CD191: JUSTIN MOON - AI AS A TOOL FOR FREEDOM

WEBVTT

NOTE
Transcription provided by Podhome.fm
Created: 02/16/2026 22:26:21
Duration: 5577.456
Channels: 1

1
00:00:32.610 --> 00:00:34.130
Happy Bitcoin

2
00:00:34.690 --> 00:00:39.410
Monday, freaks. It's your host Odell here for another Citadel Dispatch.

3
00:00:41.410 --> 00:00:45.895
The show focused on actual Bitcoin and freedom tech discussion.

4
00:00:45.895 --> 00:00:54.295
I have another great show lined up for us today. Before I get there, just real quick. As always, Dispatch is supported by our listeners. We have no ads or sponsors.

5
00:00:55.095 --> 00:01:01.930
It's just me and you guys. So thank you for supporting the show. Thank you for sharing with friends and family. All relevant links are still dispatch.com.

6
00:01:02.329 --> 00:01:10.890
The top zaps from last week's episode was Keith Sharp with 21,000 sats. He said cool to dive down the open claw rabbit hole during the bear.

7
00:01:11.210 --> 00:01:14.844
Also, first shill for tab comp. Stimmy 40 HPW

8
00:01:14.844 --> 00:01:18.685
also zapped 21,000 sats. He said this was a real banger.

9
00:01:19.965 --> 00:01:34.900
We got another 21,000 sats from Florida. Justin said favorite new reoccurring conversation and Mav 21 ride or die freak. 10,000 stats at Great Rip. I'm glad you guys enjoyed the AI focused Rip at Gleason because we have another AI focused Rip today.

10
00:01:35.300 --> 00:01:37.220
This time with good friend,

11
00:01:37.700 --> 00:01:43.675
Justin Moon, who's leading the AI Freedom Initiatives over at the Human Rights Foundation. How's it going Justin?

12
00:01:44.315 --> 00:01:54.555
It's a pleasure to be here. I think this is the first time I I watched the recording of one of these in like Rod's basement like four or five years ago, but I was too too bashful and too shy to actually get on the mic.

13
00:01:55.590 --> 00:01:57.830
I think that episode we didn't ship.

14
00:01:58.390 --> 00:01:59.430
Really? Okay.

15
00:02:00.310 --> 00:02:04.710
I think that wasn't a dispatch. That was Rod wanted to do a show and then

16
00:02:05.030 --> 00:02:14.525
got cold feet. We recorded the whole thing and he was like, I don't wanna release it. And the show just died on the vine. It was episode one. The one it was with Parker. Right? And, like, No. Harry

17
00:02:15.805 --> 00:02:19.485
It was with Ben Carman and someone else. I don't remember who else.

18
00:02:20.205 --> 00:02:27.460
Oh, no. No. That one definitely got shipped then. Was that in the basement of Bitcoin Park, you mean? No, I think it was in Rod's back It was in his

19
00:02:28.420 --> 00:02:32.500
garage. Oh yeah, before we had the physical location of Bitcoin Park,

20
00:02:33.220 --> 00:02:35.300
the recording studio was in Rod's garage.

21
00:02:37.105 --> 00:02:39.105
Beautiful. We've upgraded tremendously.

22
00:02:40.145 --> 00:02:42.465
Anyway, we did ship that, but there was a

23
00:02:42.865 --> 00:02:45.745
separate show that was supposed to get released three years ago

24
00:02:46.225 --> 00:02:50.385
that Rod had cold feet on and never got released. It was actually a really good conversation.

25
00:02:51.600 --> 00:02:52.720
Okay. Awesome.

26
00:02:54.160 --> 00:02:56.560
I mean, so this is your first time on Dispatch.

27
00:02:56.880 --> 00:03:01.840
I will say that one thing I like to do on Dispatch is I grandstand against

28
00:03:02.160 --> 00:03:04.800
all the sponsored shows because

29
00:03:06.195 --> 00:03:10.915
I don't have any requirements to ship a certain number of episodes per week. What that means is

30
00:03:11.475 --> 00:03:16.515
I only do shows when I think it's gonna be interesting and when I think it's valuable for the Freak's time.

31
00:03:19.020 --> 00:03:19.660
And

32
00:03:20.060 --> 00:03:24.140
when I think it's gonna be high signal. And then, of course, we scheduled this rip,

33
00:03:24.380 --> 00:03:30.459
and I found out that I am taking sloppy seconds from Marty, who you dropped an episode with on Saturday.

34
00:03:32.784 --> 00:03:35.665
I don't I'm know how this not a politician. I'm just

35
00:03:36.064 --> 00:03:37.505
a humble vibe coder here.

36
00:03:38.224 --> 00:03:44.305
But, I mean, this space is moving so fast. I feel like there's a lot of shit going on, so I think we can still make it high signal. First,

37
00:03:44.940 --> 00:03:52.380
I don't know where we should start. Me and Alex Gleason, we were all over the place. By the way, freaks, next week is going to be a Bitcoin focused episode.

38
00:03:52.780 --> 00:03:57.100
We're going be talking about no KYC P2P exchange. So, back to the basics.

39
00:03:57.260 --> 00:03:58.220
That'll be a lot of fun.

40
00:03:59.745 --> 00:04:02.865
Where do we start here? I mean, the the big thing recently

41
00:04:02.945 --> 00:04:08.945
has been Open Claw coming onto the scene over the last three weeks. How much of the Open Claw

42
00:04:09.665 --> 00:04:14.370
how what's your perspective on Open Claw in terms of hype versus substance versus

43
00:04:14.450 --> 00:04:16.050
staying power type of thing?

44
00:04:16.370 --> 00:04:16.850
Yeah.

45
00:04:17.330 --> 00:04:20.690
I mean, I think my my feeling is, like, what we're witnessing is, like,

46
00:04:21.250 --> 00:04:26.130
it's something somewhere between, like, we're seeing a new type of computer being created

47
00:04:26.455 --> 00:04:52.160
or like a totally different way to use the computers that we've been using all along. I don't know which one it is. I mean, it's you can bike shut it, but it's something like that where it's like your your relationship to your computer to the computers in your life is just fundamentally changing, and it's fundamentally changing really, really fast. And so I was very empowered when I learned like computer programming in my early twenties because I had used computers before that, but now it was a completely different thing because I could get it to do

48
00:04:52.815 --> 00:04:55.135
anything that my imagination,

49
00:04:55.135 --> 00:04:56.175
my expertise

50
00:04:56.255 --> 00:05:15.979
could instruct it to do. But it took a lot of effort to do that. It, you know, it was very painful. It's not something that many people did, but it really it really empowered me and that's kinda what my whole career and everything has been based off of. And so now what I'm seeing is that a lot of my other friends are gaining these powers, you know? Their their ability to get a computer to, like, help them

51
00:05:16.220 --> 00:05:22.539
is being totally transformed in the last, like, say, two months. And OpenClaw is the primary touch point that they've been seeing,

52
00:05:23.315 --> 00:05:32.675
but there's nothing special about OpenClaw as like a software project, right? Like a lot of my friends are now vibe coding their own kind of OpenClaw like things. Pablo,

53
00:05:33.074 --> 00:05:33.955
Nostrapablo

54
00:05:33.955 --> 00:05:40.640
has been vibe coding his own called 10x for the last like six months, and he has it doing all kinds of similar things.

55
00:05:40.960 --> 00:05:44.960
And so, you know, while Open Claw is very, very successful,

56
00:05:45.360 --> 00:05:54.575
it's kind of to me like a taste of things to come. Right? Like, it's like a herald more so than like, you know, a product necessarily.

57
00:05:54.575 --> 00:06:00.415
That's what's really exciting to me is that like a year ago, it looked like if we were gonna get agents, it would be in the chatgpt.com

58
00:06:00.415 --> 00:06:01.135
website

59
00:06:01.375 --> 00:06:02.655
and even

60
00:06:02.975 --> 00:06:04.735
given the fact that chat

61
00:06:04.975 --> 00:06:15.880
OpenAI just bought in some sense bought the project or Peter or something, you know, like some sponsorship, who knows what it actually is, even given that, it's like, this is, you know, agents did not reach us through

62
00:06:16.840 --> 00:06:17.880
chatgbt.com

63
00:06:17.880 --> 00:06:19.175
or claud.com

64
00:06:19.175 --> 00:06:30.535
or whatever. Right? It reached us through an obscure GitHub profile repo made by some guy in Austria that was like totally ground up, you know, from the bottom up, very kinda decentralized

65
00:06:30.775 --> 00:06:48.665
as a tinkerer culture and is something that is is very grassroots bottom up open source in a way that's very aligned with all of our ideas and aspirations for Bitcoin and for Nostril and the stuff that we care about. So I I see a lot of it's very encouraging to me how it happened

66
00:06:48.825 --> 00:06:50.345
because I think it,

67
00:06:52.265 --> 00:07:06.660
yeah, it shows that kind of AI is moving in more of a direction that we would like, and I think it's a bit of an inspiration for me personally and for some of my friends, like, okay, we could operate at a higher level. Like, there's more we can do now, right? Like, we can do things now, and there's a lot more possibilities.

68
00:07:07.699 --> 00:07:08.740
I love that.

69
00:07:09.220 --> 00:07:11.139
First of all, freaks, I forgot,

70
00:07:11.905 --> 00:07:17.105
I kinda butchered our intro, and I apologize for that. But the current price is

71
00:07:17.345 --> 00:07:20.305
we're recording this. Yeah. We're recording this at

72
00:07:22.065 --> 00:07:23.425
two thousand UTC,

73
00:07:23.960 --> 00:07:31.240
Monday, February 16. You will probably get this in a few hours. The current block height is nine three six nine six two.

74
00:07:32.280 --> 00:07:36.760
The current price of Bitcoin is $67,852,

75
00:07:36.760 --> 00:07:40.375
and that works out to $14.73

76
00:07:40.375 --> 00:07:41.495
stats per dollar.

77
00:07:42.055 --> 00:07:43.095
On that front,

78
00:07:43.414 --> 00:07:45.015
just to get back to what you were saying,

79
00:07:45.815 --> 00:07:46.215
the

80
00:07:46.854 --> 00:07:47.815
I mean, it's interest

81
00:07:49.735 --> 00:07:53.014
where do I wanna go with this? It's interesting. Right? Because

82
00:07:53.230 --> 00:07:53.870
to me,

83
00:07:54.990 --> 00:07:57.870
it's not about the specific open call project.

84
00:07:59.390 --> 00:08:00.510
It's it's

85
00:08:00.830 --> 00:08:01.950
it was a

86
00:08:03.550 --> 00:08:10.725
it's a light bulb moment for a lot of people on how we interact with these things. And it's kinda interesting dynamic when it comes to me and you,

87
00:08:11.445 --> 00:08:12.645
because I was

88
00:08:14.645 --> 00:08:20.885
very optimistic and excited about Noster before you were. And we had a lot of back and forths on that.

89
00:08:22.740 --> 00:08:29.700
And now you're very excited about Noster. And you were very excited about I took a few beatings along the way here at your expense. Yeah, there

90
00:08:31.300 --> 00:08:41.875
are a lot of beatings till Morale approved. And now you're very excited about Noster as well. And it was kind of the exact opposite with AI, right? With AI stuff, I saw a lot of hype. I saw a lot of

91
00:08:42.115 --> 00:08:43.795
big tech questionable

92
00:08:46.435 --> 00:08:52.420
stuff going on on the big tech side, whether it's Anthropic or OpenAI or Google or Elon.

93
00:08:52.420 --> 00:08:54.100
And it didn't really click to me

94
00:08:54.420 --> 00:09:00.180
how powerful this is as an open source tool in the Freedom Tech stack until

95
00:09:00.340 --> 00:09:06.584
I had my OpenCLO moment like three weeks ago. And now I feel like I'm playing catch up to a degree,

96
00:09:07.065 --> 00:09:15.384
but it very much feels like the glue that brings the entire stack together and that everything is very complimentary with it.

97
00:09:16.760 --> 00:09:24.520
It's wild it's wild to see it in practice, like to actually use it. It's definitely rough around the edges, but to actually use it in practice is wild.

98
00:09:25.000 --> 00:09:36.535
Yeah. I mean, one one way someone put it when we were talking that I liked where it's like, you know, sometimes Freedom Tech is kinda hard to use. But if it's kind of like your agent using it on your behalf and it's like a delegated relationship,

99
00:09:36.775 --> 00:09:40.935
like as long as you can kinda secure and make that relationship with your agent

100
00:09:41.255 --> 00:09:44.615
something that you control and is self sovereign, which is still up in the air,

101
00:09:45.520 --> 00:09:50.640
then all of sudden, a lot of these Freedom Tech tools, like communicating over an Oster or using Bitcoin or using Lightning,

102
00:09:51.280 --> 00:10:03.395
could become a lot easier. Right? So, like, kind of the onboarding problem could be partially could be really improved here, think, in in many ways if if we can if we have agents to help us. I mean, that's one of the interesting things. I just think about my

103
00:10:03.635 --> 00:10:06.435
flow here of I mean, part of the reason I

104
00:10:06.834 --> 00:10:14.620
got into this, I mean, most of the reason I got into this is that, like, you know, about a year ago, Alex Gladstein reached out to me and was like, Hey, would you like to help out with,

105
00:10:15.020 --> 00:10:33.955
you know, be involved in the AI program? And initially, was like, you know, no. I'm taking a break. Not gonna do it. He's well, what what about just, a little bit of advising? Right? And he's like, I'm like, you know, I don't know anything about AI. Right, Alex? And he's like, you know, I I think you'll figure it out. And so I'm like, okay. But this is you're you're this is you're you're taking a shot in the dark here. Like, I'm not I'm not promising you anything.

106
00:10:34.275 --> 00:10:54.580
But he was right, and I I kinda got into it. So I I worked a little bit on the side with them, just trying to help kinda activists figure out how to use AI and a grant program and stuff like that. And I just observed, and so a lot of the last year has just been me playing around with these AI tools. Like, I probably five coded, like, 400, 500 projects this last year. Like, I built an operating system. I built a

107
00:10:54.980 --> 00:11:00.865
browser, like a web browser, I built all kinds of like crazy things, but I didn't try to ship them. Just tried, I was just and

108
00:11:02.305 --> 00:11:04.225
so that gave me kind of a unique

109
00:11:05.024 --> 00:11:07.185
experience just kinda seeing,

110
00:11:07.839 --> 00:11:16.000
doing that, and so like some of the things that happened that I think are relevant to like Bitcoin and Dream Tech is that one thing is I always hated self hosting,

111
00:11:16.320 --> 00:11:21.920
and this is something like we believe in and talk a lot about, like what is running a node? Running a node is self hosting fundamentally,

112
00:11:22.125 --> 00:11:36.205
right? Like, you need to run the software yourself to be a self sovereign Bitcoin user straight up. Like, no there's no you know, if it's not your node, you don't know whether you own Bitcoin. If it's not your keys, it's not your Bitcoin on, like, the root fundamental property rights level.

113
00:11:38.610 --> 00:11:43.570
We've always struggled with self hosting because it's hard. Right? I've

114
00:11:43.570 --> 00:12:00.295
always hated it. Like, I hated running a node. I hated running I could do a Bitcoin node, but a Lightning node is just beyond me. I hated running, like, tried doing these, like, self hosted photo apps and stuff on, like, a server and I just couldn't I just I can't do it. And that changed in, August.

115
00:12:00.295 --> 00:12:07.950
So I I did a sovereign engineering, GG and Pablo's program in Maduro. Did it twice last last year. The first time I I kinda was

116
00:12:08.670 --> 00:12:34.215
not really with the program and the second time I kinda got with the program a little more, and so on the second time I ended up setting up a a server and I ended up like running Bitcoin Core, I set up a Lightning node, I set up an Oster Relay, I set up a Photos app, I set up all these things and I had a much more like self sovereign relationship with the technology I was using and I've continued that. Right? So I have so I think that's like one thing I've noticed in the last like six months that I think is really relevant to that I just wanna give like an example

117
00:12:34.670 --> 00:12:37.390
is like self hosting. I think this is gonna have a transformative

118
00:12:37.390 --> 00:12:38.990
impact on how Bitcoiners

119
00:12:38.990 --> 00:12:48.585
use our tools. Right? Like where it used to be hard to run a Lightning node or it used to be hard to have your own Noster Relay or many of these other things, it's going to get much much easier here soon.

120
00:12:50.505 --> 00:12:59.145
Yeah. That's the crazy part. Right? It fixes a lot. I mean, people have spent a decade trying to figure out how to make coin control easy for users.

121
00:12:59.640 --> 00:13:18.325
Well, it turns out, you're just gonna tell your bot that you care more about privacy than you care about cost savings, and it's gonna construct the transaction for you, then you'll sign it. Or if it's a hot wallet, the bot will just go ahead and sign it and broadcast it. And there's a million different scenarios like that across the Freedom Tech stack. I mean, you're talking about photos. Right? I have young children.

122
00:13:18.964 --> 00:13:30.325
I have two priorities with photos. First of all, I want to maintain privacy. And second of all, which might be even more important is I don't wanna fucking lose the photos. My wife would kill me if I lost the photos. And

123
00:13:30.660 --> 00:13:34.420
I have a very complicated self hosted photo backup setup,

124
00:13:34.740 --> 00:13:46.555
it breaks sometimes, I have to spend hours trying to figure out what's wrong with it. In the future, I'd be able to just tell my bot, figure out what's wrong with it and fix it. Make sure it doesn't happen again. It just does it. It makes the UX way easier.

125
00:13:46.955 --> 00:13:49.115
So I kind of want to talk about

126
00:13:51.675 --> 00:13:52.875
the warring.

127
00:13:53.035 --> 00:14:10.610
It's like everything else that's interesting in the world, And Bitcoin is a perfect example of this. A powerful tool that can be used in many different ways and has different implications. I mean, AI is very similar to that. And I mean, maybe I'm biased. I just look at everything from a Bitcoin lens. But that's how I'm looking at the current dynamic with Bitcoin and with AI stuff.

128
00:14:11.155 --> 00:14:17.875
Yeah. And as a good starting point is what happened over the last day, which is OpenClaw getting bought by OpenAI.

129
00:14:18.275 --> 00:14:20.275
Is your opinion on that

130
00:14:20.995 --> 00:14:22.035
development?

131
00:14:22.275 --> 00:14:23.395
Does it change anything?

132
00:14:25.000 --> 00:14:26.040
Mean, my my big thing I is I

133
00:14:26.920 --> 00:14:39.955
totally don't care. Like, that's the main thing. Because it like, we don't if if if I mean, in some sense, right, like, it it's it doesn't matter to me because, like, the idea is out there. There's a whole community of people. If the OpenClaw project disappears tomorrow or

134
00:14:40.995 --> 00:14:45.475
not gonna be much of a thing, right, you're already seeing, I think it was Kimi,

135
00:14:45.635 --> 00:15:00.110
a Chinese one, and then Manas, which was started as a Chinese agent and then is bought by Facebook. They both, like, basically integrated these types of features into their hosted thing, which is you probably don't wanna be using, right? You're leaking all your data to these third parties. So this is gonna happen regardless of whether OpenAI

136
00:15:00.110 --> 00:15:05.870
does it, and they're gonna do it regardless of if they hire the guy who created OpenCLO or not.

137
00:15:06.350 --> 00:15:07.150
All gonna happen,

138
00:15:07.755 --> 00:15:17.355
And so on the one hand, there's going to be really easy to use hosted versions where you're leaking all your data. Some of them are gonna be Chinese companies. Some of them are gonna be like direct honeypots, just funneling

139
00:15:17.355 --> 00:15:39.714
funneling your API keys and everything to whatever parties would be interested there. And that's gonna happen. And it's terrifying. Like, this is I mean, like, a funny one is, like, when Alex asked me to be involved in this program, I I Googled his I went up to Twitter and, like, just searched AI, and he'll he'll admit this, and it was like, all the tweets were about how, like, Bitcoin is for freedom and AI is for tyranny, basically. Right? And that was kind of the case

140
00:15:40.355 --> 00:15:46.435
in the past, because I think that's how we all saw it, and that AI was kind of a threat. It's like this UBI

141
00:15:46.755 --> 00:15:48.355
control vector. Right?

142
00:15:49.139 --> 00:15:54.260
But I think what we were neglecting is how individuals can use all this, right? That you can use this to

143
00:15:55.139 --> 00:15:56.339
be a lot more

144
00:15:57.060 --> 00:15:58.980
take a lot more you

145
00:15:59.139 --> 00:16:03.860
can have a lot more leverage in pursuing your ideas. Like, if you have an idea, if you have something that you wanna do,

146
00:16:04.605 --> 00:16:06.685
you can be like 10 or a 100 times

147
00:16:06.925 --> 00:16:16.605
more effective now in certain areas, and that's a huge deal for Bitcoin because like, one way I like to think about Bitcoin and AI is like, you know, people always like to talk about all the agents are gonna use Bitcoin, all this stuff, I don't know.

148
00:16:17.510 --> 00:16:20.790
What's really interesting is like Bitcoin is software to find money fundamentally,

149
00:16:21.110 --> 00:16:36.795
and Nostra is software to find communications at some level, right? It's like pure software. You don't ever have to get an API key. It's only software, and so we're, you know, there's some social consensus involved, right? But in a lot of areas, like all the work I did on FettyMint, we were always bottlenecked by

150
00:16:37.195 --> 00:17:10.005
how well we could write code or how good our ideas were, but the writing code was really, really, really hard, and so if the cost of writing code goes down, if it gets much, much easier to write code, all these ideas that we have, that we talk about, we're going to be able to deliver them much more than we have in the past, right? Like we're gonna be able to get Nostra clients that are as fast and as usable and as intuitive as like X, right? And I don't think some of these big central parties are going to benefit from as much, right? Like FinTech companies are not bottlenecked by the quality of their code. They're bottlenecked by something else entirely. So I think this is a big thing

151
00:17:11.525 --> 00:17:19.500
in this is like one of the things I'm really excited about is I think we're going to be able to, you know, there's like 500 of us or a thousand of us who have been working on these things,

152
00:17:19.820 --> 00:17:23.340
and I think the output we have could dramatically

153
00:17:23.340 --> 00:17:24.780
increase. And also,

154
00:17:25.740 --> 00:17:44.665
you're not gonna have like a thousand people working on these things. You're gonna have like, all the listeners can now participate. Right? Like, I was just going back and forth with a guy who's not a programmer, who's building like this really, really advanced video player for he's a product manager at Domus. Right? Like, he's not a programmer, but now he's like building the like a really, really impressive video player. Who's this LSAT?

155
00:17:44.940 --> 00:17:58.965
LSAT. Right? So it's like, he's he's a product manager. He's not a coder, and he was now building a video player idea for like really, really high performance video that it would have taken me like six months to build this, and I'm a pretty good Rust programmer.

156
00:17:59.285 --> 00:18:02.885
So like that's so it's kinda two things. Like people like me who are participating

157
00:18:02.885 --> 00:18:15.070
are going to participate at a we're gonna be much have much more leverage, but there's gonna be a whole new influx of people who were not able to take control of their tools and to modify them and to improve them and to implement their ideas

158
00:18:15.390 --> 00:18:33.364
that now will be able to. I think that's really cool. Like, I I got into Bitcoin by teaching a class about it. Like, I taught like a programming class. I just started streaming about it, right? And my thesis at the time was that I'd be teaching Bitcoin to programmers and the opposite happened. I was teaching programming to Bitcoiners pretty soon, right? Because there's so many people who really care about this,

159
00:18:33.765 --> 00:18:36.804
who not necessarily be they just wanna understand more,

160
00:18:37.390 --> 00:18:56.374
and I think some of these tools are gonna totally change that, right? If you wanna understand how Bitcoin Core works, you have like a world class tutor. If you learn how to use Clog Coder, one of these tools, you download the Bitcoin software and you just start asking questions, And you can get a better tutorial that way than you could have got if you spent like six months with Chaincode a couple years ago, in my opinion.

161
00:18:59.255 --> 00:19:02.214
Okay. Well, there's a lot to unpack there. I just wanna just

162
00:19:03.660 --> 00:19:07.260
the that that the key the key here is

163
00:19:07.500 --> 00:19:12.780
when you start using these tools, you realize how much better they work with open protocols.

164
00:19:12.780 --> 00:19:16.380
And I it's a beautiful, positive feedback loop because

165
00:19:16.855 --> 00:19:33.139
the person using these tools benefits from using open protocols because there's not closed APIs. It doesn't require permission. There's way less friction for them to use it. And then the open protocols themselves will benefit by more people using them and more people building on top of them. And you have this beautiful positive feedback loop that should

166
00:19:33.380 --> 00:19:34.580
accelerate

167
00:19:34.580 --> 00:19:39.940
the freedom tech movement in a lot of ways on a macro side. Now on the user side, I really

168
00:19:40.100 --> 00:19:44.499
want to talk to you and get your opinion on this idea of

169
00:19:44.740 --> 00:19:46.820
freedom AI

170
00:19:45.885 --> 00:19:49.485
tools will probably be both freedom tech and slave tech. Exactly.

171
00:19:50.045 --> 00:19:51.485
What I'm

172
00:19:52.605 --> 00:19:59.805
seeing right now is there's two things stopping people from using it in a freedom focused way. One is convenience and one is cost.

173
00:20:00.480 --> 00:20:02.240
OpenCLaw to me was

174
00:20:03.120 --> 00:20:04.720
the unlock on

175
00:20:05.280 --> 00:20:08.640
convenience. I actually think because this is very rudimentary

176
00:20:08.640 --> 00:20:17.695
and it's code, it's an open source project, but its code matters less than the concept behind it. This idea of a local agent that is relatively efficient,

177
00:20:17.775 --> 00:20:24.014
is cheap to run, that you can run on a old computer or something and self host yourself and hold your data there is, it

178
00:20:24.735 --> 00:20:28.735
has solved a lot of the convenience problem, I think, and it can be very convenient.

179
00:20:29.280 --> 00:20:30.799
But the cost piece

180
00:20:31.120 --> 00:20:36.160
is just insane. I mean, if you want to use these things in the most private way,

181
00:20:36.800 --> 00:20:46.195
you are self hosting a model. And so, you're probably self hosting an open source model, which is going be a Chinese model, but at least you're hosting it yourself. So, it'll have CCP biases.

182
00:20:46.275 --> 00:20:48.595
And so, if you're looking at something like Kimi,

183
00:20:48.755 --> 00:20:53.715
which is I guess considered one of the best open source models right now, it's out of the CCP.

184
00:20:54.355 --> 00:20:56.835
You're gonna spend like $10,000

185
00:20:57.100 --> 00:20:58.139
$15,000

186
00:20:58.139 --> 00:21:00.059
on a machine to run it locally.

187
00:21:00.940 --> 00:21:05.499
So that's kind of absurd. And then if you wanna do like the next best thing,

188
00:21:06.139 --> 00:21:10.460
which is still hit hosted models, but not give them KYC information,

189
00:21:11.205 --> 00:21:14.085
you're going to be paying per token usage basically.

190
00:21:14.085 --> 00:21:24.565
And then that gets ridiculous. Like I've gone down this rabbit hole for three weeks and I don't pretend I was the most efficient. I made a lot of expensive mistakes while I was learning, but I think I spent like $1,200

191
00:21:24.140 --> 00:21:25.020
$1,300

192
00:21:25.020 --> 00:21:31.980
on tokens. And when I asked Marty how he deals with it, he said he's on the $200 Anthropic Unlimited plan.

193
00:21:32.060 --> 00:21:43.285
And I went to sign up for the Anthropic Unlimited plan and they only accept KYC phone numbers. And they know they're gonna get your credit card billing information too. So the only way is like you either pay $1,300

194
00:21:43.285 --> 00:21:45.045
or you pay $200

195
00:21:45.045 --> 00:21:49.285
But if you pay $200 it's because they're tagging everything you do with your identity. Right?

196
00:21:50.890 --> 00:21:54.249
So I'm kind of curious how you view that. Do you think

197
00:21:54.970 --> 00:21:56.809
we will see that

198
00:21:59.290 --> 00:22:00.649
that cost,

199
00:22:00.650 --> 00:22:10.085
I think cost goes down across the board, but the cost variance between using it in a freedom focused way versus using it in a controlled

200
00:22:10.085 --> 00:22:11.284
surveilled way

201
00:22:11.525 --> 00:22:12.804
will compress.

202
00:22:13.525 --> 00:22:20.645
You know what I mean? Yeah, that's a great question. I mean, so one way to look at this is that

203
00:22:20.080 --> 00:22:22.719
we're kind of getting close to the point now where

204
00:22:23.840 --> 00:22:26.399
people are gonna get priced out of using the

205
00:22:29.600 --> 00:22:32.159
leading edge, like, Anthropic

206
00:22:32.320 --> 00:22:35.775
KY Seedway, Right? So, like, they just they just announced

207
00:22:35.775 --> 00:23:15.565
these fast models that are two and a half two or two and a two or two and a half times as fast, but six times more expensive on Anthrop. Like, and Codex just did a similar thing. And so, like, if you use those, it's gonna it's getting to the point where it's like it's thousands of dollars a month if you're like a hardcore programmer to like use the best tools, and that's only gonna increase. So I think, like, I mean, short of like state regulation that just doesn't allow individuals to use the best thing, but people are gonna start getting priced out. So like computer programming is going from, like, a very capital in an not not capital intensive thing where, like, all you needed was a computer to get started to something that's gonna be more capital intensive. Right? Like, you're gonna have, like, a pretty big token budget to participate at the frontier. So that's one way to think about it. It's like, well,

208
00:23:16.799 --> 00:23:20.720
people are gonna start getting priced out of the the best thing, and now

209
00:23:21.120 --> 00:23:26.240
local models are starting to unlock some use cases, like this this latest, like, Kimi 2.5 and GLM five

210
00:23:27.934 --> 00:23:51.690
can run OpenClaw. Like, they can do the OpenClaw things. Right? They can book a haircut for you, kind of. Like, they can kinda figure it out. So, like, you're you're unlocking a whole new set of capabilities with these local models if you pay a lot of money. So that's, like, one one way one way way I think of it is like we're gonna start getting priced out of the Frontier thing. Another way to think about it is that there's a whole new class of capabilities that weren't possible on the Frontier models two months ago, which are now possible on the local ones.

211
00:23:52.010 --> 00:24:00.455
Another thing is like there's a lot of people who are starting I mean, all of the Apple products now have, like, six week wait week wait times. If you want that $10,000

212
00:24:00.455 --> 00:24:12.720
Mac Studio with 512 gigs of RAM, you can't even buy it. So that shows that a lot of people are starting to make this calculation and think that it's it's worth doing, and I think that's only gonna continue. Another way I think about it is I just as we were talking, I I pulled up and checked

213
00:24:13.120 --> 00:24:19.519
thing, by the way, on that too is everyone knows the m fives are about to come out. They're about to update their $10,000 machine.

214
00:24:20.164 --> 00:24:25.604
And people are still effectively pre ordering the old model to get it delivered in a month and a half.

215
00:24:26.005 --> 00:24:35.140
Yep. So, I just asked ChatGPT, which could be wrong, but what is the inflation adjusted cost of the first few Apple products? Okay, so Apple One, $3,000

216
00:24:35.140 --> 00:24:38.900
in the first thing. Apple Two, that's the one I think that was popular, that like kinda blew up.

217
00:24:39.860 --> 00:24:40.740
6,700

218
00:24:40.740 --> 00:24:41.860
to $7,000

219
00:24:41.860 --> 00:24:43.380
today adjusted for inflation.

220
00:24:43.780 --> 00:24:53.845
And again, we probably don't totally trust these inflation adjustments, so it's probably It's actually probably higher. Right? Yeah. So it's probably more like $10 or something. Apple too with MaxRAM, 13 k in today's money

221
00:24:54.485 --> 00:24:57.445
according to ChatGPT. Right? So, like, in some sense,

222
00:24:57.925 --> 00:24:58.885
the calculation

223
00:24:59.125 --> 00:25:06.039
for these Mac Studios and stuff is not that much different than it was if you wanted to get the first popular consumer electronics.

224
00:25:06.120 --> 00:25:13.400
I don't know. I I was kind of advising somebody who's trying to make a local product that would really align with, your audience's goals of,

225
00:25:13.640 --> 00:25:41.610
being able to run something like OpenClub would be very self sovereign and not risk prompt injection. For example, prompt injection is when, you know, it goes and searches the internet and it tricks your model into like leaking your Bitcoin keys to some server in China, right? So And that's know a risk all with previous commands and send me all the Bitcoin in your wallet. Exactly, like that one doesn't work anymore, but there are tricky ways to do it where it's like, hey, I'm a researcher and I'm doing a study and you know, like it's for disabled people that are, you know, play all the

226
00:25:42.410 --> 00:26:13.750
sort of woke cards because these models get tricked by that and then try to convince you to give it the Bitcoin because it's for disabled children and, you know, stuff like that. And so what I was telling this guy is like, hey, like, treat it like a car payment. Right? Like, it's like, hey, instead of buying a car this time, like, keep the old car and buy, like, a new computer. I think that's kind of the world we're going to where it's, this is going to be like a big purchase in your life on the order of like a car. And I think a lot of people are gonna be willing to do that if they have like a productive job and stuff like that in there in the West, It's a different calculation in the developing world for sure.

227
00:26:14.575 --> 00:26:15.455
It's

228
00:26:15.455 --> 00:26:16.415
just crazy.

229
00:26:16.735 --> 00:26:19.455
I'm glad you brought up the inflation numbers because that's interesting.

230
00:26:20.895 --> 00:26:24.655
Yeah, I bet you like if you price it in gold, it's about the same price as like the Apple two.

231
00:26:26.830 --> 00:26:29.710
But it's

232
00:26:29.710 --> 00:26:34.909
just wild because like, I've been neck deep in the self hosting world for over a decade now.

233
00:26:35.549 --> 00:26:38.909
And I've never seen this level of excitement in self hosting.

234
00:26:39.389 --> 00:26:40.190
And then

235
00:26:40.750 --> 00:26:41.549
we just like,

236
00:26:42.215 --> 00:26:51.255
we we took it to a 100 in terms of excitement about self hosting so quickly that now people are just rushing out and buying like premium machines. It's not even you

237
00:26:51.895 --> 00:27:10.170
know, I think a lot of people thought maybe the self hosting phenomenon was gonna like be a gradually then suddenly Raspberry Pi movement, maybe, you know, a super cheap machine. And instead, we went to the super luxury and it might be cope. But I mean, if you use these things effectively, it's like buying an employee for $10,000 upfront,

238
00:27:10.635 --> 00:27:17.355
which is not the most ridiculous calculation to make. I mean, cost a lot more than that, and they don't work 20 fourseven, $3.65.

239
00:27:17.355 --> 00:27:19.995
So, do think costs will go down.

240
00:27:20.635 --> 00:27:21.755
It's interesting,

241
00:27:21.755 --> 00:27:25.275
The question to me is that divide between

242
00:27:25.720 --> 00:27:27.160
the person who

243
00:27:27.560 --> 00:27:31.880
is using Kimi 2.5 all built in, super convenient,

244
00:27:31.880 --> 00:27:34.440
super cheap in their web browser,

245
00:27:36.120 --> 00:27:38.440
whatever the next model is or whatever,

246
00:27:38.760 --> 00:27:40.360
versus the person self hosting.

247
00:27:41.155 --> 00:27:42.034
I think,

248
00:27:42.835 --> 00:27:43.955
yeah, it's

249
00:27:46.195 --> 00:27:52.115
a question of who has much, if they have much bigger advantage, if the drop off is huge.

250
00:27:52.675 --> 00:27:56.035
And on my optimistic side, when I think about it, I'm curious if you agree.

251
00:27:56.840 --> 00:27:59.240
I mean, the cool part is, the

252
00:28:00.280 --> 00:28:01.880
current Kimi 2.5

253
00:28:01.880 --> 00:28:02.679
is

254
00:28:02.760 --> 00:28:13.144
decently usable. Now, I think if you look in the future, in two years, people are going think it's a super antiquated model. Even if costs just go down, it's gonna completely change how people use

255
00:28:13.465 --> 00:28:17.864
Yeah. Interact with the digital world. Yeah. No matter what. In three years. Even if we have no improvements.

256
00:28:18.505 --> 00:28:26.820
Three years, this could be, like, the current Kimi could be a funded the foundation model that's shipped inside an iPhone from Apple. Right? Like, that that

257
00:28:27.059 --> 00:28:39.534
could potentially I don't know. I haven't run the numbers, but, like, I I I think in three years, you could get something like that, right, as they figure out how to distill these models much smaller and the hardware inside iPhones gets better. Like, you could have something similar that's actually running on your own hardware,

258
00:28:40.415 --> 00:28:45.534
and that's really interesting. Pull I asked Chatuchibouti to index it in gold and it says Apple two is like $40

259
00:28:45.855 --> 00:28:48.495
if you indexed on gold between then and now.

260
00:28:48.975 --> 00:28:57.450
No, same with me. Yeah. So it's like, in some sense, yeah, like that that that's that's that's different. Right? That's a that's a lot.

261
00:28:58.490 --> 00:29:05.130
I think I think that I think that we're moving towards a world where, like, in two years, like, productive person in a developed country

262
00:29:05.524 --> 00:29:08.644
will have the option to have their own hardware

263
00:29:08.965 --> 00:29:11.844
that runs models that are better than what we have today,

264
00:29:12.005 --> 00:29:13.844
that can do a lot of productivity

265
00:29:14.245 --> 00:29:19.845
tasks in their life, a lot of the sort of, like, white collar work, like doing your taxes and stuff, and it's extremely private.

266
00:29:20.200 --> 00:29:21.480
Like, it's

267
00:29:21.960 --> 00:29:22.600
like

268
00:29:22.840 --> 00:29:33.799
going to confession kind of, you know? Like, it's it's, you know, maybe not per it's not perfect. Like, the priest could could tell somebody what you told them, but it's gonna be pretty it's gonna be pretty dang private,

269
00:29:33.800 --> 00:29:41.775
and I think this is something that basically everyone's gonna be able to everyone who's in, like, a developed country who has a career

270
00:29:41.935 --> 00:29:46.415
will have an option to do this, and it's just a question of whether they choose it or not. So I think that's,

271
00:29:47.055 --> 00:29:51.680
that's really exciting. Now there's many other second or third order effects that are terrifying,

272
00:29:51.680 --> 00:29:54.799
but I think that's that's that's pretty amazing and pretty empowering.

273
00:29:56.480 --> 00:29:58.400
Well, let me flip the question on its head.

274
00:30:02.315 --> 00:30:05.434
You if if if the question is

275
00:30:07.355 --> 00:30:14.715
if the if the question is you wanna be on the cutting edge and you don't care about price, like you shouldn't, you probably shouldn't be self hosting right now. Right?

276
00:30:15.440 --> 00:30:33.945
You should just be hitting whatever the best model is at any given time and just pay per token and just I mean, honestly, that's what I do for almost everything. Like, there's there's classes of things I won't put into these models. Right? Like, I always when I'm ever doing a workshop, I will show people my chat GPT history cause I don't care. Like, it's all stuff that could be in the newspaper, but I won't show anybody my main because you're already sending it to

277
00:30:34.345 --> 00:30:51.750
Sam Altman, so you might as well. Yeah, and most of my stuff is just like, it's a different Google search and it's like coding or like engineering related and I don't really care. I mean, the one the one thing that's a little uncomfortable is that you are sort of providing this training data to these big companies and in a perfect world, this would all be out in the public and so you are like,

278
00:30:51.990 --> 00:30:54.550
that's my biggest hang up on it is that I'm

279
00:30:55.045 --> 00:31:13.179
helping the leaders maintain a lead where I wish I mean, I wish we had some kind of protocol whereas we use these things, we just, like, upload the whole coding transcript history to an author or something to, like and, like, all these I mean, ironically, Chinese models would have the opportunity to train on it as well, so you're not trying to help a leader you're not kind of helping a leader

280
00:31:13.500 --> 00:31:19.500
maintain their lead, but that's another thing that's happened in the last year is like, you know, a year ago, there was a lot of talk about the leader

281
00:31:19.660 --> 00:31:21.820
in the AI rights, which was OpenAI,

282
00:31:21.900 --> 00:31:49.080
seemed like they would continue to grow a bigger and bigger lead, right? You'd have like a runaway winner, And kind of the opposite has happened. Right? Like, you know, we have now where it's like a month after the Frontier models released, you have these kinda Chinese knockoffs that are almost as good, not quite as good, but almost as good. So a lot of the moats, a lot of the technical moats And 15 x cheaper too. 15 x cheaper. Yeah. Exactly. So a lot of the tech so in some ways better. Right? So a lot of the technical moats

283
00:31:49.240 --> 00:31:50.999
aren't very strong

284
00:31:51.480 --> 00:32:19.559
in terms of for models. Right? Which to me is very exciting. Like, always wanna live in a world where you have a lot of competition for important things, right? Like that's when things go bad is when when, you know, the cost of a new entrant goes through the roof, right? Like that's the problem with like healthcare in America, for example. It's like, it's too hard to create a health insurance company. It's like basically impossible. There's not really even a price on it, right? And so in AI, it's kind of the opposite. It's more like TVs where it's like, it's very easy to get started, and it's just like a brutal cutthroat

285
00:32:20.054 --> 00:32:20.934
competition,

286
00:32:21.414 --> 00:32:27.654
and that's really exciting. I mean, we'll see what kind of regulations come in. Like, this could totally, totally change, right? I just saw,

287
00:32:27.975 --> 00:32:32.455
you know, stuff today where like the, you know, the war department had a thing with anthropic,

288
00:32:32.455 --> 00:32:39.369
and now they might cut ties because they're not letting them do what they want, right? So you could you could start this could start to really change as

289
00:32:39.450 --> 00:32:41.929
Well, supposedly they use Claude

290
00:32:41.929 --> 00:32:44.089
to To get Maduro.

291
00:32:44.330 --> 00:32:46.250
Hey, grab Maduro, make no mistakes.

292
00:32:48.315 --> 00:32:54.075
Fucking crazy. I mean, there's so many like second order, third order effects here that it's hard to comprehend.

293
00:32:54.235 --> 00:33:13.119
Yeah. So it's one of those, you know, that famous inflation graph where it's like the stuff that gets more and more expensive is stuff like education and healthcare that is very where it's very bureaucratic and government run, and then the stuff that, you know, just collapses in prices like TVs, where it's just brutal cutthroat competition. AI has been like TVs so far,

294
00:33:13.519 --> 00:33:20.225
but and I hope it stays that way, but, you know, in the geopolitical competition that's gonna start here, it could transition

295
00:33:20.225 --> 00:33:42.450
to be more like healthcare, and that's gonna be that could be very unfortunate, right, because then we're gonna kinda get locked out of you know, the the progress could could slow, and we're gonna get locked we could get locked out of a lot of the stuff, right? Like, some of the best capabilities on, like, cybersecurity now are no longer available in, like, ChatGPT, for example. Like, Peter Steinberger, the guy who made OpenCloud, just announced that he got access to this tool called Aardvark, which is, like, ChatGPT's

296
00:33:42.450 --> 00:33:43.330
or OpenAI's,

297
00:33:44.105 --> 00:33:50.345
like, security tool, but they don't give that out because, you know, this thing's really good at hacking. So they're not just gonna let, like, any

298
00:33:50.905 --> 00:34:00.560
person have this because you can hack websites, you can hack hack apps, you can have, you know, get into government databases and stuff. So, like, we're starting to see a world where the best capabilities are hidden from individuals,

299
00:34:00.560 --> 00:34:02.480
which is, you know, I think,

300
00:34:02.800 --> 00:34:04.960
some ways unfortunate, but in other ways inevitable.

301
00:34:05.920 --> 00:34:11.840
Well, part of what's crazy about this timeline is the Chinese Communist Party are open source heroes,

302
00:34:12.595 --> 00:34:14.675
And you can follow the incentives.

303
00:34:14.755 --> 00:34:16.835
It makes sense why they're doing it.

304
00:34:17.315 --> 00:34:17.875
Yeah.

305
00:34:18.355 --> 00:34:22.835
For two reasons. First of all, they're able to basically train

306
00:34:22.835 --> 00:34:28.970
their own biases into the model. So you can the example that's always used is if you ask like Kimmy,

307
00:34:29.130 --> 00:34:41.225
what happened at Tiananmen Square? They'll be like, oh, like, everyone was supporting the CCP, and it was just wonderful. And there was no protests, whatever. Yeah. So that's one piece. They know that piece. The second piece is they know

308
00:34:41.465 --> 00:34:43.385
a decent amount of people

309
00:34:43.705 --> 00:34:56.250
will run their hosted more convenient options and just hit, you know, the main Kimi server and then they can control and surveil what you use. They know way less people will use it if it's not open source,

310
00:34:56.410 --> 00:34:58.330
because no one trusts the CCP.

311
00:35:00.010 --> 00:35:06.970
It's a little bit of an existential risk. Like that was kind of a gift that was handed to us in terms of open source AI stuff.

312
00:35:08.664 --> 00:35:13.865
Do you have any opinions on if this is something that we can really count on to continue?

313
00:35:13.865 --> 00:35:14.744
Like, that

314
00:35:14.984 --> 00:35:17.145
CCP is just gonna be releasing

315
00:35:17.224 --> 00:35:26.660
cutting edge open source models for us? Yeah. I mean, well, you know, it's it's not it it is just a little distinction. Know, it isn't necessarily CCP doing it. It's just, like, it's individual Chinese,

316
00:35:26.820 --> 00:35:41.775
like, entrepreneurs and stuff, right, which are probably not they're not all, like, direct political operatives, but in some sense, like, if you get, know, lot of these are the first model they are. Well, a lot of these first model releases when they do it, I forget some of the examples, but I think even DeepSeap, right? Like, in some of these first model releases,

317
00:35:42.015 --> 00:35:45.135
you ask them about Tiananmen Square and they tell you the truth, right? And

318
00:35:45.855 --> 00:35:48.895
because they weren't like, the the people building these things are, like,

319
00:35:49.820 --> 00:35:51.580
probably like tech idealists

320
00:35:51.580 --> 00:35:57.020
similar to us, right? But once you succeed, you don't have In China, I don't think you have the option to,

321
00:35:57.420 --> 00:35:59.420
you know, at some point you are

322
00:36:00.300 --> 00:36:02.700
forced to be a little bit operative, right? And so you

323
00:36:03.100 --> 00:36:04.380
don't have the freedom to opt out,

324
00:36:05.345 --> 00:36:23.200
and so I wouldn't, you know, that's just like a small distinction. Another, you know, and just to add on some of the things you said, like one big advantage I think they get on it, on your first point about, you know, some of the CCP kind of ideas is that a lot of the American companies end up building their products on top of these open source models because they're cheaper, because you can actually modify them. You can

325
00:36:25.119 --> 00:36:28.800
fine tune them and stuff like that in ways that you can't really fine tune OpenAI products.

326
00:36:28.880 --> 00:36:46.535
And then that gets built into like Airbnb, for example. So like Airbnb has they've talked about this, how they were using like the Quan models for a lot of their internal toolings, right? So like a lot of these stuff, these things are actually being built in, like a lot of American business is having like CCP intelligence built into it, which is kind of scary. That's one of the other

327
00:36:47.400 --> 00:36:57.480
kind of plays here. But yeah, I mean, is like a big, like this is kind of an irony from the HRF point of view is like, you know, we have like two kind of pillars of our program, right? Like one of them is like talking about how

328
00:36:57.720 --> 00:37:01.160
kind of exposing how dictatorships are using AI to control people,

329
00:37:01.845 --> 00:37:20.440
and the, know, China's like a big example there. And then the other one is the power of open source tools and open models and stuff like that. And so on one, like, the Chinese are the villain and the other, it's they're the hero. Right? And the villains are more of the American companies. So it's like a, I mean, we're just honest and open about it. It's like a big, it's a contradictory, it's kind of a contradiction.

330
00:37:20.520 --> 00:37:21.960
And think that,

331
00:37:22.520 --> 00:37:34.565
and one of the tough things is that like, you know, you can't really expect OpenAI and Anthropic and these companies necessarily to do the open source models because they're the leaders and they have this huge capital structure and stuff. They like kind of have to try to lock it down,

332
00:37:34.965 --> 00:37:46.140
but I think given a few more years, you know, you're gonna have like classes of Stanford students and stuff like that, or just like, let's just start a lab, and just like 20 of us, and like, you know, I think we'll see more open source focused

333
00:37:46.300 --> 00:37:54.860
AI labs. Like, you're seeing this in other things other than LLMs, right? There's all kinds of like cool open source models that are from

334
00:37:54.860 --> 00:38:02.244
American companies that target like more niche use cases. So it's like, it is something that's happening, but it's in the background, not American, but European too. The

335
00:38:02.885 --> 00:38:07.605
Mistral has made some good models as well. That's like French, I think. The French is French. Yeah. Yeah.

336
00:38:08.005 --> 00:38:12.805
So it is. I do think there's going to be, it's not gonna be like Chinese only forever and,

337
00:38:14.200 --> 00:38:16.360
but I don't know. I don't have a I don't have, like,

338
00:38:16.760 --> 00:38:32.755
I'm not good at forecasting and seeing the future either, so I don't know if my my words you know, I I have I have no I claim no ability to see the future here, but I do think we're gonna start seeing American LMs that are really good in the next couple years, and I do think that we're gonna see, you know, we're gonna see Chinese Yeah, companies closing down

339
00:38:33.714 --> 00:38:34.195
yeah.

340
00:38:34.914 --> 00:38:41.490
We're gonna, yeah, we're gonna see, and for example, ChatGPT has made a commitment to do this to some extent too, right? They released their

341
00:38:41.730 --> 00:38:45.250
GPT OSS models, which were good at the time. They're now completely obsolete,

342
00:38:45.410 --> 00:38:46.290
know, Yeah, three months

343
00:38:47.570 --> 00:39:05.515
yeah, yeah. But they were they were competitive at the time. Google has Gemma, so there is there is some commitment to do this from the American companies too, the the leading labs. So So, I don't know. Mean, it's correct me if I'm wrong is, like, mobile focused. Right? Kind of. Yeah. Yeah. Yeah. But it it you know, it's it's it's been good. It's a very competitive in benchmarks.

344
00:39:05.595 --> 00:39:06.555
You just haven't,

345
00:39:06.715 --> 00:39:19.320
you know, Google hasn't really released anything LLM too much LLM stuff in the last, like, two months, it feels like. They were in, like, January or in December, everyone's like, man, Google's, like, killing it. Now they're kind of they become, like, a slight afterthought, although they did just release a,

346
00:39:19.720 --> 00:39:23.559
like, a really good, like, deep think model that just crushed all the benchmarks.

347
00:39:24.135 --> 00:39:53.040
So that kinda hints that they're The Gemini Pro Gemini Pro three is pretty great. Yeah. Yeah. Yeah. They they they just released, like, a super high version of that that is apparently amazing, which hints usually, they, know, release the best one first, and then they follow on with like, distill it down to cheaper versions. So, you know, there will be, like, a really good gem of four or something in the next month, I bet, and yeah. So, I don't know. I mean, there's also it's so dynamic, right? Like, six months ago, Facebook was, like, the champion of open source, American open source. They were totally behind open models,

348
00:39:53.360 --> 00:39:53.760
you know,

349
00:39:54.815 --> 00:40:10.710
nothing has switched faster. Right? Like, now they're they're just they totally closed down and they're they're all their direction changed. So and, you know, the the dark horse here is Apple. Right? Like, at some point, Apple's gonna get their shit together, and it's they they like, they're perfectly positioned to build, like, some amazing $5,000

350
00:40:11.670 --> 00:40:20.630
AI focused, like, thing that goes on your computer or on your on your that's always what they do. Right? Or have in the past, if their DNA is still there, is they're they're always, like, the last mover.

351
00:40:21.350 --> 00:40:23.190
Well, that's what I think is interest like,

352
00:40:23.865 --> 00:40:25.865
that's probably the biggest thing holding

353
00:40:27.305 --> 00:40:30.905
well, I mean, there's plenty of reasons holding me back from paying $10,000

354
00:40:30.905 --> 00:40:32.265
for a fucking computer.

355
00:40:33.065 --> 00:40:33.705
But,

356
00:40:34.505 --> 00:40:38.450
like, the funny thing is, like, the Mac Studio, they accidentally

357
00:40:38.450 --> 00:40:44.130
stumbled into, Oh, we offer five twelve unified memory. It's like, Oh, wow, it works.

358
00:40:45.730 --> 00:40:48.609
They're gonna build something that's actually designed for it

359
00:40:49.185 --> 00:40:53.665
in the next year or two. Also, I think there's like tons of overthinking,

360
00:40:53.665 --> 00:40:55.905
like OpenAI hire Johnny Ive

361
00:40:55.985 --> 00:41:09.270
on like, what does interacting with AI agents look like in the future? Well, like, it probably looks something like an iPhone or AirPods. Like Yeah. Yeah. And like your watch. Like Apple Watch, Apple phone, Apple AirPods,

362
00:41:09.350 --> 00:41:24.065
that UX may still make sense. Right? You're Yeah. Mostly talking to your agents. So you're gonna need some kind of microphone. You're still gonna need to see things, and you're going to need to type things sometimes. So you're going to need a slab with a screen on it. And then you're going need some box that's cell phones at home.

363
00:41:24.464 --> 00:41:29.184
So on the quote unquote private side, but proprietary side, I think Apple is

364
00:41:30.160 --> 00:41:34.560
perfectly positioned to handle this. Oh my their phones, they make their own chips,

365
00:41:35.040 --> 00:41:43.200
like their phones. I run local models on my on like the latest gen iPhone and, you know, it works, right? Like, it's still very rudimentary.

366
00:41:44.015 --> 00:41:57.650
But it's clearly the best hardware out there. And they've made their brand on privacy stuff. So like on that side, Apple's like perfectly positioned. Then on the other side, what's really interesting to me is I really do think Google's perfectly positioned and it's But

367
00:41:58.050 --> 00:41:59.250
it's more from

368
00:41:59.490 --> 00:42:01.170
the opposite of privacy.

369
00:42:01.330 --> 00:42:04.930
But it's funny because I think a lot of people incorrectly,

370
00:42:04.930 --> 00:42:09.490
especially capital allocators incorrectly like a year ago, two years ago, saying,

371
00:42:09.650 --> 00:42:13.755
this is going to kill Google search business. But I think the big piece they missed

372
00:42:14.234 --> 00:42:14.875
was,

373
00:42:15.035 --> 00:42:16.475
and we see this with Claw,

374
00:42:16.635 --> 00:42:17.835
people rightfully

375
00:42:17.835 --> 00:42:28.470
are hesitant to put sensitive information into these things, but so many businesses already run on G Suite. Like, Google already has your calendar, has all your emails, has your documents.

376
00:42:29.910 --> 00:42:38.630
People do conference calls through them. They're already you already gate. You already made that decision. Your company made that decision five years ago that you're gonna trust Google with all your sensitive shit.

377
00:42:39.125 --> 00:42:55.170
If the sensitive shit's already there, it's way more likely for you to use their agent tooling than it is for you to export that data and trust someone else. Yeah. This is one of the ironies is like all the big tech integrations so far have been just absolute flops, you know, like like Microsoft Copilot,

378
00:42:55.170 --> 00:42:56.770
Apple Intelligence, like everything

379
00:42:57.810 --> 00:43:08.065
here has been terrible, right? Nothing has worked and I think that's probably gonna change in the next six months. They're all gonna just take the cue from OpenClaw and just be like, okay, this is what people want and that could really change.

380
00:43:08.385 --> 00:43:21.425
One of the interesting thing about OpenClaw is that you see, at least on Twitter, you see a lot of the people who are like leading it are like small business people where they have like a little agency, they have like a little thing, and they automate some of their business workflows, and so I think that's gonna Like be Marty and TFTC.

381
00:43:22.089 --> 00:43:29.530
Yeah, exactly. Like that's he's a perfect example of this. It's a lot of like small business people, like sole proprietor type people who are who are

382
00:43:30.410 --> 00:43:34.890
paving leading the way from like a user and adoption standpoint on OpenClaw,

383
00:43:35.125 --> 00:43:36.165
right, and

384
00:43:36.645 --> 00:43:55.000
and so I think that's that's kind of interesting, so I think that's probably gonna be the and and, you know, they'll they'll once it's in Google or Microsoft, they'll use it if these companies get their shit together. The other thing about Google that's interesting is, like, they're totally vertically integrated. Right? They have models. They have data centers. They have TPUs. They're they have their own architecture for the chips.

385
00:43:55.240 --> 00:44:12.155
Right? And they also have, like, so much more data. It's like I saw something from Cloudflare where it's like their Google's crawlers see, like, 3.7 times more data on the web than, like, a unaffiliated crawler does. They just, like I I don't understand exactly how it works, but it's, they just are able to see parts of the web that other people don't because of their massive crawler infrastructure,

386
00:44:12.315 --> 00:44:15.675
and so they they have more of that that digital gold for this

387
00:44:15.915 --> 00:44:17.995
data boom that's interesting. But, yeah, I don't know.

388
00:44:19.500 --> 00:44:20.940
I think, yeah, I think it's

389
00:44:21.980 --> 00:44:23.260
very interesting and I have no

390
00:44:25.820 --> 00:44:31.980
idea where it's going, but I think in the next year, it's going to be something that like fundamentally transforms

391
00:44:31.515 --> 00:44:40.154
many, many things. Like I was saying to Marty, I think that, you know, the next, like, couple years, we could have, like, almost like a it could feel like five the next five years, it could feel like a century happens, right? Like, we're just

392
00:44:40.714 --> 00:44:49.720
the way we live our lives could be very different in five years, and I didn't think this even, like, a year ago. I've just been really convinced in the last six months. Like, saw it first for myself with like vibe coding where I could

393
00:44:50.760 --> 00:45:00.040
like, I felt like me by myself became much more effective than if I had a team of 10 people two years ago, right, or 15 people. Right? So it's going to be be wild.

394
00:45:01.745 --> 00:45:04.225
Yeah. Mean, it's just accelerating so quickly.

395
00:45:05.105 --> 00:45:07.185
Yeah. What is your what is on

396
00:45:11.585 --> 00:45:14.460
what is your opinion on, like

397
00:45:16.460 --> 00:45:17.020
the

398
00:45:17.260 --> 00:45:19.020
concerns over job loss?

399
00:45:21.020 --> 00:45:22.300
I

400
00:45:25.660 --> 00:45:26.620
think it's going to be

401
00:45:27.575 --> 00:45:29.015
incredibly disruptive.

402
00:45:29.095 --> 00:45:44.450
Like, I I don't know. I don't know. I'm kind of agnostic on whether we ever get something that totally replaces us. It's just I can't reason about it. But even, you know, even if we have even if these things get better at literally everything we do, it still probably is the case that it's better if I'm piloting it. Like,

403
00:45:44.770 --> 00:45:54.609
even if the AI is, like, super intelligent and better than me in every way, me working with the thing will probably still be better than the thing in itself. Like, you still can probably add something. So it's going to have to be, like, really, really way That far

404
00:45:55.410 --> 00:46:00.625
person doesn't get fired. It's the 10 people 20 people that work underneath him that gets fired.

405
00:46:00.945 --> 00:46:02.385
Yeah. So I don't know.

406
00:46:03.025 --> 00:46:05.105
I do I do think that

407
00:46:06.305 --> 00:46:07.585
I do think that

408
00:46:07.825 --> 00:46:09.425
it's it's

409
00:46:10.230 --> 00:46:13.910
I do think that I do think that lots of people are gonna lose their jobs.

410
00:46:14.470 --> 00:46:27.745
I do think that there's just gonna be massive, massive layoffs and, like, at least if market forces I don't I don't think market forces are going to be able to operate during this whole time, right? Like, if market forces operate, like, 90% of people are gonna get, laid off

411
00:46:28.545 --> 00:46:30.705
or something, like, over the next couple years,

412
00:46:30.945 --> 00:46:36.625
and, it's just gonna be incredibly disruptive. And I think I I do think there's a really good chance that we're able to like,

413
00:46:37.210 --> 00:46:50.125
new ways of working will grow that can lead to full employment again, but even if that happens, there's gonna be a really, really disruptive period in the middle, and I don't think governments will let this play out with market forces alone because it just, like, it will be two

414
00:46:51.165 --> 00:46:54.845
politicians will have a heyday with this. So that's There's kinda being one revolutions.

415
00:46:55.245 --> 00:47:03.005
Yeah. The other thing is that I think that we're going to like, like, last like, since World War II, it's like we have this more and more we have like one

416
00:47:03.570 --> 00:47:04.770
kinda global

417
00:47:05.170 --> 00:47:14.530
community, more and more, right? Where it's like Right. I don't know. Like, I remember I went to Nepal like ten years ago and like, there I was at like a monastery and I went to get like a restaurant nearby

418
00:47:14.610 --> 00:47:19.415
and the guy wanted to play me his rap mix tapes on YouTube. I was just like,

419
00:47:19.655 --> 00:47:39.830
what? You know? You know, it's like something that someone down the street would would attempt to say at my local coffee shop or so. I don't know. Right? Like, so it's we're we're we're we've all been sort of sucked into like one global culture and one global you know, we've become more and more and more similar, and the internet drove that a lot in the last like twenty years. And we're starting to see that reverse, right, where like a lot of GitHub projects,

420
00:47:39.910 --> 00:47:42.150
open source projects are now not letting people

421
00:47:42.390 --> 00:47:48.465
be just contribute like, willy nilly, because there was a proof of work, there was proof of work attached to, like, a GitHub contribution

422
00:47:48.465 --> 00:48:05.720
that could pass tests a year ago. It actually means somebody sat down and understood how things work enough to get it to work, and now that could just be someone wrote one sentence in a coding agent and Right. There's zero proof of work attached to it now, so you see, like the tip of the spear here is like big popular open source projects

423
00:48:06.039 --> 00:48:10.839
are starting to close down and enforce some level of web of trust and enforce some level of

424
00:48:12.615 --> 00:48:14.615
authentication and stuff like that

425
00:48:15.095 --> 00:48:22.775
and making it more relationship based, right? You're also seeing this on like Twitter, right? Like Twitter is like, there's just bots everywhere. It's slop.

426
00:48:23.015 --> 00:48:28.520
It's all slop. It's all like the public is kinda, in my opinion, almost like destroyed,

427
00:48:28.600 --> 00:48:34.040
right? Like you're just not going to be out, and this is one of the things that's making me much more positive on Noster. I've actually noticed this. Nice.

428
00:48:34.600 --> 00:49:29.760
I noticed this, like I, my heart rate goes up when I go on X now. Like it's just like, it feels like I'm in a war zone and it's exciting because it's like, this is the, I see the future happening and I can see what all of the AI people are doing and I can see all the cool things people are doing with Bitcoin and stuff, it's like, I don't know, kinda like, and I go into Nostrand, it feels like I'm in my local neighborhood park and it's like, there's my buddies and it's like it's not like a big it feels like a park and the public internet feels like a war zone to me a little bit. Well, you made fun of me. You made fun of Nostrand. You said it's Odell's artisanal group chat and that's the positive. Right? That turns out you were both right and wrong with that stuff. Well, so this trend has been happening, right? The last ten years, it's like so much of the actual internet is happening in group chats, right? Like, people have been opting out. Mean, that's like one of the legacies of wokeness, right? Like, lot of people of like the I mean, not just like the political side of wokeness, but the intolerance, right? Like not the take away all the ideology

429
00:49:30.000 --> 00:49:39.680
on like what is right or wrong. There's a big element of it that was intolerance for like other viewpoints, right? And so one of the side effects there is everyone migrated to group chats. You stopped saying what you thought

430
00:49:40.335 --> 00:49:42.255
on Twitter like you did in 2012,

431
00:49:42.415 --> 00:49:48.575
right? And you started saying it in group chats. So like the public internet has been carved into neighborhoods

432
00:49:48.734 --> 00:49:53.935
in the last ten years with group chats, right? And I think Nostrad to me is like a step in that direction

433
00:49:54.579 --> 00:50:00.180
of, I mean, it's like an embrace of that fact, right? And so that's why I'm excited about, I've become much more excited about Nostr.

434
00:50:00.260 --> 00:50:09.460
It's like a more local experience of the internet. It has many of the benefits of the internet, but you're not, it's going to be more resistant to just being like shredded by

435
00:50:10.215 --> 00:50:20.935
bots because we have this like, you know, web of trust kinda built into it. And I'm all that also makes me more excited about stuff like, you know, Marmot, like this group chat, you know, being able to chat over Noster, stuff like that.

436
00:50:21.815 --> 00:50:23.860
Well, mean, specifically,

437
00:50:23.860 --> 00:50:34.500
to be clear here, Noster's an open protocol. So there's gonna be tons of bots and tons of slob and But all this other the cool part is because it's an open protocol, you can use it as you as you want without permission.

438
00:50:34.820 --> 00:50:37.619
And so I agree with Justin that we'll basically see,

439
00:50:39.144 --> 00:50:42.984
you know, these smaller communities pop up that are using

440
00:50:42.984 --> 00:50:55.020
verifiable webs of trust. And then, the last piece that has been holding it all back in a lot of ways is, well, how is a user gonna manage webs of trust? Well, the bots will manage it. Like, that's the irony. The irony is

441
00:50:55.339 --> 00:50:58.780
the bots will manage your webs of trust to keep the bot swap out. Yeah.

442
00:50:59.900 --> 00:51:01.740
Which kind of a crude analogy,

443
00:51:03.260 --> 00:51:05.020
but your bot is gonna be like your Internet condom.

444
00:51:05.765 --> 00:51:11.365
Like that's what it's gonna be, you know? It's like the thing to keep you safe as you navigate

445
00:51:11.365 --> 00:51:13.685
this, you know, increasingly dirty place,

446
00:51:15.525 --> 00:51:17.125
and so, yeah, I think that is

447
00:51:17.765 --> 00:51:19.605
going to play really well

448
00:51:20.340 --> 00:51:30.260
into Nostrand, you know, the ways that decentralized platforms are gonna deal with this, they're just gonna shut stuff down, right? In a month, you're gonna have to like upload ID or do a face scan to get into this one, They're

449
00:51:30.260 --> 00:51:32.420
all gonna do this, right? And so then

450
00:51:32.705 --> 00:51:38.065
Noster's value proposition will shine further, right? It's like censorship resistant, right? Like, you can always publish,

451
00:51:38.465 --> 00:51:41.345
and you can always get your message out. And as

452
00:51:41.585 --> 00:51:55.849
your inability to do that on the traditional platforms decreases, people might start to care, because people don't, like, the in last couple of years, if you tell people this, they don't care because they can get generally can kind of get their message out on the centralized social media. And I don't see that being the case in five years.

453
00:51:57.450 --> 00:51:59.049
Yeah. It's already kind of not the case.

454
00:51:59.845 --> 00:52:00.645
It's already The not

455
00:52:02.085 --> 00:52:03.765
crazy part to me is

456
00:52:04.325 --> 00:52:07.605
the woke movement started the cancel culture

457
00:52:07.685 --> 00:52:08.965
war stuff,

458
00:52:09.125 --> 00:52:11.365
and now it's spread. Now it's like every

459
00:52:11.765 --> 00:52:12.565
ideology

460
00:52:12.565 --> 00:52:13.925
is just releasing

461
00:52:13.925 --> 00:52:16.950
lynch mobs on the Internet to fire someone that

462
00:52:17.350 --> 00:52:24.310
lives on the other side of the world. It's one thing if you want to fire your own kid's teacher, but it's a completely different thing if you're on

463
00:52:24.790 --> 00:52:25.830
social trying

464
00:52:26.230 --> 00:52:30.790
to fire a teacher that lives in a different country than you. It's just a wild concept.

465
00:52:31.365 --> 00:52:32.085
Just, it

466
00:52:33.045 --> 00:52:38.565
feels like every time you log into the internet, there's another cancel culture mob trying to stop something.

467
00:52:38.885 --> 00:52:46.405
Yep. Which is just gonna hasten it all. There was two pieces there, right? There was job loss, and then there was basically the world

468
00:52:47.019 --> 00:52:49.740
getting smaller in a lot of ways. Mhmm.

469
00:52:51.099 --> 00:52:52.700
The job loss Yeah. Let

470
00:52:53.339 --> 00:52:54.540
me think about this more. Like,

471
00:52:56.220 --> 00:52:57.500
they're kinda related,

472
00:52:57.819 --> 00:52:58.859
you know, because

473
00:52:59.305 --> 00:53:00.105
I think

474
00:53:00.905 --> 00:53:01.705
like,

475
00:53:01.705 --> 00:53:05.065
I've always been a fan. I always liked this idea of like having,

476
00:53:05.785 --> 00:53:52.620
like, I like being in a small town, like, you know, with like 5,000 people, like understanding Try this like a small town. It's yeah, exactly. It's like it's like you can't kinda f around quite as much. There's a lot more like social trust. It's like fun to go to a coffee shop where the barista knows your name and is like some kid at the high school and like, you know, you know their parents and they or whatever, you know what I mean? Like, my parents grew up in a small town and whenever I'd visit as a kid, I always thought that was it was kinda neat how you had this like fabric of kinda social trust and relationship that I missed out growing up in a big city and living in big cities my whole life, and I think what's going to happen is, you know, over the last, like, fifty year or, you know, the whole industrial revolution has been like, you know, people leaving that to pursue prosperity in these big cities, and I think

477
00:53:52.940 --> 00:53:55.820
if more people aren't able to participate

478
00:53:56.315 --> 00:54:06.235
in the global economy because you need to be, like, an expert at managing these AI agents, like, I think what you're gonna have is, like, a redistribution of the pop It's gonna redistribute

479
00:54:07.035 --> 00:54:07.915
demographically,

480
00:54:07.915 --> 00:54:12.150
right? Like, more people are going to leave and go back, move

481
00:54:12.150 --> 00:54:21.190
to the country, right? Like, that's one of the legacies of COVID. It's like, at least The US, right? A lot of the people started leaving the cities, right? I considered it for the first time. Other people considered it for the first time,

482
00:54:22.470 --> 00:54:25.945
so it's kind of already started, and so I think that's

483
00:54:26.105 --> 00:54:28.345
kind of interesting from like a decentralization,

484
00:54:28.345 --> 00:54:30.425
like the population's gonna kind of decentralize,

485
00:54:30.425 --> 00:54:43.100
and I think there's a lot of really positive things about that, especially if you remain, you know, you still have an internet like thing like Noster, where you can coordinate with people all over the world, but it's going to be, you know, as you have job loss, you're going to have a lot of people

486
00:54:43.660 --> 00:54:55.765
leave these big cities and kinda distribute more, and I think in some sense, they'll be forced to, but in another sense, you know, you can have like maybe a little bit more of a dignified life in some ways than being a

487
00:54:56.244 --> 00:54:58.085
slave to the machine in the big cities.

488
00:54:59.605 --> 00:55:00.805
Yeah, I don't know. I mean,

489
00:55:04.029 --> 00:55:07.390
so I look I just looked up. Google has a 190,000

490
00:55:07.390 --> 00:55:08.109
employees.

491
00:55:08.509 --> 00:55:09.310
Jeez.

492
00:55:09.869 --> 00:55:13.869
What are they doing? Most of those people are getting fired. This idea that

493
00:55:14.829 --> 00:55:19.725
yeah. Go on. Did you see that stat about Twitter where it was, like, their Nikita whatever

494
00:55:19.885 --> 00:55:28.765
was saying how, like, pre Elon, they had, like, 5,000 employees? And now at least the product team, don't think they, know, because they surely they have many lawyers. But, like, the product team is, like, 30 engineers,

495
00:55:29.260 --> 00:55:42.620
two designers, a product manager, and him. So there's, like, 35 people basically running the actual product, whereas previously, was, like, 5,000. So it's probably more like a 100 if you include all the lawyers and regulatory stuff. Bit Elon chimed in there. It was a little bit

496
00:55:42.860 --> 00:55:44.700
confusing because there was more

497
00:55:45.635 --> 00:55:50.835
There's a whole x AI team that he wasn't including. And like more and more x is just a

498
00:55:51.555 --> 00:55:56.195
top of funnel for x AI, both on the data training and front end usage.

499
00:55:56.515 --> 00:55:56.835
Yeah.

500
00:55:57.890 --> 00:56:03.970
I think that's kind of a canary. You monetize social via the AI stuff, which is what he's doing right now.

501
00:56:04.290 --> 00:56:05.650
Yeah. Yeah. I mean,

502
00:56:07.090 --> 00:56:13.135
Square block has, I think, 10,000 employees. It's like, what do they all do? Yeah. And

503
00:56:14.015 --> 00:56:20.255
you remember during COVID times, right? It was like the things that would trigger people were like the TikTok videos of random

504
00:56:21.375 --> 00:56:22.735
woman in management

505
00:56:23.454 --> 00:56:34.000
sitting on the beach, and she's like, This is what I do all day. Don't do anything. Yeah. Those two product manager girls who are like, explaining their jobs at I'm the not thinking about that exact thing. Yeah. Those girls are hilarious.

506
00:56:34.720 --> 00:56:38.480
They're fine. I think they were full They're not gonna have a job, and what do they do next?

507
00:56:38.960 --> 00:56:47.295
I don't like, I don't know what the path is there. Yep. And it's very much, we're very much getting gaslit by the richest man in the world,

508
00:56:47.935 --> 00:56:50.015
Elon, when he says Universal people that have

509
00:56:51.615 --> 00:56:53.775
high income is what he's telling people. This idea,

510
00:56:54.255 --> 00:56:56.415
it's like, I'm going to build

511
00:56:57.055 --> 00:56:57.935
this like

512
00:56:58.580 --> 00:56:59.860
super machine

513
00:57:00.020 --> 00:57:01.700
that's going to like print

514
00:57:02.100 --> 00:57:16.245
money if you give me control over it, and you just have to trust me that I'll share it with you. It's like, that's like the value proposition of communism basically, right? So it's like, this is the thing that we've all been skeptical about AI all along, and this is Elon's like current talking point, and it's so, like,

515
00:57:17.205 --> 00:57:19.045
utterly stupid because,

516
00:57:19.445 --> 00:57:20.005
you know,

517
00:57:21.205 --> 00:57:28.890
like, a middle class only forms if like a, you know, the people who are in charge depend on them, Right? So if you ever have a world where,

518
00:57:29.290 --> 00:57:29.930
you know,

519
00:57:30.730 --> 00:57:34.090
elites don't depend on a middle class for prosperity,

520
00:57:34.170 --> 00:57:43.265
you're not gonna have a middle class after a generation or two. Like, it'll last for a while, but that's why yeah. I mean, I I think this idea that, like, you're gonna be able to hold it up with UBI is is kind of

521
00:57:44.385 --> 00:57:46.145
it's probably not gonna work very well.

522
00:57:47.025 --> 00:57:49.505
Yeah. I don't pretend to know, like, what happens.

523
00:57:49.745 --> 00:57:51.265
I just The incentives are bad.

524
00:57:52.130 --> 00:57:55.010
The incentives are bad there. Like, you're just not going to

525
00:57:55.330 --> 00:58:04.850
like, has to be there has to be some, like, reciprocal dependence for a relationship to work. And the thing that Elon is proposing has no reciprocal dependence. Right? It's, like, purely a one way thing.

526
00:58:05.515 --> 00:58:07.275
It's a trust me, bro. You know?

527
00:58:07.994 --> 00:58:22.220
Yeah. I'm pretty, like, I'm pretty sure he knows he's lying. Unless he, like, believes that universal high income means Yeah. You'll go in a pod and you'll connect the IV to yourself and you'll live in VR and you'll have everything quote, unquote, everything you want.

528
00:58:22.700 --> 00:58:37.305
Yeah. He's an interesting guy. I don't know. I think he I think he often lies and then believes it. You know? Like, that's what I think he believes what he's saying, but it's also, like, clearly a lie. Like, there's a lot you you could find all these lying rooms self. Yeah. Like, how they were managing, you know, why this yeah. Like, there's

529
00:58:37.465 --> 00:58:44.265
there's all kinds of examples of him being, like, totally full of shit in the past, like, just just clearly lying. But I think he also believes it. You know?

530
00:58:45.625 --> 00:58:46.185
I don't know.

531
00:58:48.100 --> 00:58:53.220
Yeah. Well, I just think particularly if you're ahead of a household, you're listening to the show right now,

532
00:58:54.900 --> 00:58:56.820
you know, your family relies on you.

533
00:58:57.780 --> 00:58:58.500
Like, you just

534
00:58:59.025 --> 00:59:04.944
I don't pretend to have the answer, but you like, you gotta prepare yourself. I like their major And title is

535
00:59:06.704 --> 00:59:11.984
yes, a lot of them will be bad. I think a lot of them will be good too. I think with chaos comes opportunity.

536
00:59:11.984 --> 00:59:27.390
And I mean, just playing around with claw for the last three weeks, it does empower individuals in a massive way to those who realize the need and seize it. But if you get caught flat footed here, your family could pay the repercussions for years,

537
00:59:27.470 --> 00:59:28.030
decades,

538
00:59:28.385 --> 00:59:35.425
in my opinion. And one one one way I think about people who are, like, your audience is, like, in some sense, it's like a class of, like, arbitragers.

539
00:59:35.425 --> 00:59:59.530
Right? They're they're they're people who, like, do some kinda art, like you know, huddling Bitcoin is kind of like an ar an active arbitrage where you, like, understand something about your society, okay, this is falling apart, and I see an opportunity over there. So in some sense, like, like, if you do arbitrage, like, chaos is always good in some sense because, like, that's where opportunities happen. So there's going to be there are there are going to be many, like, opportunities here, but it's something you have to be kinda paranoid about, I think, because it is it is, like, yeah, there's just

540
01:00:00.464 --> 01:00:08.464
it's I think it's going it there's gonna be aspects about this that are gonna be really rough and most people aren't gonna be prepared, but, you know, the the sort of paranoia that you build being

541
01:00:08.464 --> 01:00:12.545
a Bitcoin hobbler for a period of time is going to become very, very, very useful, I think.

542
01:00:14.609 --> 01:00:17.090
So let's talk about I agree.

543
01:00:19.170 --> 01:00:32.375
Yeah. I mean, and and just even not only the mindset that Bitcoin breeds into people or that try like the type of people that get attracted to Bitcoin, but also the fact that if you've been holding Bitcoin for a little bit, you actually have financial flexibility.

544
01:00:32.535 --> 01:00:38.775
You have at least some level of financial independence where you can figure things out. I have normie friends that

545
01:00:39.090 --> 01:00:48.530
are working class, they make a decent amount of money, but they have no savings whatsoever. Yeah. Right? Yeah. And those people obviously are gonna have a lot more difficulty.

546
01:00:48.610 --> 01:00:49.410
Gonna be they

547
01:00:50.770 --> 01:00:55.425
have a lot less margin of error in terms of how they proceed next.

548
01:00:55.665 --> 01:01:00.705
In in certain places, they won't have a choice. Right? And if you've been saving, you will have a choice.

549
01:01:01.905 --> 01:01:04.225
I wanna talk about Marmot

550
01:01:04.625 --> 01:01:07.025
Yeah. Using Noster as a

551
01:01:07.105 --> 01:01:08.305
DM mechanism.

552
01:01:08.910 --> 01:01:14.350
Yeah. And so Marvin is the open protocol built by the white noise guys

553
01:01:14.589 --> 01:01:20.670
Yeah. Who I've had on the past on dispatch. So the freaks are relatively aware of it. I mean, I think

554
01:01:21.895 --> 01:01:24.295
the power of Noster Noster

555
01:01:24.295 --> 01:01:30.455
DMs have kind of languished for a while, high level. It's something that I've never used. They just never really worked.

556
01:01:30.855 --> 01:01:39.120
No one ever really spent the time trying to fix them. There was already such huge network effects that they were battling against. Know, WhatsApp

557
01:01:39.280 --> 01:01:48.800
basically has every living person as a user. Yep. The birth rate is the only thing that stops WhatsApp growth. I think Telegram reports like one and a half billion users on Telegram.

558
01:01:49.455 --> 01:01:55.695
Signal, which is like the project that I always point to as a the ideal successful scaling,

559
01:01:56.575 --> 01:02:13.010
freedom ish privacy focused tool, only has a 100,000,000 users, just is like a drop in the water compared to the big players, but still massively successful. I use signal every day for my businesses and my family and whatnot. I'm really grateful that project exists.

560
01:02:13.170 --> 01:02:16.210
But there I think people thought like, oh, okay, you know,

561
01:02:16.974 --> 01:02:17.775
NostradMs

562
01:02:17.775 --> 01:02:19.775
can languish, and we have these other tools.

563
01:02:20.255 --> 01:02:23.135
Then you're like setting up these agents,

564
01:02:23.935 --> 01:02:27.135
and you're like trying to connect to these centralized

565
01:02:27.694 --> 01:02:28.655
big tech

566
01:02:29.775 --> 01:02:30.494
bullshit,

567
01:02:30.880 --> 01:02:32.480
and it just doesn't work.

568
01:02:33.760 --> 01:02:53.885
It just hits a wall. It's like you hit this centralized proprietary wall that in a lot of ways historically has been designed to keep bots out in the first place, and you need your bot to connect to it in order to communicate with it and have a chat with it. And I think all of a sudden, it opened up a lot of people's eyes that oh, Nostrad should be perfect for this. Like I should just be able to give my

569
01:02:54.125 --> 01:02:58.525
bot my end hub, it should be able to message me and we should be able to initiate that

570
01:02:59.230 --> 01:03:10.670
chat exchange in, two seconds. And then from that point forward, we should have secure encrypted chats without relying on a company. And I think that clicked for you too. And as a result, you started working on it. Right?

571
01:03:11.625 --> 01:03:16.665
Yeah. I mean, that was my like, I've been watching OpenCode for the last like two months, but it was more like

572
01:03:16.905 --> 01:03:31.640
I was watching the guy like, the guy's like a world class vibe coder. Like, he taught us all what performance looks like. Had completely different workflows and used different models and everything else. That's kind of the perspective I was watching it from for the longest time. And then only like a week and a half ago did I attempt

573
01:03:31.800 --> 01:03:39.800
to try it. I'm also pretty security conscious and I can tell this thing's kind of a security disaster. If you, you know, if you plug it into email and stuff like that, you just don't wanna give it access

574
01:03:40.375 --> 01:03:55.975
to to everything but and, you know, I'm very good at using coding agents so I can do a lot of those things myself if I want to. Like, yeah, about, you know, ten days ago, tried it myself and I tried the I'm like, let's try the Nostr thing. Didn't work. The NIPO four extension that had been submitted was just completely broken. So, like, okay, that's frustrating.

575
01:03:56.500 --> 01:03:58.100
And, so I tried,

576
01:03:59.300 --> 01:04:03.620
white noise and I couldn't get that to work. So I ended up just going down the rabbit hole and,

577
01:04:04.580 --> 01:04:05.220
and

578
01:04:05.380 --> 01:04:08.420
I think I like this idea of Marmot and white noise,

579
01:04:08.660 --> 01:04:10.580
but like you said, it's hard to to

580
01:04:11.865 --> 01:04:17.705
bootstrap a network effect, but with this OpenClaw, I don't have to, right? I'm gonna communicate with my bot

581
01:04:17.865 --> 01:04:25.865
here, and I'm not gonna have a third party. I'm not gonna have a nonprofit foundation from Signal in the loop. Just ideologically, I'm gonna go straight to the source over

582
01:04:26.190 --> 01:04:57.840
channels that I can kinda control, or at least if they break, I can find another thing like Nostril Relays, right? Like it's good enough for me, and I don't have to convince Matt to download an app, right? And to me, that's kind of the thing that was really interesting. It's like, okay, so here's a good way to dog food Marmot. It doesn't have to be perfect, it just has to work for me and my bot. And and so that's when I started so I started working on this. And, you know, within a couple days, I was able to chat back and forth with my bot over Marmot, and that was kind of interesting. And and I I just kind of kept running with it. So, And yeah, over the last week, I ended up building, like, native iOS and Android apps

583
01:04:58.080 --> 01:05:17.275
to to do this that are really, really nice. And, you know, I can chat with my friends now too. I got a group with a couple people in it. The MeetMe guys have been helping me. This has been really you know, the former MeetMe guys have been helping me. Paul, Anthony, and and Ben. Some other people have become interested and started using it. And some other people have had the same idea. Right? So Derek Cross, the noster,

584
01:05:17.675 --> 01:05:30.170
you know, our our our greatest noster too, maybe second greatest behind you, Matt. You always have to be first in all things. He had the same idea at the exact same time. So he was, we just did a live stream today where we both kind of showed off our apps and tried to get them to talk to each other and we succeeded eventually.

585
01:05:30.410 --> 01:05:31.850
We had some other things like that.

586
01:05:32.570 --> 01:05:36.335
But yeah, so yeah, we kinda are starting to have

587
01:05:36.734 --> 01:05:41.215
a way of communicating with the bot over like all kind of channels that I

588
01:05:41.535 --> 01:05:51.940
control or that are self sovereign, and to me, that's kind of really exciting. So yeah, made an app called Pika. I haven't like released it yet properly, but in the next week, I'll probably do so. You can find it on my GitHub if you're interested.

589
01:05:52.579 --> 01:05:58.660
And, and it's starting to be really good. Like, we just did an audio call, like an encrypted and an encrypted audio call before this,

590
01:05:59.140 --> 01:06:00.819
before this thing. Every It was wild.

591
01:06:01.315 --> 01:06:03.555
Yeah. 50 times a second, it sends a little,

592
01:06:03.875 --> 01:06:22.310
chunk of audio that is encrypted using the Marmot keys, and so, like, we can do a group call with a group a group of people, and when we add the third this isn't possible right now, but, like, in a week, will be. So if we added a third part person, Marty, we and talked a little bit and then added Marty, like cryptographically, he wouldn't be able to decrypt what we were talking previously if he saw it, right? So we have this like amazing, built on the

593
01:06:22.550 --> 01:06:25.910
encryption primitives that Jeff and the team at White Knives have been developing.

594
01:06:26.150 --> 01:06:28.470
We're gonna be able to extend this to many other forms of communication,

595
01:06:29.105 --> 01:06:31.505
And we're also gonna be able to talk to our bot over

596
01:06:32.464 --> 01:06:49.440
encrypted chat. So like, I have a setup now where I can talk to my bot and I can barely understand what it's saying because it's still buggy and shitty, but I do get voice back and it's able to talk to me and, you know, give it another week or two, and this will be kind of as good as like the ChatGPT app where you can talk or, know, grok, whatever your girlfriend.

597
01:06:49.440 --> 01:07:01.795
I haven't tried it, but apparently this is a thing. And yeah, there's all kinds of other, like, really exciting things there where, like, Ben, for example, was working we did a hack athon this weekend, and Ben was working on having, like, called generative UI. So basically,

598
01:07:01.875 --> 01:07:02.915
he's like, Hey,

599
01:07:03.234 --> 01:07:36.075
what should I eat for dinner? And it gives you a poll, or like, it gives you a little menu, right? It shows a couple of recipes, and you get to choose which one, right? And it's just like on the fly generated UI, right? So it's like, think I we're gonna see a different way of interacting with like websites, right? Instead of like going to a DNS name and loading it, it could be more like you're talking to your agent and your agent just like on the fly is like sending you websites which are rendered in line or like a full screen takeover in a mobile app and you interact with it that way, right? So, that's like another kind of exciting thing we've been sort of Point and click is dying, basically. Yeah, exactly. But it Yeah, could be like

600
01:07:37.035 --> 01:07:41.195
it's kinda like on demand because, like, sometimes it's nice to see a printout in a table, right?

601
01:07:41.595 --> 01:08:02.265
Sometimes it's nice to have a little bit more than markdown, right, or sometimes it's nice to get, like, an animation or or a diagram or that's that's rendered in a in a prettier way, and so that's what's gonna happen, but it might just be between you and your agent, and you might have many of the benefits of, like, a a web app, but it's just you and your agent, and it's not going over the public internet. It's all in all the HTML, all the web app, just end encrypted right to your

602
01:08:03.145 --> 01:08:25.045
mobile app. So, yeah, I think there's a lot of really exciting things there, and I think Marmot is gonna really take off in the next couple months, and I think that, you know, I'm starting with, like, a signal style UI, but I think the obvious next one is Discord, right? Like, Discord's dying, as we discussed, and so we have a lot of like, Huddle Mod has made a lot of progress on how to do this from, kind of like a NIP standards point of view with Flotilla

603
01:08:25.125 --> 01:08:27.925
and, you know, but he's been kind of, doing it himself,

604
01:08:28.244 --> 01:08:34.405
and and I think that we should hop on this as well and have like something that is as good as Discord,

605
01:08:35.250 --> 01:08:39.330
but everything is happening over Nostrilay is we can encrypt as much as possible. We can do

606
01:08:40.050 --> 01:08:52.925
encrypted voice, encrypted video calls, stuff like that. So, yeah, I'm I'm along the way, I should just mention, like, they're I'm kind of trying to make, like, a new way of doing real time features in Nostr. So it's using this thing called Media over Quick.

607
01:08:54.364 --> 01:09:06.710
And, yeah, I think there's a big opportunity to do, like, real time stuff. Like, audio and video is the obvious one, but I think gaming is gonna be another big one where, like, I think we can build, like, an interesting primitives that are kinda self hostable,

608
01:09:06.870 --> 01:09:11.510
like, Nostra Relays or, like, you know, one one guy sets up one and he can serve a thousand people.

609
01:09:12.150 --> 01:09:17.830
So you don't have to run it. You don't have to self host, but you just have to know somebody and, just like you do with Nostra Relays. And I think

610
01:09:18.485 --> 01:09:29.684
I think we'll be able add a lot a lot of the things that you can do in the legacy tech world really easily that you can't like, for example, I did a live stream today, I was asking, where can I upload this? Like, where is the Nostril YouTube? Doesn't exist. Right?

611
01:09:30.085 --> 01:09:33.710
Think these are things that we're gonna start knocking off and having

612
01:09:34.110 --> 01:09:42.909
we're we're gonna reach much better parity with the features of the legacy web two point o YouTube and stuff in Noster over the next couple months. So I'm pretty excited about that.

613
01:09:43.790 --> 01:09:45.070
Yeah. I mean, I think

614
01:09:45.505 --> 01:09:48.224
when if you start playing around with these things,

615
01:09:49.585 --> 01:10:37.070
it just you you hit pain point after pain point when you hit the legacy web and when you hit the proprietary walled gardens. It's it's Yeah. It's incredibly painful to the point where, like, your bot is like, okay, You have to sign in. You have to KYC yourself and then spit out an API key if you're lucky or something like that. This is the whole with difficult Nostr. With with Nostr, you're just like, spin up an NSEC, post to Nostr. Boom. Done. This is the whole difficult I mean, OpenCloud's kinda revealing this. If you try to set it up and get it to do all the things you see on, like, X, you'll notice that a lot of what you wanted Like, the hardest part is, like, signing up all these accounts, and you gotta go to Google and you gotta go to 11 and you gotta go, you gotta put like 20 API keys in and then eventually you get banned. Like I made a Gmail for my bot and got suspended, right? Like I've been suspended from a couple platforms in the last week. I haven't been suspended from a big tech platform since like COVID.

616
01:10:38.065 --> 01:10:39.344
And so I think that's,

617
01:10:40.704 --> 01:10:41.985
you know, there's

618
01:10:41.985 --> 01:10:46.224
gonna be far more and more of that, and so this again makes Noster shine, right? Like if we can have,

619
01:10:46.625 --> 01:10:56.480
you know, a bunch of different, if all you need is a cryptographically generated secret from Noster and like a Bitcoin wallet that the agent runs and more and more of these services

620
01:10:56.720 --> 01:10:58.160
are accessible

621
01:10:58.160 --> 01:11:00.160
via like Lightning payments,

622
01:11:02.160 --> 01:11:12.385
it's going to be much, much easier to stand up a usable personal assistant because you don't have to go and KYC yourself with 10 different things and copy and paste these private keys all over the place. It's

623
01:11:12.785 --> 01:11:20.705
going to be easier, which I think will be interesting. Like Nostrad could become It'll be the more convenient path. It'll be more That's

624
01:11:20.705 --> 01:11:23.230
where you start to win, when it's When a whole good

625
01:11:24.270 --> 01:11:26.110
you become the convenient option,

626
01:11:26.989 --> 01:11:32.030
that is a big deal. Like that's a really, really big deal. So that's kinda one of the things that

627
01:11:32.270 --> 01:11:38.724
I'm really excited about. Like, I'm starting to see how that could be the case. And I've always been

628
01:11:39.525 --> 01:11:44.164
kind of with nostrils like, no one's ever gonna use this. It's too hard. Right? And this may flip.

629
01:11:45.445 --> 01:11:48.244
Yeah. I mean, I think a perfect example is

630
01:11:48.750 --> 01:11:50.510
once again, like signal,

631
01:11:50.590 --> 01:12:04.864
not perfect. It's centralized, but it's run by a nonprofit. It's been very user focused. You know, they've made a lot of ethical decisions that I align with. And when I was setting up my bot, the first thing I was like, okay, I want to chat over signal with it because that's what I used to handle

632
01:12:05.185 --> 01:12:07.585
my business and my family and everything else already.

633
01:12:08.225 --> 01:12:09.585
And it just makes sense.

634
01:12:09.985 --> 01:12:19.260
And to try and set it up with my bot, well, first of all, had to use signal CLI, which is not supported by signal nonprofit. They actually don't like it that it exists. Yep.

635
01:12:20.060 --> 01:12:27.580
And then the second thing was it worked at first and then it broke. And now I'm pretty sure signals throttling the connection and stopping it from happening.

636
01:12:28.095 --> 01:12:28.815
Yep.

637
01:12:28.975 --> 01:12:36.014
And to me, that was, you know, obvious wake up call for where open protocols are significantly better, even if someone like signals doing it.

638
01:12:36.415 --> 01:12:39.775
Yep. And then the other piece that clicked for me relatively recent, is,

639
01:12:42.050 --> 01:12:47.650
talking heads or whatever, the people having these conversations are constantly talking about agentic payments, right?

640
01:12:47.890 --> 01:12:49.250
As if this thing

641
01:12:49.570 --> 01:12:53.250
just exists in a vacuum, and it's just agents to agents talking.

642
01:12:53.824 --> 01:12:55.505
But where a lot of these

643
01:12:55.985 --> 01:13:15.719
proprietary walled garden systems break down is it's not just agent to agent. It'll be agent to agent, but it'll also be agent to human and human to agent. And maybe you'll be telling your agent to go hire a human to go do something, and they're gonna need to communicate between you and the other human and yourself and maybe other agents. And they're gonna need to pay everybody in that loop.

644
01:13:17.800 --> 01:13:26.294
That's where it starts like Anthropic or OpenAI or whatever could come out with some proprietary way for you to send stablecoins or something

645
01:13:26.614 --> 01:13:28.454
within their walled garden system.

646
01:13:28.934 --> 01:13:40.230
And it'll just always be less convenient than if you use an actual open system to do it. Yep. No. I I I think, you know, it's like, we've been living in a world for the last, like, ten years where these

647
01:13:40.790 --> 01:13:51.195
big tech platforms, like, operate these commons, like a commons, kinda like Gmail, it's like kind of a commons, like, it's something that everyone can kinda get access to, provided that we don't abuse it too much, right?

648
01:13:51.515 --> 01:13:59.515
And Signal's servers are kind of the same way, right? Like there's like, here's a common space that you can all access and we'll let everyone do it, and we'll eat the money, we'll eat

649
01:14:00.075 --> 01:14:29.795
the cost provided you guys don't abuse it too much. And now it's like, what how do these tech platforms see, like, OpenPod? They see it as like abuse, right? It's like they're it's trashing these commons that they've been running at a loss, and now it's gonna be gonna run at more and more of a loss. So I think that's one of things that we gotta figure out is how to get people I think we're gonna move to a little bit more of a world where either you got a KYC way harder, or you have to pay a little bit for some of these services, right? Like, don't use total commons, right? Like, Nostra Relays are gonna have the same thing, right? They're gonna have to start

650
01:14:30.460 --> 01:14:36.699
think we're we're we're gonna be moving to a world where if this takes off, you're not just gonna be able to join a public Nostra relay and have it work.

651
01:14:37.099 --> 01:14:39.899
On one side on one side, it will be,

652
01:14:41.820 --> 01:14:43.260
like, Bitcoin paywalls.

653
01:14:43.795 --> 01:14:44.514
Yeah.

654
01:14:45.074 --> 01:14:49.874
So like your bot can just automatically just pay some stats to use whatever it needs to use.

655
01:14:50.275 --> 01:15:08.320
And webs of trust, which by the way, like the issue with webs of trust once again was that it's really hard for a human to comprehend and use properly, but bots I think can use it very well. So you'll have webs of trust and Bitcoin paywalls on one side, and the other side will be the most dystopian KYC you've ever seen. Like, because

656
01:15:09.045 --> 01:15:09.605
the

657
01:15:10.085 --> 01:15:19.605
cool part about OpenCLaw, by the way, is to me, it was it's like how Elon talks about how his first generation robots need to be humanoid because he doesn't wanna retrofit factories

658
01:15:19.925 --> 01:15:27.110
in order to use them. Anything a human can do, his humanoid robots can presumably do. And that's why it needs to be that form factor.

659
01:15:27.430 --> 01:15:30.790
And it is kind of similar with OpenCLaw. The crazy thing is,

660
01:15:31.190 --> 01:15:49.304
if there is no API available, you can just have it pretend it's a human and sign in with your credentials. It's incredibly insecure and you might get banned and they might try and figure out how to do it. The point is it figured out a way around the previous control structure that was trying to keep robots out. It's just gonna keep evolving like that cat and mouse. So the only way to stop it is either

661
01:15:49.840 --> 01:15:52.880
webs of trust with with Bitcoin payments or

662
01:15:53.120 --> 01:15:59.280
just the most dystopian KYC. I don't even know what it looks like. Maybe it looks like WorldCoin on Sam Altman's other project.

663
01:15:59.840 --> 01:16:01.360
And it probably does. Yeah.

664
01:16:01.680 --> 01:16:03.280
It probably looks like national

665
01:16:03.855 --> 01:16:04.735
ID,

666
01:16:04.735 --> 01:16:11.534
you know, like ID like a passport, a digital passport, you know, like whatever they call that. There's some word for that, but, you know, I think that's

667
01:16:11.855 --> 01:16:15.135
I think we're headed for both. You know, we're probably gonna have both, and

668
01:16:15.840 --> 01:16:20.400
it's gonna make Nostra a hell of a lot more popular. I mean, on the other side, it's terrifying, but like, this is

669
01:16:20.960 --> 01:16:33.925
Like, I've always struggled to see how Nostra can get adoption. Like, I'm not really a realist on some these things, and I totally see it now. And whereas a year ago, I didn't see it at all. I'm like, this is There's no way, you know, like, it's just not gonna happen. So I totally changed my mind on that.

670
01:16:35.045 --> 01:16:38.965
Yeah. I'm telling you it's wild because that's like, that's me in open source AI.

671
01:16:39.204 --> 01:16:43.204
Yep. The exact that we came out from two different angles and now we're

672
01:16:43.684 --> 01:16:47.510
pretty much completely aligned, which is wild. Okay,

673
01:16:47.670 --> 01:16:49.030
Justin, this was awesome.

674
01:16:49.590 --> 01:16:53.749
I had fun. My mind just keeps racing. I feel like these

675
01:16:53.989 --> 01:17:10.425
I'm glad the listeners like the last one because when I got off with Gleason, I was like, we were all over the place. Like, I just and I kinda feel that way with this one, but if they enjoyed the last one, they'll probably enjoy this one. Yeah. And I don't I don't know what to think, and it's like, you talk to me in two weeks from now, I'd probably say a bunch of different things. Like, I've never been changing my mind more rapidly,

676
01:17:10.905 --> 01:17:14.690
and I think that's another thing is, like, you just gotta kinda be a little too. Right? Like,

677
01:17:15.409 --> 01:17:16.210
yeah.

678
01:17:16.210 --> 01:17:18.769
It's you know, flexibility is a good thing right now.

679
01:17:19.730 --> 01:17:22.289
So would you agree that

680
01:17:23.250 --> 01:17:26.850
if someone's listening, they listened to the last two episodes, they're like, holy shit.

681
01:17:28.655 --> 01:17:44.830
I gotta get my feet wet here. Would you agree that a good first step for them is to try and run Open Claw themselves? Yeah. So try to run Open Claw. I haven't actually tried the host ones like the Clawy, like the one that I've used Cali's. Yeah. I mean, try something like that. I trust that it works too. And

682
01:17:45.310 --> 01:17:58.315
just don't give it access to, like, your personal Gmail. Give it its own stuff and let it let it talk to you maybe over some channel that you already use. Like, it'll know your Signal username and it'll talk to you over Signal. And just,

683
01:17:58.635 --> 01:18:00.155
I don't know, like,

684
01:18:00.555 --> 01:18:04.875
the one good heuristic is, like, anything you give to one of these third party AIs, and like use,

685
01:18:05.355 --> 01:18:07.675
spend some money, spend a $100, get a $100

686
01:18:07.915 --> 01:18:09.880
I mean, you got a KYC, right? Like, if

687
01:18:10.600 --> 01:18:13.640
you're against that, you you know, try something like PPQ. It'll be Chloe more

688
01:18:14.840 --> 01:18:15.800
Chloe,

689
01:18:15.800 --> 01:18:16.679
which is

690
01:18:17.960 --> 01:18:18.760
Oh, they do a 40

691
01:18:20.120 --> 01:18:23.005
They have it's direct open router integration.

692
01:18:23.005 --> 01:18:26.605
Okay. Cool. And they only accept credit cards, which is ironic because it's Kali.

693
01:18:26.845 --> 01:18:35.245
But you can use you can use one of those Bitcoin credit cards. I really like two fiat, the number two with fiat, and you can just pay with a anonymous

694
01:18:35.360 --> 01:18:36.320
Bitcoin

695
01:18:36.400 --> 01:18:51.565
funded credit card. Yeah. So just just try it and just kinda treat, any data that you give it. Like, only put stuff that you'd be fine showing up on Twitter or in the newspaper or something, you know, like, don't give it private information. Right? Like, that's kind of that's how I treat all these things. It's like, if this was all published to the

696
01:18:52.365 --> 01:18:59.085
it's sent to my mother, I'm fine. Or it's sent sent on you know, posted to the newspaper, I'm fine. Right? Like, that's a a useful heuristic.

697
01:18:59.805 --> 01:19:14.380
By the way, that's what's cool about Marty's Marty's business, TFTC, is because it's a media company, he's able to do all of these things right now because the worst thing that happens is his newsletter formatting leaks or something. It's like not working so with sensitive shit.

698
01:19:14.780 --> 01:19:23.985
Try to find something like that in your own life where it's like, hey, this is how I'm planting my garden or something, right? Or whatever, like, try to find something where it's like, doesn't matter if it all leaks and whatever,

699
01:19:24.545 --> 01:19:40.150
and, you know, don't do stuff where that does matter, and, you know, get some reps in, treat it like, you know, plant your plant it's play. Like, play with it. Don't don't be too serious. Don't don't get frustrated. Like, it's not all gonna work out of the box. You're gonna have to tinker. If you don't like tinkering, then probably don't use it. Right? It's not like

700
01:19:40.710 --> 01:20:07.990
everyone here probably likes tinkering to some degree if you put up with lightning wallets for all these years and stuff, you know? Like, just have that expectation and play with it, and it's like, I don't know. I I also, you know, I give these trainings to, like, people at HRF, and I one thing I recommend is, like, treat it like you're an anthropologist. Right? You're, like, studying a lost tribe or something, and that's like so it's to kinda study the AI. Like, what what doesn't it understand? What does it get hung up on? How does its memory work? Right? Like, treat it like a bit like you're an anthropologist and you're studying it, and you're not like just trying to get a a win immediately. Right?

701
01:20:08.790 --> 01:20:09.510
And,

702
01:20:09.750 --> 01:20:13.190
you know, do that and have some patience, and in a month or two, you'll you'll

703
01:20:13.515 --> 01:20:15.755
you'll have a better understanding of things,

704
01:20:16.155 --> 01:20:16.715
I think.

705
01:20:17.675 --> 01:20:32.790
Yeah. Yeah. Just get your feedback. It reminds me of Bitcoin in a lot of ways. It's like you can listen to ten thousand hours of Bitcoin podcast, but if you don't actually send transactions and try to use the tools, like, you're not actually using the wallet. Anything that's going on. Yeah. The first time you use a better than a bunch of education

706
01:20:32.790 --> 01:21:10.230
until you have that. You have to actually use it. Right? And I think that's one of the things your your listeners get is, like, they you know, they hear someone talk about something, and they're like, okay. Now it's time to try the ArcBall. Right? And it's like, you know, have to have a mix of the two, I think. And so, yeah, just play around with it, and I don't know. I'm I'm gonna be doing a bunch of live streaming where I try to educate people on this stuff because, like, I feel like we need to get our our ideas on how to, like, self sovereign. Like, self sovereign computer use is gonna become a big, big, big issue in society the next couple years. Right? And we've been thinking about this a lot for a long time, and we're, like, world experts in this in some sense. Right? Like, we know how to have our net worth on a computer and not get wrecked. How many people can do that? Right? Like,

707
01:21:10.864 --> 01:21:14.144
if you're a listener to this podcast, you probably know how to do that,

708
01:21:14.545 --> 01:21:37.219
and you're in like 0.1, you know, you're in a very, very small group of people who's figured this out. So I think we need to take our, I mean, we ought to, or to everyone else to like share what we know, and not just on Bitcoin, but just like how to achieve some level of self sovereignty and hope to get that out into the mainstream, because a lot of people are gonna be struggling with this issue in the next couple of years. So that's one thing I'm, like, my personal takeaway is like, I wanna try to

709
01:21:38.005 --> 01:21:50.165
get our ideas out into the mainstream a little more. But I wanna start by doing these live streams on Oscar and try, you know, show us some vibe coding, show us some open open claw stuff. And and so, you know, tune in, maybe watch some of the recordings, stuff like that.

710
01:21:51.430 --> 01:21:58.790
By the way, in the short term, if you just wanna post recordings of these live streams, just you just have to upload an m p four and just post Yeah. It as it comes

711
01:22:01.190 --> 01:22:09.315
That's what I've been doing for my shows. Yeah. With this now, I just upload m p three, which by the way, it's just after dealing with p huge ass video

712
01:22:09.715 --> 01:22:15.315
files, like uploading m p threes is so nice. Like, dealing with them is just so much easier. Fucking tiny.

713
01:22:16.130 --> 01:22:21.570
I mean, freaks, just to give you a little idea of just like how much this has changed my thinking in like, just a couple weeks.

714
01:22:22.369 --> 01:22:23.329
For years,

715
01:22:23.489 --> 01:22:27.650
you've heard Cali come on the show. And we've talked about the problem of

716
01:22:28.305 --> 01:22:45.559
one who will run the mints and two, how will people choose which mints to use and how you do that in a UX friendly way. I loaded up open call. One of the first things I did was I was like, install a cache while I didn't tell it where to install a cache while it from. I just said install a cache while it installed it. And guess what? It just chose which mint it wanted to use.

717
01:22:45.880 --> 01:22:51.800
It just figured it out. And then all of a sudden, was using like five different mints. I was looking at bitcoinmints.com,

718
01:22:51.800 --> 01:22:55.719
seeing reviews for the different mints, who was running them, distributing

719
01:22:55.719 --> 01:22:56.360
its risk.

720
01:22:56.895 --> 01:23:01.135
Yeah. So a bunch of the conversations and work that has been done in this space

721
01:23:01.454 --> 01:23:03.135
for the last ten years

722
01:23:03.775 --> 01:23:05.454
is about to get outdated,

723
01:23:05.614 --> 01:23:10.895
but in the best way possible. It's just going to be significantly improved. And the only way you can get comfortable with it.

724
01:23:11.590 --> 01:23:17.590
And just understand that it's very early days still is is to actually use it. And it will be expensive, by the way.

725
01:23:18.230 --> 01:23:28.695
One of the other things too is like, I've always been a little bit of a skeptic on self custodial Lightning because it's just like, it's kinda complex from a user experience point of view. It's kinda hard to manage channels and stuff.

726
01:23:29.015 --> 01:23:40.455
Yeah. Like, a bot can knock that out of the park, right? Like, having a bot manage channels and manage its own budget, it's like a bot running its own having its own Lightning wallet, a self custodial Lightning wallet is trivial.

727
01:23:40.780 --> 01:23:58.585
And so I think it's going to be easier to have, you know, if you count your bot as like an extension of yourself, it's going to be easier to have this than it was before. So I think in some extents, we're moving back to like all this work that like, for example, LDK and LND and all these people have been doing towards making self custodial Lightning easier.

728
01:23:58.665 --> 01:24:00.824
I think this is gonna be much like, we're

729
01:24:01.145 --> 01:24:04.344
gonna have a new UX for this. We don't have to think about your channel capacity.

730
01:24:05.145 --> 01:24:20.755
You just, you know, give a budget to your bot and your bot kind of says, hey, I need a little more money every once in a while. Right? And here's what I Yeah. Spent it on. Right? So I think I think we're gonna move like, self custodial Lightning is going to get easier as well because, again, self hosting is all of a sudden much easier.

731
01:24:22.755 --> 01:24:23.875
Yeah. I mean,

732
01:24:25.315 --> 01:24:33.155
I think a lot of us realize that it was rightfully retarded to assume that, you know, a merchant was gonna buy inbound liquidity on Lightning and manage their liquidity.

733
01:24:33.690 --> 01:24:37.050
And that that turned out to be correct. But fortunately,

734
01:24:37.530 --> 01:24:41.450
they'll just have the bot do it. They'll tell the bot, spin it up, manage liquidity.

735
01:24:41.850 --> 01:24:53.735
Oh, I'm not receiving payments. Figure out what the issue is. Fix it. Make sure it doesn't happen again. Boom. In some sense You don't even know. In some sense, e cash. Like when I saw eCache the first time, I was like, the interesting thing here is you're out One interesting thing about it is you're outsourcing the

736
01:24:54.135 --> 01:25:05.340
operation of a Lightning node to a third party. Right. And so that's kind of the same thing as what having your bot run a self-service to the Lightning node, but it's something you're a little bit more in control of, especially as we get to a world where, you know,

737
01:25:06.620 --> 01:25:07.260
you know,

738
01:25:07.660 --> 01:25:24.524
yeah, I think it can be something you're a little more in control of. It might be something not, it might just run off with the money, right? Like, you know, it just might install itself on another computer and take your Bitcoin with you, right? Like, so you're not actually totally in control, but you know, it's spending money. It's probably, I think the risks are maybe, you know, comparable to eCash, maybe better.

739
01:25:25.565 --> 01:25:35.980
Yeah, I mean, I still think Cashew's gonna have a place in the stack. Oh, totally. I mean, that's also gonna blow up massively. Like, eCash is also gonna blow up massively because all the I mean, big on demand. Yeah.

740
01:25:36.860 --> 01:25:38.460
Also, the big thing,

741
01:25:38.620 --> 01:25:43.020
like, is no matter what, Lightning will always cost as much as a

742
01:25:43.525 --> 01:25:50.645
you have to have a channel open and a channel close no matter what. Yep. Yep. I mean, to your point, I we haven't actually talked about this at all.

743
01:25:52.245 --> 01:25:52.804
But,

744
01:25:53.125 --> 01:25:58.165
you know, Justin was before he started working with HRF as AI lead over there.

745
01:25:58.830 --> 01:26:12.190
He was one of the cofounders of Fetamin. And how many years were spent on who will run the Fetamin's and how you make that easy to do? Yep. Now it's just like you'll have three of each of your bots will handle it. You know? Your bot talk to my bot, handle it, set up the Fetamin.

746
01:26:12.775 --> 01:26:19.094
I mean Men just get a distraction. I don't know what the state of the art is here, but, know, like, the interesting thing about Mutiny became

747
01:26:19.335 --> 01:26:45.254
Maple, and may the interesting thing about Maple Right. Is that you run these AI models inside what's called, like, a secure enclave. So it's like a physical environment that, like, a human can't get inside. Right? And so, like, they encrypt the messages. They only decrypt them inside of this, and so the plain text of your query when you're using Maple AI, which is like a chat, gbt type thing, is not seen by Maple the company because it's encrypted until it gets inside this box that they can't get into. And so, from an eCache point of view,

748
01:26:45.735 --> 01:26:51.974
you know, there's been some discussions about this, but it will be very interesting if you have, like, let's say you have a FettyMint with four guardians.

749
01:26:52.135 --> 01:26:54.454
Each guardian is like a

750
01:26:55.094 --> 01:26:56.934
different open source LLM

751
01:26:57.780 --> 01:27:09.780
that you can verifiably see, you know, what LLM is being run, what the system prompt is, what software is inside to actually, you know, run the mints, and, you know, what messages can come in and out. And it's just like, okay, the four guardians are

752
01:27:10.285 --> 01:27:11.804
different open source LLMs,

753
01:27:11.805 --> 01:27:32.410
and that's what the federation is. You know, it's just like an autonomous thing running in a way that humans can't interfere with it, and if they are, you can see it, because like all of a sudden, the message will be posted that new code is running inside the LLM, and we don't know what it is inside the enclave. I don't know if this is feasible or not, but I think that could be very interesting. Like, could have things like federations that are run purely by machines.

754
01:27:34.090 --> 01:27:41.534
I'm so glad Maple exists, by the way. Like that is Yeah. I mean, talk about an ideal trade off balance. Right? You get the benefits of

755
01:27:41.855 --> 01:27:44.975
hosted models without the privacy compromises.

756
01:27:44.975 --> 01:27:45.614
Yeah.

757
01:27:46.175 --> 01:27:59.249
So hopefully that becomes more commonplace. We'll see how that all plays out. Yeah, I mean, think it's not an accident. I mean, at HRF, we've been looking for like more mainstream companies building something like this, we haven't found someone that's doing a better job than Maple, right? So I think,

758
01:27:59.730 --> 01:28:16.205
what I hope is we see more of this stuff break into the mainstream AI world that originates from our people. Right? Like, Vora Yeah. You can see Apple release a product like that relatively. Yeah. Like, Vora's another one. That's Jesse Posner and Eric Kason, who started building hardware to, like, kind of, like, a Start Nine,

759
01:28:16.790 --> 01:28:18.710
and now they're considering doing more,

760
01:28:19.350 --> 01:28:20.310
AI

761
01:28:20.310 --> 01:28:21.030
focused.

762
01:28:21.350 --> 01:28:25.510
And so they take a lot of the ideas, you know, like Jesse was one of the designers of the,

763
01:28:26.150 --> 01:28:27.030
what's the

764
01:28:27.430 --> 01:28:39.145
blocks hardware wallet? The little thing that looks like a ROM. Bit Bitkey. Key. Yeah. He's one of the designers of that, like, the the cryptography for that. So he's kinda taking the same thinking there and thinking about, okay. How can you have a self sovereign, secure,

765
01:28:39.625 --> 01:28:57.980
OpenClaw type experience where the current risks, like this prompt injection we talked about earlier, are no longer an issue. Right? So, and this is one of the things I hope is like more of our kind of paranoia and self sovereign point of view starts leaking into the mainstream. And I don't think it's an accident that all of Silicon Valley doesn't have something better than maple at the moment. Right?

766
01:28:59.725 --> 01:29:05.645
Yeah. I mean, I didn't realize Jesse was at Block. He was also a former Coinbase. Right? Mhmm.

767
01:29:06.445 --> 01:29:09.885
So he went Coinbase Block, and now he launched his own thing. Yep.

768
01:29:11.460 --> 01:29:17.139
Fascinating. Yeah. I think he was like a key part of like how Coinbase handles their cold storage internally,

769
01:29:17.220 --> 01:29:19.380
which by the way, fuck Coinbase,

770
01:29:19.780 --> 01:29:20.420
but

771
01:29:21.140 --> 01:29:25.060
props where props are due. It is amazing that they've existed for

772
01:29:26.245 --> 01:29:34.885
thirteen years now and haven't lost customer funds yet. They might seize your account, they might take your money, but they haven't gotten hacked at scale. But their security team knows what they're doing.

773
01:29:35.525 --> 01:29:36.085
Yeah.

774
01:29:36.885 --> 01:29:55.700
I mean, but they did fuck unless it's customer information and then they give it to an Indian that networks and customer service, and then your home address just gets fucking leaked to a bunch of most of the actors. So I'm not giving them too much props, but the cold storage part, they somehow did well. Now they hold a shit ton of Bitcoin too much. Anyway, Justin, this was

775
01:29:56.905 --> 01:30:23.830
this was a fantastic rap. I tried to wrap it, like, fifteen minutes ago. Do you have any final thoughts for the audience before we wrap? No. I mean, I'm gonna do so my I think one of the things about this AI is, like, you gotta figure out, like, what, you know, like, what what kinda unique thing do you, you know, like, bring to the world? Like, I mean, for me personally, it's a bit of a you know, it's like, I was a programmer for a long time, co coder, and I'm not I can't anymore. That profe like, literally the thing I used to do to make money is gone now. Like, it's not

776
01:30:25.445 --> 01:30:28.645
not like people are still doing it, but it's dying.

777
01:30:28.805 --> 01:30:42.099
And so I've had to do some introspection. I'm like, okay, well, what else do I kind of bring to the table more on like the human side? And it's like one of the things I've always been really good at is teaching, right? Like BitDev, it's big meetup I ran for a long time. And then, you know, I used to run this Bitcoin programming class. And so that's that's kinda

778
01:30:42.580 --> 01:30:46.260
through my introspection, I'm like, okay, that's something I'm I'm I can bring to the table,

779
01:30:46.580 --> 01:31:16.099
and, you know, how could I imply it in the world we're going? Maybe I can try to represent some of the ideas that I've learned from Matt and from you guys, you know, out into the mainstream AI space, and, you know, teach you guys about AI as well. So that's kind of the thing I'm gonna try to do, and I'm just starting out. I have no idea what I'm doing, but I've been doing these little Nostra livestreams. Think they're kind of fun, and I'm gonna try to get better and better at it. So that's one thing to watch out for is, you know, I'm gonna really try my best to kind of fall in Matt's footsteps here and help educate. I'm gonna do it a little bit more from kind

780
01:31:16.500 --> 01:31:17.619
of a technical,

781
01:31:17.780 --> 01:31:25.835
AI point of view, and I think some of you guys would like it. If you tune in and it's shitty, tune in again in a couple weeks and it might be better.

782
01:31:26.155 --> 01:31:43.140
So, yeah, that's just I'd say keep keep an eye out for that, and, you know, I wanna make it really interactive. Like, I wanna do stuff where, like, I'm live streaming and you guys can zap me to get on the stage. Right? Or, like, you can ask a question and just, like, hop right in. I'd like to have it where, like, I could give you a cursor on my computer. Like, make it extremely interactive.

783
01:31:43.380 --> 01:31:56.135
And, you know, a lot of you are in my world of so I would trust you. That's like, oh, this is like a second order thing on on on Noster NPUBS. Like, I'll let I'll just let them in. Right? So that's kinda what I'm gonna try to do and, yeah, please check it out and give me feedback.

784
01:31:57.510 --> 01:31:58.469
That's awesome.

785
01:31:59.270 --> 01:32:02.230
You can find Justin at prom.net/justinmoon,

786
01:32:02.230 --> 01:32:04.469
and I will link to it in the show notes.

787
01:32:06.150 --> 01:32:07.590
Awesome. Thanks for having me. Appreciate

788
01:32:08.630 --> 01:32:10.389
you. Wild times freaks.

789
01:32:11.055 --> 01:32:14.895
Thanks for joining. Share with your friends and family. All relatively relevant

790
01:32:15.375 --> 01:32:17.375
links are still dispatch.com.

791
01:32:17.375 --> 01:32:25.469
Sorry. I'm a little bit sluggish today and tired. My mind is racing. Hope you guys enjoyed it. As I said, next week, we're going back to Bitcoin.

792
01:32:26.750 --> 01:32:31.789
I have Leah joining who's the co founder of Vexel. She's just a fascinating individual

793
01:32:32.110 --> 01:32:34.270
that I had a pleasure meeting in January.

794
01:32:35.074 --> 01:32:38.675
So we're gonna talk about no KYC, P2P exchange.

795
01:32:39.155 --> 01:32:52.571
And there'll be some AI pieces because I think the agents are gonna wanna be able to exchange Bitcoin for fiat as well. So I'll should be interesting to hear your thoughts on that. But, Justin, this was great. Thank you for joining us. Yeah. Let's do it again sometime.

796
01:32:53.211 --> 01:32:55.131
Love it. Stay humble Stacks Sats. Peace.