralph Corduroy. Is that, like, Ralph Corduroy, or is that your actual name? Ralph Corduroy. Nice to dip into the show live for a few minutes. We'll listen to the whole recording later. Thanks for the show. Much appreciated.

We've got the unknown discipline. Can you use a gaming rig to run a local model? I think we kind of answered that already. Yes. And, like, just think about how you separate things out. Bong again, the reason I ask is because for for example,

Maple AI agent yesterday had full shell access to my computer. Then he said it was prompting me to accept,

deny its next move. To be honest, it was super long command, and I truly didn't know exactly what the fuck it was doing. Mate, that's me every every time I use. It's just like, put this into the shell, press enter, do this, do that. Yeah. So you're not alone there.

Yeah. Yeah. I mean yeah. Again, that it is scary, and it is why we need to be cautious what we allow in our machines because they also like, something non malicious also causes problems. Like like, when I was working on this, this presentation, at one point, I told it to delete one specific line,

and it just deleted the whole presentation.

Thankfully, I was able to recover it all. Like, well, I I say I. It was able to recover everything exactly as it was. So it wasn't the end of the world, but that I mean, you see the stories all the time of people having having a

clot or something accidentally delete all of their host, or they're using it to work on a database, and it just wipes the whole database. They they make mistakes. Even the top tier, frontier level models make mistakes too. So, yeah, you do want to be cautious what you give it access to, but, again, that more relies on

like, if I I have a mini PC running here that runs all of the the core of what I do, most of it, all my personal stuff, I run off of it. So if it wipes itself,

okay, it's not the end of the world. It would be a pain in the ass because it's where all, like, my self hosted stuff also runs from, but it's not where the the important data lives. It's not the end of the world, but it's like it wiped my work Mac. Would be a lot more painful.

And so I'm a little bit more cautious what I let run there, or I only run work stuff where if me doing a work thing deletes it, obviously, it's still a pain in the ass, but that's more justifiable than if I was doing something for personal reasons and it wiped my Mac and I lose work time as a result.

There there is good reason to have separation of duties, but I don't think we have to be overly paranoid about it. I I think that's something that, like, over the coming weeks and months, we just keep chipping away at. I think that

sort of us less technical people will just start to use and play and test and, you know, we we start to understand how to do this in in a less sort of basic way. But I don't wanna derail us too much, like, getting into the weeds with it. I wanna wanna get your thoughts on, as I say, like, coming back from there.

Mhmm. Actually, I'm I'm derailing us again because you were halfway through a presentation. You're good. I know. I I didn't wanna get the whole thing. I just wanted to to show people some kind of expectations for for the different tiers of hardware.

Okay. Which I think we we mostly talked through. I mean, the own the last thing I'll really kinda cover is the more RAM you have, the more intelligent model you can run. And so it's not necessarily about speed. Like, you see the thirty ninety here is Mhmm. Super fast, but it's running a model that's less intelligent. It's like Opus 4.6

intelligence. But the fact that you can run that at a crazy speed, like, that speed is is amazing. It'll be more than enough for anyone, is awesome. But if you need something much more intelligent, like, that's why the sparks are really good, where I can run two sparks together and I get something that's, like, Opus 4.8 at home,

but faster. And that's lot of users. So it's a bit like Top Trump's cards if you ever used to play those. You're talking about the the the number there, the two fifty six. Mhmm. The RAM, that's how intelligent

a model you can run. So Generally. Yeah. So if you if you had a two fifty six or a a 500 or whatever, you you can run a lot, but it's not necessarily fast depending on what you what you go for. Would not I think, like so I would prefer really smart.

Mhmm. But I don't care how how long it takes for for my workflow. But some people want quick prompt, like chatbot style back and forth stuff. It doesn't have to be too smart. So it's kind of you you've gotta pick your hardware depending on how you use it is what you're saying. Yeah. Yeah. Like, if I was running

like, if I was running Hermes, and I just use that as, like, a personal assistant, and I'm constantly interacting with it. Like, something that's super fast but a little bit less intelligent like this, quint 3.827 beats. That's gonna be great, and it's gonna feel better because it's gonna be much, much faster to actually

give me responses to generate output to write code. It it won't be as good at large, like, long horizon tasks.

But, again, if I'm not using it for that, that's that's fine. There's there's trade offs every way where the Spark's like they give you a lot of memory capacity really quickly, but they don't have the best memory bandwidth. And so that's why they can run large, really intelligent models, but they're not super fast.

I think they're fast enough. And depending on the model you run, they're they can be quite fast. But they they let you run something like GLM 5.3 Flash, which you see here, which is OPUS 4.8, which in many ways is better than OPUS five in real world usage if you use those models.

And, again, I can run that home with no limits and quite quick at the same time, which makes a big difference. So so given all of that and, like, we've talked about maybe financing and all that kind of stuff, do you think like, are you confident in

us going in a direction? And when I say us, I don't necessarily mean the whole world, but, like, people who care about their freedom and their their personal info. Because I think a lot of people just literally couldn't care less.

Do you think we're going in the right direction where most people who are involved in ungovernable and, like, care about Bitcoin and Freedom Tech and stuff will be able to have and it will be normal for them to have their own self hosted open source AI and that they're not gonna be going,

oh, this is just too fucking difficult, or the the the paid for models, the the anthropics of the world are just so much better that I feel like I'm losing out. Will it be a stage where, yeah, you gotta buy the hardware, expensive, and you've got a bit of a learning curve like we all did with with cryptocurrency and other things, but

you don't feel like like, I I I never feel when I send a Bitcoin transaction, like, oh, fuck. That would have been easier with a bank. Like, I wish I'd have I wish I mean, I always go, oh god. At least if that wasn't a pain in the ass because the bank is so much worse.

Whereas I think early days, people were like, oh, it would just be easier to use a bank because, like, I I can't do all this in command line and fuck around and all this. So you get what I'm what I'm getting at here. Like, are we gonna get to that stage?

I mean, I I really think we're already there. Really? That's been the the huge breakthrough, I think, over the last, like, six weeks where like, it it really started with DeepSeek v four Flash, which was good enough and fast enough that you could use it for everyday tasks without much compromise.

And things have snowballed drastically since then. Like, even on the, like, on the the one that the model that I showed on the sparks right there, GLM 5.3 flash. Like, it's

everyone I know, including myself, I would much prefer working with it over any model except the very, very, very highest tier ones that are crazy expensive, like Fable 5.1 and Astra. And with the way limits are right now, you don't get much usage with those large models anyways. So you end up you're using OPUS five or you're using

Luna or Terra or Sol heavily, and it's competitive at least with those, but with all the benefits of no limits of privacy of everything else. And I I really think we're actually there now where you're getting an intelligence level that for 99.9% of people is more than enough and at a speed that is, for 99.9%

of people, more than enough. Obviously, if you're a incredibly high level developer and you are a heavy user of this stuff and you really understand it, maybe, yes, you also still need a you need a fable or something like that. But the the other thing that people forget is it doesn't have to be one or the other for every task.

Like, you can use your local AI for 99% of tasks, and then for the for the 1% you need Fable for, guess what? You can pair them together. I can use Fable over NanoGPT or something as the the orchestrator or the planner, so let it control the long horizon stuff that it's really good at,

and then let it push the actual grunt work off to my local models. So I'm only paying for the bare minimum of Fable that I need, and then I'm doing all of the work itself locally on something where I already paid for the hardware, and I don't have any per token costs.

Like, the it's not it doesn't have to be one or the other. I think for a lot of people, it can be only local AI already, but it also can remain both. And the tools around combining those two are getting really good, or you can use something like an an OpenGo desktop app, or there's a lot of options to to manage using those together.

Do you know what I've been thinking about? And you might have seen this in San Francisco. A little bit like these node packages came out in Bitcoin, and it became pretty easy to install and run and, like, know, especially with something like Umbral, it kinda feels like, oh, you know, I'm just, like, using Apple here pretty much.

Whereas early days, you might have to get into command line and do all this clever stuff that, like, 99% of people would be like, oh, no. Fuck that. I'm really used to using Claude and probably using it in quite a dumb way, like, just in the chat.

Like, oh, I wanna build a website. Can you give me a mock up of this? I wanna I wanna create a logo. Can you do this? And I'm not using it in a smart way and got GitHub and got, like, different things open or, like, giving it access to all my files and that kind of clever stuff that other people might do.

Is there gonna be, or is there already, like, a wrapper, like, where you could say, yeah. Go and buy yourself this box from Apple or go and buy a couple of the Sparks, and then you just install, like, an umbrellesque thing that's, like, makes every like, makes using all that power and ability easy.

Do you do you kinda get what I mean? Like a like a OS for for this stuff? Yeah. So the I would say it's not quite to that level yet, but the thing that makes that not matter is just use AI to deploy it. To build it.

Yeah. If if you're if you're wanting to do this, you already have a subscription to something, or you're already using a NanoGBT or something like that. Like, you're already an AI user if you wanna do local AI.

So why would you do it all yourself anyways? Like, I I'm not I'm not hand managing any of these even though I absolutely could because it's drastically faster and takes less of my time for me to say, hey. Here's a new recipe for this this new model that just dropped.

Get it running. Let me know the performance and compare it to the last model that I was running before. Or, hey. There was an update to this recipe that says that it's much faster now. Can you test it, deploy it, and roll back if it fails? And you just let it do it for you. It's just it's not like, we don't

for a lot of the things we need, we don't need and I'm not saying this because I don't like them or the companies, but we don't need a NumPril or a StartOS or whatever. Like, if you if you have access to AI, it can manage this stuff for you. And the this these infrastructure type tasks are, like

it's amazing at it. And I say that as someone who, like, my career was infrastructure. It was systems administration. It was site reliability. It was cybersecurity engineering. Like, that is what I trained for and did professionally for a long time, and it's better than me and faster than me. And it it's just simpler.

Does that piss you off a bit? Like, I know it's cool, but are you a bit like, you fucking AI prick. Like, I studied really hard for that, you just just do it like it's nothing. If I was still maybe if that was still my day job, I would definitely like, I I'm understanding more and more why people feel threatened by it. But

as someone who did it for long enough that I got really bored and tired of it, the fact that I can do it now but better and with very little interaction from myself is awesome. Like, I I love things that make me more efficient or that make Cake Wallet

more reliable or more useful. Like, that's what I love. I don't really care that it's replaced the old things that I used to do. Also, because just like anything else, you get you get bored and tired of doing stuff Yeah.

Stuff yourself over the long run. So, no, I'd say it doesn't piss me off, but it I'd like, if you're a site reliability engineer or an IT engineer right now, I should be worried. You you should be worried because I I can do

the vast majority of that simply while I'm also doing a thousand other things. So Do do you know what makes me laugh is, you know, that whole learn to code movement that was going on. Mhmm. And I remember coming into Bitcoin and trying

to learn to code because I was like, oh, well, I've gotta fucking know how this stuff works. I've gotta so I, like, did a few courses

and was, like, trying to learn how to build websites and then trying to learn, like, learn all this stuff. And I remember just being like, I'm very bad at this. Like, I'm I'm not built for it. Feeling like, oh, I wish I wish I could be better at this. And now I'm thinking, thank god I didn't

try any harder. And thank god I gave up on this idea of learning to code because this stuff just does it a 100 times better than I ever could. So it's just it's weird. It is it is a a strange world.

We I asked you this before a few weeks ago or maybe a few months ago. I said, if you could flick a switch now to keep all of this on, AI on, and let it continue to roll out the good, the bad, the ugly, all of it, or turn it off Mhmm. What would you choose? And last time you said, I'd let it run.

Is there anything that you've learned recently, anything that's scaring you a little bit where you're going, maybe not?

I think it's hard. Like, I'm a very pragmatic person. It's hard for me to think theoretically when I know the reality is there's just there's no putting the genie back in the bottle. Yeah. So, like, even with all this, like, this let's slow down AI garbage that's happening right now by all the doomers in The US. Like, okay.

So let's say you even get The US to slow down. So you just gave China China. Even more of an advantage? Like, I'm not they're not gonna agree to slow down. That's not happening. So, like, it's hard for me to put that aside. If I could, I still would not redo it because, like, something that I also love about this is, like, as a parent,

as someone who is running a company, as someone who has a lot of things on his plate, the ability for me to amplify the amount of work I do and to do work when I'm not working, I just like, I can't go back. Like, it's so satisfying and

rewarding that when I wrap up my day, I can fire off a bunch of agents to do things that I need, but I know it would take a while, and then I just wanna be able to review after. I can do that. I can go downstairs. I can play with my kids. I can make dinner. I can put them to bed. I can do all of that stuff.

And then when I log back on that night or the next morning, I have a whole host of new things to review, to iterate on. And it that was all happening when I wasn't at my computer. And, like, one of the like, as someone who really is very efficiency minded and loves to be productive,

like, one of the hardest things for me as a parent was, like, I love my kids. I wanna spend time with them, but when I'm spending time with them, I'm not working.

And I I need to figure out that balance, and that is a hard thing to do when you love your job, and you know that the more you can do, the better your job and your company can be. And this merges that gap a lot. Every every father Yeah. In the every every good father in the world gonna say. Like,

is is thinking, I wanna spend time with my family, but I also don't wanna spend time with my family because I want my family to have the best. You know what I mean? Like, there's there's that constant trade off. So Yeah. What you're effectively saying is, like, you wouldn't because it's an extension of yourself,

which allows you to be so much more productive, that allows you to spend the time doing the things that really truly matter in life. And even if there are some, like, potentially scary things, you'd still take that option.

Yeah. Yeah. I think so. Is there anything that worries you in the sense of, like not that we can stop it, but, like Mhmm. Anything worrying on the other side of, like, with with this great power, you can do great things or you can do really bad things. Like, we've talked a little bit I talked with

Zach about, like, oh, you know, you could effectively check your own DNA and, like, make adjustments and do all sorts of

crazy things to dial in your health. But then on the other side, it's like, well, awesome fucking weird and scary stuff can happen as well. Like, that's that's one thing I go to and think about. What what about you? Is there anything where you're like, fuck. This could be bad.

Yeah. I mean, there's a lot of areas. Like, we've already seen the one in terms of security when it comes to Bitcoin. Like, obviously, we've we've beaten out that dead horse many times, so I won't Oh, yeah. Won't recap that even more, but we've seen those downsides.

From a surveillance perspective, we're already seeing those downsides of how that can be leveraged by authoritarians and dictators and even, quote, unquote, good nation states to Mhmm. Abuse the access that they have to personal information and and enhance their surveillance capabilities.

I I think just like any new breakthrough technology, it's going to be an arms race between the people who

choose to use it for good and know how to do that, and the people who choose to use it for bad and know how to do that. And that like, that that's not new. Like, the Internet did that as well. Bitcoin has done that as well. There's been a lot of people using Bitcoin for really bad things.

It doesn't remove the overall good, but it it puts some pressure on us to figure out how do we weaponize this for good. How do we make this something that is an extremely aggressive offensive weapon for people actually fighting for freedom?

Mhmm. Because it it it is an arms race. And, I mean, just like even the the infrastructure self hosting stuff I talked about, like, all of the things that I used to spend so much time making to make it easier for people to run things and self hosting and all of that, like, you can run and build

the tools that you uniquely need for better security, for better privacy, for better autonomy,

the things specifically your wife wants, specifically that your kids need in a way that just was never possible before. And so it it can also be that weapon to push back against surveillance, to build the the privacy tools you need, or to be able to self host stuff you would never be able to technically

figure out how to do in the past. Now you can do it because AI gives you tooling that you didn't have before. So,

I think that it, yeah, it puts more pressure on us to figure out good ways to use this stuff. But, yeah, there are a whole host of of evil ways that this is going to be used, and there's not a way that we can necessarily stop that.

We just have to make sure that that we're armed with what we need and that we're arming others with what they need to push back in in this in this fight.

And I guess the the one of the biggest things you can do is just protecting your as usual, protecting your privacy, protecting your information because any information that's out can be used against you and and all of that then the answer to a lot of that is, well, run these things locally.

Mhmm. And so, yeah, it's exciting. I'm excited to see how these things develop, and I I'm still on the sidelines very much and sort of watching you and Q do some incredible things, but, I do hope to get some soon.

Was there anything before we close out, we're getting close to time. Mhmm. Was there anything that really jumped out to you from San Francisco? Any companies that you met, any people that you met that you thought were really great? I know you you touched on. We haven't really gone into this slide maker or whatever it was called.

Mhmm. What what stood out? Anyone that should come on the show? Anyone that you think is gonna do great things? Yeah. I I think I mean, there were there were a bunch of things. One really interesting talk was covering kind of how quickly

intelligence is scaling down to where you can fit it in more and more reasonable hardware. And one thing that I think will balance out the the cost of hardware that will continue to increase, I think, is that the the intelligence of models you can run on any given amount of RAM is getting better and better and better over time.

So that one was really interesting, and it's this, like it's called the densification law, which is this new new AI thing of how they get smarter and smarter at smaller sizes, over time, which is really important for local AI to be sustainable long term.

So that was really interesting. I I think the the biggest one that I'll leave people with, and just because it resonated with me as, like, a an alternative name and a manifesto around the idea of Freedom Tech, was something called resonant Computing.

The guy who wrote this manifesto, it's just resonantcomputing.org. The guy who wrote this was there and did a fireside chat. And it he he told essentially the story of, like, how he had worked at Google for a long time and had been a part of that rat race and seen how

they were incentivized and pushed by management to build things that were not evil, but built to, like, milk Enslave people for all money. Yeah. Yeah. Like, yeah, really. Like, to to enslave you specifically because the the more enslaved you are, the better their profit margins are. Yeah.

And so he came up with this this idea of resonant computing as the alternative to that, which is is hardware and software that not only is good in the moment, but leaves you feeling nourished and satisfied after you use it. And that,

that has these five principles that it covered, that it's private. Obviously, we understand that and agree with that, that it's dedicated, that it's it's built to work for you, not for someone else, which is really the the core of how we've defined Freedom Tech over the years.

Yeah. That it's plural. There's not one person or entity or company that controls it and gets to decide who uses it, how much they get to use it, that it's adaptable, that you can tweak it to fit your needs, even better,

and that it's prosocial, which is it's it's supposed to enable connection and coordination. It's supposed to help you be be a better person, but also to to build real relationships. I would definitely recommend, yeah, I would definitely recommend going and reading the manifesto.

You can sign it as well. I went ahead and signed it after the conference, because I think it so beautifully aligns with this idea of freedom tech and puts different words around it and a lot more a lot more weight behind it than I've seen before. So that honestly, that was it was not technical at all, but that was my favorite,

content, I guess, for lack of a better word, at the conference because it was a great philosophical foundation for all the things that we build, all the things that we do, all the things we talk about here on Freedom Tech Friday fit in line with that, and it put it put good good language behind it, which I really, really appreciated.

Yeah. I like that. I I like the the idea of not being a product, and you you're saying that you walk away from using it not feeling like shit, and that's Yeah. That's not actually you said it much more eloquently than than that. But

but, you know, we I think we're all know this. You you step off Twitter, and you're like, fuck. I feel like a worse person for that. Like, a lot of this stuff that you use. So I love that idea of using these tools to make a better life for your family and and not being the product and all of the things that he said there. That's that's,

like, a a very nice idea of how to use this, and it seems like you're already kind of doing it, and we're we're not that far away from it. So that's exciting. Yeah. For sure. Very nice. Well, we are now at time.

Sorry, everyone. I somehow fucked this up, and we still have kill your customer, the Bitcoin brief brief 90 at the bottom. I don't know how I did change it. It's even there at the top left. Can AI be freedom tech? It's what I've named it. I don't know why I've done that. Q will Q will tell me.

But, yeah, thanks, Seth, for going through all of that. And I think this is just gonna be a a constantly fast paced evolving topic, and I'm hoping that a lot of the ungovernables join along on this journey in in their different ways, their different speeds, and,

you know, maybe you and Kur running off ahead, but we're all gonna start catching up and learning how to use this and interact with each other. So it's exciting. It's it's very, very cool. For sure. For sure.

Right. Thank you to everyone for tuning in. We will catch you on the next one. We've got the node runners coming in on the sixteenth. We've got Roy from Breeze coming up. I think we might have Pavel Luptak as well. So we got some really good Freedom Tech Fridays coming up for you all soon.

Stay ungovernable.

Thank you for listening to Freedom Tech Friday. To everyone who boosted, asked questions, and participated in the show, we appreciate you all. Make sure to join us next week on Friday at 9AM EST and 2PM London.

Thanks to Seth, Max, and Q for keeping it ungovernable. And thank you to Cake Wallet Foundation and my Nim Box for keeping the Ungovernable Misfits going. Make sure to check out ungovernablemisfits.com to see mister Crown's incredible skills and artwork.

Listen to the other shows in the feed to hear Kareem's world class editing skills. Thanks to expatriotic for keeping us up to date with Boost's XMR chats and sending in topics. John, great name and great guy, never change and never stop keeping us up to date with mining news or

continuing to grow the Mestadel. Finally, a big thanks to the unsung hero, our Canadian overlord Jordan, for trying to keep the ungovernable in check and for the endless work he puts in behind the scenes.

We love you all. Stay ungovernable.