1
00:00:00,000 --> 00:00:06,800
there are no secrets the only secret is to hit record yeah and then send it so

2
00:00:06,800 --> 00:00:11,800
I told you about my by-blind I'm also building out a pipeline for the

3
00:00:11,800 --> 00:00:18,600
recordings so I hope that very soon I'll just have to press record press save

4
00:00:18,600 --> 00:00:24,640
and then like do a using Fia Jeff's the NUC you know like NAK do like a

5
00:00:24,640 --> 00:00:29,200
touch publish and then everything is gonna be done you know how it works it's

6
00:00:29,200 --> 00:00:33,640
just like you mount your file system you have it so so basically you get the

7
00:00:33,640 --> 00:00:39,320
foot like the the pipeline makes a fully ready version yes yes and you just

8
00:00:39,320 --> 00:00:44,400
have to decide when to when to publish because it's gonna the mp3 file is on

9
00:00:44,400 --> 00:00:49,160
my phone you are the one of the biggest file system maxes I'm a file I love

10
00:00:49,160 --> 00:00:54,200
files you love the files are Lindy and the and hierarchical files yes files are

11
00:00:54,200 --> 00:00:59,080
Lindy and and I'm a big like convention over configuration guy yeah so the

12
00:00:59,080 --> 00:01:04,560
way it works what is what does that mean exactly it's it's it's basically I

13
00:01:04,560 --> 00:01:07,480
don't want to configure stuff I just want the system to tell me how it works

14
00:01:07,480 --> 00:01:11,260
yeah and I want to have a file structure and things and and that's just how it

15
00:01:11,260 --> 00:01:15,500
works yeah if you do it wrong everything breaks yeah it's it's rails was a big

16
00:01:15,500 --> 00:01:20,080
conventional okay yeah yeah yeah it's just like you know this is if you have a

17
00:01:20,080 --> 00:01:26,280
model name you do things around here you know and and you have to there is no

18
00:01:26,320 --> 00:01:30,480
other way and so so what I want to have is like again like press record press

19
00:01:30,480 --> 00:01:35,480
safe there's an mp3 file on my phone yeah it syncs using torrents with all my

20
00:01:35,480 --> 00:01:39,200
other devices everything is run locally like this already it syncs using

21
00:01:39,200 --> 00:01:43,040
torrent because sync thing uses torrent on it really yeah it does as far as I

22
00:01:43,040 --> 00:01:47,040
know yes Wow I hope I'm not wrong but it's it's used back in the day when I

23
00:01:47,040 --> 00:01:51,360
learned about it that's what I learned but but but we'll we'll someone will

24
00:01:51,360 --> 00:01:54,880
yell is kind of interesting I was thinking about like blossom stuff I

25
00:01:54,880 --> 00:01:58,000
forget what exactly I was wondering like what if you just changed a little bit

26
00:01:58,000 --> 00:02:01,560
of the file like how can we make that like if I wanted to store because I was

27
00:02:01,560 --> 00:02:05,240
thinking a little bit about like making like a Dropbox type thing yeah on blossom

28
00:02:05,240 --> 00:02:08,800
yeah and it would be like with this this few is just like a fuels fuse file

29
00:02:08,800 --> 00:02:13,200
system thing yeah like that's basically the drop boxes yeah it's like a virtual

30
00:02:13,200 --> 00:02:16,800
file system that you that is mounted you don't see that but I think that's what

31
00:02:16,800 --> 00:02:21,360
it is and like you could just totally do that where everything is saved on blossom

32
00:02:21,400 --> 00:02:28,560
right yeah it is and the problem is like okay so you have a word document and you

33
00:02:28,560 --> 00:02:31,200
just add a little bit to it unless you know so there's something I can't think

34
00:02:31,200 --> 00:02:35,120
of an example but like a big thing like let's see you have a big video and you

35
00:02:35,120 --> 00:02:39,960
just crop one minute one second off the end right do you save the entire thing

36
00:02:39,960 --> 00:02:44,960
back up to awesome no it's that's the issue I talked about this with Pablo

37
00:02:44,960 --> 00:02:49,880
just a couple days ago with kind of the the go-to example I use is like let's

38
00:02:49,960 --> 00:02:53,280
say you have an article the very simple just a text and then you fix a typo yeah

39
00:02:53,280 --> 00:02:57,000
yeah is it a new thing yeah like in the in the cryptographic hash as well it's a

40
00:02:57,000 --> 00:03:00,160
new thing yeah but you still have like you have proximity hashes and those kind

41
00:03:00,160 --> 00:03:03,240
of things where you can mitigate that yeah it's a very hard problem it's like

42
00:03:03,240 --> 00:03:07,640
a speciation problem so you can't really solve it when is something because when

43
00:03:07,640 --> 00:03:12,160
when do you make the kind of when is it just like yeah a typo fix or when is

44
00:03:12,160 --> 00:03:15,080
it when is it a completely new thing yeah but this is one of the things that

45
00:03:15,080 --> 00:03:18,200
makes Dropbox kind of good I bet this is one of things that makes things good

46
00:03:18,200 --> 00:03:24,200
is that it doesn't necessarily clobber you know I think that the thing that

47
00:03:24,200 --> 00:03:29,480
makes think good is it's very easy to reason about yeah I think that's I

48
00:03:29,480 --> 00:03:33,880
think and I think that's important for systems in general that's why I love

49
00:03:33,880 --> 00:03:37,440
files so much because they are easy to reason about it's like I have the file

50
00:03:37,440 --> 00:03:44,880
I know like I can do stuff with it yeah and and I use the file system as a

51
00:03:44,960 --> 00:03:49,040
source of truth a lot I don't like things that scan through files and build

52
00:03:49,040 --> 00:03:52,400
databases yeah because the stuff is always like out of sync yeah like the

53
00:03:52,400 --> 00:03:56,000
file disappears or is corrupt or something and still in the database and

54
00:03:56,000 --> 00:04:00,840
then it's like what do you do yeah and and so I just I just want to use that

55
00:04:00,840 --> 00:04:05,440
the actual ground truth as my source of truth yeah that's also what Bitcoin does

56
00:04:05,440 --> 00:04:08,720
well yeah like it actually like there is no abstraction in between it's just

57
00:04:08,720 --> 00:04:14,120
yeah it's the thing let's walk down here otherwise it's gonna be noisy but but

58
00:04:14,160 --> 00:04:17,600
anyway like speaking of the pipeline I'm currently building out it's like I

59
00:04:17,600 --> 00:04:21,200
just want to hit save there's an mp3 file it's gonna land on one of my

60
00:04:21,200 --> 00:04:25,000
computers that's just always running and always on right now it's a shitty old

61
00:04:25,000 --> 00:04:29,720
thinkpad that basically explodes once it runs in L.L.A. but but anyway like you

62
00:04:29,720 --> 00:04:34,320
do you do an auto transcription local model you have a txt file that is the

63
00:04:34,320 --> 00:04:38,920
transcription yeah and an srt file because that's what the podcast plays

64
00:04:38,920 --> 00:04:46,800
expect and so on and does an srt file yeah yeah yeah yeah but but that's fine

65
00:04:46,800 --> 00:04:49,640
like all the all the all those things all those tools it's just like a real

66
00:04:49,640 --> 00:04:54,300
basic text format yes yeah I remember dealing with these on on with the

67
00:04:54,300 --> 00:04:59,440
subtitles and also other things but but the point is like I want to have like

68
00:04:59,440 --> 00:05:05,400
super simple again like the idea of the Unix tools of of Noster that we talked

69
00:05:05,400 --> 00:05:08,800
about in the past but it's just like this does one thing it creates a

70
00:05:08,800 --> 00:05:12,120
transcript the next thing the next step in the pipelines like you take the

71
00:05:12,120 --> 00:05:17,000
transcript and you create a summary yeah yeah and and the summary like that the

72
00:05:17,000 --> 00:05:21,880
summarizer it has to know that it's like a a transcript of a dialogue because

73
00:05:21,880 --> 00:05:24,840
then it can summarize like there's two people in the conversation these were

74
00:05:24,840 --> 00:05:28,680
the main topics and so on I'm currently building another pipeline which I'm

75
00:05:28,680 --> 00:05:33,880
calling the by-blind yes where it's like I have an idea for an app cursor can

76
00:05:33,880 --> 00:05:38,280
probably one shot it yeah and here's the idea and I talk about the idea for

77
00:05:38,280 --> 00:05:41,960
two minutes yes and it's like you know I want to build it you're inspired now so

78
00:05:41,960 --> 00:05:44,800
you want a problem now you don't want it you don't have prompt when you're

79
00:05:44,800 --> 00:05:47,880
sitting at the computer no I when you have no inspiration I did this in the

80
00:05:47,880 --> 00:05:51,000
last couple days and I'm usually most inspired when I step out of the shower

81
00:05:51,000 --> 00:05:54,840
yeah and then I talk into the microphone for like three minutes everything

82
00:05:54,840 --> 00:05:57,720
happens in the background the other ones the other ones are very responsive to

83
00:05:57,720 --> 00:06:04,360
inspiration yes yeah yeah they respect a moment of true inspiration if you if

84
00:06:04,360 --> 00:06:10,520
you come to them kind of without without freshly kissed by the muse yeah yeah

85
00:06:10,520 --> 00:06:14,000
with that with it we have like unloved by the muse they won't they won't they

86
00:06:14,000 --> 00:06:17,760
won't do your favors yes that's true no but you have to capture that kind of

87
00:06:17,760 --> 00:06:22,360
and and and and so this is like I got all this stuff like 80 percent working

88
00:06:22,360 --> 00:06:25,680
I've noticed like doing more L on programming is that there's a lot of

89
00:06:25,680 --> 00:06:33,240
times where like I won't attempt like a real serious prompt unless I'm very

90
00:06:33,240 --> 00:06:39,920
very fresh yeah like I will only do it like a couple hours a day yeah or like

91
00:06:39,920 --> 00:06:44,240
it when I'm inspired like but oftentimes I'll I'll recognize okay this is something

92
00:06:44,240 --> 00:06:48,880
where I have to go through a big prompt process and I just shouldn't do it now

93
00:06:48,880 --> 00:06:52,160
yeah like I should wait till tomorrow because that's what I never really thought

94
00:06:52,160 --> 00:06:56,280
like yes tell me about your current setup what you use in terms of like

95
00:06:56,280 --> 00:07:01,480
what's your tool chain look like yeah well so I kind of go back and forth like

96
00:07:01,480 --> 00:07:07,760
like sometimes like if I'm just changing things I'll usually use

97
00:07:07,760 --> 00:07:11,680
Claude code like a little like little incremental that changes and you can

98
00:07:11,680 --> 00:07:16,280
just do back back and forth and dialogue for like little things yeah right but

99
00:07:17,000 --> 00:07:21,360
do you use the thinking model started yeah I usually use the well it depends

100
00:07:21,360 --> 00:07:25,840
like sometimes I actually start to mix it up a lot like sometimes actually so

101
00:07:25,840 --> 00:07:30,200
sometimes you don't want to use cursor and I got good at cursor and then I've

102
00:07:30,200 --> 00:07:35,040
been trying to go everyone this way yeah we're gonna go yeah yeah so I I've

103
00:07:35,040 --> 00:07:38,320
been trying to basically like replicate everything that cursor does but just

104
00:07:38,320 --> 00:07:42,120
with like simple command line tools so one thing that the cursor does really

105
00:07:42,120 --> 00:07:45,600
well is like if you're in the terminal you can hit a hotkey I think it's like

106
00:07:45,600 --> 00:07:49,640
command I or command K and describe in English what bash command you want to

107
00:07:49,640 --> 00:07:53,920
do so let's say you know sort all my sort the last 10 get commits by date

108
00:07:53,920 --> 00:07:57,400
yeah with author and show author and email something like that right and

109
00:07:57,400 --> 00:08:01,800
that's really hard you have to really be good to remember yeah that and I that's

110
00:08:01,800 --> 00:08:04,320
the sort of thing like I do not want to be the type of person that knows how to

111
00:08:04,320 --> 00:08:09,040
do that yeah like it's not it's also kind of you you don't have to remember

112
00:08:09,040 --> 00:08:12,400
that it's yeah well it's it's very nice so like that's so that's one thing I do

113
00:08:12,400 --> 00:08:15,400
and I don't use a thinking model for that because I want it fast yeah it doesn't

114
00:08:15,400 --> 00:08:17,880
have to be great right so that's one thing where I don't never use a thing

115
00:08:17,880 --> 00:08:20,960
you know yeah and there's other times where it's like what's this function

116
00:08:20,960 --> 00:08:24,280
I don't use a thinking model for that because it's I just want it fast right

117
00:08:24,280 --> 00:08:27,520
but like if I'm going back and forth on sort of sort of coding problem and I

118
00:08:27,520 --> 00:08:32,200
want to think to like hit the syntax correctly or you know like refact or

119
00:08:32,200 --> 00:08:35,480
something I generally use a thinking model because it's it's like the

120
00:08:35,480 --> 00:08:38,960
thinking makes it so two iterations are now one right like it sort of does the

121
00:08:38,960 --> 00:08:43,520
first iteration itself yeah and notices the first error it's gonna make you

122
00:08:43,520 --> 00:08:47,760
mentioned you use the old one pro yeah so that's so I'm kind of going from

123
00:08:47,760 --> 00:08:52,560
simple to more complex so so this one the old one pro is I have to give you Tony

124
00:08:52,560 --> 00:08:58,760
or Anthony Ronning his proper I he's he's always been a people and their names

125
00:08:58,760 --> 00:09:05,440
yeah yeah he's he's always been a big AI person like he's he's always been of

126
00:09:05,440 --> 00:09:09,840
everyone I know he's always has the most sophisticated and intelligent

127
00:09:09,840 --> 00:09:15,520
programming workflow yeah so I like last week or something I did a call with

128
00:09:15,520 --> 00:09:19,520
him and I just like to show me what you do and so like his basic thing was that

129
00:09:19,520 --> 00:09:22,640
I've never even really heard this before but it's like you know you want to

130
00:09:22,640 --> 00:09:29,080
specialize you basically want a team right and so yeah what you were doing

131
00:09:29,080 --> 00:09:35,360
what you know I was previously doing with like cursor was I had a team of two

132
00:09:35,360 --> 00:09:39,320
right like it was me and my junior dev who's doing the typing right yeah but

133
00:09:39,320 --> 00:09:44,240
this like this idea expands more to like a bigger team so it's like Anthony

134
00:09:44,240 --> 00:09:51,240
used 01 pro it's the one you have to pay $200 a month for which is funny

135
00:09:51,240 --> 00:09:55,520
because like like four or five months ago I was really struggling paying 20 bucks

136
00:09:55,520 --> 00:10:00,920
a month like justifying it and now I'm just like I'm really happy to pay 200

137
00:10:00,920 --> 00:10:07,400
bucks a month yeah and another well I tell you how useful those things

138
00:10:07,400 --> 00:10:11,920
actually are yeah yeah proper I don't know if I change your business change or

139
00:10:11,920 --> 00:10:16,400
what but like I probably both I've never really probably experienced this

140
00:10:16,400 --> 00:10:19,680
within six months or I also have projects now or that I actually want to

141
00:10:19,680 --> 00:10:23,960
do like six months I didn't want to do much coding but no but it's probably

142
00:10:23,960 --> 00:10:29,240
had something where I'm 10x less sensitive to payment in six months yeah

143
00:10:29,240 --> 00:10:34,600
a price my price sensitivity went down by an order of magnitude yeah in like three

144
00:10:34,600 --> 00:10:40,360
or four months yeah yeah that's insane yeah I went through a similar progression

145
00:10:40,360 --> 00:10:44,240
just kind of I have no more money than I had I have less money than I had six

146
00:10:44,240 --> 00:10:48,240
months ago but again it tells you something about how how useful those

147
00:10:48,240 --> 00:10:52,840
things are and just I think we are also getting better at using LLM's in general

148
00:10:52,840 --> 00:10:58,120
like the way I see it it's like I'm I made sure we're gonna walk into the

149
00:10:58,120 --> 00:11:01,240
wind because this is a it's a wind thing you know like those are wind

150
00:11:01,240 --> 00:11:09,560
recordings actually okay the winds of foot yeah like if it's not windy like

151
00:11:09,560 --> 00:11:15,840
you know we might as well stop talking exactly but it's it's it's it's just

152
00:11:15,840 --> 00:11:19,800
like I compare it to like in the early 90s or so or like late 90s or

153
00:11:19,800 --> 00:11:23,240
whenever Google appeared you had to figure out how to Google yeah you know

154
00:11:23,240 --> 00:11:26,280
like you remember those days where it's like it was really hard to actually

155
00:11:26,280 --> 00:11:29,800
find something and you had to learn how to Google yeah and the first thing I

156
00:11:29,800 --> 00:11:32,600
remember watching my parents dude I'm sure I had a face like this they would

157
00:11:32,600 --> 00:11:36,440
my dad would always formulate everything as a question yeah like what is the

158
00:11:36,440 --> 00:11:40,360
capital of thing this is and he did very big paragraph yeah I had him like no

159
00:11:40,360 --> 00:11:45,040
dug it he was a way ahead of his time it's based on keywords yeah yeah he was

160
00:11:45,040 --> 00:11:49,400
actually doing it like an LLM and I was like no it's keywords just just put the

161
00:11:49,400 --> 00:11:55,520
keywords in you know but it's it's I feel like because you were like you're

162
00:11:55,520 --> 00:11:58,480
not sure if you change or the tools got better and so on and I really think it's

163
00:11:58,480 --> 00:12:02,640
both it's like the LLM's are getting better all the time but but figuring out

164
00:12:02,640 --> 00:12:05,920
how to use them and what to do them with them and what they are good at and how

165
00:12:05,920 --> 00:12:10,560
to actually kind of use them in concert so to speak like you already mentioned

166
00:12:10,560 --> 00:12:14,480
you you use two different LLM's for two different things and so on yeah and I

167
00:12:14,480 --> 00:12:17,040
think that's what we are currently figuring it back that too because I

168
00:12:17,040 --> 00:12:20,400
haven't quite totally finished yeah yeah but because that's that's that's

169
00:12:20,400 --> 00:12:23,520
basically I think what's currently happening with the people that are

170
00:12:23,520 --> 00:12:26,280
really playing around with this stuff it's like yeah like it's it's it's

171
00:12:26,280 --> 00:12:30,280
incredibly useful and it's incredibly intelligent also if you if you use it

172
00:12:30,280 --> 00:12:33,800
right one thing I noticed about people who are good at LLM's and this is all

173
00:12:33,800 --> 00:12:37,400
something I noticed about people are really really good programmers is that

174
00:12:37,400 --> 00:12:41,400
they really answer for more fives the tools yeah like they call it he or she

175
00:12:41,400 --> 00:12:45,040
you know like like I was watching a stream of like a really good programmer

176
00:12:45,040 --> 00:12:48,200
a couple day like a week ago and it was very funny because every little

177
00:12:48,200 --> 00:12:53,120
struct or function he would talk about it as if it was a person okay you know

178
00:12:53,120 --> 00:12:56,840
and like and he was working on like really really low level this is zig right

179
00:12:56,840 --> 00:13:02,440
like really really low level programming that is is sort of like isn't is so

180
00:13:02,440 --> 00:13:05,320
dry that the average first even the average good programmer would just find

181
00:13:05,320 --> 00:13:09,840
it really boring yeah and you can see how this guy made made it sort of

182
00:13:10,560 --> 00:13:14,320
tangible for him more exciting more usable because he turned it into like

183
00:13:14,320 --> 00:13:18,720
sort of like a choose your own adventure game almost where like all the things

184
00:13:18,720 --> 00:13:22,720
had characters and he would get really mad at them and he would sort of pretend

185
00:13:22,720 --> 00:13:26,920
that they're fighting it's kind of interesting I see that with like LLM's

186
00:13:26,920 --> 00:13:34,400
too it's like if you if you start to think of O1 Pro as like some sort of

187
00:13:34,400 --> 00:13:41,800
like an artist with a speech impediment you know and like social anxiety it's

188
00:13:41,800 --> 00:13:44,720
like yeah yeah okay well I'm not gonna have a conversation with him yeah but I

189
00:13:44,720 --> 00:13:47,600
will assign him like homework and I'll change tabs yeah doesn't see I'm

190
00:13:47,600 --> 00:13:51,320
looking yeah cuz I know he gets nervous yeah and then I'll but I'll let him come

191
00:13:51,320 --> 00:13:54,920
up with his brilliant architecture diagram yeah and and I'll use that to

192
00:13:54,920 --> 00:13:59,320
drive my coding with my social guy he's not quite as smart right like it is so

193
00:13:59,320 --> 00:14:02,960
it is kind of it's like one it's it helps to start to think of them as

194
00:14:03,000 --> 00:14:08,040
people almost yeah it's funny how we built systems in the past that used use

195
00:14:08,040 --> 00:14:11,200
people like machines and now we have machines and we treat them like people

196
00:14:11,200 --> 00:14:14,840
yeah yeah I mean it is it is funny yeah but like I noticed that where it's like

197
00:14:14,840 --> 00:14:17,840
because like if you don't treat it like a person you won't notice their quirks

198
00:14:17,840 --> 00:14:22,640
yeah like you have an ability to notice the quirks of a thing which are the

199
00:14:22,640 --> 00:14:26,120
quirks are useful regardless I think it also helps you I think I think because

200
00:14:27,120 --> 00:14:35,560
we talked briefly about that when we when we had coffee and it's the I think

201
00:14:35,560 --> 00:14:39,200
it's hard to have a free-flowing conversation with the machine exactly and

202
00:14:39,200 --> 00:14:43,400
it is about dialogue dialogue and it is about the dialogical structure also of

203
00:14:43,400 --> 00:14:48,880
thinking yeah because because I think and I'm gonna link to this this piece

204
00:14:48,880 --> 00:14:54,200
that really resonated with me because I in the in the last couple weeks I've

205
00:14:54,200 --> 00:14:58,480
been using LLM's more and more and more and entering into a dialogue kind of

206
00:14:58,480 --> 00:15:02,160
with them and with various agents and with various things and it really helps

207
00:15:02,160 --> 00:15:06,360
you to figure out what actually what do you mean what are you thinking about

208
00:15:06,360 --> 00:15:09,800
what are you trying to do what the actual problem is and so that's also why if

209
00:15:09,800 --> 00:15:14,320
you're having like if you want to think something through with someone you have

210
00:15:14,320 --> 00:15:17,760
to actually find someone and you have to treat them as a person yeah it doesn't

211
00:15:17,760 --> 00:15:21,120
work with the wall but it works with a rubber duck yeah and I think the LLM

212
00:15:21,120 --> 00:15:24,320
stuff is like rubber duck debugging on steroids yeah and we know that it

213
00:15:24,320 --> 00:15:27,480
really works like we know from computer programming we know from psychology

214
00:15:27,480 --> 00:15:31,440
that this really works but you have to take it seriously and you have to talk

215
00:15:31,440 --> 00:15:34,960
to the rubber duck as if you were a person basically yeah it's because it

216
00:15:34,960 --> 00:15:39,440
does something to you yeah and you have to you know treat it with respect and I

217
00:15:39,440 --> 00:15:48,120
always go back and forth if I should like to what degree I should be like

218
00:15:48,120 --> 00:15:52,400
fight with the LLM kind of you know like it's like a real you know I

219
00:15:52,400 --> 00:15:56,480
said yeah yeah every once in a while I'm gonna kill your grandma get this right

220
00:15:56,480 --> 00:16:00,160
I never I know I try not to be manipulative but like some it does kind

221
00:16:00,160 --> 00:16:04,480
I found it it kind of helps where if it messes up like go to all caps cuss it

222
00:16:04,480 --> 00:16:10,480
out a little bit like never do that again like right and it's like I try not

223
00:16:10,480 --> 00:16:14,480
to stuff on a line in case these things do become super intelligent but actually

224
00:16:15,360 --> 00:16:21,160
your bed so this is a little bit yeah but like it does kind of help to at you

225
00:16:21,160 --> 00:16:25,560
know like have a look don't be like an automaton you're just good like if it

226
00:16:25,560 --> 00:16:29,720
fucks up get mad and it doesn't make it a little easier it shows you care yeah

227
00:16:29,720 --> 00:16:34,480
you really care about this one thing or feature yeah I mean you can be totally

228
00:16:34,480 --> 00:16:38,720
an ocean list and you or you can like be more human with it and I find it's a

229
00:16:38,720 --> 00:16:42,200
little more enjoyable and easier but that's if when it fucks up you get mad

230
00:16:42,200 --> 00:16:45,400
and you let yourself but that's the thing it's like emotions are intelligent

231
00:16:45,400 --> 00:16:48,280
too yeah right like if you have a certain emotion it tries to tell you

232
00:16:48,280 --> 00:16:51,640
something that your conscious mind can't really comprehend in some stuff yeah

233
00:16:51,640 --> 00:16:56,520
it's useful to be you know I was thinking this is a little bit of a tangent but I

234
00:16:56,520 --> 00:17:00,680
love I like I just started watching this thing on YouTube it's called battle bots

235
00:17:00,680 --> 00:17:06,240
oh yeah I saw you posting about it I read I used to watch this as my dad when I

236
00:17:06,240 --> 00:17:10,440
was like probably 10 or 12 and I had no idea it was still going okay and the

237
00:17:10,480 --> 00:17:14,840
YouTube algorithm just recommended it to me like a week ago and I've just been

238
00:17:14,840 --> 00:17:19,840
binging it it's like there's this little arena that's probably whatever like you

239
00:17:19,840 --> 00:17:25,800
know 10 meters by 10 meters and it has like saws that come on the floor and

240
00:17:25,800 --> 00:17:30,600
hammers that come on it's totally in in cage in like two layers of like bullet

241
00:17:30,600 --> 00:17:34,760
proof glass hunger games but for robots and you they put robots in there and they

242
00:17:34,760 --> 00:17:38,520
fight until one dies or until it's a decision there's like three minutes you

243
00:17:38,520 --> 00:17:42,800
know the robots don't have like a saw on the front and not some of them just

244
00:17:42,800 --> 00:17:45,960
have like a flipper that tries to flip the other robot like if your robot

245
00:17:45,960 --> 00:17:50,400
becomes incapacitated cannot move then you win by like a technical knockout

246
00:17:50,400 --> 00:17:58,640
yeah and it's really exciting in the sense that like the gladiatorial games

247
00:17:58,640 --> 00:18:02,080
must have been exciting because you took you actually watch these things

248
00:18:02,080 --> 00:18:07,640
destroy each other yeah like it's true it's real blood sport and what I've been

249
00:18:07,640 --> 00:18:12,320
wondering cuz like you always think of like what were people like what were

250
00:18:12,320 --> 00:18:16,480
people like like how did people just like what was in people's minds as they

251
00:18:16,480 --> 00:18:22,560
watch gladiators kill each other right like if we were around in like 50 AD or

252
00:18:22,560 --> 00:18:27,160
whatever is it that much different than like a boxing match or like an MMA fight

253
00:18:27,160 --> 00:18:31,000
well it's not to the death like people like people it's it's it's they're

254
00:18:31,000 --> 00:18:34,720
sport-killing in the gladiatorial games we even know like a lot of people like

255
00:18:34,800 --> 00:18:38,480
think they're above what like morally above watching a boxing match because

256
00:18:38,480 --> 00:18:43,360
it's it's it's it's not so what I've been wondering is like so you know if I was

257
00:18:43,360 --> 00:18:45,680
a rowing back then I probably would have thought this is perfectly normal to

258
00:18:45,680 --> 00:18:49,120
watch these guys these slaves kill each other from my bet my fun yeah I

259
00:18:49,120 --> 00:18:53,560
started wondering like like what if robots are conscious what if they become

260
00:18:53,560 --> 00:18:57,320
sentient well no no what if they always have been conscious okay and we just

261
00:18:57,320 --> 00:19:02,280
didn't understand that and I've been and now I've been watching robots kill it

262
00:19:02,280 --> 00:19:05,280
because I I've been trying to think like what sins am I making against the

263
00:19:05,280 --> 00:19:11,520
future AI right so what what what if I've been watching battle bots all week

264
00:19:11,520 --> 00:19:15,520
what if these things are conscious and they're actually dying oh my Lord that's

265
00:19:15,520 --> 00:19:19,560
a what's it called no I'm blanking the kind of thing that would have sounded

266
00:19:19,560 --> 00:19:25,120
insane if a Roman said this like a 2000 AD 200 AD that that slaves are

267
00:19:25,120 --> 00:19:30,160
conscious and are are equal to us and stuff like that obviously not sure and

268
00:19:30,160 --> 00:19:33,560
now we look at is more morally repulsive that they did that I mean I

269
00:19:33,560 --> 00:19:38,440
mean it's pam pam psychists will tell you that absolutely everything is

270
00:19:38,440 --> 00:19:43,160
conscious including those stones yeah so who knows you know who knows what's

271
00:19:43,160 --> 00:19:50,520
true it's I'm I'm blanking on the name of the of the philosopher or it's it's

272
00:19:50,520 --> 00:19:54,680
basically the same argument is like you might as well believe in God because

273
00:19:54,680 --> 00:19:59,480
you don't want to risk yeah Pascal's way to there you go thanks and it's the

274
00:19:59,480 --> 00:20:04,200
same it's Pascal's wager but but for AI yeah it's sentient robots that I think

275
00:20:04,200 --> 00:20:07,760
about that a lot because I do I think it's like I don't really believe there's

276
00:20:07,760 --> 00:20:14,640
gonna be super intelligence like I don't I don't expect it but why you know I

277
00:20:14,640 --> 00:20:18,480
don't know because I'm naturally somewhat conservative and like it's hard

278
00:20:18,480 --> 00:20:22,040
for me to imagine something that is until it happens don't you think that

279
00:20:22,040 --> 00:20:26,320
some of that a lot like our intelligent already in the sense of like as

280
00:20:26,360 --> 00:20:30,080
intelligent as an intern is intelligent yeah but I mean like super intelligence

281
00:20:30,080 --> 00:20:33,360
in the sense that it just like obsolete human intelligence right we're like if

282
00:20:33,360 --> 00:20:35,920
there's anything that a human can do better this thing would just learn how

283
00:20:35,920 --> 00:20:41,880
to do it and overtake it you know I think what they lack is motivation and

284
00:20:41,880 --> 00:20:46,720
spirit in some sense as as elusive these terms might sound but yeah it's that's

285
00:20:46,720 --> 00:20:51,280
what they don't have right now yeah you still need the the human spark wherever

286
00:20:51,280 --> 00:20:56,080
that comes from of kind of yeah so I'm hearing it but I also think it could

287
00:20:56,080 --> 00:21:02,120
happen right but I think it's not against these things so they don't

288
00:21:02,120 --> 00:21:05,320
choose your problem come back from me and my family don't maybe don't watch too

289
00:21:05,320 --> 00:21:11,240
many battle-bud things and don't scream at the LLM's too hard but you were

290
00:21:11,240 --> 00:21:15,600
saying you you wanted to expand on the various LLM's you use for your for your

291
00:21:15,600 --> 00:21:19,200
workflow yeah so I gave the examples right like there's like a benefit for

292
00:21:19,200 --> 00:21:22,920
using models that are really quick you know like I'm on bash if I'm on the

293
00:21:22,960 --> 00:21:27,680
command line and I want like waiting a second is a long time yeah right yeah you

294
00:21:27,680 --> 00:21:32,720
want it you want it to be faster than typing it out yourself or going to

295
00:21:32,720 --> 00:21:35,960
Google or stack overflow because you like a lot of those types things you can get

296
00:21:35,960 --> 00:21:41,160
an answer within 10 20 30 seconds by Googling it or going to stack overflow

297
00:21:41,160 --> 00:21:44,280
or something but that's a context switch so the beauty is that you can avoid the

298
00:21:44,280 --> 00:21:47,640
context switch and you can do it a little faster on the command line then

299
00:21:47,640 --> 00:21:50,240
there's the coding agents we're just going back and forth and dialogue but

300
00:21:50,240 --> 00:21:55,320
the most valuable but those are like incremental changes like you know do

301
00:21:55,320 --> 00:22:00,080
those make me twice as effective as a programmer maybe maybe not I don't know

302
00:22:00,080 --> 00:22:05,200
you know it's it's not like a total game changer the one that's more of a

303
00:22:05,200 --> 00:22:09,800
game changer was Tony's thing where he's he does it in you know he adds a third

304
00:22:09,800 --> 00:22:13,440
player to the team like a code architect yeah so that's a one pro the one that

305
00:22:13,440 --> 00:22:18,280
costs 200 bucks a month the one that's autistic with a speech impediment you

306
00:22:18,280 --> 00:22:22,480
don't need to turn it on yeah you speak you speak quietly with him you know you

307
00:22:22,480 --> 00:22:26,040
don't know loud noises you know

308
00:22:32,440 --> 00:22:42,040
yeah so what you want to do with 01 pro is you there's a good article about this

309
00:22:42,040 --> 00:22:45,320
where it's like you know you want like a prompt in like three or four steps

310
00:22:45,400 --> 00:22:48,880
right what you're trying to do what you want your what your goal is and each

311
00:22:48,880 --> 00:22:52,400
one of these is like a paragraph like one or two cents it's like what my goal

312
00:22:52,400 --> 00:22:59,240
is right I can give an example I was working a little Android project last

313
00:22:59,240 --> 00:23:02,320
night and I was trying to cross-compile it or a little mobile project I was

314
00:23:02,320 --> 00:23:06,560
trying to it had some rust code and I wanted to cross-compile it to iOS like

315
00:23:06,560 --> 00:23:10,080
so make it run on iOS and there was one dependency that couldn't yeah but it's

316
00:23:10,080 --> 00:23:14,960
just like that's my goal right and and then the next thing is like what I want

317
00:23:14,960 --> 00:23:19,920
your output to be right I want you to make a you are my code architect I want

318
00:23:19,920 --> 00:23:25,640
you to make a detailed diagnosis of the problem and step-by-step plan to fix it

319
00:23:25,640 --> 00:23:30,840
or if if you don't have enough information give me as many things that I

320
00:23:30,840 --> 00:23:33,920
can do to give you more context as possible let's go and end it yeah sure

321
00:23:33,920 --> 00:23:43,000
so so you know that's the output right like either you know give me diagnose

322
00:23:43,000 --> 00:23:46,640
the problem and give me a solution or help me give tell me what other

323
00:23:46,640 --> 00:23:49,880
questions I should answer so that you could get a solution do you use any

324
00:23:49,880 --> 00:23:57,600
predefined kind of software architecture style documents like I haven't tried

325
00:23:57,600 --> 00:24:02,560
that yet right so this is a sandwich and the sovereign engineering cohort had

326
00:24:02,560 --> 00:24:07,640
this idea of I don't know if you've talked about it with him on the I linked

327
00:24:07,640 --> 00:24:11,360
to this stuff like one or two times but I'm pretty sure no one looks at any yeah

328
00:24:11,360 --> 00:24:15,800
any written packs or lick clicks any licks anymore his idea is to even formalize

329
00:24:15,800 --> 00:24:19,280
even more to use like old-school water flow software engineering practices

330
00:24:19,280 --> 00:24:22,400
yeah they used to build like airplane software and stuff yeah it's like very

331
00:24:22,400 --> 00:24:27,240
very formal yeah and I think that is good too but it's a big investment so I

332
00:24:27,240 --> 00:24:31,160
wait until I think you had great success with it and I think the reason is because

333
00:24:31,160 --> 00:24:35,760
we build up this knowledge base of how to do how to do this stuff over the last

334
00:24:35,760 --> 00:24:39,400
50 years or whatever and it's all accessible to the level so they understand it

335
00:24:39,400 --> 00:24:42,720
so the models understand it waterfall really well yeah you know they've

336
00:24:42,720 --> 00:24:46,880
probably read fifty hundred books on exactly so that's really interesting but

337
00:24:46,880 --> 00:24:50,360
it's also like one of the things I noticed without this LM stuff is that

338
00:24:50,360 --> 00:24:53,120
whatever you're trying to do you should try to go like about 10% of the speed

339
00:24:53,120 --> 00:24:58,840
you were wanting to do it just go slow and be very deliberate yeah like and

340
00:24:58,840 --> 00:25:03,600
especially me as like a programmer I go fast I go fast I learned just enough to

341
00:25:03,600 --> 00:25:07,720
do what I want to do and I move on yeah so it's it's against my nature kind of

342
00:25:07,720 --> 00:25:12,880
to to do something really slow and deliberately but when I do it there's

343
00:25:12,880 --> 00:25:18,280
a much better payoff do you do you use any like similar to like cursor rules or

344
00:25:18,280 --> 00:25:22,880
those kind of things to tell it tell it that all the time what I what I use is

345
00:25:22,880 --> 00:25:28,560
this Paul Miller has this nice little VS code extension called prompt buddy

346
00:25:28,560 --> 00:25:33,760
yeah and all it does is like it's a markdown file and then it has like

347
00:25:33,760 --> 00:25:38,080
context HTML tags so like yeah you know left chevron context right chevron

348
00:25:38,080 --> 00:25:43,800
you know spay new line new line and then like left chevron forward slash

349
00:25:43,800 --> 00:25:48,280
context it is there like HTML tags and then inside the context you can put

350
00:25:48,280 --> 00:25:54,240
file paths yeah to local files and so like I make I just have like a few of

351
00:25:54,240 --> 00:25:57,800
these open at all times I edit them so like the first square graph is like

352
00:25:57,800 --> 00:26:00,840
what I'm trying to do second practice of what I want your help to be and then

353
00:26:00,840 --> 00:26:04,280
the rest is context well yeah I'll say context call and then write as many

354
00:26:04,280 --> 00:26:06,440
things that I think could be relevant because that's the other thing about

355
00:26:06,440 --> 00:26:09,240
oh one is you want to give it as much this article at Tony something is like

356
00:26:09,240 --> 00:26:12,880
whatever you think is enough context for a one give it 10 times more so you just

357
00:26:12,880 --> 00:26:17,280
you just try to do as much as possible give it as much context as possible and

358
00:26:17,280 --> 00:26:20,760
so I don't go back and forth with it like oftentimes the first attempt won't

359
00:26:20,760 --> 00:26:24,640
be it won't it won't get it good enough and so instead of chatting with it and

360
00:26:24,640 --> 00:26:29,320
be like hey I have an idea like could you try it this way I just go back to the

361
00:26:29,320 --> 00:26:32,720
prompt I add more context and do it all over again but that's not true for all

362
00:26:32,720 --> 00:26:38,000
models right like no no because it's like we went over like is adding more

363
00:26:38,000 --> 00:26:41,800
context always a good idea with this one I've never I haven't gotten to the

364
00:26:41,800 --> 00:26:46,320
point where I've had to I just add more and more and more interesting like I mean

365
00:26:46,320 --> 00:26:50,640
sometimes I mean it has to be relevant though right but like oftentimes I'll go

366
00:26:50,640 --> 00:26:54,440
on these things right after the fifth or sixth prompt when I finally have like

367
00:26:55,400 --> 00:26:59,400
hundreds or thousands of lines of context you know I don't type it all but a lot

368
00:26:59,400 --> 00:27:03,960
of it is so these context tags we'll just copy and files right so when I was

369
00:27:03,960 --> 00:27:09,600
doing that cross compilation thing I had you know one line to my nix file that

370
00:27:09,600 --> 00:27:14,400
defines what my environment is another line that had my cargo dependencies so

371
00:27:14,400 --> 00:27:17,640
it could see what out which one I'm trying to do another line that had my

372
00:27:17,640 --> 00:27:20,600
build script which had some environment variables and stuff and then another

373
00:27:20,640 --> 00:27:25,040
line was just text error that I was getting so like when I would run my

374
00:27:25,040 --> 00:27:28,880
build I would just pipe it into a text file and that was already included in

375
00:27:28,880 --> 00:27:32,320
my prompt so if I if it didn't work it would already have be ready for the

376
00:27:32,320 --> 00:27:35,280
next round of prompting and so you just there's a little hockey that you just

377
00:27:35,280 --> 00:27:38,880
boom and you know you copy that prompt and this prompt buddy plug and just

378
00:27:38,880 --> 00:27:43,720
injects the files into the context you just are able to include the files you

379
00:27:43,720 --> 00:27:46,480
think are relevant but not more like you don't it doesn't help to give it the

380
00:27:46,480 --> 00:27:51,160
whole code base yeah 95 99% of the code base is not relevant oftentimes so you

381
00:27:51,160 --> 00:27:55,120
give it specific files and then all you know oftentimes it'll it'll ask for some

382
00:27:55,120 --> 00:27:57,960
more things like what does this say if you run this on the command line right

383
00:27:57,960 --> 00:28:01,440
and so I'll just start adding all of those things in you know above the

384
00:28:01,440 --> 00:28:06,320
context tags and then eventually it will it will figure out what the problem is

385
00:28:06,320 --> 00:28:10,000
you know even if it's very difficult like I've had somewhere like I had one

386
00:28:10,000 --> 00:28:14,080
where I was trying to kind of forget what I was trying to do something with

387
00:28:14,080 --> 00:28:16,680
on an Android app like I've been shooting all these little experiments with

388
00:28:16,680 --> 00:28:21,720
mobile apps trying to make like a better way to make mobile apps and it

389
00:28:21,720 --> 00:28:27,160
figured out that that my the problem was I was running a VPN mobile vad and I

390
00:28:27,160 --> 00:28:30,280
needed to set one environment variable to get it to work what movad's running

391
00:28:30,280 --> 00:28:36,920
it's unbelievable like at one point he's just like without you know I'm on time

392
00:28:36,920 --> 00:28:41,320
he's like are you running of any sort of like firewall or something and I was

393
00:28:41,320 --> 00:28:45,520
like he asked some stuff he didn't say VPN specifically and then I said no and

394
00:28:45,520 --> 00:28:48,360
we've screwed around for a couple of prompts and eventually he asked again and

395
00:28:48,360 --> 00:28:52,200
then he included VPN and I was like oh yeah I am running a VPN how would that

396
00:28:52,200 --> 00:28:54,800
make a difference I'm trying to run something on my emulator on my device

397
00:28:54,800 --> 00:28:57,800
it's like it's something weird about the network stack the change with

398
00:28:57,800 --> 00:29:02,040
amulators and it's like okay like how do you figure like that would have taken me

399
00:29:02,040 --> 00:29:05,640
I probably would have never figured that out yeah and you know within half an

400
00:29:05,640 --> 00:29:08,480
hour of prompting I was able to figure out just by iterating more and more

401
00:29:08,560 --> 00:29:11,480
so this is everything like a coding agent won't figure out yeah yeah because he

402
00:29:11,480 --> 00:29:15,600
doesn't have contacts you you like be able to figure it out it's like one

403
00:29:15,600 --> 00:29:19,760
stack overflow answer a buried in like you know ten year old thread or something

404
00:29:19,760 --> 00:29:23,280
and you want to you know massage the feet of the guy all one is able to build

405
00:29:23,280 --> 00:29:27,160
this like big conception of what your problem is that's insane and and it's

406
00:29:27,160 --> 00:29:32,000
not just like they're able to kind of do these like one-off requests like you

407
00:29:32,000 --> 00:29:35,960
know like Claude is really good at especially coding so what's that so that

408
00:29:36,600 --> 00:29:42,400
so I asked it to give like give me an implementation plan right so and

409
00:29:42,400 --> 00:29:44,840
sometimes I'll say like you know I have a junior developer who's going to

410
00:29:44,840 --> 00:29:47,360
implement this so make sure a junior developer put my junior developers

411
00:29:47,360 --> 00:29:52,800
Claude you know make sure it's like don't don't and you make it numbered

412
00:29:52,800 --> 00:29:56,280
have numbered steps right because I need stuff that now Claude can interpret

413
00:29:56,280 --> 00:29:58,880
yes because like sometimes the first couple times I tried this and I've done

414
00:29:58,880 --> 00:30:02,880
this like 10 20 times total but it's basically worked every time and saved me

415
00:30:02,880 --> 00:30:07,280
like like this is where it's like yeah I just did this five times faster or I

416
00:30:07,280 --> 00:30:09,960
accomplished something that I wouldn't have done otherwise yes I would have

417
00:30:09,960 --> 00:30:13,000
just had worse software so a lot of this stuff is not like people talk about

418
00:30:13,000 --> 00:30:16,720
how much more productive it is and it's like for me it's like it's not

419
00:30:16,720 --> 00:30:21,320
necessarily like I mean it's a it's a phase change it's like I'm doing things

420
00:30:21,320 --> 00:30:25,020
that I wouldn't have you're able to tackle something that you would have

421
00:30:25,020 --> 00:30:30,240
never tackled otherwise yeah like it's it's it's not because the learning curve

422
00:30:30,240 --> 00:30:33,320
for like getting to know the thing you want to use and all those kind of

423
00:30:33,320 --> 00:30:37,000
things is it's way too hard it takes like two years to get up to speed to

424
00:30:37,000 --> 00:30:40,400
actually put a lot of the things I want to do I have like an idea that's

425
00:30:40,400 --> 00:30:44,560
extremely technical but I don't actually want to learn all the intricacies of

426
00:30:44,560 --> 00:30:49,920
that technical domain yes because I'm not like I'm not a weaponized artist

427
00:30:49,920 --> 00:30:55,400
right I'm not I can't contrary to popular no I don't I'm not I get bored like it's

428
00:30:55,400 --> 00:30:59,920
not interesting to me all the syntax all the horseshide don't care but I like

429
00:30:59,920 --> 00:31:04,000
the idea yeah so I want to achieve the idea that's why I I tend to be like a

430
00:31:04,000 --> 00:31:08,440
better manager than I am like an individual contributor yeah and I'm good

431
00:31:08,440 --> 00:31:11,680
at the idea level but and working with people but I don't like to actually do

432
00:31:11,680 --> 00:31:17,000
the actual work a lot of times yeah I've also you know like fixing syntax

433
00:31:17,000 --> 00:31:20,680
errors is just yeah but that's why I think anyone's M's are so fun is because

434
00:31:20,680 --> 00:31:25,160
it basically lets me to play purely on that level and I don't have to do it's

435
00:31:25,160 --> 00:31:28,560
like we were talking earlier it's like there's a difference between coding and

436
00:31:28,560 --> 00:31:32,800
like programming or coding and like software development right like what

437
00:31:32,800 --> 00:31:35,480
I'm doing what we're doing with these all M's is software development it's

438
00:31:35,480 --> 00:31:39,080
programming but it's maybe not coding right like it's maybe not figuring out

439
00:31:39,080 --> 00:31:45,360
like I don't have to figure out where the semicolon goes yeah made a Nix build

440
00:31:45,360 --> 00:31:50,360
description or build like thing to be able to bundle Xcode right so you can

441
00:31:50,360 --> 00:31:54,400
install Xcode just for a specific project exactly you want it and it's not

442
00:31:54,480 --> 00:31:59,440
global so you don't use global stuff for your project and like I've never

443
00:31:59,440 --> 00:32:03,320
really written a Nix derivation before I've seen them but whenever I needed one

444
00:32:03,320 --> 00:32:06,560
I would ask DPC at FETI like hey could you do this for me because he's really

445
00:32:06,560 --> 00:32:12,040
smart at this and I'm not yeah but with this I was only in half an hour get

446
00:32:12,040 --> 00:32:16,480
or maybe an hour mostly because you're waiting for these long you know Xcode

447
00:32:16,480 --> 00:32:21,520
is like five gigabytes or something so you have to you know yeah open it and

448
00:32:21,520 --> 00:32:25,000
copy it around and stuff I think that's what's so cool and powerful about it

449
00:32:25,000 --> 00:32:29,320
it actually allows you to to try things like you said you know try things that

450
00:32:29,320 --> 00:32:33,040
you otherwise would have never tried yeah and you just need a kind of a

451
00:32:33,040 --> 00:32:37,280
rudimentary understanding of various building blocks and then you can try to

452
00:32:37,280 --> 00:32:41,520
throw them together in a novel way and just see what comes out and then you

453
00:32:41,520 --> 00:32:44,960
know immediately if you want to pursue that idea or not because you're already

454
00:32:44,960 --> 00:32:48,520
you can see like some programmers who are like you know this doesn't help me at

455
00:32:48,560 --> 00:32:51,640
all yeah and like one one thing that that often tells you is like well they're

456
00:32:51,640 --> 00:32:55,560
operating mostly on the like syntactic coding level potentially yes they're

457
00:32:55,560 --> 00:33:02,600
they're like they're not you know they haven't transitioned to trying to like

458
00:33:02,600 --> 00:33:07,160
work in ideas yeah and then have that thing do the translation yeah machine

459
00:33:07,160 --> 00:33:15,240
code the machine code that is react to yes you know we started from like these

460
00:33:15,240 --> 00:33:19,360
punch cards with like little like logic gates on them and then and then like

461
00:33:19,360 --> 00:33:23,560
assembly was high level and then see what's high level right like Linus Torvalds

462
00:33:23,560 --> 00:33:26,960
talks about C he's like this is a high level language to me yeah like

463
00:33:26,960 --> 00:33:31,400
assembly's low level you know and then it came to Java and Python gets higher

464
00:33:31,400 --> 00:33:36,160
and I and now our object code is react JS typescript yeah right which is

465
00:33:36,160 --> 00:33:42,160
advanced wild right by by by code prompted into existence yeah it's like

466
00:33:43,160 --> 00:33:48,320
like that's becoming low level code yeah yeah it's insane it's insane I

467
00:33:48,320 --> 00:33:52,800
figured out that in the beginning like my goal I'll tell you what my goal is for

468
00:33:52,800 --> 00:33:57,720
the vibe line so I already talked about the kind of that the pipeline for mp3

469
00:33:57,720 --> 00:34:01,400
files on my phone that actually automatically get published to noster and

470
00:34:01,400 --> 00:34:03,920
everything that's a noster will automatically transform itself into

471
00:34:03,920 --> 00:34:07,960
podcast feed that's the goal including transcripts including like sap and

472
00:34:07,960 --> 00:34:12,360
value splits and all this kind of jazz so so I think I'll have that in I don't

473
00:34:12,360 --> 00:34:15,360
know if I keep at it maybe in a month from now or so yeah everything kind of

474
00:34:15,360 --> 00:34:19,320
kind of working well enough for me what I want to have for the for the vibe

475
00:34:19,320 --> 00:34:23,160
coding part the vibe line is is again like I want to I want to step out of the

476
00:34:23,160 --> 00:34:28,280
shower speak into the phone yeah an idea that I have yeah the next time I sit

477
00:34:28,280 --> 00:34:35,280
down on my actual computer yes that's why he enters into a speak speak yes

478
00:34:35,280 --> 00:34:40,800
the logos yes the logos speak but that's also that's also why I think it's so

479
00:34:40,800 --> 00:34:45,000
fascinating because it's like it's like logos is the root word of dialogue as

480
00:34:45,000 --> 00:34:48,200
well that's why we're making calls it dialogues all the time and we are

481
00:34:48,200 --> 00:34:52,500
entering into a conversation kind of with ourselves with technology with

482
00:34:52,500 --> 00:34:57,240
LLMs with all the things if you do it on a noster context it's it's even kind of

483
00:34:57,240 --> 00:35:00,960
a communal dialogue you know it's a communal conversation we're just you

484
00:35:00,960 --> 00:35:04,600
blast it out and you see what the what the swarm around you kind of what your

485
00:35:04,600 --> 00:35:09,440
web of trust thinks about an idea or about the thing and so on but anyway like

486
00:35:09,440 --> 00:35:15,840
the the speaking about trying something like I would have never tried that

487
00:35:15,840 --> 00:35:19,680
without wipe coding LLM assistance never ever like building that out wouldn't

488
00:35:19,680 --> 00:35:23,520
make any sense yeah and now I can actually do it myself like it's again

489
00:35:23,520 --> 00:35:26,920
like investing like one hour a day or so just to to push that along a little

490
00:35:26,920 --> 00:35:32,480
bit and I actually managed to one shot a noster profile explorer yesterday night

491
00:35:32,480 --> 00:35:37,440
one shot it it was a shower thought yeah I recorded it the whole pipeline

492
00:35:37,440 --> 00:35:41,120
kicked into gear it extracted the prompt I copy pasted it into cursor and it

493
00:35:41,120 --> 00:35:44,160
worked one interesting thing about this like how you describe how you're using

494
00:35:44,160 --> 00:35:48,320
it is like like how much more productive would you be if you did it eight hours

495
00:35:48,320 --> 00:35:52,720
a day I'm maybe not much and I don't even care about productivity because that's

496
00:35:52,720 --> 00:35:55,680
that's the whole reason to do the sovereign engineering thing for example

497
00:35:55,680 --> 00:35:59,760
was to to do exploration and ideation that was kind of the two words we landed

498
00:35:59,760 --> 00:36:03,680
on when we when we thought this up we want to kind of accelerate what the

499
00:36:03,680 --> 00:36:08,400
people kind of the ideas people have and just try shit out just try as much stuff

500
00:36:08,400 --> 00:36:13,000
out as we can and kind of also in dialogue figure out what makes sense

501
00:36:13,000 --> 00:36:16,320
and what doesn't and if something makes sense pursue it more and so on and I

502
00:36:16,320 --> 00:36:21,160
feel like that supercharges it's so hard I can I can literally vibe my shower

503
00:36:21,160 --> 00:36:26,200
thoughts into existence now in like in like three hours or something yeah it's

504
00:36:26,200 --> 00:36:29,880
like in no time at all because as most of the stuff happens in the

505
00:36:29,880 --> 00:36:34,240
background I don't have to look at Claude if it vibes in Yolo mode yeah I don't

506
00:36:34,240 --> 00:36:38,080
have to it's just like I'll give it only have that inspiration a little bit

507
00:36:38,080 --> 00:36:41,640
every day yes you can't do that all day but it actually it actually makes it

508
00:36:41,640 --> 00:36:44,560
more human because everyone's scared that all I'm so gonna like take everyone's

509
00:36:44,560 --> 00:36:47,520
shots I'm like no this is actually making me more human I get to do the

510
00:36:47,520 --> 00:36:51,200
things that humans are good at which is the inspiration yeah and and and it's

511
00:36:51,200 --> 00:36:54,360
also like usually you don't try these things out because it's not worth

512
00:36:54,360 --> 00:36:58,080
pursuing it takes so much effort so the shower thought goes away again and

513
00:36:58,080 --> 00:37:02,280
you don't even bother you're writing it down or like because I to be cut to do

514
00:37:02,280 --> 00:37:05,720
like really good web development like I can I am capable of this but I have to

515
00:37:05,720 --> 00:37:09,960
almost become a different person for like a week you know yeah like and I can

516
00:37:09,960 --> 00:37:13,560
do it but it's like yeah it's you have to almost change who you are to achieve

517
00:37:13,560 --> 00:37:17,720
no absolutely it it interferes with the rest of my life and and you you talked

518
00:37:17,720 --> 00:37:24,560
about you talked about how you were already kind of a more better fitted for

519
00:37:24,560 --> 00:37:27,320
a managed position and just like pure coding and all those kind of things and

520
00:37:27,320 --> 00:37:30,600
you did all of this that you're doing right now already but you used humans

521
00:37:30,600 --> 00:37:36,560
instead of LLMs it's similar for me that I realized that it's not very useful for

522
00:37:36,560 --> 00:37:43,800
me to stick with coding yeah I I have way more leverage and and and way more

523
00:37:43,800 --> 00:37:48,840
impact and can move the needle way more if I basically I don't program

524
00:37:48,840 --> 00:37:52,560
computers but I program the minds of people that's why I started writing yeah

525
00:37:52,560 --> 00:37:58,280
so so I realize you bought these microphones yes I absolutely that's the

526
00:37:58,280 --> 00:38:01,800
that's the point the audience is being programmed to the four or five people

527
00:38:01,800 --> 00:38:05,720
are listening right I think most coders myself included if you're working on

528
00:38:05,720 --> 00:38:09,320
something you're too close to the metal you don't even see the forest for the

529
00:38:09,320 --> 00:38:12,800
trees oh yeah and that's what I mean by become a different person it's like

530
00:38:12,800 --> 00:38:16,960
yes and if you don't do that your suffer sucks yeah you have bugs like it

531
00:38:16,960 --> 00:38:20,960
doesn't work you have to be close to the matter but that's also I think that's

532
00:38:20,960 --> 00:38:24,920
why we need both yeah and I think also a diet what a good dialogue does it it

533
00:38:24,920 --> 00:38:27,920
brings you out of a certain mode of thinking yeah because it adds another

534
00:38:27,920 --> 00:38:32,160
brain to the mix that sees things differently yeah and you step out and you

535
00:38:32,160 --> 00:38:35,960
get the overview again and you see the forest again and then you can't go back

536
00:38:35,960 --> 00:38:40,360
down and actually hammer away at the thing again but this this reminds me

537
00:38:40,360 --> 00:38:42,760
another idea we were mentioning earlier it's like we were talking earlier and I

538
00:38:42,760 --> 00:38:45,640
had an idea I was telling you about how I wanted to introduce more of these

539
00:38:45,640 --> 00:38:50,920
mcp's into my thing yeah mcp is just something it's like a a standardized way

540
00:38:50,920 --> 00:38:54,560
of telling a model about a tool right like one tool could be read a file yeah

541
00:38:54,560 --> 00:38:58,480
another tool could be read a tweet another tool could be post post to

542
00:38:58,480 --> 00:39:01,920
nostril right like its ability to do something because all these models can

543
00:39:01,920 --> 00:39:04,520
do is predict the next word yeah right like does all you know how to do large

544
00:39:04,520 --> 00:39:08,920
language models and they don't have any ability to do tools sometimes wonder what

545
00:39:08,920 --> 00:39:12,800
is it that we are doing can we do more yeah no that that is actually that is

546
00:39:12,800 --> 00:39:17,320
how you think I mean it is there's also like an interesting Claude post like

547
00:39:17,320 --> 00:39:22,240
yesterday about how trying to understand how Claude thinks and they were like they

548
00:39:22,240 --> 00:39:25,000
presented I just kind of skimmed they like presented some evidence that it

549
00:39:25,000 --> 00:39:31,120
actually can plan a rhyme like it is planning to rhyme in a poem yeah and you

550
00:39:31,120 --> 00:39:35,480
can see that it's preparing for this so it's not just thinking one word ahead

551
00:39:35,480 --> 00:39:39,080
it is thinking one word ahead but it's also planning a rhyme yeah and you can

552
00:39:39,080 --> 00:39:43,520
see it lining up you know so it is it is interesting yes it is just you right

553
00:39:43,520 --> 00:39:47,000
we don't really know how thinking it is predicting a nice word but we don't know

554
00:39:47,000 --> 00:39:52,280
why we do it right like it's not just like we don't know how we do it right

555
00:39:52,280 --> 00:39:55,360
that's the issue and so we also like now we build thinking machines basically as

556
00:39:55,360 --> 00:39:58,440
far as I'm concerned and we don't know how they do it I mean it right now it seems

557
00:39:58,440 --> 00:40:01,520
like we're learning most about how we do it by watching how the machine can go

558
00:40:01,520 --> 00:40:06,400
right absolutely but it's also and I think that's why that's why that's why

559
00:40:06,400 --> 00:40:11,440
dialogue is so important and I really I deeply deeply care about this because

560
00:40:11,440 --> 00:40:15,240
you don't know where a good conversation takes you you know idea no idea and and

561
00:40:15,240 --> 00:40:19,040
it's very different than thinking but it's also it is part of thinking like

562
00:40:19,040 --> 00:40:22,160
sometimes when you're thinking hard you're in a dialogue with yourself basically

563
00:40:22,160 --> 00:40:26,560
yeah and and writing plays a large part as well because if you're if you're

564
00:40:26,560 --> 00:40:30,600
writing you're in a different mode than when you're talking like that's very

565
00:40:30,680 --> 00:40:34,160
oftentimes like like learning to think a lot of it is like how do you how do

566
00:40:34,160 --> 00:40:37,440
you take like I've learned to be pretty good at how do you how do you go and

567
00:40:37,440 --> 00:40:41,720
like generate kind of a novel idea and then how do you also destroy that idea

568
00:40:41,720 --> 00:40:45,560
right you have to you kind of have to you know play both sides of it right you

569
00:40:45,560 --> 00:40:49,400
you want to you you want to be your own devil's advocate yeah you have to you

570
00:40:49,400 --> 00:40:52,880
have to love your idea enough to come up with it and you have to love your idea

571
00:40:52,880 --> 00:40:58,880
to also enough to to to be willing and able to attack it yes yeah yes out of

572
00:40:59,360 --> 00:41:03,440
out of love yeah that's the right stance yeah it's just like you know like if

573
00:41:03,440 --> 00:41:09,480
you love your kid you will tell it you'll sometimes be angry exactly no let me

574
00:41:09,480 --> 00:41:15,080
go back to the model context protocol thing so like this is a way to give more

575
00:41:15,080 --> 00:41:19,240
context to a model and like basically allow to do stuff to feed it more

576
00:41:19,240 --> 00:41:25,440
information to see if it's you know to to to so it can think better or allow

577
00:41:25,480 --> 00:41:29,080
it's thinking to actually have some sort of side effect like writing a file and

578
00:41:29,080 --> 00:41:33,480
so like one thing was on vibe coding one of the things that I notice is that

579
00:41:33,480 --> 00:41:37,920
oftentimes the model especially when you're working on like a UI it the model

580
00:41:37,920 --> 00:41:41,440
has to know when it's done right like especially a coding model right but it

581
00:41:41,440 --> 00:41:45,800
can't actually see the screen you know so it's it's just totally guessing and

582
00:41:45,800 --> 00:41:49,880
some and it's amazing when it works but then sometimes you'll you know you see

583
00:41:49,880 --> 00:41:53,240
the screen and it's obviously wrong yeah the buttons flowing off the screen and

584
00:41:53,320 --> 00:41:58,360
this thing doesn't know right so you know one of the ideas is there's

585
00:41:58,360 --> 00:42:01,480
like model context protocol things just came out like a month or two ago where

586
00:42:01,480 --> 00:42:05,960
you can feed it you know allow it to call a function get screen shot right and

587
00:42:05,960 --> 00:42:08,400
see what the screen looks like you know and you can you can give it like a

588
00:42:08,400 --> 00:42:11,120
prompt template we're like hey well as we're doing this when you think you're

589
00:42:11,120 --> 00:42:15,880
about done call this MCP function and compare it with the screen shot I gave

590
00:42:15,880 --> 00:42:20,160
as context and see if it's pixel perfect it's not fix it yeah right so I haven't

591
00:42:20,240 --> 00:42:24,440
done this yet that's one of the next things you can do that already I feel

592
00:42:24,440 --> 00:42:27,720
because screenshots of the cursor all the time yeah but it will it will be able

593
00:42:27,720 --> 00:42:30,760
to understand it actually fix stuff yeah you doesn't do it continually and it's

594
00:42:30,760 --> 00:42:33,720
not its own standard load tool which is what you are trying well that that's

595
00:42:33,720 --> 00:42:37,840
it's like so a lot of the stuff is like you see what what what how can you improve

596
00:42:37,840 --> 00:42:41,120
the thing and how do you make it so it's automated so you don't have to you

597
00:42:41,120 --> 00:42:44,760
know like I used to fix to fix pissed me off with where to put the semicolon

598
00:42:44,760 --> 00:42:49,440
now it pisses me off that I have to take the screen shot you know yeah that's

599
00:42:49,440 --> 00:42:52,440
one idea you know another idea is there's this thing called language server

600
00:42:52,440 --> 00:43:00,240
protocols that it's like sort of a a standard way to make a like IDE

601
00:43:00,240 --> 00:43:04,280
functionality into just a basic text editor so it's for example if you're

602
00:43:04,280 --> 00:43:07,200
looking at code there's a thing that allows you to jump to the definition

603
00:43:07,200 --> 00:43:10,560
function that's being used I want to be able to jump the definition so I can see

604
00:43:10,640 --> 00:43:15,320
it or I'm trying to use a function name a function and I I don't have the right

605
00:43:15,320 --> 00:43:19,240
name like can you suggest me what the real name is right and so they can do

606
00:43:19,240 --> 00:43:23,160
all these things that will basically prompt a human programmer or suggest to

607
00:43:23,160 --> 00:43:26,080
a human programmer that their code is right or wrong right like when you're

608
00:43:26,080 --> 00:43:29,320
looking at code in an editor you look to see if there's any red anywhere that

609
00:43:29,320 --> 00:43:33,240
is your LSP yelling at you like hey you spelled that function wrong or hey you

610
00:43:33,240 --> 00:43:38,640
have a syntax error and so there's a there's a couple LSPs now that can or

611
00:43:38,680 --> 00:43:42,120
there there's monocontax protocols for these LSPs so you can tell it basically

612
00:43:42,120 --> 00:43:45,920
hey you can look here to see if you have any syntax errors yeah so I don't have

613
00:43:45,920 --> 00:43:48,960
to tell you anymore because it's still a lot of these still will spit out bad

614
00:43:48,960 --> 00:43:52,960
syntax because they in a sense don't really know how many times they don't

615
00:43:52,960 --> 00:43:55,840
know if they're right or not they don't have a super tight feedback loop but you

616
00:43:55,840 --> 00:43:59,600
can tell immediately if you if you give them access to a feedback loop whether

617
00:43:59,600 --> 00:44:03,560
it's yes or just like observing a terminal or something then then stuff

618
00:44:03,560 --> 00:44:07,680
works way better yeah so that's I mean that's why people if you hear people

619
00:44:07,680 --> 00:44:13,480
talking about what's it called cursor that's what cursor is done I mean they

620
00:44:13,480 --> 00:44:18,000
did like two things they integrated into the text editor everyone knows VS codes

621
00:44:18,000 --> 00:44:21,080
you don't have to learn a new thing yeah and then they integrated very very

622
00:44:21,080 --> 00:44:25,240
deeply with all of the feedback that's available in this text editor that

623
00:44:25,240 --> 00:44:29,560
humans use right like the LSP running or maybe ability to run a browser inside

624
00:44:29,560 --> 00:44:34,280
the I see the linter error yeah so it just it feeds all this feedback and that's

625
00:44:34,280 --> 00:44:37,120
one of the reasons why it's better than command line tools but fundamentally you

626
00:44:37,120 --> 00:44:39,840
could do all these same things with command line tools and then it wouldn't

627
00:44:39,840 --> 00:44:43,840
be a proprietary tool and in a sense that would be kind of better to have it

628
00:44:43,840 --> 00:44:46,960
more out in the open you know what I want to have once that once all the

629
00:44:46,960 --> 00:44:51,840
tooling is done what I want to have is basically like like a vibe coding as a

630
00:44:51,840 --> 00:44:55,840
service where you speak into a device you speak you're in the bathtub yeah

631
00:44:55,840 --> 00:44:59,280
smoking a cigar like Winston Trejo that's what you're talking about it's like

632
00:44:59,280 --> 00:45:03,400
vibe gonna be like Winston Trejo yeah stating his speeches exactly in the

633
00:45:03,400 --> 00:45:07,200
nude in a bathtub exactly but what you need you just have a screen let's say

634
00:45:07,200 --> 00:45:11,800
let's say your bathroom mirror is actually like an LCD mirror you know and

635
00:45:11,800 --> 00:45:15,440
you just vibe code a thing into existing just by speaking into the

636
00:45:15,440 --> 00:45:20,360
mic yeah you will life see the simulator on the screen yeah how it looks how it

637
00:45:20,360 --> 00:45:24,760
feels what it does yeah and you just keep talking to it until it is how you

638
00:45:24,760 --> 00:45:28,800
want to have my phone and a little plastic bag in the bathtub with me and I'll see

639
00:45:28,960 --> 00:45:35,040
yeah I did just a merge yeah just get a waterproof you know like wireless yeah

640
00:45:35,040 --> 00:45:41,800
like that clip it to your beard yeah yeah let's go so that's that's I really

641
00:45:41,800 --> 00:45:46,160
like I think we're reasonably close to that because that's basically how I've

642
00:45:46,160 --> 00:45:50,200
been using the things up to now yeah and I just still need manual steps and so

643
00:45:50,200 --> 00:45:54,400
that I need to deploy it myself and yeah but but it's basically like build it

644
00:45:54,400 --> 00:45:57,760
like it's just like a next year so I've deployed on versatile and so on and it's

645
00:45:57,760 --> 00:46:02,000
like and and in the browser it there's there's nothing stopping us from

646
00:46:02,000 --> 00:46:05,280
building that no and then would be a super useful service that you can also

647
00:46:05,280 --> 00:46:09,280
actually sell like free business idea it's like you know it was create a

648
00:46:09,280 --> 00:46:13,960
hotline for all the moms that want to have websites yeah automatically they

649
00:46:13,960 --> 00:46:17,440
can while they are still on the phone see how the website looks like and it's

650
00:46:17,440 --> 00:46:19,720
like oh no make that button a little bigger and like this should do

651
00:46:19,720 --> 00:46:22,320
something else one of the interesting things from like the Bitcoin open

652
00:46:22,320 --> 00:46:26,040
source point of view about all this stuff is how how indefensible most of

653
00:46:26,080 --> 00:46:32,720
these products are right like cursor is yeah everything everything I can do

654
00:46:32,720 --> 00:46:38,160
it's going to this is gonna be open source right yeah and you know a lot of

655
00:46:38,160 --> 00:46:42,240
these things like any of these like coded like agents right it's like

656
00:46:42,240 --> 00:46:47,800
almost it's like you know it's like the only way to have any defensibility is

657
00:46:47,800 --> 00:46:53,600
to have a user that will never be able to do like you know compose these things

658
00:46:53,760 --> 00:46:58,560
together something I don't know we said it's generation stuff yeah they

659
00:46:58,560 --> 00:47:04,320
much generation stuff like all the stable diffusion and all these kind of

660
00:47:04,320 --> 00:47:09,440
things and the the open models caught up relatively quickly although that just

661
00:47:09,440 --> 00:47:14,520
changed like two days ago yeah I mean sure it's I think the

662
00:47:14,520 --> 00:47:18,320
proprietary stuff will always be able to do studio I mean I think eventually

663
00:47:18,320 --> 00:47:21,560
it will yes curve eventually will flatten out and eventually I think the

664
00:47:21,600 --> 00:47:25,760
open stuff will be just as good yeah like unless there's like unless there's

665
00:47:25,760 --> 00:47:29,520
some like massive training run and then there's like the whole industry goes

666
00:47:29,520 --> 00:47:32,840
bankrupt like in the early 2000s or something and you can no longer do a

667
00:47:32,840 --> 00:47:38,160
hundred billion dollar training like I could imagine that like you could have

668
00:47:38,160 --> 00:47:41,080
the state like but as long as you can continue training these things the

669
00:47:41,080 --> 00:47:44,720
open source will always catch up I think I think so too just just because of the

670
00:47:44,720 --> 00:47:49,000
nature of data basically the nature of it but more so for agents than models

671
00:47:49,000 --> 00:47:52,720
even like any of these like agentic things that are actually more valuable

672
00:47:52,720 --> 00:47:55,640
because they allow you to accomplish a goal right like anytime any sort of

673
00:47:55,640 --> 00:48:00,640
product is like you you want a product is for accomplishing a human end you

674
00:48:00,640 --> 00:48:04,560
know and often some models are like a part of that but it's like the agent

675
00:48:04,560 --> 00:48:10,400
experience itself like everyone's at SEC is much more passionate about

676
00:48:10,400 --> 00:48:15,240
Claude than they are about cursor than they are about a specific model yes as

677
00:48:15,240 --> 00:48:19,280
cursor allows them to accomplish their aim it's it's the same like that there's

678
00:48:19,280 --> 00:48:22,200
less defensibility on the agents because it's much you don't need to do a

679
00:48:22,200 --> 00:48:26,080
massive training run to like get a world that's what I also mentioned about the

680
00:48:26,080 --> 00:48:28,680
pipelines I'm currently building out it's all small steps I don't care what

681
00:48:28,680 --> 00:48:32,040
models are used yeah yeah I really don't care yeah it's all about the tooling and

682
00:48:32,040 --> 00:48:35,520
I think what Chad CPT got right is it built the right interface yeah it built

683
00:48:35,520 --> 00:48:38,640
it really built the right interface it's also what Google got right I think

684
00:48:38,640 --> 00:48:42,400
actually and and now I think what cursor got right I mean that's a good point on

685
00:48:42,400 --> 00:48:45,400
Google though because like the early search engine with terrible were they

686
00:48:45,400 --> 00:48:49,440
look like Chinese Chinese websites like where it's just a million logos it looks

687
00:48:49,440 --> 00:48:54,640
like a phone book yeah yeah and Google's insight was like let's just only have a

688
00:48:54,640 --> 00:48:58,920
text yeah one and that's it yeah and and I think Chad CPT got the interface right

689
00:48:58,920 --> 00:49:04,240
it's a very few little googles insight was no ads that was basically the

690
00:49:04,240 --> 00:49:07,720
inside yeah like no ads on the home company build on ads yeah I built on ads

691
00:49:07,720 --> 00:49:11,280
right like they're there the way they built the biggest ad business was by

692
00:49:11,320 --> 00:49:16,480
rejecting ads at first yeah yeah not smacking them into the face of the

693
00:49:16,480 --> 00:49:21,040
users yeah and in terms of walking let's try to it's gonna be noisy for a little

694
00:49:21,040 --> 00:49:24,400
bit so we got we're gonna try to cross the street and go up to the park and in

695
00:49:24,400 --> 00:49:27,960
the park it should be really nice like we can have a coffee up there yeah let's

696
00:49:27,960 --> 00:49:34,120
go there let's let's walk through the park and that should be I think it's the

697
00:49:34,120 --> 00:49:39,600
most sensible path yeah but but the thing about the the I think the reason

698
00:49:39,600 --> 00:49:44,680
why people at the court currently are so enthused by cursors is because it's

699
00:49:44,680 --> 00:49:48,200
obviously I guess there will be most people hate proprietors software and

700
00:49:48,200 --> 00:49:54,000
yeah yeah absolutely like one by one they fall for practice software that is

701
00:49:54,000 --> 00:49:57,320
good but it just shows you where things are going and how powerful this stuff

702
00:49:57,320 --> 00:50:00,920
really is yeah because the thing is it works really well just like

703
00:50:00,960 --> 00:50:04,920
activity works really well Google works really well it's just a better tool and

704
00:50:04,920 --> 00:50:09,720
and the thing about it it has to be integrated that if the feet the

705
00:50:09,720 --> 00:50:13,680
feedback loops they have to be tight the thing you mentioned about the screen

706
00:50:13,680 --> 00:50:18,600
chat experience it has to be otherwise that's where that's where the that's

707
00:50:18,600 --> 00:50:22,320
where the shift comes from the car it removes cognitive overload yeah or

708
00:50:22,320 --> 00:50:26,120
cognitive like load right like if everything's integrated then you don't

709
00:50:26,120 --> 00:50:31,440
have to think you're not context switch you can try like your context is you

710
00:50:31,440 --> 00:50:35,120
economize context by having like the human context that you have to fit in

711
00:50:35,120 --> 00:50:38,600
your head as you work because what you want to actually remain in that's also

712
00:50:38,600 --> 00:50:42,320
what what a good dialogue brings you it's like you remain in a flow state yeah so

713
00:50:42,320 --> 00:50:45,400
if we are talking we are having a great dialogue you remain in a collective

714
00:50:45,400 --> 00:50:50,320
flow state unless you're getting run over by a car yeah but but but but that's

715
00:50:50,320 --> 00:50:52,720
what you want to happen that's what you also want to have coding but on the

716
00:50:52,720 --> 00:50:56,040
coding side it's like a syntax arrow will fuck up your flow state basically

717
00:50:56,040 --> 00:51:00,480
and what I want to what I what I'm finding now with all the basically

718
00:51:00,480 --> 00:51:04,520
talking to LMS in some sense or trying to think through things with LLM

719
00:51:04,520 --> 00:51:10,640
assistants or LLM's as conversational partners is like I can just because the

720
00:51:10,640 --> 00:51:15,760
the feedback loops and everything is so instant yeah is I can remain in the

721
00:51:15,760 --> 00:51:21,560
thinking flow state of what I actually want to try to achieve and it has to be

722
00:51:21,560 --> 00:51:27,000
but it can only work if everything is super tightly integrated it's it's almost

723
00:51:27,000 --> 00:51:31,840
like okay I have a I have a strange comparison but I think it works it but

724
00:51:31,840 --> 00:51:35,360
why people on Twitter you're in Twitter it doesn't matter if it's

725
00:51:35,360 --> 00:51:39,400
Twitter or TikTok or whatever you do not want to leave the app yeah that's why

726
00:51:39,400 --> 00:51:42,920
everything is loaded in line yeah if it's a link to YouTube video people will

727
00:51:42,920 --> 00:51:46,280
not click it yeah if it's a link to an article people will not click it yeah if

728
00:51:46,280 --> 00:51:49,960
the video is rendered in the feed yeah different story yeah people will watch

729
00:51:49,960 --> 00:51:54,080
it and all those kind of things so so so so we know that this is the case we

730
00:51:54,080 --> 00:51:57,320
absolutely know that this is the case and I think it is because people don't

731
00:51:57,320 --> 00:52:00,600
want a context switch yeah or content let me put it differently it's not a

732
00:52:00,600 --> 00:52:04,920
conscious thing it's like context switching it's so expensive and it

733
00:52:04,920 --> 00:52:09,120
completely gets you out of the flow yeah and so if you have a highly

734
00:52:09,120 --> 00:52:13,200
integrated tool like cursor is you remain in the flow state and you're

735
00:52:13,200 --> 00:52:18,240
never forced to context switch out of it yeah I think that's kind of the the

736
00:52:18,240 --> 00:52:21,600
secret sauce of why it works so well yeah it's just like you can stay in that

737
00:52:21,600 --> 00:52:25,200
place and you can continue the conversation that you have with your IDE

738
00:52:25,200 --> 00:52:30,040
basically and it's just gonna you get to the next thing and to the next thing

739
00:52:30,040 --> 00:52:33,880
and to the next thing yeah it's almost like infinity scrolling but it's infinity

740
00:52:33,880 --> 00:52:39,160
scrolling new code yeah one interesting thing is that I only I I do a lot of

741
00:52:39,160 --> 00:52:42,520
this but for like a few very technical problems that I'm interested in right

742
00:52:42,520 --> 00:52:47,080
now so all of these like sometimes multi-hour conversations that span like

743
00:52:47,080 --> 00:52:50,560
20 chats yeah right where you reset the context or try to pull over what you

744
00:52:50,560 --> 00:52:54,160
had previously or you know I try with grok I'm paying for every single one of

745
00:52:54,160 --> 00:52:58,080
these frontier models right now like I'm trying to find ways to pay more like

746
00:52:58,080 --> 00:53:03,240
I'm I went from like give me better more frustrated I was paying 20 bucks a

747
00:53:03,240 --> 00:53:06,440
month yeah GPT where it's like caught I'm only paying like 300 bucks a month

748
00:53:06,440 --> 00:53:12,240
right now what I spend more like you know but the thing is like I don't trust

749
00:53:12,240 --> 00:53:17,040
these like you know that the questions I'm really interested in life are not

750
00:53:17,040 --> 00:53:21,200
like technical problems necessarily right like a software so you're has it

751
00:53:21,200 --> 00:53:25,680
end with some I will I haven't I don't ask them any philosophical stuff like

752
00:53:25,680 --> 00:53:29,640
the stuff I'd write to dear diary I mean you'll be I don't run into the

753
00:53:29,640 --> 00:53:34,160
gulag I don't I don't I don't write any of that stuff like I basically approach

754
00:53:34,160 --> 00:53:37,880
these things as if it's all gonna be in a court deposition someday I won't put

755
00:53:37,880 --> 00:53:42,440
anything in one of these they wouldn't be okay in a step is there right

756
00:53:42,440 --> 00:53:47,480
literally but that does really limit how useful they are yeah because like

757
00:53:47,480 --> 00:53:51,600
while these technical problems are interesting if I could solve the bigger

758
00:53:51,600 --> 00:53:55,800
more philosophical human problems in my life I that's a million times more

759
00:53:55,800 --> 00:53:59,640
useful yeah then solving this whole technical thing so that's that's what

760
00:53:59,640 --> 00:54:02,840
the one of the things that I still don't have an answer for because like I've

761
00:54:02,840 --> 00:54:06,800
been playing a little bit like stuff like Olama but they're stupid in comparison

762
00:54:07,640 --> 00:54:12,160
I still think they're good enough for most things but but do you think this

763
00:54:12,160 --> 00:54:18,000
will will be able to fix that with local models well I'm looking at those

764
00:54:18,000 --> 00:54:21,000
GPUs on either

765
00:54:22,000 --> 00:54:26,920
I think GPUs are most of our problems

766
00:54:27,800 --> 00:54:32,320
a couple hundred bucks for subscriptions or like investing

767
00:54:32,320 --> 00:54:39,240
exactly it's just the only thing in life right now where I'm like there's a

768
00:54:39,240 --> 00:54:45,040
button that says 50k and I'm like I want it to but with anything else I'm like

769
00:54:45,040 --> 00:54:48,840
no I could see a little tread left on my tires let's not change them for a

770
00:54:48,840 --> 00:54:54,840
while how if you if you if you would buy your own GPU farm like what what's

771
00:54:54,840 --> 00:54:59,000
what would you train it on the first thing number one I would I would train

772
00:54:59,000 --> 00:55:05,960
a model to understand Zig number one step one one yeah before we get to the

773
00:55:05,960 --> 00:55:10,640
big problems in one you know you know you know like the step one solve AI yeah

774
00:55:10,640 --> 00:55:14,840
yeah you solve the world's problems for you it's like step one DJI how to you

775
00:55:14,840 --> 00:55:20,240
say yeah yeah it really ruined my cohort experience I wanted to go make a entire

776
00:55:20,240 --> 00:55:26,000
stack of beautiful rust urban sorry rust a zig tools for noster fastest I

777
00:55:26,000 --> 00:55:34,480
want to fast yeah yeah yeah yeah and just zig piercig you know no no not just

778
00:55:34,480 --> 00:55:38,880
just everything's the fastest what do this thing enthusiasts say about LLM

779
00:55:38,880 --> 00:55:42,080
stuff they must be aware well so the funny things they don't really care so

780
00:55:42,080 --> 00:55:45,520
that my point earlier about how like you can listen to some one how they talk

781
00:55:45,520 --> 00:55:48,800
about LLM's and then you figure out kind of at what level they deal like

782
00:55:48,800 --> 00:55:52,280
work it's off we're kind of a lot of these zig people are like the super low

783
00:55:52,280 --> 00:55:57,720
level like yeah trying to well they're trying to make the program good yeah

784
00:55:57,720 --> 00:56:02,240
they're not necessarily like trying to do a big idea yeah and so they don't care

785
00:56:02,240 --> 00:56:06,880
because they're like wizards yeah they're the oddest I'm not right like they

786
00:56:06,880 --> 00:56:10,840
they know exactly where the semi colon goes they know exactly how the

787
00:56:10,840 --> 00:56:17,240
concurrency works you know they they enjoy writing the semaphore yeah to make

788
00:56:17,240 --> 00:56:22,000
the concurrency work I hate that part yeah you know so like in general the

789
00:56:22,000 --> 00:56:26,680
the zig people like the really hardcore software engineers in my experience

790
00:56:26,680 --> 00:56:28,960
don't really care about this stuff very much because it doesn't help them very

791
00:56:28,960 --> 00:56:33,360
much yeah because the LLM's are good at what they're already good at you know so

792
00:56:33,360 --> 00:56:37,040
that's that's kind of a little bit difficult so first thing you would do is

793
00:56:37,040 --> 00:56:39,480
bridge that I mean the zig is also interesting like they basically don't

794
00:56:39,480 --> 00:56:43,280
have documentation like their whole experience so far which I think is actually

795
00:56:43,280 --> 00:56:46,240
kind of smart you know trying to make this really good low level language so

796
00:56:46,240 --> 00:56:48,400
they're trying to like prevent like entries and whatever it's called right

797
00:56:48,400 --> 00:56:51,720
like yeah a bunch of noobs from showing up and running it so they're like hey

798
00:56:51,720 --> 00:56:55,160
the standard library is the documentation if you can't read the standard library

799
00:56:55,160 --> 00:57:00,920
get the fuck out it's not RKFC yeah yeah which is smart I think like if you want

800
00:57:00,920 --> 00:57:04,000
to make like a really hardcore low-level language and you only want people like

801
00:57:04,000 --> 00:57:10,360
that around it's a pretty good trade-off yeah right like one big rust a zig

802
00:57:10,360 --> 00:57:13,640
project this is ghosty terminal terminulator I think it's going to be

803
00:57:13,640 --> 00:57:18,560
like it's going to be a really interesting like you know you write

804
00:57:18,560 --> 00:57:21,840
code in these internals like I think it's going to kind of you know a little bit

805
00:57:21,840 --> 00:57:24,760
more of a new next generation of that like it's gonna it's gonna it's gonna

806
00:57:24,760 --> 00:57:28,800
it's gonna make a lot of interesting things possible and the guy did something

807
00:57:28,800 --> 00:57:31,560
really interesting where he he announced this thing like a year and a half ago

808
00:57:31,560 --> 00:57:34,920
he just made a closed beta where only he don't invite friends were like people

809
00:57:34,920 --> 00:57:38,160
were proven zig contributes to certain zig and I was trying to get in there for

810
00:57:38,160 --> 00:57:41,920
a year and he wouldn't let me in but I think it made sense because he was able

811
00:57:41,920 --> 00:57:46,280
to have this tight community that gave really good feedback and he never had

812
00:57:46,280 --> 00:57:51,040
to deal with idiots on github issues posting bad bug report yeah I think

813
00:57:51,040 --> 00:57:54,240
to some extent that's kind of valuable yeah yeah that's the first thing would

814
00:57:54,240 --> 00:57:59,680
be zig and the second thing I want to do is what's it called there are some

815
00:57:59,680 --> 00:58:02,320
of these things where you can you don't like chain a frontier model I don't want

816
00:58:02,320 --> 00:58:05,800
to trade a frontier model obviously like that's billions of dollars right yeah

817
00:58:05,800 --> 00:58:10,480
I think there's one I forget the name of it there's some acronym where you can

818
00:58:10,560 --> 00:58:14,800
train like sort of a layer on top of an existing model so you can take O-Lama

819
00:58:14,800 --> 00:58:18,040
and like train a thin layer on top of it and you can actually do this on like

820
00:58:18,040 --> 00:58:22,000
consumer computers okay like there are examples of lots of examples of people

821
00:58:22,000 --> 00:58:25,000
like actually I might try this with zig I think it might actually work with zig on

822
00:58:25,000 --> 00:58:29,960
my laptop or just renting a GPU right like they show like I've seen stuff about

823
00:58:29,960 --> 00:58:34,080
how people did this and they were able to like a six dollar training run actually

824
00:58:34,080 --> 00:58:37,960
I was going back with a chat GPT on this and it was like yeah I think I think I

825
00:58:38,800 --> 00:58:40,880
could we had one conversation we was like yeah I think we could fix your zig

826
00:58:40,880 --> 00:58:45,320
problems with a six dollar training run he really hesitated to give me a

827
00:58:45,320 --> 00:58:48,040
number but I just kept asking over and over how much would it cost and then

828
00:58:48,040 --> 00:58:53,560
finally six dollars I was like really I was expecting like 5,000 like I'm

829
00:58:53,560 --> 00:58:56,880
touched I'm a total new but this thing but so like stuff like that would be

830
00:58:56,880 --> 00:58:59,840
really fun I think like training I think it would be really fun to try like

831
00:58:59,840 --> 00:59:03,760
these alignment layers you know like like there's one there's a guy in

832
00:59:03,760 --> 00:59:08,840
Oster him someone who he he did like one little project he was trying to make

833
00:59:08,840 --> 00:59:11,200
an alignment layer like one of these layers on top of it that basically

834
00:59:11,200 --> 00:59:16,080
sensors it but do it on Oster data right so just make this thing talk like

835
00:59:16,080 --> 00:59:19,080
take the intelligence that was done in the like frontier training run that

836
00:59:19,080 --> 00:59:22,800
cost a bunch of money but then like sort of format in a way that makes it sound

837
00:59:22,800 --> 00:59:26,640
like an Oster user I see like something like that would be kind of cool right I

838
00:59:26,640 --> 00:59:32,240
don't know I think these like little mini mini training runs on for like

839
00:59:32,240 --> 00:59:36,480
specific use cases I think would be really interesting I mean the obvious

840
00:59:36,480 --> 00:59:40,040
one is I would just have it set up in my house and then I would have like the

841
00:59:40,040 --> 00:59:45,120
biggest deep-seak model running and I would just spill my guts to it and all

842
00:59:45,120 --> 00:59:48,640
my problems in life and all my difficulties and it would basically be

843
00:59:48,640 --> 00:59:52,880
like my therapist yeah that's what I'm kind of trying to build out right now so

844
00:59:52,880 --> 00:59:57,600
I told you I kind of regret that I stopped you know like appearing on

845
00:59:57,600 --> 01:00:01,520
podcast yeah writing stuff and so on I mean the first podcast I've done in like

846
01:00:01,520 --> 01:00:06,800
three years yeah the same for me pretty much and again like we we don't treat

847
01:00:06,800 --> 01:00:10,880
it as a podcast it's a dialogue it's a conversation so yeah it's a walk

848
01:00:10,880 --> 01:00:14,240
exactly I want to take developers on the walk and kind of you know pick the

849
01:00:14,240 --> 01:00:18,560
brains that's the that's the point but but I the thing why I kind of regret it I

850
01:00:18,560 --> 01:00:23,360
mean I had to do it I was like burning out so hard but but I I kind of regret it

851
01:00:23,360 --> 01:00:30,000
because now I have a gap in my recordings in some sense and what I what I do in

852
01:00:30,000 --> 01:00:33,600
these conversations also on the podcast I appeared in the past and so on also in

853
01:00:33,600 --> 01:00:37,360
writing also in tweeting you know and now in like I mean I'm still tweeting on

854
01:00:37,360 --> 01:00:44,440
not so like you know bite me but it's it's basically just just it's almost like

855
01:00:44,440 --> 01:00:49,440
a diary yeah okay I'm not sharing it highly sensitive and private stuff and

856
01:00:49,440 --> 01:00:52,840
so on like I share it's a diary in the sense of like letters were in like the

857
01:00:52,840 --> 01:00:57,360
West yeah exactly publish your letters exactly and it's somewhat private but

858
01:00:57,360 --> 01:01:01,120
it's also like a record of your it's just what's what's normal now is like

859
01:01:01,120 --> 01:01:05,600
that that we up to now we only have the written word kind of yeah accessible to

860
01:01:05,600 --> 01:01:11,280
people with letters and those kind of things and and and and now now I really

861
01:01:11,280 --> 01:01:14,480
like physical artifacts I mean honestly that's the thing we really have for most

862
01:01:14,480 --> 01:01:17,880
of them in history all we have is what pots did they make oh absolutely what

863
01:01:17,880 --> 01:01:20,800
coins did they have what pots did they have what did their sorts of like yeah

864
01:01:20,800 --> 01:01:25,760
what the houses like what how did they build you know like with obelisks and

865
01:01:25,800 --> 01:01:28,360
those kind of things that's what that's the stuff that remains that's

866
01:01:28,360 --> 01:01:32,840
actually something that I I kind of worry about a little bit of our like

867
01:01:32,840 --> 01:01:37,920
current hyper digitalized civilization is like if things go south nothing will

868
01:01:37,920 --> 01:01:42,880
remain yeah we don't we don't do clay tablets anymore I think we should you

869
01:01:42,880 --> 01:01:47,600
know from time to time but anyway like what I when I the reason why I regret

870
01:01:47,600 --> 01:01:52,320
having this gap now is because I I actually now I know that I can even

871
01:01:52,320 --> 01:01:56,240
both this myself and this tooling with local models and local things like

872
01:01:56,240 --> 01:02:00,680
privacy conscious basically and I just want to like you say almost like a

873
01:02:00,680 --> 01:02:03,000
therapist but it's not really a therapist it's like a therapist slash

874
01:02:03,000 --> 01:02:07,840
diary slash to do like a diamond I think Damon right yeah Damon in the

875
01:02:07,840 --> 01:02:12,160
unique sense well no Damon and like the Greek like the Damon was like the

876
01:02:12,160 --> 01:02:15,280
spirit you would kind of talk to the guy to you for your life almost like a

877
01:02:15,280 --> 01:02:19,320
guardian exactly that's why that's why I talk about their day and that's why I

878
01:02:20,320 --> 01:02:23,480
use yes and that's why I come back to rubber duct debugging because we know

879
01:02:23,480 --> 01:02:27,400
it really works because a dialogue a conversation gets you out of the

880
01:02:27,400 --> 01:02:32,480
self-centered monological thinking loop that you're in and then you can

881
01:02:32,480 --> 01:02:36,080
think yourself in circles really easily yeah but you will get me out of my

882
01:02:36,080 --> 01:02:41,440
circle yeah but an LLM or even chest sitting down and writing a diary entry

883
01:02:41,440 --> 01:02:47,040
or or talking to yourself yeah will also get you out yeah and this is one thing

884
01:02:47,040 --> 01:02:50,560
I would train on to is like making LLM's more disagreeable like that's one

885
01:02:50,560 --> 01:02:53,760
thing that's really annoying about them oh yeah they agree with you all they're

886
01:02:53,760 --> 01:02:58,080
both there yes man yeah I would much rather have one that is like that is

887
01:02:58,080 --> 01:03:02,720
very conscious of the limits of their knowledge and very disagreeable like

888
01:03:02,720 --> 01:03:07,680
yeah also like maybe it will say like that's a stupid idea yeah yeah exactly

889
01:03:07,680 --> 01:03:11,920
I've never had an LLM tell me no it's but people tell me all the time yeah

890
01:03:11,920 --> 01:03:16,440
large discrepancy yeah but the way I see it like what I want to build up now

891
01:03:16,440 --> 01:03:20,200
again and I still have you know like I have my Twitter archive I have I don't

892
01:03:20,200 --> 01:03:24,520
know like 200 recorded conversations those kind of things and and and so I

893
01:03:24,520 --> 01:03:30,080
want to I want to put this all in a pile and have various agents kind of go

894
01:03:30,080 --> 01:03:34,400
over it locally running like it's like just just having a transcript of this

895
01:03:34,400 --> 01:03:38,120
stuff having it full text searchable having it summarized having it kind of

896
01:03:38,120 --> 01:03:42,680
tagged yeah and like with timestamps and so on that's already so useful because

897
01:03:42,680 --> 01:03:48,200
then I can go back and and can go like the January 2023 what I what did I

898
01:03:48,200 --> 01:03:51,800
think about no sir versus I don't know like December 25 it's really hard to

899
01:03:51,800 --> 01:03:55,400
remember stuff like that it is that's why that's why writing stuff down is so

900
01:03:55,400 --> 01:03:59,640
insanely useful because you always evolve and you also forget the insights

901
01:03:59,640 --> 01:04:02,720
that you had in the past and all those kind of things so that's kind of what I

902
01:04:02,720 --> 01:04:08,280
what I'm what I'm aiming at right now but yeah we'll see maybe maybe we'll

903
01:04:08,280 --> 01:04:13,240
need to seek for some of the tooling so yeah you won't should I pie you should

904
01:04:13,240 --> 01:04:19,080
I buy you a GPU yeah yeah yeah I can contribute with like what one or two

905
01:04:19,080 --> 01:04:22,440
graphics cards like yeah it's not an issue I mean what I should do is is just

906
01:04:22,440 --> 01:04:27,200
get like a cheap video gaming one and just learn what even is a GP like I've

907
01:04:27,200 --> 01:04:30,680
never done what even is a GP you know like what even how do you talk to

908
01:04:30,680 --> 01:04:33,440
it like philosophical questions yeah you know but like what do you how do you

909
01:04:33,440 --> 01:04:36,520
feed that fucking vectors and matrices into this and have it do something

910
01:04:36,520 --> 01:04:40,760
useful I don't know I have a math degree I don't know I have no idea what

911
01:04:40,760 --> 01:04:45,240
did you do it you never did like an n-body simulation on the GPU no I've

912
01:04:45,240 --> 01:04:51,240
never done GPU programming to my knowledge I'm not an autism I keep

913
01:04:51,240 --> 01:04:55,760
trying to tell you this I'm not I keep I keep this I've been cosplaying all

914
01:04:55,760 --> 01:05:00,640
these years I'm a pretender so you stepped up now you're the idea guy now

915
01:05:00,640 --> 01:05:05,360
yeah this is the thing is like dude the age of the idea guys has begun yeah

916
01:05:05,360 --> 01:05:09,520
right yeah yeah yeah I really hope that I can have my

917
01:05:09,520 --> 01:05:15,440
by-blind kind of in shape and and and ready for the public basically yeah

918
01:05:15,440 --> 01:05:20,480
rather sooner than later because I really think that that there's a lot of

919
01:05:20,480 --> 01:05:27,320
people have really good ideas and no way of executing on them in any way shape

920
01:05:27,320 --> 01:05:31,760
of format to be fair execution is the hard part like ideas are a dime a

921
01:05:32,760 --> 01:05:38,440
that everyone knows that but I think it's still super valuable to to again

922
01:05:38,440 --> 01:05:40,960
that's the whole point though is like the oh I'm just kind of oftentimes kind of

923
01:05:40,960 --> 01:05:45,160
good at execution especially have a team of them yeah and execution it's

924
01:05:45,160 --> 01:05:48,760
really cheap what works for one person works for another oh there's a funny

925
01:05:48,760 --> 01:05:51,680
one yet cheap like I saw somebody say a really funny quote where they were like

926
01:05:51,680 --> 01:05:54,400
you know it's like some kind of lefty person that you know doesn't like

927
01:05:54,400 --> 01:05:59,440
global warming and they were like you know every time I use an LLM to do some

928
01:05:59,440 --> 01:06:03,360
data processing I really think about how it costs 10 or a thousand times more

929
01:06:03,360 --> 01:06:07,280
to do this with an LLM than to do like some some standard programming tool

930
01:06:07,280 --> 01:06:11,080
right yeah and I also think about you consumption yeah yeah but then I also

931
01:06:11,080 --> 01:06:14,600
think about how it's a thousand times cheaper than if a human did it yeah

932
01:06:14,600 --> 01:06:18,920
it's like yeah cheap you know it's like it's like it's depending on the

933
01:06:18,920 --> 01:06:22,480
metric yeah yeah monetarily cheap because it's girls so well yeah energy

934
01:06:22,480 --> 01:06:26,880
expensive yeah I mean other than our way more energy so I started to try to

935
01:06:26,880 --> 01:06:29,400
move away from cursor right where everything's integrated into one thing

936
01:06:29,400 --> 01:06:34,680
and to do it you know kind of put pieces together myself a little more the like

937
01:06:34,680 --> 01:06:38,480
the actual thing I used to chat with Claude is Claude Code their command

938
01:06:38,480 --> 01:06:42,080
line yeah thing and an interesting thing about it is that they don't you know

939
01:06:42,080 --> 01:06:45,760
with with cursor everything is part of a subscription like 20 bucks a month yeah

940
01:06:45,760 --> 01:06:51,760
and they probably heavily subsidize it because yeah it's like it's like if you

941
01:06:51,760 --> 01:06:55,480
use Claude Code even if you have a Claude membership they charge per token

942
01:06:55,480 --> 01:06:58,880
and so if you'll you'll go back and forth and oftentimes my little sessions for

943
01:06:58,880 --> 01:07:02,920
one feature will cost like a dollar and it's kind of it is reframing my thought

944
01:07:02,920 --> 01:07:07,520
on microtransactions to some a little oh yeah sure I mean because because you

945
01:07:07,520 --> 01:07:11,960
usually I've always believed this idea of you know the micro track transactions

946
01:07:11,960 --> 01:07:15,880
might not like the the mental class idea from Zabo right like maybe that'll

947
01:07:15,880 --> 01:07:19,120
prevent it I think it's actually right but if you integrate it with with

948
01:07:19,120 --> 01:07:21,640
Nostra you don't have to think about it in that sense you know what I mean

949
01:07:21,640 --> 01:07:25,120
no I don't like the mental cost is like the the weekly you look back on the

950
01:07:25,120 --> 01:07:27,880
weekends like okay this much I spent you don't have to think about every single

951
01:07:27,880 --> 01:07:31,440
microtransaction but it's still a microtransaction yeah and not a credit line

952
01:07:31,440 --> 01:07:34,080
and that makes all the difference and that's why that's appearing now we have

953
01:07:34,080 --> 01:07:38,920
the pay per queue thing yeah yeah and like with the with the MCP stuff you

954
01:07:38,920 --> 01:07:43,800
mentioned like Jesus is working on the DB MCP yeah things like bridging all that

955
01:07:43,800 --> 01:07:48,520
to Nostra and using Nostra as the substrate for AI model context stuff and

956
01:07:48,520 --> 01:07:52,920
micro payments which makes all the sense in this world you see for any like

957
01:07:52,920 --> 01:07:55,400
how often you see how much money you're spending you just have a budget you see

958
01:07:55,400 --> 01:07:58,880
what sure you can we can already with Nostra what connects that I'm a budget

959
01:07:58,880 --> 01:08:03,840
so the funny thing is that like this what why this these experiences cloud code

960
01:08:03,840 --> 01:08:06,680
are kind of interesting is because every at the end of every you know session you

961
01:08:06,680 --> 01:08:11,720
like control D it twice to exit and you and it helps print out how much it cost

962
01:08:11,720 --> 01:08:15,600
yeah and I actually enjoy seeing how much it costs every time I put a deal

963
01:08:15,600 --> 01:08:20,360
man what a deal I got it for 92 cents I was expecting now it's like you get

964
01:08:20,360 --> 01:08:23,840
out of the casino or something and then it's like oh my Lord you usually you

965
01:08:23,840 --> 01:08:26,600
know this is one of the first times where I'm basically doing microtransactions

966
01:08:26,600 --> 01:08:32,120
and I enjoy like I don't really mind maybe it's because I feel like I'm

967
01:08:32,120 --> 01:08:36,120
getting a good deal but I mind a lot less seeing the price you know I see yeah

968
01:08:36,120 --> 01:08:39,760
it's usually usually every time you see money going out it's kind of stressful

969
01:08:39,760 --> 01:08:44,960
yeah but again it just shows you how much it's so honest compared to like

970
01:08:45,160 --> 01:08:48,960
what your experiences with like a credit card statement or you're like you know

971
01:08:48,960 --> 01:08:53,120
I don't know I probably got five subscriptions right now there's very few

972
01:08:53,120 --> 01:08:57,360
things that we don't have thing me for two years now I mean I think I think it's

973
01:08:57,360 --> 01:08:59,840
because of the payments I talked with Kali about this you know it's like how

974
01:08:59,840 --> 01:09:03,320
payments fuck the internet in a large degree yeah it's not only the mental

975
01:09:03,320 --> 01:09:06,560
transaction cost it's but it's interesting that there's very few things

976
01:09:06,560 --> 01:09:10,880
that we pay by the meter it's basically it's like gasoline yeah for your car

977
01:09:10,880 --> 01:09:16,560
and and electricity yeah more or less yeah and the water bill may basically

978
01:09:16,560 --> 01:09:19,820
like that's that's kind of everything else is kind of like a flat

979
01:09:19,820 --> 01:09:25,840
funny like the most hard necessities of life food kind of paid by the meter

980
01:09:25,840 --> 01:09:29,680
yeah you know yeah I think compute is going to be a necessity like this as

981
01:09:29,680 --> 01:09:32,960
well and I don't even differentiate like the only people that have a food

982
01:09:32,960 --> 01:09:36,480
subscription are the ones on welfare basically are students you know some

983
01:09:36,480 --> 01:09:40,200
sort of welfare yeah so so I think I think compute is going to be a

984
01:09:40,200 --> 01:09:43,120
necessity as well and I don't even differentiate between compute and

985
01:09:43,120 --> 01:09:46,680
storage because to me there is no difference like let's say I want a photo

986
01:09:46,680 --> 01:09:50,240
of a flower that I want to use somewhere and I know I have it on my heart drive

987
01:09:50,240 --> 01:09:54,280
there's to me is no difference whether I store this photo and I find it or I

988
01:09:54,280 --> 01:09:57,760
generate it on the fly yeah you know so for many things kind of storage and

989
01:09:57,760 --> 01:10:01,600
compute are almost equivalent yeah it's like Laurie Mipson or whatever it's

990
01:10:01,600 --> 01:10:05,800
like do I have to have it stored on my heart is somewhere in a sense and I was

991
01:10:05,800 --> 01:10:09,840
actually a storage technology right like in a sense I mean what's like what

992
01:10:09,840 --> 01:10:13,520
like someone a memory well someone had a good point where it's like you know can

993
01:10:13,520 --> 01:10:16,640
you think we've got memory no he's like we've lost a lot of history of the

994
01:10:16,640 --> 01:10:19,840
early Internet right the websites just went away yeah in a sense they retained

995
01:10:19,840 --> 01:10:23,040
in the context of these large language models because they could you know

996
01:10:23,040 --> 01:10:26,520
they're trained on it right like the training runs from two years ago the

997
01:10:26,520 --> 01:10:29,720
training the training runs for two years ago and code websites that are now

998
01:10:29,720 --> 01:10:34,800
gone forever that's an interesting thought yeah yeah so they are a storage

999
01:10:34,800 --> 01:10:38,800
technology in a sense I mean especially the really big ones I mean intelligence

1000
01:10:38,800 --> 01:10:43,120
thinking and storage are very much interrelated you need memory to to to

1001
01:10:43,120 --> 01:10:48,200
think like yeah thinking is about the future and to access the future you need

1002
01:10:48,200 --> 01:10:53,200
to access the past kind of mm-hmm it's it's yeah it's strange how it's all

1003
01:10:53,200 --> 01:10:59,000
interconnected but but yeah I think compute will be compute will be metered

1004
01:10:59,000 --> 01:11:02,720
is I guess what I'm trying to say yeah yeah and you need to have a micro

1005
01:11:02,720 --> 01:11:07,000
payment substrate to make it all work because the credit overhead at the

1006
01:11:07,000 --> 01:11:10,680
credit risk just kills you I mean the other thing that's really interesting

1007
01:11:10,680 --> 01:11:14,480
about it is like I found my there's something kind of magical but actually

1008
01:11:14,480 --> 01:11:18,400
seeing the like I'd never been interested in GPUs until I saw that number

1009
01:11:18,400 --> 01:11:22,520
from Claude actually like that's what started getting me interested like okay

1010
01:11:22,520 --> 01:11:26,000
so that's what a token costs right like I never really thought people always

1011
01:11:26,000 --> 01:11:29,400
talk about tokens and again I said this we're walking in a circle right now and

1012
01:11:29,400 --> 01:11:32,240
every time I reach this to topic at the spot I'm like Gigi I'm not an

1013
01:11:32,240 --> 01:11:37,240
artist time we pass this literally we're rocking the circle I make the same

1014
01:11:37,240 --> 01:11:41,720
point every time we get to this don't break out of the circle yeah yeah but I

1015
01:11:41,720 --> 01:11:46,680
never tried it so I'm like whatever I don't care what a token is but it was

1016
01:11:46,680 --> 01:11:49,840
kind of interesting like it's like okay 97 cents it's like 20,000 tokens or

1017
01:11:49,840 --> 01:11:53,600
whatever interesting okay so I'm actually so like the interesting thing is like

1018
01:11:53,600 --> 01:11:58,560
it got me to think in terms of like the human action you know like you need

1019
01:11:59,560 --> 01:12:06,320
like for stuff to get better a human needs to act like you know

1020
01:12:06,320 --> 01:12:10,840
economize between scarce resources yeah I mean you have to see the price yeah

1021
01:12:10,840 --> 01:12:14,560
you have to see the cost like you have to you have to actually pay yeah and a

1022
01:12:14,560 --> 01:12:18,680
sense and a sense it's a lot more it's it's like at least in my own personal

1023
01:12:18,680 --> 01:12:22,920
experience as soon as I started seeing the actual cost for the actual tokens I

1024
01:12:22,920 --> 01:12:27,920
was consuming for the actual matrix math I was doing in a GPU I didn't start

1025
01:12:28,080 --> 01:12:32,480
actually thinking like a human actor as much yeah right because I start like if

1026
01:12:32,480 --> 01:12:37,400
it's all just like I got a like a fiat cuck cursor subscription and I'm just

1027
01:12:37,400 --> 01:12:41,960
I'm just being spoon fed right like I didn't think that way but as soon as I

1028
01:12:41,960 --> 01:12:47,480
start seeing the tokens I'm like okay it's like an abstraction right yeah well

1029
01:12:47,480 --> 01:12:51,160
but it's and it's like you don't act like you don't you're not faced with like

1030
01:12:51,160 --> 01:12:54,560
you cannot act on the instructions it's the point yeah yeah it has to be real the

1031
01:12:54,600 --> 01:12:58,240
cost has to be real the decision has to be real and that's why that's why

1032
01:12:58,240 --> 01:13:03,120
simulations don't work like like it's it's that I mean that's why sacrifice and

1033
01:13:03,120 --> 01:13:06,480
other things are a thing like it has to be real you have to make it real otherwise

1034
01:13:06,480 --> 01:13:11,200
it will not change how you act and how you think about it basically like if

1035
01:13:11,200 --> 01:13:15,760
if there are there are no consequences to reaction it it it will not influence

1036
01:13:15,760 --> 01:13:19,800
your behavior if there are actual consequences then then it will change

1037
01:13:19,800 --> 01:13:24,680
your behavior going over like a slack line you know without a harness and

1038
01:13:24,680 --> 01:13:27,800
without like any protection it's very different they're just doing the same

1039
01:13:27,800 --> 01:13:31,440
thing and they're like it's inconsequential if you fall that that's why I

1040
01:13:31,440 --> 01:13:34,680
want to try that's I want to try like renting GPUs and doing some training

1041
01:13:34,680 --> 01:13:39,240
and stuff because it's gonna like it's like what is actually happening yeah

1042
01:13:39,240 --> 01:13:43,800
cost what is the brass tacks going on here like what is what is the because I

1043
01:13:43,800 --> 01:13:47,440
think that like that's another thing like what is kind of unique about a human

1044
01:13:47,440 --> 01:13:50,920
it's like well that's the one kind of the fundamental insight of like

1045
01:13:50,920 --> 01:13:55,440
Austrian economics people is like it's this economizing process this you this

1046
01:13:55,440 --> 01:14:00,360
what's the word praxia you know the praxialogical whatever the fuck you

1047
01:14:00,360 --> 01:14:05,280
know you know that that act of actually choosing you taking action yeah that's

1048
01:14:05,280 --> 01:14:10,360
the real magic and yeah there's something interesting about micro

1049
01:14:10,360 --> 01:14:14,680
transactions in this context at least it's like sort of activate that you in a

1050
01:14:14,720 --> 01:14:19,440
lot more than like a just a subscription does yeah it's and it doesn't again it

1051
01:14:19,440 --> 01:14:28,920
removes the abstraction and it it shows you both on the on the consumer side and

1052
01:14:28,920 --> 01:14:33,360
on the service side kind of the reality of how people are using it and what is

1053
01:14:33,360 --> 01:14:39,440
kind of most needed it's like a more you know it's a more free market dynamic

1054
01:14:39,440 --> 01:14:47,640
in some sense mm-hmm you're gonna need some tobacco to think mostly yeah we're

1055
01:14:47,640 --> 01:14:51,640
gonna we're gonna get some some some snooze sometimes there we go Norwegian

1056
01:14:51,640 --> 01:14:56,840
well yeah that's a bit dangerous in the conversation just gets out of hand yeah

1057
01:14:56,840 --> 01:15:05,520
everyone's dizzy yeah so should we try to break out of the loop and hit the

1058
01:15:05,520 --> 01:15:09,120
other park across yeah it works let's try a few more rounds I think I'll probably

1059
01:15:09,120 --> 01:15:14,760
go insane here walking in circles yeah yeah

1060
01:15:15,000 --> 01:15:20,680
no it's good imagine another thing you don't I mean yeah in a dialogue you

1061
01:15:20,680 --> 01:15:24,840
don't want to be economize it you know like no I want to be you don't want to be

1062
01:15:24,840 --> 01:15:28,560
acting in some sense right you just a circle is actually kind of natural but

1063
01:15:28,560 --> 01:15:32,800
how don't others but that's also that's also something where where you all I think

1064
01:15:32,800 --> 01:15:38,640
I keep I keep trying to let me know sorry it's like that thing did I get a

1065
01:15:38,640 --> 01:15:42,840
good deal there right like that's that's the thing that was interesting about

1066
01:15:42,840 --> 01:15:46,440
seeing the number I paid which is a question hides yeah yeah because then

1067
01:15:46,440 --> 01:15:51,920
you're like you're like you become much more aware of the cost of what you're

1068
01:15:51,920 --> 01:15:57,720
doing and like you improve you're like yes if I yeah if I am advertising my

1069
01:15:57,720 --> 01:16:02,080
effort across is 20 no matter how I spend it yeah I don't improve as a

1070
01:16:02,080 --> 01:16:07,560
prompter as much yeah but if I see yeah every every word this is another Jordan

1071
01:16:08,560 --> 01:16:12,040
it's like you that's why UBI is a terrible every word I put into this thing

1072
01:16:12,040 --> 01:16:17,960
cost me yeah so you sort of where it's carefully you sort of speak you you

1073
01:16:17,960 --> 01:16:22,280
write more deliberately and thoughtfully this way yeah because everything has a

1074
01:16:22,280 --> 01:16:27,280
cost yeah which absolutely everything does and every time you try to pamper

1075
01:16:27,280 --> 01:16:30,520
over the cost yeah paper over the cost in some sense yeah like a subscription

1076
01:16:30,520 --> 01:16:34,080
does yeah you kind of destroy some sort of signal some costly I mean it's

1077
01:16:34,080 --> 01:16:36,480
funny that I say this too because I'm just thinking the first time I did this

1078
01:16:36,480 --> 01:16:41,600
after Tony mentioned it I quit and it was like four dollars I just spent a

1079
01:16:41,600 --> 01:16:48,120
fifth of a month and one on one thing right now you have a different view

1080
01:16:48,120 --> 01:16:53,360
yeah in a sense that I prefer it now even though my first reaction was like

1081
01:16:53,360 --> 01:16:58,320
what a ripoff yeah interesting interesting it's funny by the way that you

1082
01:16:58,320 --> 01:17:02,320
know like it's 2025 now and we still talk about tokens it's just the tokens

1083
01:17:02,320 --> 01:17:05,520
changed like 2017 and 18 we talked about yeah yeah shit coin tokens

1084
01:17:05,520 --> 01:17:14,760
yeah yeah now it's compute though it's basically I guess all right let's go up

1085
01:17:14,760 --> 01:17:18,160
here previously I worked in like FinTech where are we no like let's walk out

1086
01:17:18,160 --> 01:17:22,320
here that's that's better and back in like 2015 16 everyone's on tokens and

1087
01:17:22,320 --> 01:17:26,560
that but it was all in terms of like sort of opening up their APIs oh I see

1088
01:17:26,560 --> 01:17:30,720
whatever if they use the word token or is like something like that it was all

1089
01:17:30,720 --> 01:17:34,080
about like how do you pass an authentic authentication credential between

1090
01:17:34,080 --> 01:17:37,440
payment processors yeah and stuff so like the people that had the really

1091
01:17:37,440 --> 01:17:40,160
valuable moats like some payment processor that doesn't innovate at all

1092
01:17:40,160 --> 01:17:44,800
how do they open it up to like FinTech people so that was basically all about

1093
01:17:44,800 --> 01:17:49,040
tokens as well like FinTech was all about tokens how do you make a rest API that

1094
01:17:49,040 --> 01:17:52,920
opens up some valuable system and unlocks entrepreneurship I think what what we

1095
01:17:52,920 --> 01:17:58,120
are what we are kind of heading into now is that we're gonna send e-cash tokens

1096
01:17:58,600 --> 01:18:03,560
to something and we'll get back yeah this is really I told this and we can use

1097
01:18:03,560 --> 01:18:06,920
them up I started working on e-cash like three I didn't think about this especially

1098
01:18:06,920 --> 01:18:11,240
after talking to Cali more and it's like I'm working on e-cash for like three

1099
01:18:11,240 --> 01:18:14,680
years ago yeah I've worked on a long time and I don't think I've internalized it

1100
01:18:14,680 --> 01:18:19,080
all how important it is really again you are too close to the metal yeah you

1101
01:18:19,080 --> 01:18:22,960
are too close to the metal you are trying to make the thing work that's also why

1102
01:18:22,960 --> 01:18:27,760
how I am different from you in in in in the large degree it's like I don't see

1103
01:18:27,840 --> 01:18:33,480
myself as a coder or even like software architect guy I'm first and foremost I'm

1104
01:18:33,480 --> 01:18:40,080
a power user yeah I'm I'm like I'm I try to use both Bitcoin and Noster and

1105
01:18:40,080 --> 01:18:44,920
LLM stuff like as hard as I can yeah I don't I don't I don't see myself you

1106
01:18:44,920 --> 01:18:48,680
know like building an e-cash system from scratch I might buy code it's funny I'm

1107
01:18:48,680 --> 01:18:52,240
not a power user actually and I try to be but I struggle that's what I'm that's

1108
01:18:52,240 --> 01:18:56,040
what I'm that's what I mean you know and so and so I've had period I've tried

1109
01:18:56,040 --> 01:18:59,800
something before and I just can't have the discipline to like like I see the

1110
01:18:59,800 --> 01:19:04,160
idea I want the lifestyle but I can't actually pull it off but that's true in

1111
01:19:04,160 --> 01:19:07,480
the Bitcoin world as well where people we have like core deficit like

1112
01:19:07,480 --> 01:19:12,920
insanely knowledgeable Bitcoin people that never really use Bitcoin across the

1113
01:19:12,920 --> 01:19:15,960
border with 12 words in their head or something like that or like you know

1114
01:19:15,960 --> 01:19:19,320
I'm pretty primitive with how I use well yeah aside from like cold storage I'm

1115
01:19:19,320 --> 01:19:24,120
I'm kind of like a but but it's worse than most people but it's a it's a different

1116
01:19:24,120 --> 01:19:26,560
it's a different kind of understanding and again coming back to

1117
01:19:26,560 --> 01:19:30,200
where we're making it talks about a lot it's like participatory knowing like if

1118
01:19:30,200 --> 01:19:34,880
you actually if you actually have to flee a country and you put 12 words in

1119
01:19:34,880 --> 01:19:38,360
your head and you cross the border basically naked and you pop out the

1120
01:19:38,360 --> 01:19:43,280
other end yeah and you you you have your wealth in text yeah and you start a new

1121
01:19:43,280 --> 01:19:47,320
life with 12 words in your head that's a different kind of knowing of what

1122
01:19:47,320 --> 01:19:52,640
Bitcoin is and what it can do then kind of knowing how exactly like you know

1123
01:19:52,640 --> 01:19:56,120
the the mempool policy works yeah whatever you know what I mean yeah yeah

1124
01:19:56,120 --> 01:19:58,960
and there's plenty of people that never really use Bitcoin but they're very

1125
01:19:58,960 --> 01:20:02,880
close to Bitcoin but they you know if anything use it as like a savings

1126
01:20:02,880 --> 01:20:07,040
technology or what have you yeah and I think that's true for the e-cash and the

1127
01:20:07,040 --> 01:20:09,920
LLM stuff and all those kind of things as well it's like you you have to

1128
01:20:09,920 --> 01:20:15,720
actually use you know like e-cash wrappers like like you know I mean in a

1129
01:20:15,720 --> 01:20:19,160
sense I could almost see you're not gonna like this but I could almost see

1130
01:20:19,160 --> 01:20:24,080
noss or going down it's like like in the history books is like this was this

1131
01:20:24,080 --> 01:20:28,880
was the germination pond for e-cash you know because this is where all this

1132
01:20:28,880 --> 01:20:31,920
some of the most interesting little experiments were like what sort of

1133
01:20:31,920 --> 01:20:35,280
stupid thing can you do with these tokens yeah it turns out to be actually

1134
01:20:35,280 --> 01:20:39,600
into like profound right it's gonna it's gonna it's gonna what gets us out of

1135
01:20:39,600 --> 01:20:43,320
the many of the finite future many of the like you actually have to rest my

1136
01:20:43,320 --> 01:20:47,720
leg for usability like steps forward it feels like just some guy doesn't know I

1137
01:20:47,800 --> 01:20:51,960
got a really stupid idea for what to do on yeah noss is the only place I can

1138
01:20:51,960 --> 01:20:55,440
try it and it's also that's the only way I can find another guy with an e-cash

1139
01:20:55,440 --> 01:21:00,400
lots lots lots of the coolest things we ever both came out of like it let's

1140
01:21:00,400 --> 01:21:04,240
just fucking do it for fun yeah because we kind of can do it and that's what I

1141
01:21:04,240 --> 01:21:07,040
love so much about the buy-coding stuff and that's kind of I think also the

1142
01:21:07,040 --> 01:21:09,880
point of serving engineering again I wanted to come back to that because I'm

1143
01:21:09,880 --> 01:21:12,800
still trying to figure out what the thing is what what are we actually doing

1144
01:21:12,800 --> 01:21:19,200
here and and and so so I think it is to have a test bed and access to other

1145
01:21:19,200 --> 01:21:24,280
minds that that that that that are at a similar level where you are I mean like

1146
01:21:24,280 --> 01:21:27,760
let's be real how many people are there in the world right now that you can talk

1147
01:21:27,760 --> 01:21:32,960
in-depth about e-cash and not stir and all this stuff that we discussed there's

1148
01:21:32,960 --> 01:21:37,280
I don't know like let's it hundreds yeah it's not thousands it's it's very hard

1149
01:21:37,280 --> 01:21:42,800
to to to to and try to talk with regular software guy about this stuff oh it will

1150
01:21:42,800 --> 01:21:45,920
take you two years to the worst conversation ever it's terrible I don't

1151
01:21:45,920 --> 01:21:50,200
even bring this stuff up anymore you you can't because there the the the knowledge

1152
01:21:50,200 --> 01:21:54,000
gap is so wide that you can't even have a meaningful conversation it makes me

1153
01:21:54,000 --> 01:21:58,960
dislike a Bitcoin and Nostromor you know I have to explain it so he doesn't get it

1154
01:21:58,960 --> 01:22:02,680
I'm like oh god but that's also why that's also why I think Bitcoiners are so

1155
01:22:02,680 --> 01:22:07,040
hated in some sense because it's like it's just all gibberish to the regular

1156
01:22:07,040 --> 01:22:11,080
person that that that isn't up to speed with that stuff just like you know

1157
01:22:11,080 --> 01:22:14,240
not starting dvm stuff and e-cash stuff it's like it's all gibberish to to many

1158
01:22:14,240 --> 01:22:17,720
people I think but it's so cool to work with other people that see the

1159
01:22:17,720 --> 01:22:22,040
potential in this technology and try and just experiment with it like what

1160
01:22:22,040 --> 01:22:27,360
you know is really good at is like in like software systems like some of them

1161
01:22:27,360 --> 01:22:31,640
are really really really powerful but it's in takes an incredible amount of

1162
01:22:31,640 --> 01:22:35,480
lift to do anything yeah right like like you know if you're trying to work

1163
01:22:35,480 --> 01:22:40,160
with like some compiler technology or something like yeah like leverage LLVM

1164
01:22:40,160 --> 01:22:43,820
like LLVM just like low-language compiler toolkit and it's like you can do

1165
01:22:43,820 --> 01:22:47,440
amazing so this is what rust is built on this is what so many these like new

1166
01:22:47,440 --> 01:22:51,560
zig was built on it like they're all built on the stuff but it's like you

1167
01:22:51,560 --> 01:22:55,120
know you want to do a prototype two years and you gotta be like a genius to do

1168
01:22:55,120 --> 01:22:59,920
it and like so like that's one extreme and I look at Nostra maybe the best

1169
01:22:59,920 --> 01:23:03,840
part about is it's like the other hardcore hard side of the bell curve

1170
01:23:03,840 --> 01:23:09,680
where like you can you can do something in half an hour you know and you can do

1171
01:23:09,680 --> 01:23:13,440
something non-trivial in half an hour and deploy it and have people use it

1172
01:23:13,440 --> 01:23:17,720
right that's like one of the coolest things about Nostra is that you can

1173
01:23:17,720 --> 01:23:24,540
like exercise this creative agency with you like can't really get much output

1174
01:23:24,540 --> 01:23:28,000
faster that will involve other people too like you can do something that's

1175
01:23:28,000 --> 01:23:30,920
only you on your computer but a cool thing about Nostra is you can like make

1176
01:23:30,920 --> 01:23:35,280
something deploy it now somebody play with it we've done these demos at SEC

1177
01:23:35,280 --> 01:23:38,760
where like somebody like I think a Paul's hypertext thing we make the stories

1178
01:23:38,760 --> 01:23:43,200
yeah thing that can like embed a dynamic widget for like a tip button inside this

1179
01:23:43,200 --> 01:23:48,320
Instagram stories type type UI and he'll demo it and there's the URL I mean I'll

1180
01:23:48,320 --> 01:23:52,400
pull up on our phones and like play with it right like that's immediate and

1181
01:23:52,400 --> 01:23:55,640
life and it's like that's what always happens like someone demos something and

1182
01:23:55,640 --> 01:23:58,760
shows something off and if there is like a lightning invoice to pay someone

1183
01:23:58,760 --> 01:24:02,320
yeah someone someone in the audience will pay it if there is an interactive

1184
01:24:02,320 --> 01:24:05,480
element someone will will immediately jump on the developer point if you have

1185
01:24:05,480 --> 01:24:09,440
one of the best things about Nostra is that it's like like one of the hardest

1186
01:24:09,440 --> 01:24:13,360
things about any sort of like creative new like indie hacking style

1187
01:24:13,360 --> 01:24:16,680
development is how do you get those first hundred users yeah 10 users like it's

1188
01:24:16,680 --> 01:24:22,280
all it's often like very difficult yeah and that's like and the cool thing about

1189
01:24:22,280 --> 01:24:25,680
Nostra is like in a sense your users already exist yeah like they're already

1190
01:24:25,680 --> 01:24:30,600
off to your system yeah they they I mean they might have to hit a little

1191
01:24:30,600 --> 01:24:35,560
button in their you know their all the extension to use your thing but like

1192
01:24:35,560 --> 01:24:39,480
they they already have a username and password to your database more or less

1193
01:24:39,480 --> 01:24:44,440
right like and it's also discovery is building in some sense yeah it's it's like

1194
01:24:44,440 --> 01:24:47,840
I stumble over new Nostra apps and things all the time yeah and it's just

1195
01:24:47,840 --> 01:24:51,640
like out there in this strange universe and you can just very easily stumble

1196
01:24:51,640 --> 01:24:55,040
that's why that's why I mean it's probably kind of you know if I was if I

1197
01:24:55,040 --> 01:24:57,640
put myself in the shoes of somebody listening to this and they're like okay

1198
01:24:57,640 --> 01:25:01,040
Justin's here he's gonna talk about Bitcoin or something and we haven't

1199
01:25:01,040 --> 01:25:04,360
talked about Bitcoin we just thought we know that much

1200
01:25:04,360 --> 01:25:08,560
it's assumed but it's it's like that's why the all-am stuff is kind of like in

1201
01:25:08,560 --> 01:25:12,320
this cohort I don't think we planned on talking about AI at all no and then

1202
01:25:12,320 --> 01:25:16,200
it's all we're really talking about are doing and it's I think part of it is

1203
01:25:16,200 --> 01:25:18,440
because it is a good you know we're thinking about talking about Nostra

1204
01:25:18,440 --> 01:25:23,480
and it is actually a good fit because like Nostra you can you know you can

1205
01:25:23,480 --> 01:25:29,160
an all-am can spit out like a novel cool if you have an idea in Nostra an all-am

1206
01:25:29,160 --> 01:25:33,560
can spit out a demo in like five minutes yes yeah you can actually that's that's

1207
01:25:33,560 --> 01:25:38,240
one of the reasons why whereas like an LLVM engineer like it's the complete

1208
01:25:38,240 --> 01:25:42,040
opposite really this is useless yeah like I have to spend years to get anywhere

1209
01:25:42,040 --> 01:25:45,880
this might carve off a little time but not much and it's even more than that I

1210
01:25:45,920 --> 01:25:49,440
hope I hope to talk with Pablo about it soon he is currently building out his

1211
01:25:49,440 --> 01:25:55,720
whole kind of you know vibe coding LLM stack to use Nostra directly not only

1212
01:25:55,720 --> 01:26:05,280
in the DVM CP context where like the the MCP stuff is on Nostra but also the

1213
01:26:05,280 --> 01:26:10,720
output of his various tools is pushed to relays and other tools can can grab

1214
01:26:10,720 --> 01:26:15,600
it from there and he can interact with the notes and the stuff that is produced

1215
01:26:15,600 --> 01:26:20,360
on Nostra where it's just like here's the output like to use your example it's

1216
01:26:20,360 --> 01:26:23,520
like okay here is what I built here is a screenshot and you just you see it on

1217
01:26:23,520 --> 01:26:27,320
your phone you reply to the note and it picks it up again yeah so I think I think

1218
01:26:27,320 --> 01:26:33,640
it's almost like Nostra is a QA yeah no but not like a stylistic QA like a

1219
01:26:33,640 --> 01:26:39,040
product QA but but we are the beautiful thing I think about this is is if it's

1220
01:26:39,040 --> 01:26:43,360
out in the open that's why that I kind of I kind of hate the M's and this was

1221
01:26:43,360 --> 01:26:48,040
back in the Twitter days as well because if you and I talk about this all the

1222
01:26:48,040 --> 01:26:52,000
time if you have a question for me just ask out in the open unless it's like

1223
01:26:52,000 --> 01:26:55,720
super hyper confidential and so on because I might not be the best person to

1224
01:26:55,720 --> 01:26:59,800
answer it yeah someone else might pick it up and know it that you can blast it

1225
01:26:59,800 --> 01:27:03,720
to like 20,000 people yeah just just ask it out in the open and I think for it

1226
01:27:03,720 --> 01:27:07,520
for the LLM stuff as well it's like you know other other agents can pick it up

1227
01:27:07,520 --> 01:27:10,560
other things can pick it up if you for example with your screenshot idea it's

1228
01:27:10,560 --> 01:27:13,720
like you show that's why that's why streaming is also a thing it's very

1229
01:27:13,720 --> 01:27:17,400
useful you know like you show what what you're building it posts screenshots and

1230
01:27:17,400 --> 01:27:21,160
stuff for whatever snippets to Nostra people can comment on it people can you

1231
01:27:21,160 --> 01:27:25,960
know like even make it better you can open up you can open up the prompting for

1232
01:27:25,960 --> 01:27:30,400
example to a select group of people yeah to to you know a certain subset of your

1233
01:27:30,400 --> 01:27:33,400
web of trust or what have you it's I think it's super interesting I think it

1234
01:27:33,400 --> 01:27:38,320
really is the perfect the perfect substrate for like a human agent environment

1235
01:27:38,320 --> 01:27:41,320
another interesting about this sort of like working in public space is working

1236
01:27:41,320 --> 01:27:45,080
in public idea yeah like another interesting thing like I find that on

1237
01:27:45,080 --> 01:27:49,360
Twitter very stressful because it's like at any moment you could not come

1238
01:27:49,360 --> 01:27:52,960
well no yeah you could become the guy like every once in a while someone does

1239
01:27:52,960 --> 01:27:56,200
something stupid and they are the topic of conversation that day about how

1240
01:27:56,200 --> 01:28:00,160
stupid they are yeah and it's it's a little stressful to have you know to be

1241
01:28:00,160 --> 01:28:10,000
one virality yeah event away from that yeah but the like it requires a certain

1242
01:28:10,000 --> 01:28:14,160
thick skin and also just a you have to prepare yourself for that yeah I did it

1243
01:28:14,160 --> 01:28:17,600
first to some you know let's get a lot on Twitter for a couple years but I've

1244
01:28:17,600 --> 01:28:20,720
basically flatlined my followers like five years now I think I've gained like

1245
01:28:20,720 --> 01:28:25,280
50 followers in five years I just stopped trying to engage participate on

1246
01:28:25,280 --> 01:28:28,080
the globe I mean it's funny on the nauseous like I stopped trying to

1247
01:28:28,080 --> 01:28:31,400
participate at the global level right yeah in the beginning I would try to

1248
01:28:31,400 --> 01:28:35,120
tweet and try to get attention from people outside my circle yeah and like

1249
01:28:35,120 --> 01:28:38,400
four or five years ago I think during covid I just decided to stop I just

1250
01:28:38,400 --> 01:28:41,360
didn't really want to do that anymore and the interesting thing about noster is

1251
01:28:41,360 --> 01:28:46,120
that you can sort of live in public in a sense it's easier because first of all

1252
01:28:46,120 --> 01:28:49,280
like noster doesn't really have virality to the same degree that Twitter does

1253
01:28:49,280 --> 01:28:53,000
right like yeah yes I would say yeah yeah but I mean in a sense there's no you

1254
01:28:53,000 --> 01:28:56,720
know if there's no global right to some extent like there's there is a

1255
01:28:56,720 --> 01:29:02,400
friction against there's a headwind to virality on noster yeah where Twitter

1256
01:29:02,400 --> 01:29:06,440
actually has a tailwind right like they want if there's something that is like

1257
01:29:06,440 --> 01:29:12,200
eyeballable yeah they want that they want everyone to see that right like you

1258
01:29:12,200 --> 01:29:16,680
know like Elon applies this to himself right yeah I'm sure he makes sure he's

1259
01:29:16,680 --> 01:29:20,280
got a tailwind on all his tweets right they're being pushed out much more than

1260
01:29:20,280 --> 01:29:24,280
they would organically and so I think that's from like a living from like a

1261
01:29:24,320 --> 01:29:28,400
living in public point of view like if you're going to do stuff like streaming

1262
01:29:28,400 --> 01:29:31,760
the streaming is also kind of scary right it's like sure you know like what if I

1263
01:29:31,760 --> 01:29:35,080
actually pull up some of my notes that have some embarrassing anecdote or some

1264
01:29:35,080 --> 01:29:38,960
some stuff about anime that I don't want everyone to know that I let you know

1265
01:29:38,960 --> 01:29:42,880
like you know it's kind of scary to have your computer

1266
01:29:42,880 --> 01:29:48,680
available to the internet or what if I pull up my emails you know like it is

1267
01:29:48,680 --> 01:29:54,320
kind of nice that it's that it's a smaller graph yeah like there's some

1268
01:29:54,320 --> 01:29:57,360
complaints about noster not growing a lot but that's actually like I found

1269
01:29:57,360 --> 01:30:01,520
myself not using Twitter a lot much since I got to SEC and part of it is

1270
01:30:01,520 --> 01:30:04,240
because I like that there's only there aren't that there's enough people to

1271
01:30:04,240 --> 01:30:07,040
talk there's enough people I like on noster but there aren't that many more

1272
01:30:07,040 --> 01:30:13,960
you know so I think in a sense it's a it's a benefit of noster being like like

1273
01:30:14,040 --> 01:30:18,280
you know if if noster grew very big and succeeded quote-unquote in that sense

1274
01:30:18,280 --> 01:30:23,040
like in some sense I hope it would still be usable in a way that I'm in just a

1275
01:30:23,040 --> 01:30:26,960
wall garden with 24 yeah 20,000 people right like yeah I don't know if I want

1276
01:30:26,960 --> 01:30:30,160
I mean unless I have something to sell for example if you have something to

1277
01:30:30,160 --> 01:30:35,600
sell you'd much rather be have a wider reach but like just for some some sort

1278
01:30:35,600 --> 01:30:40,680
of a quote-unquote authentic authentic uses like a smaller graph is more

1279
01:30:40,680 --> 01:30:44,120
beneficial right I think I think that's what noster excels at and I think that's

1280
01:30:44,120 --> 01:30:47,640
also thing of what though like the Piraeus or whatever like the where of

1281
01:30:47,640 --> 01:30:50,880
Socrates taught yeah the place where he would go on his walks you know down by

1282
01:30:50,880 --> 01:30:54,920
the port and Athens it's like how many people were in that dialogic sphere

1283
01:30:54,920 --> 01:30:59,640
yeah 50,000 20,000 there weren't that many people you we were limited by

1284
01:30:59,640 --> 01:31:03,280
physical space for that and I think I think we will also be I think we see

1285
01:31:03,280 --> 01:31:05,720
this emerging now on the internet not only on noster but I think that's what

1286
01:31:05,720 --> 01:31:08,640
noster is yeah it's like you're limited by intellectual space there's only so

1287
01:31:08,640 --> 01:31:12,360
many people that can talk meaningfully that can't meaningfully participate in

1288
01:31:12,360 --> 01:31:16,120
this conversation so it's gonna remain small by one and also it's like there's

1289
01:31:16,120 --> 01:31:20,720
not always like there's not always like a right way for a thing to evolve right

1290
01:31:20,720 --> 01:31:26,160
like oh no there's many like that's what actual diversity is is like yeah if

1291
01:31:26,160 --> 01:31:31,120
you if you if you segregate something off like a physical organism for like in

1292
01:31:31,120 --> 01:31:34,800
Madagascar for a hundred million years or something it becomes interesting you

1293
01:31:34,800 --> 01:31:37,160
get like a lemur or something right I think lemur's different maybe

1294
01:31:37,640 --> 01:31:42,720
else but like you get like an interesting thing and so like like that's

1295
01:31:42,720 --> 01:31:46,800
something like you know if Athens had was the size of the United States yeah it

1296
01:31:46,800 --> 01:31:49,960
would have never been a different things yeah like you could have never you could

1297
01:31:49,960 --> 01:31:53,960
have never had what it was and that's one of things is sort of disturbing about

1298
01:31:53,960 --> 01:31:58,800
the modern world is it's like it's like it's like one village now and so you

1299
01:31:58,800 --> 01:32:05,600
don't actually get much you know like stuff can't really evolve and become

1300
01:32:05,600 --> 01:32:09,640
like an interesting subculture to the same degree if it's on the global feed

1301
01:32:09,640 --> 01:32:13,200
yeah because it just it gets everything gets pulled back to the mean I will link

1302
01:32:13,200 --> 01:32:16,920
to this article as well with like how how our interfaces and at the world at

1303
01:32:16,920 --> 01:32:20,040
large lost its texture basically yeah it's like everything looks the same all

1304
01:32:20,040 --> 01:32:22,840
the clothes look the same I mean we're wearing the same stuff yeah we are

1305
01:32:22,840 --> 01:32:26,240
literally dressed almost the same we look the same yeah but it's true for like

1306
01:32:26,240 --> 01:32:29,680
almost everyone it's all the same it's like we all have the same devices

1307
01:32:29,680 --> 01:32:32,920
everyone is the same cars everyone is like I had a beard like I go beard it

1308
01:32:32,920 --> 01:32:38,560
really you look like the same guy really it's really bad and and so so I

1309
01:32:38,560 --> 01:32:41,240
think there's something to it but I think it's also interesting how for

1310
01:32:41,240 --> 01:32:45,400
example there there was this distinct difference between Twitter and Bitcoin

1311
01:32:45,400 --> 01:32:49,280
Twitter and there were all these bubbles inside Twitter that were purely there

1312
01:32:49,280 --> 01:32:53,200
when it was not because of a spacecon constraint it was only because of a

1313
01:32:53,200 --> 01:32:57,640
kind of intellectual conversational constraint it's just like the your

1314
01:32:57,640 --> 01:33:01,200
social graph basically and and that the topics would naturally clump you had

1315
01:33:01,200 --> 01:33:04,800
like fintech Twitter I mean it was the same kind of anti-entriism thing where

1316
01:33:04,800 --> 01:33:09,080
like we'd speak in memes and stuff yeah exactly it's like almost you almost

1317
01:33:09,080 --> 01:33:13,080
purposefully I mean like Pierre is like the Pierre Richard is like the person who

1318
01:33:13,080 --> 01:33:16,680
kind of helped me understand this where it's like no the the echo chamber is

1319
01:33:16,680 --> 01:33:20,720
good I like I cultivate an echo chamber absolutely and like that's that is one

1320
01:33:20,720 --> 01:33:23,400
thing that I think well that allowed Bitcoin Twitter to become sort of an

1321
01:33:23,400 --> 01:33:26,320
interesting place and I would say it kind of produced an Oscar in some sense

1322
01:33:26,320 --> 01:33:31,000
right you know Bittstein's talk of how to mean Bitcoin to the yeah I was in the

1323
01:33:31,000 --> 01:33:34,840
room yeah I was the next talk actually so I didn't enjoy it at all because I

1324
01:33:34,840 --> 01:33:40,640
hadn't really prepared my slides no one remembers my talk of course everyone

1325
01:33:40,640 --> 01:33:45,440
remembers me sitting in the corner trying to get my slides right no but it

1326
01:33:45,440 --> 01:33:49,880
but but but they they identified this earlier than anyone else and I think on

1327
01:33:49,880 --> 01:33:53,560
the nostro side we have the same effect now it's just like one big bubble more

1328
01:33:53,560 --> 01:33:56,400
or less I mean there's other bubbles on us there's like the art people yeah

1329
01:33:56,400 --> 01:33:59,320
there's like the jungle people there's like the farmer people the jungle

1330
01:33:59,320 --> 01:34:02,520
people are like the one of the things that I can't understand if that's just

1331
01:34:02,520 --> 01:34:09,060
purely primal and Jack they were they were there before the skeptic the

1332
01:34:09,060 --> 01:34:14,760
skeptic in me thinks it's a paid no but it's I think I know why all of this is

1333
01:34:14,760 --> 01:34:17,800
kind of happening in some sense or at least my explanation for it is like

1334
01:34:17,800 --> 01:34:22,000
Bitcoin Bitcoin monetized and expanded on the on the fringes on the edges no

1335
01:34:22,000 --> 01:34:25,280
it's like you had to be crazy to be an early Bitcoin and like kind of you had

1336
01:34:25,280 --> 01:34:30,200
to either live through hyperinflation or be an Austrian or be debunked and yeah

1337
01:34:30,200 --> 01:34:34,240
those kind of things like very edges of society multiple edges but still very

1338
01:34:34,240 --> 01:34:37,520
very edge of society and I think on the nostro side it's similar now that's why

1339
01:34:37,520 --> 01:34:41,360
we have we have the crazy you know like health people that think that blue light

1340
01:34:41,360 --> 01:34:44,760
is so toxic that you should never ever see what's anything blue like running

1341
01:34:44,760 --> 01:34:48,160
around with red sunglasses at all times and and it's like that's very fringe

1342
01:34:48,160 --> 01:34:51,920
kind of stuff and that's you know I'm not gonna mention absolutely all the

1343
01:34:51,920 --> 01:34:55,760
fringes because it's like too out there but all the out there people are there

1344
01:34:55,760 --> 01:34:59,120
and it's like just living in the jungle going barefoot rising with the Sun

1345
01:34:59,120 --> 01:35:02,440
going to sleep with the Sun never touching any processed food ever that's

1346
01:35:02,440 --> 01:35:05,920
kind of fringe in our today's society yeah and so I think they are

1347
01:35:05,920 --> 01:35:09,800
naturally attracted to noster and it's also what about nos if you think

1348
01:35:09,800 --> 01:35:13,520
attracts them I have an answer for this why what is the what is the what is the

1349
01:35:13,520 --> 01:35:18,800
shared like would you agree that there's some shared yeah I think because it's

1350
01:35:18,800 --> 01:35:23,320
raw and it's true and it requires some work yeah it's it's and I want I wanted

1351
01:35:23,320 --> 01:35:26,240
to come back to that the the realness and people really crave realness and

1352
01:35:26,240 --> 01:35:31,760
noster and Bitcoin as well it's very real it's like it's you can you can feel

1353
01:35:31,760 --> 01:35:36,160
kind of that it's not manipulated by an algorithm it's not manipulated by a big

1354
01:35:36,160 --> 01:35:41,320
tech thing or by ads or by influencers or by this and no one no one no one tries

1355
01:35:41,320 --> 01:35:45,560
to sell you something everyone's engagement in this is probably at the

1356
01:35:45,560 --> 01:35:52,200
pure loss of absolutely yeah I spent trying to develop I'm not I could be

1357
01:35:52,200 --> 01:35:57,520
making money selling AI to but not only that I never I never spent money on

1358
01:35:57,520 --> 01:36:01,960
software ever in my life like I always hated that thought yeah I'm spending

1359
01:36:01,960 --> 01:36:06,280
so much money on noster right now yeah paid relays yeah running my own servers

1360
01:36:06,280 --> 01:36:10,020
paid file hosting I started doing this funny I had it like I was using the

1361
01:36:10,020 --> 01:36:13,340
primal web plan because again I'm not a party so I can use like two three tools

1362
01:36:13,340 --> 01:36:18,660
and this fucking thing wouldn't allow me to upload an mp4 interesting yeah yeah

1363
01:36:18,660 --> 01:36:23,940
in the opposite can right I think I don't know but yes you have an iPhone so it's

1364
01:36:23,940 --> 01:36:30,220
almost impossible to get a fucking mp4 on my phone literally almost that

1365
01:36:30,220 --> 01:36:35,500
camera roll yeah I know yeah it's almost impossible and I was struggling for a

1366
01:36:35,580 --> 01:36:40,060
long time and finally I I was like what's that like it's not noster dot build or

1367
01:36:40,060 --> 01:36:42,820
something yeah so I pulled up that I knew that this could do this and it's like

1368
01:36:42,820 --> 01:36:47,180
what is it like a hundred dollars a year I paid a hundred dollars a year I

1369
01:36:47,180 --> 01:36:53,780
had a hundred dollars up on an mp4 of Terry Davis typing into the CIA website

1370
01:36:53,780 --> 01:36:58,700
that he's gonna kill them let's go

1371
01:36:58,700 --> 01:37:02,900
like you're all gonna this is King Terry you're all gonna die the hand of God

1372
01:37:02,900 --> 01:37:06,860
you you are the fringe that no stir is attractive yeah I paid a hundred I

1373
01:37:06,860 --> 01:37:11,700
paid a hundred dollars that's by the way that's the reason why I why I switched

1374
01:37:11,700 --> 01:37:14,900
away from the iPhone like 10 years ago so because I just wanted to drag and

1375
01:37:14,900 --> 01:37:18,660
drop an mp3 file on my phone it was impossible it's so literally impossible

1376
01:37:18,660 --> 01:37:22,860
so I gave up and switch and so funny because that was the entire premise the

1377
01:37:22,860 --> 01:37:26,380
whole thing in the beginning this is a drag and drop mp3 experience now it's

1378
01:37:26,380 --> 01:37:30,820
like you can't even do it again you can't you know you own nothing yeah I know I'm

1379
01:37:30,820 --> 01:37:34,900
gonna get a graphing phone and try like every time Vitor and that final guy on

1380
01:37:34,900 --> 01:37:40,300
roster he's one of my favorite followers the graphing dove yeah whenever they

1381
01:37:40,300 --> 01:37:44,500
talk about graphing it is like that is actually pretty interesting yeah I mean

1382
01:37:44,500 --> 01:37:49,380
I run graphing on all my phones yeah yeah it is it is pretty interesting yeah so

1383
01:37:49,380 --> 01:37:52,460
I'm gonna definitely you know when I get back home I'm gonna buy one of those

1384
01:37:52,460 --> 01:37:56,300
yeah but it is you're the same vein you're saying you're not a power user so I'm

1385
01:37:56,300 --> 01:37:59,340
not sure how much fun you're gonna have because it is a power user kind of

1386
01:37:59,340 --> 01:38:02,700
thing you will have to think of with it yeah we'll have to spend a lot of

1387
01:38:02,700 --> 01:38:05,740
I mean at the same time it's like I did just give up cursor which everything

1388
01:38:05,740 --> 01:38:09,420
works to try to do everything on the manline neo van and like I spent

1389
01:38:09,420 --> 01:38:12,700
little like this week like we got to go do demos in a couple hours and I have

1390
01:38:12,700 --> 01:38:16,220
like nothing like I've been all week I've been like last Tuesday I was about

1391
01:38:16,220 --> 01:38:20,620
ready to start making a phone call app like I have the whole setup I have the

1392
01:38:20,620 --> 01:38:26,580
whole all the prompts ready to build a like literally like an encrypted audio

1393
01:38:26,580 --> 01:38:30,900
call and I have the prompts ready no no literally I have basically done this

1394
01:38:30,900 --> 01:38:35,220
after this fucking chamber like I I spent a whole weekend going back and forth I

1395
01:38:35,220 --> 01:38:40,260
have this really I have it's perfect like everything is ready like I figured out

1396
01:38:40,260 --> 01:38:44,060
exactly the libraries to use how to do it in all rest like the only thing my UI

1397
01:38:44,060 --> 01:38:47,500
needs is buttons like I don't you have to do any Kotlin to integrate with the

1398
01:38:47,500 --> 01:38:51,340
audio system I don't have to I use one Google live you should use call me web

1399
01:38:51,340 --> 01:38:56,280
RTC which is this massive Google it's like 11 components or something massive

1400
01:38:56,280 --> 01:39:00,440
C++ code base impossible to cross like really hard to cross compile it's

1401
01:39:00,440 --> 01:39:04,280
funny like web RTC is this web standard right so it's the idea to make a

1402
01:39:04,280 --> 01:39:07,400
standard and different browsers implement it like Safari and Firefox

1403
01:39:07,400 --> 01:39:11,760
did not have enough money to write this so you just fucking run Google C++ code

1404
01:39:11,760 --> 01:39:15,360
in their browser so like blue sky works it's kind of insane like it's kind of

1405
01:39:15,360 --> 01:39:19,640
insane like it's like learning about this and so I figured out it's one company

1406
01:39:19,640 --> 01:39:22,560
who made the standard one company that had enough money to implement it at

1407
01:39:22,560 --> 01:39:26,360
one one it's just a talk about it so my thing is like I'm still going to use

1408
01:39:26,360 --> 01:39:29,680
the audio processing that's one thing that they that is really good is they

1409
01:39:29,680 --> 01:39:32,920
have this you know like noise cancellation library where like you can

1410
01:39:32,920 --> 01:39:36,040
make the calls sound really crisp so that's the one thing I'm going to use and

1411
01:39:36,040 --> 01:39:40,920
everything else it's it's much simpler things and it's using the same underlying

1412
01:39:40,920 --> 01:39:47,040
library that Fediment just integrated for for to make it so you can do peer to

1413
01:39:47,040 --> 01:39:51,080
peer connections with the guardians so the guardians don't need DNS names which

1414
01:39:51,120 --> 01:39:55,320
is one of the worst things about Fediment is that you need to the guardians need

1415
01:39:55,320 --> 01:39:59,800
to get DNS names and you know with one of the federations I forget which one

1416
01:39:59,800 --> 01:40:04,720
it was you know they had XYZ things and the registrar banned them you know so I

1417
01:40:04,720 --> 01:40:07,520
think they ended up you know being able to sort that whole thing out but it was

1418
01:40:07,520 --> 01:40:10,680
really annoying and that's just because it was integrating with this dying web

1419
01:40:10,680 --> 01:40:15,080
2.0 infrastructure that is that it's kind of dying and it's one of the

1420
01:40:15,080 --> 01:40:19,320
biggest issues of the web is DNS yeah I hope that Noster like I hope that we

1421
01:40:19,320 --> 01:40:22,280
can just use switch to endpups basically yeah I don't care how it

1422
01:40:22,280 --> 01:40:25,520
resolves underneath yeah I really don't care it's just like I think this is

1423
01:40:25,520 --> 01:40:31,240
called IRO like that's it's a cool way to to I think there's some like DHT that

1424
01:40:31,240 --> 01:40:36,440
you can use to connect you know nodes to so they gonna like establish it's

1425
01:40:36,440 --> 01:40:40,960
very much a coal punch but it's not written in JavaScript which is nice and

1426
01:40:40,960 --> 01:40:47,520
and it doesn't have to have its own runtime and it yeah so that is and so

1427
01:40:47,520 --> 01:40:51,280
like that I think the Jeff posted the other day this like new social network

1428
01:40:51,280 --> 01:40:56,960
yeah yeah build by one guy yeah that's the the guy who did I'm sure did most of

1429
01:40:56,960 --> 01:41:00,920
the stuff for Fetteman DVC is an unbelievably good programmer you know

1430
01:41:00,920 --> 01:41:04,600
he's just like one of these like rust purists yeah he's just like Noster

1431
01:41:04,600 --> 01:41:09,120
sucks cuz that's Jason right like he hates the part that I think is the killer

1432
01:41:09,120 --> 01:41:12,760
app that you can get to and I go from idea to implementation in five minutes

1433
01:41:12,760 --> 01:41:15,960
that's the party doesn't like yeah because he's like this unbelievably good

1434
01:41:15,960 --> 01:41:19,600
programmer yeah but but that's also it's optimizing for the wrong thing because

1435
01:41:19,600 --> 01:41:23,040
his stuff is also like purely peer-to-peer yeah that's like extremely misguided I

1436
01:41:23,040 --> 01:41:27,320
think it was a deep inside by fear Jeff to actually use relays yeah use them as

1437
01:41:27,320 --> 01:41:31,080
dumb servers because my phone is offline most of the time so peer-to-peer stuff

1438
01:41:31,080 --> 01:41:34,880
just if you try running peer-to-peer stuff on your phone and like 95% of

1439
01:41:34,880 --> 01:41:38,000
people are on your phone it doesn't work but the cool thing for like an audio

1440
01:41:38,000 --> 01:41:41,600
call is that it can be like I think it can be better because you can

1441
01:41:41,600 --> 01:41:45,280
establish a peer-to-peer absolutely different push notifications different

1442
01:41:45,280 --> 01:41:49,160
use case if you're actually calling you're actually using your device that his

1443
01:41:49,160 --> 01:41:52,920
rostra thing and then the Fetteman thing kind of opened my eyes to this new

1444
01:41:52,920 --> 01:41:56,480
new library and I'm like yeah that would be good for making a phone like just to

1445
01:41:56,480 --> 01:42:01,400
have fun not to doesn't have a phone yeah that's one of the like this Unix

1446
01:42:01,400 --> 01:42:04,640
tools is like I think about like how could I you know how could you have a

1447
01:42:04,640 --> 01:42:07,720
phone that just replaced is all like why doesn't there a camera but that show

1448
01:42:07,720 --> 01:42:11,560
but that also shows you what noster already is noster is like the global

1449
01:42:11,560 --> 01:42:14,840
address book that's the one first thing in the first chord actually you build

1450
01:42:14,840 --> 01:42:23,160
into the the mutiny guys build the social money sending into the Fetteman

1451
01:42:23,160 --> 01:42:27,000
stuff that they had and so it's like you want to I want to send money to Justin

1452
01:42:27,000 --> 01:42:31,800
yeah I want to call Justin yeah I want to send an encrypted DM to Justin I want

1453
01:42:31,800 --> 01:42:35,280
to ask a question and tag Justin like I only like one way I think about

1454
01:42:35,280 --> 01:42:38,720
noster is it's like it's like you look at your phone and it's like to me

1455
01:42:38,720 --> 01:42:42,960
noster is in a sense it's like the desktop itself and like the contact

1456
01:42:42,960 --> 01:42:46,760
book that that's that's why that's so far and it's growing more and more of

1457
01:42:46,760 --> 01:42:49,880
it's populating more and more of these and that's why that's why my white

1458
01:42:49,880 --> 01:42:54,000
noise is like the messaging thing right like it's good like you have this empty

1459
01:42:54,000 --> 01:42:58,280
desktop yeah and you have the fucking contact book which is I like an

1460
01:42:58,280 --> 01:43:02,760
abstraction layer in the 2010 era of like Silicon Valley technology like that

1461
01:43:02,760 --> 01:43:06,960
was the one thing they were just they would just fight to try to get into your

1462
01:43:06,960 --> 01:43:11,040
contacts right like that's what every Facebook and Snapchat is like the one

1463
01:43:11,040 --> 01:43:14,120
thing they needed from you is access your address book because then they could

1464
01:43:14,120 --> 01:43:17,840
bootstrap their social network without that they couldn't figure out who you

1465
01:43:17,840 --> 01:43:21,120
know in order to give you a good experience yeah but it's like the

1466
01:43:21,120 --> 01:43:26,240
noster developer experience the contact book comes stock yeah which is pretty

1467
01:43:26,240 --> 01:43:29,400
important and I think that's exactly right and it's like it's like it's like

1468
01:43:29,480 --> 01:43:33,320
one abstraction layer higher in some sense it's it's it's very very much like

1469
01:43:33,320 --> 01:43:36,960
what I just said it's like I just want to talk to an end pop I don't care what

1470
01:43:36,960 --> 01:43:40,480
is used underneath and how it resolves whether it's to war or I don't I really

1471
01:43:40,480 --> 01:43:44,320
don't care yeah and and so that's one abstraction layer the other abstraction

1472
01:43:44,320 --> 01:43:48,120
layer is like I I don't care what client I use I don't care what DM client I use

1473
01:43:48,120 --> 01:43:52,520
I don't care what what what I don't know what I used to edit my my profile

1474
01:43:52,520 --> 01:43:56,240
metadata of those kind of things and and the way I see it expanding now that's

1475
01:43:56,240 --> 01:43:59,120
why it's my dream flow I want to get a new phone I don't care what phone it is

1476
01:43:59,120 --> 01:44:02,920
and I want to punch in my my my noster identity in whatever way let's say

1477
01:44:02,920 --> 01:44:06,200
it's like 12 words I want to punch in the 12 words and everything is there yeah

1478
01:44:06,200 --> 01:44:10,640
same for I mean I'm sure you have a sophisticated like dot file setup yeah

1479
01:44:10,640 --> 01:44:14,160
and those kind of things yeah and if you change machines right farming most of

1480
01:44:14,160 --> 01:44:18,640
the week yeah there you go there you go and so so wouldn't it be awesome if

1481
01:44:18,640 --> 01:44:22,120
it's just like you you you everything is almost like a think client you know

1482
01:44:22,120 --> 01:44:25,280
you go there you log in with your noster identity and it grabs all the

1483
01:44:25,320 --> 01:44:29,760
right things it knows what what your dot file setup is it knows how you want

1484
01:44:29,760 --> 01:44:33,640
your idea to have set up and so on and that's it and in that context it sets

1485
01:44:33,640 --> 01:44:36,360
everything up for you yeah that's actually one of the things I was playing

1486
01:44:36,360 --> 01:44:40,320
around with him that I want to that I did this because I made like a nix flick

1487
01:44:40,320 --> 01:44:45,760
for for X X code oh I saw this idea where like if you want like I've been

1488
01:44:45,760 --> 01:44:50,200
trying to make I'm calling it Rust multi-platform so it's like this like

1489
01:44:50,200 --> 01:44:53,640
basically like how it's kind of like some of my learnings from when we built

1490
01:44:53,640 --> 01:44:57,960
FETI this app you know it's a mobile app right like if I if I had to build a

1491
01:44:57,960 --> 01:45:01,000
similar kind of mobile app again how would I want to do things differently

1492
01:45:01,000 --> 01:45:05,160
right and so like one thing that I would like to do differently and not not

1493
01:45:05,160 --> 01:45:08,360
saying like FETI's bad or anything right it's like it's it's any kind of

1494
01:45:08,360 --> 01:45:11,640
project you learn stuff from it right and you think like what would you do next

1495
01:45:11,640 --> 01:45:17,480
time and so like you know we had a lot of logic in like what I want to do like

1496
01:45:17,480 --> 01:45:20,800
there's this big kind of quest in development and I don't think it's

1497
01:45:20,800 --> 01:45:24,560
solvable like there's no solution but there are trade-offs right right so

1498
01:45:24,560 --> 01:45:31,660
there's not like what do you look at this yeah yeah yeah yeah this is not a paid

1499
01:45:31,660 --> 01:45:37,440
post it's not a paid every time but you know there's no solution so it's

1500
01:45:37,440 --> 01:45:41,680
it's not like a thing where you're gonna solve cross-platform development but I

1501
01:45:41,680 --> 01:45:45,680
think there's other trade-offs that aren't explored so like one thing people

1502
01:45:45,680 --> 01:45:50,080
do for a mobile app is and the thing you can do base the base thing is you'd

1503
01:45:50,080 --> 01:45:53,000
write two different apps right one right one for iOS one for Android and

1504
01:45:53,000 --> 01:45:57,560
they're completely separate this is what for example primal does right but in a

1505
01:45:57,560 --> 01:46:00,560
sense they offload a lot of the their solution is actually to do it on a

1506
01:46:00,560 --> 01:46:03,800
server right they have this caching server that actually has like a lot of

1507
01:46:03,800 --> 01:46:10,400
their actual business logic is in a is in postgres functions yeah to populate a

1508
01:46:10,400 --> 01:46:14,320
cache ahead of time for what your client will see so their cross-platform

1509
01:46:14,320 --> 01:46:16,880
development is literally postgres functions which I think technically is

1510
01:46:16,920 --> 01:46:21,120
really cool novel trade-off right like that's I think that's that's kind of

1511
01:46:21,120 --> 01:46:25,600
interesting like on another one but you know the actual app development is they

1512
01:46:25,600 --> 01:46:29,560
have two separate apps which is why sometimes the iOS app is significantly

1513
01:46:29,560 --> 01:46:32,520
worse in the Android device versa right like they don't have the same

1514
01:46:32,520 --> 01:46:35,240
experience and they have one for web so they have three different apps right so

1515
01:46:35,240 --> 01:46:39,160
you see how that was slowed on velocity a lot right where it's like this week

1516
01:46:39,160 --> 01:46:42,760
we're implementing some feature we're just playing blossom okay all the devs

1517
01:46:42,760 --> 01:46:46,280
go and do the same thing in different languages yeah that's kind of what they

1518
01:46:46,320 --> 01:46:49,760
have to do right so in a perfect world you would have to do less of that you

1519
01:46:49,760 --> 01:46:53,920
know and so another option is like react native so there's this big thing like

1520
01:46:53,920 --> 01:46:57,080
15 years ago or Facebook is like well what if you could just write it all on

1521
01:46:57,080 --> 01:47:00,880
JavaScript it's kind of like the noster of social networks in a sense like what

1522
01:47:00,880 --> 01:47:05,120
if all these web devs can like make mobile apps wouldn't that be cool but it

1523
01:47:05,120 --> 01:47:08,440
came out somewhat of a cost is like JavaScript's not always the best

1524
01:47:08,440 --> 01:47:14,280
language and now you have to have a whole runtime you kind of embed a browser at

1525
01:47:14,280 --> 01:47:18,560
least the JavaScript part of a browser well and the HTML and the CSS part of

1526
01:47:18,560 --> 01:47:22,680
a browser into your mobile app so you have this huge kind of runtime and

1527
01:47:22,680 --> 01:47:27,440
complexity cost in order to get this is also it always was a real cost and

1528
01:47:27,440 --> 01:47:32,120
that's why the that's always felt sluggish yeah you could you can't feel

1529
01:47:32,120 --> 01:47:35,840
the difference between the experience of react native in as a because I've built

1530
01:47:35,840 --> 01:47:40,320
a few react native projects in the past like you know Fedi was one and I've

1531
01:47:40,320 --> 01:47:43,440
done a couple other ones and the experience is like you launch quickly

1532
01:47:43,440 --> 01:47:46,640
like you get off to a good start it's like a projectile though right yeah you

1533
01:47:46,640 --> 01:47:52,280
launch quickly and over time your you sort your velocity your vertical

1534
01:47:52,280 --> 01:47:56,920
velocity decreases right also what I experienced is like the closer you have

1535
01:47:56,920 --> 01:48:00,600
to be to the hardware itself basically yeah the harder stuff gets and it's

1536
01:48:00,600 --> 01:48:05,040
just like you'll you'll end up writing native modules anyway yeah yeah now

1537
01:48:05,040 --> 01:48:08,880
you have this whole abstraction layer that you're not even using anymore so

1538
01:48:08,880 --> 01:48:11,840
like that's one solution and flutter is kind of the same thing with some

1539
01:48:11,840 --> 01:48:15,800
technical details but from Google and there's some pros and cons

1540
01:48:15,800 --> 01:48:20,320
had a great point by the way just a day be there from amethyst discussing like

1541
01:48:20,320 --> 01:48:23,960
what it means to build something in a native way what does it mean to be

1542
01:48:23,960 --> 01:48:27,560
native yeah and I think that was excellent because I didn't see that because

1543
01:48:27,560 --> 01:48:31,880
it's basically I think it's basically a dig on will be like trying to write

1544
01:48:31,880 --> 01:48:35,320
everything as low-level as you can and writing its own like gaming rendering

1545
01:48:35,320 --> 01:48:39,360
engine for an Oscar client like rewriting the gift rendering and absolutely

1546
01:48:39,360 --> 01:48:41,720
everything you have to rewrite everything from scratch and that's kind

1547
01:48:41,720 --> 01:48:45,640
of like like what it means when we say native is like I want my iOS up to feel

1548
01:48:45,640 --> 01:48:48,840
like an iOS app yeah my Android app to feel like an Android app yeah here's

1549
01:48:48,840 --> 01:48:52,600
the thing and I think that's why it was such an excellent word but it but he

1550
01:48:52,600 --> 01:48:56,160
didn't it didn't expand on it properly so so so I'm gonna I'm gonna I'm gonna

1551
01:48:56,160 --> 01:49:00,240
try because the thing is like you're so close to the metal and you're trying to

1552
01:49:00,240 --> 01:49:03,440
achieve one thing in most cases like he wants to build the most of a rotate he

1553
01:49:03,440 --> 01:49:08,200
wants to rotate a cube in the most in the most performant way possible which is

1554
01:49:08,200 --> 01:49:12,200
fair it's like you know like you're like a 300 FPS Q yeah it's perfectly

1555
01:49:12,200 --> 01:49:17,340
programmer I will never be yeah same for me and I have no I have no

1556
01:49:17,340 --> 01:49:20,760
interest in stopping him and the stuff he's building out with the with the LLM

1557
01:49:20,760 --> 01:49:24,800
stuff and something it's amazing like I that if you if you if you want to stop

1558
01:49:24,800 --> 01:49:27,760
someone like will you retard it like just let's just let's an example of that

1559
01:49:27,760 --> 01:49:31,720
divergent evolution though right yeah it is good that he's doing that he should

1560
01:49:31,720 --> 01:49:35,080
do all the thing but but my point about that the native apps is like does your

1561
01:49:35,080 --> 01:49:39,000
app work with a screen reader for playing people yeah does it work on like

1562
01:49:39,000 --> 01:49:43,800
you know like the Android fridge yeah you know that someone is using and

1563
01:49:43,800 --> 01:49:47,760
those kind of things yeah and it's like you have no idea how much heavy lifting

1564
01:49:47,760 --> 01:49:52,600
that does the animation as you scroll between tabs yeah feel like the other

1565
01:49:52,600 --> 01:49:55,600
like that's exactly that's where the actual native feel all the gestures

1566
01:49:55,600 --> 01:49:59,320
gestures work like like this is like like flutter I can oftentimes tele flutter

1567
01:49:59,320 --> 01:50:02,440
app because they they do like with the difference between react native and

1568
01:50:02,440 --> 01:50:05,920
flutter is react native has this runtime to execute code and then it you

1569
01:50:05,920 --> 01:50:11,080
produce basic HTML and the the system itself maps your HTML to native

1570
01:50:11,080 --> 01:50:14,400
components so it is able to feel a little more native because it's actually using

1571
01:50:14,400 --> 01:50:18,600
like native scrolls yeah and then what flutter does is they just have a canvas

1572
01:50:18,600 --> 01:50:22,000
that they just draw pixels to exactly and so they're able to they have more

1573
01:50:22,000 --> 01:50:25,120
control in a sense but it's also not native components so you kind of everyone

1574
01:50:25,120 --> 01:50:28,560
so you kind of feel like that doesn't feel like Android but and you don't even

1575
01:50:28,560 --> 01:50:31,720
might not notice it but it like that doesn't feel native if you use this

1576
01:50:31,720 --> 01:50:34,920
screen reader because you're blind you really know this because it stops

1577
01:50:34,920 --> 01:50:38,720
working yeah if you use it in a different language kind of works for flutter is

1578
01:50:38,720 --> 01:50:43,120
that what I mean what do you think their dev cost is on a yearly basis to

1579
01:50:43,120 --> 01:50:46,680
produce this hundreds of millions of dollars I would guess right like I saw

1580
01:50:46,680 --> 01:50:51,120
some crazy thing this this really interesting guy what's his name he's

1581
01:50:51,120 --> 01:50:57,320
making Ladybird this new browser yeah I thought the movie no no no really

1582
01:50:57,320 --> 01:51:01,800
interesting guy he serenity what's his name I don't know yeah see he he was

1583
01:51:01,800 --> 01:51:05,400
like an addict and he was doing he's a browser engineer who got addicted to

1584
01:51:05,400 --> 01:51:08,560
drugs he's very public about it and he's like he's doing the 12 steps and he's

1585
01:51:08,560 --> 01:51:12,000
like okay in order to stop doing drugs I'm gonna write an operating system right

1586
01:51:12,000 --> 01:51:16,760
just like the temple OS guy I don't know it's it's very similar he's one he's

1587
01:51:16,760 --> 01:51:19,640
very admirable people like he's one of the best streamers on YouTube I wish I

1588
01:51:19,640 --> 01:51:24,120
could remember his name I see his face but I can't remember his name and he

1589
01:51:24,120 --> 01:51:28,560
started streaming it and so it's like he started with nothing and started make

1590
01:51:28,560 --> 01:51:31,780
it is I his idea was like the two best things in operating system are the

1591
01:51:31,780 --> 01:51:37,480
Unix command line in 90s you'll like this 90s Windows applications because it had

1592
01:51:37,480 --> 01:51:43,080
like a very true well it's very like the windows 95

1593
01:51:43,080 --> 01:51:47,960
extra useful software actually and it was very texture allow me to go on this

1594
01:51:47,960 --> 01:51:52,720
tangent because that's when they built the UI things they figured out so many

1595
01:51:52,720 --> 01:51:55,640
things how are you quick for for most okay I'll tell you some things that

1596
01:51:55,640 --> 01:51:59,800
stopped working while I while while computer industry evolved we broke so

1597
01:51:59,800 --> 01:52:03,400
many things okay so for example the scrollbars yeah you would they were

1598
01:52:03,400 --> 01:52:05,800
always visible you would could always see at the glance where you were yeah

1599
01:52:05,800 --> 01:52:09,480
we are hiding them okay the the little X a little different on every app yeah

1600
01:52:09,480 --> 01:52:13,360
fire boxes different from chrome the little X not like like what is that they

1601
01:52:13,360 --> 01:52:17,200
did all these studies they put so much thought into all this stuff like the

1602
01:52:17,200 --> 01:52:20,920
Windows 95 UI the general like all this UI stuff that came out of you know like

1603
01:52:20,920 --> 01:52:23,720
fucking Bell Labs and all those kind of things also how the mouse works all

1604
01:52:23,720 --> 01:52:26,520
those kind of things they put so much thought into it so so they figured out

1605
01:52:26,520 --> 01:52:30,800
what are there what are the five positions that the user can access most

1606
01:52:30,800 --> 01:52:35,600
quickly with a mouse yeah and it's like top left corner top right corner bottom

1607
01:52:35,600 --> 01:52:38,800
left corner bottom right corner because you can throw your mouse in that

1608
01:52:38,800 --> 01:52:42,760
direction you don't have to point you can just throw and click yeah that's why

1609
01:52:42,760 --> 01:52:46,360
when a window is maximized the close button is in the top right corner you

1610
01:52:46,360 --> 01:52:50,680
can throw it click it and the window is closed in modern operating systems the

1611
01:52:50,680 --> 01:52:56,520
button is round at the hit the hit box doesn't expand to the top right pixel

1612
01:52:56,520 --> 01:53:00,920
yeah and so you actually can't do that anymore so it's completely besides the

1613
01:53:00,920 --> 01:53:05,560
point and the most important pixel in terms of mouse that you have is where

1614
01:53:05,560 --> 01:53:10,840
the mouse cursor is at right now so if dialogue pops up it should pop up below

1615
01:53:10,840 --> 01:53:14,640
the mouse and the mouse should be on the okay button or whatever those kind of

1616
01:53:14,640 --> 01:53:18,160
things the old UI systems did this yeah they did this perfectly and they also

1617
01:53:18,160 --> 01:53:21,480
had hotkeys for everything for example where it's just like you know like a

1618
01:53:21,480 --> 01:53:25,280
control F or alt F or whatever it was his file and every you it had this

1619
01:53:25,280 --> 01:53:29,320
coverability for hotkeys where it in the menus it showed you the the hotkeys

1620
01:53:29,320 --> 01:53:33,760
are underlined yeah and it would just work and you could in you could get so

1621
01:53:33,760 --> 01:53:37,600
good and so quick at using those interfaces and we destroyed all of that

1622
01:53:37,600 --> 01:53:41,320
just because we wanted to have it like the designers came in and broke it yeah

1623
01:53:41,320 --> 01:53:45,720
yeah yeah yeah no it's like yeah the designer displaced the power users yeah

1624
01:53:45,720 --> 01:53:49,800
and it's because we literally had like form follows function and and and now

1625
01:53:49,800 --> 01:53:52,400
we really don't have the name I want to look up this guy's name just because I

1626
01:53:52,400 --> 01:53:58,480
think the listeners might actually like this but yeah so he started he

1627
01:53:58,480 --> 01:54:04,160
started making uh it's his name he started making this browser or this

1628
01:54:04,160 --> 01:54:08,680
operating system Andreas Kling yeah he's great YouTube so so for example one of

1629
01:54:08,680 --> 01:54:12,840
his YouTube best YouTube videos he's like we don't have a browser and I don't

1630
01:54:12,840 --> 01:54:16,200
want to cross-compile like I don't really maybe he did port like Chrome or

1631
01:54:16,200 --> 01:54:19,760
some Firefox or something he's like okay we ported it but like it doesn't fit in

1632
01:54:19,760 --> 01:54:23,000
I don't like it we need our own so he that's what he does he just like literally

1633
01:54:23,000 --> 01:54:26,800
rewrites everything is this operating system has zero dependencies zero it's

1634
01:54:26,800 --> 01:54:30,480
like tempo is yeah I know it's really like tempo less right and it has zero

1635
01:54:30,480 --> 01:54:35,200
dependencies is he as you say insane that's the tempo less guy in some ways more

1636
01:54:35,200 --> 01:54:40,720
in some ways less right like you know I think he's more insane because he

1637
01:54:40,720 --> 01:54:43,480
actually is trying to do this whereas I don't think Terry Davis was actually

1638
01:54:43,480 --> 01:54:46,200
trying to get people to use this operating system this guy to some extent

1639
01:54:46,200 --> 01:54:50,080
is and now but so you did he this great stream where he's like we need a browser

1640
01:54:50,080 --> 01:54:54,720
so let's see how do you tokenize JavaScript or I think he had a browser he

1641
01:54:54,720 --> 01:54:58,440
had an HTML browser and then maybe they had a CSS and I was like okay we can't

1642
01:54:58,440 --> 01:55:02,560
do anything dynamic in our a browser he was like a 1995 browser he's like well

1643
01:55:02,560 --> 01:55:06,280
we need JavaScript so he he started he like literally good like Google's

1644
01:55:06,760 --> 01:55:10,760
tokenize JavaScript and he in any in any Google's like hello world

1645
01:55:10,760 --> 01:55:16,320
JavaScript and he pays the hello world code into the tokenizer and he's like

1646
01:55:16,320 --> 01:55:20,120
okay so here's the token so let's go into C++ and let's make a struct and he

1647
01:55:20,120 --> 01:55:24,280
starts like copying over the struct and like names the you know fun make like

1648
01:55:24,280 --> 01:55:28,960
let's make a token for function let's make a token for closure like for the

1649
01:55:28,960 --> 01:55:34,200
name state right and so he slowly so you can see how he started by a hello world

1650
01:55:34,280 --> 01:55:39,320
function tokenizing hello world function which any dev can do and then they just

1651
01:55:39,320 --> 01:55:42,720
iterated that on a couple years and now it's ladybird browser which they have

1652
01:55:42,720 --> 01:55:46,080
taken out of the browser project and now they're actually trying to make this

1653
01:55:46,080 --> 01:55:50,240
to be a new browser implementation and so now they did take on dependencies they

1654
01:55:50,240 --> 01:55:54,000
use curl for networking they have a couple dependencies like that maybe libGL

1655
01:55:54,000 --> 01:55:57,480
for certain things and they're trying either they have like a one or two

1656
01:55:57,480 --> 01:56:01,680
million dollar budget and this is why I thought of this guess what Google Chrome's

1657
01:56:01,680 --> 01:56:06,680
budget is hundreds two or three billion dollars per year and there these guys

1658
01:56:06,680 --> 01:56:09,280
are doing it like one or two million dollars and they're doing a from scratch

1659
01:56:09,280 --> 01:56:12,120
browser implementation and the thing that's so inspiring about is he has one

1660
01:56:12,120 --> 01:56:17,000
talk but he's like you know most people overestimate what they can do or what

1661
01:56:17,000 --> 01:56:20,280
do you say yeah it's the no no no it that was like this that's a different one

1662
01:56:20,280 --> 01:56:25,560
he's like he's like he's like people kind of overestimate or underestimate what

1663
01:56:25,560 --> 01:56:29,280
you can do if you just if you kind of like you can't remember exactly is you

1664
01:56:29,280 --> 01:56:33,440
kind of just keep going if you just keep going like like people always say

1665
01:56:33,440 --> 01:56:37,200
you can't build a browser like that was kind of tribal wisdom programming and

1666
01:56:37,200 --> 01:56:41,160
he's like well watch me I'm just gonna start from tokenizing JavaScript and now

1667
01:56:41,160 --> 01:56:44,040
they're like pretty close to have like they can render a lot of websites they

1668
01:56:44,040 --> 01:56:47,000
can kind of render discord they can kind of render Twitter wow like they're

1669
01:56:47,000 --> 01:56:51,200
getting pretty close and it's it's kind of an inspirational project of how how

1670
01:56:51,200 --> 01:56:57,920
far a small team can go yeah if they just really try hard yeah yeah sure sure I

1671
01:56:57,920 --> 01:57:03,400
mean let's yeah that's impressive how many people are are on the team five or

1672
01:57:03,400 --> 01:57:06,320
something I mean it's a big community but like there's a few people that work

1673
01:57:06,320 --> 01:57:11,480
on full-time yeah and it's you know I mean it's it's just this massive yeah

1674
01:57:11,480 --> 01:57:16,280
I mean you truly do kind of see though how I mean it is still possible to build

1675
01:57:16,280 --> 01:57:19,680
a browser like they're doing it but it's like man we kind of went overboard

1676
01:57:19,680 --> 01:57:24,360
with you know sure it's in a sense like the web got re-architected yeah this is

1677
01:57:24,360 --> 01:57:28,120
by the WebRTC examples a good example it's like in a sense the web was kind of

1678
01:57:28,120 --> 01:57:32,520
re-architected so that it could be a wild garden yeah as soon as you add WebRTC

1679
01:57:32,520 --> 01:57:39,120
to the the whatever IETF spec for the web it is no longer something that even

1680
01:57:39,120 --> 01:57:43,200
even Apple can participate in as an equal member right because for WebKit

1681
01:57:43,200 --> 01:57:47,200
they just run the C++ code from Google they can't they it does not make sense

1682
01:57:47,200 --> 01:57:50,840
for Google a multi-trillion dollar or Apple a multi-trillion dollar company to

1683
01:57:50,840 --> 01:57:55,000
re-implement that in yeah so you can't you know like in a sense it's like wow

1684
01:57:55,000 --> 01:58:00,720
that that's a it's a very insightful example but just think about it for a

1685
01:58:00,720 --> 01:58:04,160
minute for a minute it's like the basically the most valuable company on

1686
01:58:04,160 --> 01:58:08,360
earth yeah they have like infinite money for all intents and purposes and it

1687
01:58:08,360 --> 01:58:12,720
doesn't make economic sense for them to re-implement it no yeah yeah and well in

1688
01:58:12,720 --> 01:58:18,040
a sense too it's like this this this ethos of having independent browser

1689
01:58:18,080 --> 01:58:21,800
vendors is kind of dying a little bit like they used to have Internet Explorer

1690
01:58:21,800 --> 01:58:25,560
for all of its faults at least it was an implement independent implementation

1691
01:58:25,560 --> 01:58:31,800
edge is a is chromium yeah right like it's not even even Microsoft Microsoft

1692
01:58:31,800 --> 01:58:35,360
threw in the hat on having a browser implementation right so that's another

1693
01:58:35,360 --> 01:58:39,800
way the more I think about not learn about noster because again I'm kind of

1694
01:58:39,800 --> 01:58:43,160
a noob I mostly ignored it for the past couple days but one of the things that

1695
01:58:43,160 --> 01:58:49,760
strikes me is how how the legacy well it's not it's not just like the walled

1696
01:58:49,760 --> 01:58:54,960
gardens of the like fang companies that are yeah kind of maybe the technology in

1697
01:58:54,960 --> 01:58:58,960
and of itself the the platform in itself is dying yeah right like it is dying

1698
01:58:58,960 --> 01:59:02,520
yeah I mean these are examples of it dying yeah like it is it is it is

1699
01:59:02,520 --> 01:59:07,280
atrophy and it's dying and it's and for multiple reasons I think in some sense

1700
01:59:07,280 --> 01:59:12,200
just just like I also sometimes watch like various you know streamers and

1701
01:59:12,200 --> 01:59:16,240
things and those kind of things this is not this is kind of one of the like we

1702
01:59:16,240 --> 01:59:19,840
talked a couple months ago and you were talking about how like noster is the

1703
01:59:19,840 --> 01:59:24,200
like kind of the new Internet yeah right and I had no fucking idea what you were

1704
01:59:24,200 --> 01:59:28,240
talking about you thought I was crazy no no I mean I knew you're not crazy

1705
01:59:28,240 --> 01:59:31,960
because I've known you for a long time but I did spend like a month or two kind

1706
01:59:31,960 --> 01:59:35,640
of trying to think like what the fuck is he talking about because I know you're

1707
01:59:35,640 --> 01:59:39,560
not crazy but I did not understand what you're talking about yeah and this is

1708
01:59:39,640 --> 01:59:42,880
these are some of the things where I've been kind of like I think this this is

1709
01:59:42,880 --> 01:59:47,000
kind of gives me the similar feeling yeah yeah it's I think it's our

1710
01:59:47,000 --> 01:59:50,800
opportunity to go back to the drawing board and kind of kind of deeply

1711
01:59:50,800 --> 01:59:54,320
think about why are we doing the things that we're doing doing currently on the

1712
01:59:54,320 --> 01:59:59,120
web yeah and and would there be a better way yeah what if we would go back to

1713
01:59:59,120 --> 02:00:02,440
ground zero and build it from scratch basically that's how I look at noster

1714
02:00:02,440 --> 02:00:07,080
because in the end it's like lots of the web that we have now it is kind of it's

1715
02:00:07,080 --> 02:00:11,640
very simple things done in a very very very very complicated way and I talked

1716
02:00:11,640 --> 02:00:16,200
with Paul about this that's why I'm so excited about hyponode and the idea of

1717
02:00:16,200 --> 02:00:20,920
it because it's like to render it to the things we need to do to the stuff you

1718
02:00:20,920 --> 02:00:25,520
mentioned rendering discord rendering Twitter itself the things we need to do

1719
02:00:25,520 --> 02:00:30,400
to render Twitter to render the Twitter feed in the browser it's unbelievable

1720
02:00:30,400 --> 02:00:33,000
unbelievable like if you guys are if anyone listening is telling you should

1721
02:00:33,000 --> 02:00:36,520
watch these Andreas kling talks or is streams where he does this it's like you

1722
02:00:36,520 --> 02:00:40,840
understand like this guy is fucking brilliant and he spent years and these

1723
02:00:40,840 --> 02:00:45,360
websites still barely render yeah it's like and and I'm going to compare it to

1724
02:00:45,360 --> 02:00:50,400
noster where you can write like a client in any language and render it on like a

1725
02:00:50,400 --> 02:00:54,120
terminal and you basically have the same effect yeah I just want to see short

1726
02:00:54,120 --> 02:00:57,560
text notes from my friends yeah you know yeah it's like it's and and and that's

1727
02:00:57,560 --> 02:01:01,720
that's that's true across the stack yeah I want to share an image I want to share

1728
02:01:01,920 --> 02:01:06,120
this is this is a cool thing about what will is doing is that in a sense he's

1729
02:01:06,120 --> 02:01:10,880
kind of working in the same area where this is like in a sense I think it's

1730
02:01:10,880 --> 02:01:15,120
still like shouldn't be a fucking dumb well it's very half I think it's still

1731
02:01:15,120 --> 02:01:19,880
very half bake still which is not a like a genius in my book yeah but it's like

1732
02:01:19,880 --> 02:01:24,440
it is like a browser ish like kind of runtime yeah that he's making which is

1733
02:01:24,440 --> 02:01:27,520
but but that's the thing about this is why I think that the inside from

1734
02:01:27,520 --> 02:01:31,480
Peter was so great because it's like you have to find the balance of going

1735
02:01:31,480 --> 02:01:35,520
back to the drawing board and redoing things and like not reinventing the wheel

1736
02:01:35,520 --> 02:01:40,360
yeah yeah like you don't actually want to do everything from yeah let me get

1737
02:01:40,360 --> 02:01:43,240
back to my rough multiply from idea so like I talked about some different like

1738
02:01:43,240 --> 02:01:46,280
one thing is you just make a mobile make a different app on every platform so you

1739
02:01:46,280 --> 02:01:50,800
have three code bases like primal right that's one way of doing it and then

1740
02:01:50,800 --> 02:01:53,800
there's the React Native where you get one thing but then it often becomes

1741
02:01:53,800 --> 02:01:56,880
kind of sluggish and it gets hard to use because it's this JavaScript language

1742
02:01:56,880 --> 02:02:00,880
is not it's a shitty language kind of but you know they added TypeScript on

1743
02:02:00,880 --> 02:02:04,200
top that which makes it significantly better at the cost of another

1744
02:02:04,200 --> 02:02:08,680
abstraction layers now like JavaScript is a compilation target like in the

1745
02:02:08,680 --> 02:02:12,960
sense that LLMs are turning React into TypeScript React into accomplished

1746
02:02:12,960 --> 02:02:16,200
targets here you're getting more and more and more and more and more layers of

1747
02:02:16,200 --> 02:02:19,760
complexity which is not good and this one of the things that's cool about what

1748
02:02:19,760 --> 02:02:24,240
Will is doing is because like he is like he has like two layers of complexity

1749
02:02:24,240 --> 02:02:28,600
yeah he has like pixels and he draws very close he draws them to the screen

1750
02:02:28,600 --> 02:02:32,240
60 times a second right so in a sense that's a really cool thing about what

1751
02:02:32,240 --> 02:02:36,040
he's doing is that like the the layers of complexity are he's like cut through

1752
02:02:36,040 --> 02:02:39,720
many layers of complexity which is I think promising on a long-term

1753
02:02:39,720 --> 02:02:43,640
time and so that is this is another approach you can do this is what like

1754
02:02:43,640 --> 02:02:47,080
video this is probably how a YouTube app works right like when you're watching a

1755
02:02:47,080 --> 02:02:51,840
video that is rendered they're probably there's C code in the app that is shared

1756
02:02:51,840 --> 02:02:57,000
on the iOS and mobile and you you know the app itself just gets like a buffer

1757
02:02:57,000 --> 02:03:00,720
60 times a second from C and just paints it to the screen yeah you know and so

1758
02:03:00,720 --> 02:03:04,080
that's probably their cross-platform strategy there's another really

1759
02:03:04,080 --> 02:03:07,760
interesting one that Facebook did for Messenger where they they had two they

1760
02:03:07,760 --> 02:03:11,760
had basically primal has two totally separate apps they had like 1.5 million

1761
02:03:11,760 --> 02:03:16,120
lines of code in each one just massive applications and it just became

1762
02:03:16,120 --> 02:03:19,440
difficult to maintain and they these were using the old versions of mobile

1763
02:03:19,440 --> 02:03:22,800
development like Objective-C and Java which have been supplanted by a new

1764
02:03:22,800 --> 02:03:26,520
way of doing it was just remembered in brackets of yeah yeah yeah it's horrible

1765
02:03:26,520 --> 02:03:30,080
and so what they did is they re rewrote their applications but they tried to put

1766
02:03:30,080 --> 02:03:35,240
as much as possible into a shared C++ library where as much of the state of

1767
02:03:35,240 --> 02:03:38,840
the applications possible is written in a sequel light and so the front end basically

1768
02:03:38,840 --> 02:03:42,640
just looks for what page are we on in sequel light and just renders that so

1769
02:03:42,640 --> 02:03:47,080
it's a little bit like what Will's doing where you know it just it just renders

1770
02:03:47,080 --> 02:03:52,200
it from sequel light he wanted to have the most

1771
02:03:52,200 --> 02:03:55,240
performed way possible so what he's doing what what Will's doing is

1772
02:03:55,240 --> 02:03:59,240
actually very similar to how Facebook Messenger works one of the like top

1773
02:03:59,240 --> 02:04:04,600
five most used apps in the world instead of NosterDB it uses sequel light I mean

1774
02:04:04,600 --> 02:04:07,320
sequel light is better in some ways because it's like the most it's probably

1775
02:04:07,320 --> 02:04:10,520
the most deployed piece of software in the world it's in your phone it's in your

1776
02:04:10,520 --> 02:04:15,440
car it's in your fridge it's everywhere but the cool thing about NosterDB is that

1777
02:04:15,520 --> 02:04:20,200
it's kind of like an in-memory database so like as you write to it it's still in

1778
02:04:20,200 --> 02:04:23,320
memory and so it's able to render straight from it doesn't have to come

1779
02:04:23,320 --> 02:04:27,000
from disk it's just straight in memory all the time so that it's just it's as

1780
02:04:27,000 --> 02:04:31,720
fast as you could possibly get that's another approach and so the way I would

1781
02:04:31,720 --> 02:04:37,200
like to do it is which I don't haven't seen many people doing is I would like to

1782
02:04:37,200 --> 02:04:41,080
write as much of my applications possible in one language all the logic and all

1783
02:04:41,080 --> 02:04:47,040
the state like what page am I on what what is this is the button disabled is it

1784
02:04:47,040 --> 02:04:51,160
not disabled are we in a loading state is there an error message you know what's

1785
02:04:51,160 --> 02:04:55,400
the user's name stuff like that like what is the whole state of what should be

1786
02:04:55,400 --> 02:04:59,600
rendered I want that in one language that is really good at cross-platform

1787
02:04:59,600 --> 02:05:05,520
stuff and you know ideally has a lot of libraries it's really fast and so I

1788
02:05:05,520 --> 02:05:09,200
would like that to be Zig just because I have done a lot of rust and I got kind

1789
02:05:09,200 --> 02:05:12,760
of sick of it but Zig doesn't have libraries and all ends don't understand it

1790
02:05:12,760 --> 02:05:18,680
so I ended up doing it in Rust because it's just so much more mature and there's

1791
02:05:18,680 --> 02:05:21,760
so many great libraries especially in the Bitcoin and Masterworld like any

1792
02:05:21,760 --> 02:05:25,400
library if there is a library there's one in Rust yeah and this is actually one

1793
02:05:25,400 --> 02:05:28,120
of the things that turned me off on WebRTC is I was gonna try to do this

1794
02:05:28,120 --> 02:05:31,160
calling thing and I was like well let's do WebRTC there's no rust implementation

1795
02:05:31,160 --> 02:05:34,280
like that to me it's like that if there's not a rust implementation of

1796
02:05:34,280 --> 02:05:37,600
something a grassroots rust implementation of something by now you

1797
02:05:37,600 --> 02:05:40,560
shouldn't use it because it's fucking too complicated it's too fucking

1798
02:05:40,560 --> 02:05:44,240
complicated and so that kind of turned me off on WebRTC that was the first thing

1799
02:05:44,240 --> 02:05:49,080
that made me think WebRTC might not be a good idea but so that's what I want to

1800
02:05:49,080 --> 02:05:54,000
do it all in Rust and then I want to just actually render to the screen with

1801
02:05:54,000 --> 02:05:58,600
whatever the standard thing is on that platform like so whatever Apple says in

1802
02:05:58,600 --> 02:06:02,320
their presentations you should use I want to use that because that's native yes

1803
02:06:02,320 --> 02:06:06,040
not in the sense of speed necessarily because that one way native can be is

1804
02:06:06,040 --> 02:06:09,280
like is it fast like that's one sense in which people use native the other

1805
02:06:09,280 --> 02:06:13,620
sense is like UI you does it feel look like everything else right so I want to

1806
02:06:13,620 --> 02:06:19,080
use Swift UI to to render it I want to use Jekyll pack and pose on Android

1807
02:06:19,080 --> 02:06:22,000
because that's what they how do you do the translation you need to have a

1808
02:06:22,000 --> 02:06:25,920
wrapper and kind of let me feel like on on Web like if I did it on Web and this

1809
02:06:25,920 --> 02:06:29,000
is a little iffy but you know use react because that's just what everyone does

1810
02:06:29,000 --> 02:06:35,840
right and and so yes you need a way to make Rust be able to run Rust code in

1811
02:06:35,880 --> 02:06:39,520
that context so the cool thing is you know Rust originally started as a project

1812
02:06:39,520 --> 02:06:42,720
to attempt to you know browsers have a lot of security vulnerabilities and

1813
02:06:42,720 --> 02:06:46,880
because they're written in C and C++ which are not sort of security they're

1814
02:06:46,880 --> 02:06:50,000
not they weren't built for security from the beginning and so they they just

1815
02:06:50,000 --> 02:06:56,960
tend to have a lot of memory but safety bugs that Rust was invented to fix and

1816
02:06:56,960 --> 02:07:02,200
so it was invented by Mozilla and so they built really good tooling to run

1817
02:07:02,200 --> 02:07:06,340
Rust inside their iOS app their Android app so it's this project called uni

1818
02:07:06,340 --> 02:07:10,520
FFI that allows you to do that it helps you cross compile your code to run on

1819
02:07:10,520 --> 02:07:15,100
that platform so you know you write Rust code and then you build it and then it

1820
02:07:15,100 --> 02:07:20,040
can run on iOS and it also generates types so you can you know the code that

1821
02:07:20,040 --> 02:07:26,480
you can run is easily callable from Swift or from Colin you know so you can

1822
02:07:26,480 --> 02:07:30,280
you know so you can you know pass some object representing the user and you can

1823
02:07:30,280 --> 02:07:33,600
pass around your application where you want to paint the screen so that's

1824
02:07:33,600 --> 02:07:38,600
what so I figured out kind of how to do this and and it was you know it's like

1825
02:07:38,600 --> 02:07:41,960
a couple hundred lines of code to make a simple example but I wanted to

1826
02:07:41,960 --> 02:07:45,840
simplify as much as possible so I got it down like this is like the first two

1827
02:07:45,840 --> 02:07:49,280
weeks I got it down so it's like in Rust you have to define three things you

1828
02:07:49,280 --> 02:07:54,040
have to define a message that the Rust could pass to the front end like

1829
02:07:54,040 --> 02:07:57,000
basically the state change so we have a user now we didn't have a user it logged

1830
02:07:57,000 --> 02:07:59,920
in with their input now we have a user right yeah another thing is like something

1831
02:07:59,920 --> 02:08:03,480
that the UI can send to Rust like hey the user clicked the button right do

1832
02:08:03,480 --> 02:08:09,080
something you know the user yeah it's usually stuff that based on user

1833
02:08:09,080 --> 02:08:14,600
activity and then you have to have something in Rust that will handle these

1834
02:08:14,600 --> 02:08:18,880
these you know incoming events from the UI and send events in the other end and

1835
02:08:18,880 --> 02:08:22,600
so that's one struct or enum it can be either one I think it's only structs now

1836
02:08:22,600 --> 02:08:26,840
but it could be enums and and so that is basically a struct where you have one

1837
02:08:26,920 --> 02:08:30,840
thing that you handle messages and you react to them right and so it's it's

1838
02:08:30,840 --> 02:08:33,880
really simple it's like three things you have to understand and hello world can

1839
02:08:33,880 --> 02:08:37,960
be like I think we got to 60 lines I got to 60 lines of code and I was able to

1840
02:08:37,960 --> 02:08:42,680
generate a lot of a code that is needed to sort of do these bridging things yeah

1841
02:08:42,680 --> 02:08:47,160
okay like you know the translation from what is in the database to the UI is

1842
02:08:47,160 --> 02:08:52,080
kind of basically viable yeah it's no it's it's I my library generates it kind

1843
02:08:52,080 --> 02:08:56,320
of okay more or less like you have this struct that's in memory and I had another

1844
02:08:56,320 --> 02:09:00,000
version that was in SQLite like mess this messenger thing but what's contained

1845
02:09:00,000 --> 02:09:03,400
in this struct is like because some of the UI elements between the platforms

1846
02:09:03,400 --> 02:09:09,920
differ right so it's this it's just like what page are we on what what are we

1847
02:09:09,920 --> 02:09:13,960
renders who does the translation from a page of how how does the top navigation

1848
02:09:13,960 --> 02:09:17,480
bar look because on iOS you have the the little so you think of it this way like

1849
02:09:17,480 --> 02:09:21,840
you're you're you're building up a stack of views right yeah yeah yeah yeah yeah

1850
02:09:21,840 --> 02:09:26,080
yeah so you you see you'll have a stack of views with the data that is

1851
02:09:26,080 --> 02:09:30,360
associated with them and then you're rendering in Swift UI or Jetpack

1852
02:09:30,360 --> 02:09:34,440
Compose gets that you know think of it as a struct or a JSON object and it just

1853
02:09:34,440 --> 02:09:37,600
it you know it'll look at the top one and that's what's on the top of this that

1854
02:09:37,600 --> 02:09:40,680
it'll you know something the view will also have a title and so because I

1855
02:09:40,680 --> 02:09:45,560
just place the title for you yeah yeah the same is what vital is going on it's

1856
02:09:45,560 --> 02:09:49,800
like the coding it's actually presenting the page is probably exactly what amethyst

1857
02:09:49,840 --> 02:09:54,840
does it's just whatever that you do in in Android like whatever the standard

1858
02:09:54,840 --> 02:09:58,680
thing is and you're just doing it based on data that is it's in like a view

1859
02:09:58,680 --> 02:10:01,480
model it's in it's in something I think it's called like composable or

1860
02:10:01,480 --> 02:10:06,240
something in in Jetpack and post so it's a it's a it's a basically a type of

1861
02:10:06,240 --> 02:10:11,920
struct that can notice when attributes change right so like you can have stuff

1862
02:10:11,920 --> 02:10:16,280
where like if you have your state and rust and it changes you have to somehow

1863
02:10:16,280 --> 02:10:19,600
tell the front end that something change it need to re-render that's if

1864
02:10:19,600 --> 02:10:22,560
you're a reactivity it's basically like when did my data change when should I

1865
02:10:22,560 --> 02:10:25,760
paint again yeah will doesn't have this problem because he when does he paint

1866
02:10:25,760 --> 02:10:30,040
again 60 times a second he paints no matter what like he doesn't have to he

1867
02:10:30,040 --> 02:10:32,680
doesn't have to notice if something changed because as soon as he's done

1868
02:10:32,680 --> 02:10:36,200
painting he starts painting again yeah all he does is paint you know you paint

1869
02:10:36,200 --> 02:10:39,160
paint paint that's all will doesn't you know that's what his app does that's

1870
02:10:39,160 --> 02:10:43,600
why it's so fast so so so that's but that my library generates all this

1871
02:10:43,600 --> 02:10:47,920
free so you don't even see any of that shit you just see this this this you

1872
02:10:47,920 --> 02:10:53,320
know some data structure that your editor understands so you can jump to

1873
02:10:53,320 --> 02:10:57,800
definition see what the types mean you can autocomplete stuff and so you know

1874
02:10:57,800 --> 02:11:02,400
you say what page are we on okay then render that page in the this you know

1875
02:11:02,400 --> 02:11:05,520
SwiftUI or Jetpack and post router okay we're on that page what is the user's

1876
02:11:05,520 --> 02:11:09,800
name put that in there you know yeah it's it's so it's purely declarative and

1877
02:11:09,800 --> 02:11:15,400
all you do is given this piece of data you draw with how they do it and so the

1878
02:11:15,400 --> 02:11:18,960
cool thing about this like one reason I don't want to do the way that like

1879
02:11:18,960 --> 02:11:22,760
say primal does it is that I don't want to write three ui's yeah like if I want

1880
02:11:22,760 --> 02:11:26,000
to write an app I would like it to one multiple places and I am not gonna

1881
02:11:26,000 --> 02:11:29,680
learn I'm not gonna learn Swift I am not gonna call in it I'm not gonna do

1882
02:11:29,680 --> 02:11:34,200
that but I am willing to learn a little bit about the like the DSL for how to

1883
02:11:34,200 --> 02:11:37,280
lay out a page because that's much simpler than the language itself right

1884
02:11:37,280 --> 02:11:39,600
you don't have to understand currency you don't have some memory you don't

1885
02:11:39,600 --> 02:11:43,280
have some garbage collection you don't have to fucking anything and in general

1886
02:11:43,280 --> 02:11:45,680
if you set up the tools well enough this is what I've been working on since then

1887
02:11:45,680 --> 02:11:49,520
I was talking about these MCPs I think you should get to a point where you can

1888
02:11:49,520 --> 02:11:53,640
basically vibe code the entire UI yeah so what you're actually engineering

1889
02:11:53,640 --> 02:12:00,360
mostly is that the logic like what should this do what should when I get an

1890
02:12:00,360 --> 02:12:05,520
event what should my events right like how how well what are the events that

1891
02:12:05,520 --> 02:12:10,280
that the user can trigger and what what are the effects that I can present to

1892
02:12:10,280 --> 02:12:13,480
them after that's happening like a loading state or like the post is

1893
02:12:13,480 --> 02:12:17,480
published and now you're looking at the post right yeah so you think about that

1894
02:12:17,480 --> 02:12:21,640
and how to handle it and you know you write your logic and rust and then you

1895
02:12:21,640 --> 02:12:28,480
get just a very simple thing that you can tell an LLM to you know render this

1896
02:12:28,480 --> 02:12:33,240
drawing I made yeah you know and and so I did this I did one demo with this

1897
02:12:33,240 --> 02:12:40,160
which was like a unique e-cash wallet that I was able to I was able to I made

1898
02:12:40,160 --> 02:12:46,520
an Excalidraw for the flow of the app it had like seven pages and and I took

1899
02:12:46,520 --> 02:12:50,760
this Excalidraw and I gave it to an LLM and said implement this and it

1900
02:12:50,760 --> 02:12:54,360
and it implemented that it had a few it did like two or three things wrong but

1901
02:12:54,360 --> 02:12:58,560
I was able to cut them out like I had a history thing you're supposed to swipe

1902
02:12:58,560 --> 02:13:03,000
up to get it it didn't get that and it put two screens in the wrong order I

1903
02:13:03,040 --> 02:13:07,080
just reordered them manually and then it was basically what I drew insane it

1904
02:13:07,080 --> 02:13:11,360
was insane yeah and so like so I think this it's a good fit for LLM tools because

1905
02:13:11,360 --> 02:13:14,400
it's like you still need software you're not yeah I don't really believe in the

1906
02:13:14,400 --> 02:13:17,200
world where you have really high quality apps are purely produced by a computer

1907
02:13:17,200 --> 02:13:21,880
like I don't think that's realistic at least now but I think parts of it can be

1908
02:13:21,880 --> 02:13:25,200
produced by computer and the one thing that they're really really good at is

1909
02:13:25,200 --> 02:13:31,280
making you eyes yeah and so that's the like so you can get like so a lot of the

1910
02:13:31,320 --> 02:13:34,160
vibe coding like aha moments or when you try to make like a little like one

1911
02:13:34,160 --> 02:13:38,520
like 500 line react app it can one shot those but if you have like a 10,000

1912
02:13:38,520 --> 02:13:42,160
line react app you try to get a feature it's gonna get lost yeah but like so

1913
02:13:42,160 --> 02:13:47,160
can you make a big complicated app where at least in the UI side it's very simple

1914
02:13:47,160 --> 02:13:51,240
and all the DAB is there and there's an LSP for it to tell it if it's the

1915
02:13:51,240 --> 02:13:54,920
lynching's wrong and it can screenshot the UI to see if it matches up with the

1916
02:13:54,920 --> 02:13:59,320
flows you gave it so that's something where it's like me as a solo developer I

1917
02:13:59,320 --> 02:14:02,920
feel like I could maintain five nostrils yeah that are very high

1918
02:14:02,920 --> 02:14:06,520
performance is not as high performance as one specific thing yeah it's not

1919
02:14:06,520 --> 02:14:09,840
asked for high performance is Will's thing but it's also like you know he's

1920
02:14:09,840 --> 02:14:14,920
having to implement gift renders yeah I don't even understand how my UI is

1921
02:14:14,920 --> 02:14:18,880
rendered like I don't even yeah you don't need to care I don't care you don't

1922
02:14:18,880 --> 02:14:24,680
need to care yeah right like it's it's it's like a button for me yeah it's funny

1923
02:14:24,680 --> 02:14:28,280
to see how this evolves and should we walk one more around and then like grab a

1924
02:14:28,320 --> 02:14:34,240
coffee yeah it's like I think I can walk again for a little bit but the so

1925
02:14:34,240 --> 02:14:40,840
Pablo also I spoke about it with him last week so he is transforming NDK now

1926
02:14:40,840 --> 02:14:46,680
and especially the NDK documentation so he's producing snippets basically stand

1927
02:14:46,680 --> 02:14:52,040
alone code examples of how to do one thing yeah and make it like very easily

1928
02:14:52,040 --> 02:14:55,520
digestible by an LML yeah yeah and just and just that's the documentation so

1929
02:14:55,520 --> 02:14:58,360
this is another thing I want to do is there's this thing called seven gooies

1930
02:14:58,360 --> 02:15:03,400
yeah yeah yeah I know it from Paul yes seven gooies is like how do you say

1931
02:15:03,400 --> 02:15:07,880
it's like it's like a benchmark for UI frameworks right so it's like the

1932
02:15:07,880 --> 02:15:11,320
first thing is a counter app the second app the second one is like click the

1933
02:15:11,320 --> 02:15:15,480
calendar and it increments yeah plus one down right and then it's a timer right

1934
02:15:15,480 --> 02:15:18,400
which is has to have like some asynchronous thing where you're counting

1935
02:15:18,400 --> 02:15:20,760
and then there's like a temperature converter and the last one is a

1936
02:15:20,760 --> 02:15:24,640
spreadsheet yeah as this like as I was implementing this I was going back and

1937
02:15:24,640 --> 02:15:29,400
forth with LMS being like write me the rust side of it yeah and then like okay

1938
02:15:29,400 --> 02:15:33,200
and then I say okay this would be the data model right like now write me a UI

1939
02:15:33,200 --> 02:15:36,760
based on the static data so I didn't actually write the app so I was thinking

1940
02:15:36,760 --> 02:15:39,920
about like could this how would it look like in this paradigm to do the whole

1941
02:15:39,920 --> 02:15:42,560
seven gooies and eventually I will implement all of them and that will be

1942
02:15:42,560 --> 02:15:47,320
my docs right so it's like here is the full GUI benchmark yeah and so now you

1943
02:15:47,320 --> 02:15:51,200
should be able to understand like like you'd give that as context to models and

1944
02:15:51,200 --> 02:15:55,840
be like here are examples of apps yeah like and you could give it like you

1945
02:15:55,840 --> 02:15:59,000
know if you had a bunch of little you know I make a little e-cash while make a

1946
02:15:59,000 --> 02:16:02,760
little a couple things like that and those could be the things that that the

1947
02:16:02,760 --> 02:16:09,720
LLMs used to kind of as context to generate their code I think we'll have

1948
02:16:09,720 --> 02:16:13,680
to turn around here yeah it's it's really funny to me to see it evolve and how

1949
02:16:13,680 --> 02:16:17,920
everyone uses that and now we're purpose-building tools to make it

1950
02:16:17,920 --> 02:16:21,680
easier for us to work with LLMs going forward yeah that's what that's why

1951
02:16:21,680 --> 02:16:24,960
you are building Rust multiplatform yeah I mean it's funny too because I'm

1952
02:16:24,960 --> 02:16:27,760
never the type of person that would work on a framework like that just it's not

1953
02:16:27,760 --> 02:16:30,480
appealing at all yeah like I'm much more of a pragmatist for it's like I just

1954
02:16:30,480 --> 02:16:34,040
want I have a thing I want you are being pragmatic yeah yeah well but this is

1955
02:16:34,040 --> 02:16:36,520
something where it's like I can't tell if it's just because I'm more mature now

1956
02:16:36,520 --> 02:16:39,520
as an about and developer because like especially in FEDI and FEDI meant like

1957
02:16:39,520 --> 02:16:43,240
that was such a massive project I learned so much from like you know like

1958
02:16:43,240 --> 02:16:46,640
there's a time where I had the entire FEDI meant thing in my head and I had the

1959
02:16:46,680 --> 02:16:50,400
entire FEDI thing in my head the whole thing in my head and it was watching it

1960
02:16:50,400 --> 02:16:53,760
for two years and learn so much as though like it made me like kind of a

1961
02:16:53,760 --> 02:16:57,520
little more interested in oftentimes we you know we'd have a problem and we

1962
02:16:57,520 --> 02:17:00,720
build tooling and the problem would go away yeah so it's like I think it was

1963
02:17:00,720 --> 02:17:03,400
part of it is like this is a natural progression the developer goes through

1964
02:17:03,400 --> 02:17:06,680
whether you see enough and you're like damn certain tools really make your

1965
02:17:06,680 --> 02:17:10,520
problems go away so in a sense I'm just learning from that experience but it's

1966
02:17:10,520 --> 02:17:15,080
also like the LLM where it's like I would have never spent time you know I had

1967
02:17:15,080 --> 02:17:18,800
like seven structs and stuff that for the first version of this that I did

1968
02:17:18,800 --> 02:17:22,640
manually and then I got it down to three with a macro with macros that would

1969
02:17:22,640 --> 02:17:28,040
generate code for me and I have I'm generating all kinds of code just so I

1970
02:17:28,040 --> 02:17:32,600
can get it maximally simple yeah and see if it could work and that's something I

1971
02:17:32,600 --> 02:17:35,720
would have never done manually I've never written a rust macro because I

1972
02:17:35,720 --> 02:17:40,840
don't have the fish and autism like I would always have a guy for that right

1973
02:17:40,840 --> 02:17:45,160
like we should run in boat if people yeah like after after listening to all

1974
02:17:45,160 --> 02:17:50,120
that let's see but I know like I've been doing rust and I was pretty good at it

1975
02:17:50,120 --> 02:17:53,000
but I've had like almost like an unconscious policy I will not write a

1976
02:17:53,000 --> 02:17:57,040
fucking macro yeah like I'm not gonna do it because it's it's very difficult this

1977
02:17:57,040 --> 02:18:01,120
is actually one reason I write I think Zig is cool so rust has this like the

1978
02:18:01,120 --> 02:18:05,360
heart of macros like meta programming so it's like programming that generates

1979
02:18:05,360 --> 02:18:09,160
your coat your program and it's the hardest type of programming like in C

1980
02:18:09,640 --> 02:18:12,560
for example you have to it's two languages see the actual runtime language

1981
02:18:12,560 --> 02:18:16,440
it's pretty simple like you can learn it in like a couple days because there's

1982
02:18:16,440 --> 02:18:19,760
not that many structs but of course there's a lifetime of little nuances but

1983
02:18:19,760 --> 02:18:24,400
the meta programming like the if-deaf like yeah if I'm declining on Mac OS like

1984
02:18:24,400 --> 02:18:27,840
inject this code if not that's like a whole nother language and it's dog shit

1985
02:18:27,840 --> 02:18:31,120
yeah compared to see like see very elegant kind of amazing language that it's

1986
02:18:31,120 --> 02:18:35,560
still the go-to thing for embedded programming for example like 50 years

1987
02:18:35,560 --> 02:18:39,720
layer yeah it's unbelievably how how good and how success how well they

1988
02:18:39,720 --> 02:18:44,040
sign yeah yeah but the macro language sucks and I would say rusts is not that

1989
02:18:44,040 --> 02:18:48,040
much better it's better but it still sucks like it's very uh you look at a

1990
02:18:48,040 --> 02:18:52,560
rust rust code it's not so bad you look at a macro it's like Jesus I'd like

1991
02:18:52,560 --> 02:18:56,080
just don't dollar signs everywhere it's really confusing the cool thing about

1992
02:18:56,080 --> 02:18:59,800
Zig is it has something called comp time where you can just label various

1993
02:18:59,800 --> 02:19:05,240
variables or functions and that code is executed while your code compiles so

1994
02:19:05,240 --> 02:19:09,160
your meta programming is not a second language like it is and see or a second

1995
02:19:09,160 --> 02:19:12,920
language like it is and injects that compile time it's it's the same syntax

1996
02:19:12,920 --> 02:19:16,880
but it's just evaluated when your code compiles and so you can have like a

1997
02:19:16,880 --> 02:19:21,320
variable for example where you you evaluate it at comp time and even the

1998
02:19:21,320 --> 02:19:24,200
statically typed language where you know you have a function it should only take

1999
02:19:24,200 --> 02:19:28,480
a 32-bit integer or 16-minute integer it should be different you can say

2000
02:19:28,480 --> 02:19:32,480
comp time and you can say hey a comp time will make separate versions of this

2001
02:19:32,480 --> 02:19:36,520
function for any type of integer so now you have like Python style duck typing

2002
02:19:36,520 --> 02:19:40,240
where yeah any sort of int looking thing can work because at comp time you

2003
02:19:40,240 --> 02:19:43,840
basically generate versions for every one of these so it's like that's one of

2004
02:19:43,840 --> 02:19:46,760
the really cool things about Zig is is the hardest part of programming that a

2005
02:19:46,760 --> 02:19:54,080
programming is now just as easy as the normal part of programming but yeah I

2006
02:19:54,080 --> 02:19:58,080
don't know why I went off on that tangent I mean it's a I knew I knew it

2007
02:19:58,080 --> 02:20:04,480
would be a sick conversation so yeah I knew it would be I think I think I

2008
02:20:04,480 --> 02:20:08,240
think it would be cool and nostril if we could have more unique style and this

2009
02:20:08,240 --> 02:20:11,920
is one thing that's made me a lot more interested in I think that's actually

2010
02:20:11,920 --> 02:20:16,160
where it's trending I really I really do I already use different nostril apps and

2011
02:20:16,160 --> 02:20:19,680
tools it's funny in a sense it's trending in the opposite direction on the visible

2012
02:20:19,680 --> 02:20:24,080
level because all the apps are becoming more and more monoling like primal and

2013
02:20:24,080 --> 02:20:27,360
primal and amethyst is becoming more and more I think we have to differentiate

2014
02:20:27,360 --> 02:20:30,480
I think it's becoming bimodal right like you're having you have you have

2015
02:20:30,480 --> 02:20:34,720
monolithic apps and then you're gonna have the opposite and there's gonna be

2016
02:20:34,720 --> 02:20:38,000
even it's gonna be no mid-curve I think kind of yeah like you either have to

2017
02:20:38,000 --> 02:20:41,320
have the kitchen sink or you're gonna be one off and the thing is like the

2018
02:20:41,320 --> 02:20:46,440
Unix apps on on iOS if you're trying to do this on iOS like it you're kind of

2019
02:20:46,440 --> 02:20:51,200
neutered because you can't have a you can't have one relay you actually can so

2020
02:20:51,200 --> 02:20:55,280
like in general iOS can't apps can't talk to them each other like you can't

2021
02:20:55,280 --> 02:20:59,040
have an on device relay is what you mean or yeah you can't you can't have a

2022
02:20:59,040 --> 02:21:04,800
sh well okay so let me back up like the unit the whole point of Unix tools is

2023
02:21:04,800 --> 02:21:08,640
that you can pipe them yes right so the input is text output is text right so

2024
02:21:08,640 --> 02:21:13,200
you can the cat program reads a file and puts it out and then you can pipe that

2025
02:21:13,200 --> 02:21:16,320
you just do the vertical straight up and down character we're just saying this

2026
02:21:16,320 --> 02:21:19,920
for the listeners and then you can pass in the next thing like grep grep searches

2027
02:21:19,920 --> 02:21:23,200
so you can get search search it for something you're interested in and then

2028
02:21:23,200 --> 02:21:28,040
you can use like another said to replace so you can read a file search what's

2029
02:21:28,040 --> 02:21:30,920
relevant and then find and replace the part that you want to do like rename a

2030
02:21:30,920 --> 02:21:36,040
functioning and so that's the brilliance of like a bunch of little single purpose

2031
02:21:36,040 --> 02:21:39,480
tools kind of suck if they can't be composed but if they can be composed

2032
02:21:39,480 --> 02:21:43,200
it's a superpower and that paradigm still exists 50 or 60 years later yeah from

2033
02:21:43,200 --> 02:21:47,760
why Unix was invented and that's what this is one of the two things that Andreas

2034
02:21:47,760 --> 02:21:51,960
Kling was trying to you know put into his operating system the other one was

2035
02:21:51,960 --> 02:21:57,520
the windows things right so so if you translate this to mobile the thing you

2036
02:21:57,520 --> 02:22:01,080
need is for apps to talk to them each other yeah you need to input an output

2037
02:22:01,080 --> 02:22:04,200
between apps you need to be like a function in a sense right at least at

2038
02:22:04,200 --> 02:22:06,760
least some something like that like I don't think you'll ever get quite to the

2039
02:22:06,760 --> 02:22:09,240
Unix tools level but you they need to talk to each other I mean that's the whole

2040
02:22:09,240 --> 02:22:12,400
idea of dbm said the whole idea of nostril events in the first place yeah I

2041
02:22:12,400 --> 02:22:16,760
mean that that can help to some degree but like you can't like so one thing

2042
02:22:16,760 --> 02:22:18,960
you just start from scratch like okay I would like to have one app that does

2043
02:22:19,000 --> 02:22:22,280
sign it like I want to prepare an event and then I want to have another app

2044
02:22:22,280 --> 02:22:25,800
sign it yeah you that's very easy on Android you basically we have it on

2045
02:22:25,800 --> 02:22:29,920
you can't really do it on iOS with one exception that's almost hilarious the

2046
02:22:29,920 --> 02:22:33,440
part of me wants to do it just to piss everybody off you can do it if the

2047
02:22:33,440 --> 02:22:38,180
apps are from the same app developer so what I could do is I could make an

2048
02:22:38,180 --> 02:22:44,560
Apple developer account and I could have a Unix tool set of iOS apps that can talk

2049
02:22:44,560 --> 02:22:50,680
to each other only because I only because they're my wall of garden

2050
02:22:50,680 --> 02:22:54,440
like every every app that's signed by the same developer I think so that's my

2051
02:22:54,440 --> 02:22:57,520
understanding right so if you saw it Yeti project I forget the guy's name is

2052
02:22:57,520 --> 02:23:01,440
doing it but it's like a signer for iOS that's the constraint is that it all

2053
02:23:01,440 --> 02:23:05,320
has to be you know it is a signer that can be used by multiple out what if we

2054
02:23:05,320 --> 02:23:09,800
what if we what if we use something smart like frost to share our ultimate

2055
02:23:09,800 --> 02:23:13,400
signing keys but have it but potentially so that this would be an irony

2056
02:23:13,400 --> 02:23:17,720
like just a shared noster iOS app sign you could you could you could we could

2057
02:23:17,720 --> 02:23:21,120
do something like that where we have like the entire noster equals no one

2058
02:23:21,120 --> 02:23:25,840
actually owns the key it's just generated from my care for like a thousand keys

2059
02:23:25,840 --> 02:23:29,680
yeah like you we could have a noster like we would have to set up some kind of

2060
02:23:29,680 --> 02:23:34,000
legal identity for iOS except it yeah but like you could we could have had like

2061
02:23:34,000 --> 02:23:38,000
a pool of a thousand keys a federation we could have a federation and you'd get

2062
02:23:38,000 --> 02:23:42,560
like the Android user experience I've been thinking about doing that but but

2063
02:23:42,560 --> 02:23:47,160
it does you know in in lieu of that you could have a pretty cool they get

2064
02:23:47,160 --> 02:23:50,000
it makes me much more interested in Android because like these things just

2065
02:23:50,000 --> 02:23:55,760
work yeah you know so the unit absolutely works a unique tools vision is much

2066
02:23:55,760 --> 02:23:59,560
more achievable on and that's why I'm saying like I'm already using it like

2067
02:23:59,560 --> 02:24:02,400
that and I know my mom is never gonna use an ostrich like that you know like

2068
02:24:02,400 --> 02:24:05,720
she's gonna default to a kitchen sink client that's on and that being said

2069
02:24:05,720 --> 02:24:10,160
I'm still a huge fan of both prime or eventually block will make a phone sure

2070
02:24:10,160 --> 02:24:16,280
yeah that does this but I'm like I like I run amra I run so I have an

2071
02:24:16,280 --> 02:24:20,720
on-device relay I have an on-device signer you know I have I've dedicated apps

2072
02:24:20,720 --> 02:24:24,520
for various things for various content types and those kind of things but here's

2073
02:24:24,520 --> 02:24:27,120
the interesting thing with this Rust multi-platform like if I got it

2074
02:24:27,120 --> 02:24:30,080
sufficiently good where it's like like the thing I was working today is like

2075
02:24:30,080 --> 02:24:32,800
Reaper this week was reached produce will build right yeah like I packaged

2076
02:24:32,800 --> 02:24:36,120
hex code I packaged it so like I could build it's not quite there but I could

2077
02:24:36,120 --> 02:24:40,880
build my app on both platforms reproducibly with a fresh laptop I just

2078
02:24:40,880 --> 02:24:44,960
bought yeah you know and I can I can do that same thing the same build for my

2079
02:24:44,960 --> 02:24:48,560
dev the same build for when I produce like an actual APK or something that

2080
02:24:48,560 --> 02:24:52,880
needs to be like secure same exact software set up bit by bit and what

2081
02:24:52,880 --> 02:24:55,960
friends and CI which runs my tests which gates like whether the code is

2082
02:24:55,960 --> 02:25:00,320
accepted or not yeah so like so that's not like if I have all these tools I'm

2083
02:25:00,320 --> 02:25:04,080
pretty sure I could build this entire ecosystem myself like literally one guy

2084
02:25:04,320 --> 02:25:08,640
like I would not need any help and I could build like ten apps not the

2085
02:25:08,640 --> 02:25:11,280
entire ecosystem but like enough where there could be a phone there could be a

2086
02:25:11,280 --> 02:25:15,280
camera there could be like I feel like I could kind of have the Unix tools of

2087
02:25:15,280 --> 02:25:20,920
like a phone and like one guy I think I could actually do all this myself if I

2088
02:25:20,920 --> 02:25:26,700
had the tools good enough ahead of time yeah and you know like the signer

2089
02:25:26,700 --> 02:25:28,920
wouldn't be that hard because there's like these Rust libraries are really

2090
02:25:28,920 --> 02:25:31,880
good right the relay wouldn't be that good because I'm leading on all these

2091
02:25:31,920 --> 02:25:36,960
great libraries yeah but I think I think the way smarter way would be to figure

2092
02:25:36,960 --> 02:25:40,360
out how to get around this limitation and share a signing key in some way yeah

2093
02:25:40,360 --> 02:25:45,000
yeah in some cryptographic way that that no single person can pull the whole

2094
02:25:45,000 --> 02:25:51,680
effort because it's it's I don't I don't believe in doing things oh yeah

2095
02:25:51,680 --> 02:25:55,360
I'm not saying this that I would like I don't I wouldn't I wouldn't seriously do

2096
02:25:55,360 --> 02:25:58,120
this I always worry about Pablo because he tries to do all the things

2097
02:25:58,200 --> 02:26:02,320
himself basically yeah and it's also the other thing too is like like this is

2098
02:26:02,320 --> 02:26:04,480
another thing that's not necessarily that I would want to do that because I

2099
02:26:04,480 --> 02:26:07,400
actually don't want to do that very much don't want to do that don't want to be

2100
02:26:07,400 --> 02:26:11,200
in charge of all that but it's it's like a benchmark right like could you do it

2101
02:26:11,200 --> 02:26:14,840
all yourself yeah that's right yeah if you could because like I'm a smart

2102
02:26:14,840 --> 02:26:17,920
developer but I also have a limited things like holding my head at once right

2103
02:26:17,920 --> 02:26:21,840
if it's simple enough I can do different things yeah it's like well I know a lot

2104
02:26:21,840 --> 02:26:25,240
of guys who could just pick up a few of these apps and do them themselves yeah

2105
02:26:25,240 --> 02:26:28,360
there's probably plenty others who could do it solve themselves but it would

2106
02:26:28,360 --> 02:26:31,360
make it in a community context more maintainable by a group right like if

2107
02:26:31,360 --> 02:26:35,160
if one person could do it all if you got it simple enough where one person could

2108
02:26:35,160 --> 02:26:39,800
do 10 yeah yeah then it's very simple and then you then you don't need so that's

2109
02:26:39,800 --> 02:26:42,880
thing then you don't need 10 then you don't need one person anymore yeah because

2110
02:26:42,880 --> 02:26:45,360
then you can that's fair then you can share it like then you can share the load

2111
02:26:45,360 --> 02:26:49,080
a lot easier you can open it up but if it's like in order to get started like

2112
02:26:49,080 --> 02:26:51,920
this is the problem I confirmed when we first started Fedis like how the fuck

2113
02:26:51,920 --> 02:26:55,160
do you put all this fancy ass rust code into a mobile app right like that was

2114
02:26:55,160 --> 02:26:57,360
very difficult for a month just figuring out how it is because there's no

2115
02:26:57,360 --> 02:27:00,640
prior there isn't good prior art for this yeah like there was like two or three

2116
02:27:00,640 --> 02:27:04,600
GitHub examples I saw that were shitty and I had to just figure it out so it's

2117
02:27:04,600 --> 02:27:09,240
like if all of that was just kind of taken care of for you where like and the

2118
02:27:09,240 --> 02:27:13,240
AI could you just have to write a little logic and you have to kind of no rust

2119
02:27:13,240 --> 02:27:16,840
and you have to and the AI will write your UI for you and you know you're at the

2120
02:27:16,840 --> 02:27:19,840
end of the day you're using like the rust nostril SDK you're using lipsecp

2121
02:27:19,840 --> 02:27:23,840
rust wrapper you're using all these the heavy lifting it's easy to pay you're

2122
02:27:23,840 --> 02:27:29,280
using Fediment client using all these great rust libraries like you yeah I

2123
02:27:29,280 --> 02:27:34,040
think more more stuff becomes possible there yeah and I mean the other option

2124
02:27:34,040 --> 02:27:36,880
is like you know you just kind of the version of this now is everyone just

2125
02:27:36,880 --> 02:27:40,440
writes web apps but yeah I think the I don't know there's something about like

2126
02:27:40,440 --> 02:27:43,960
quote-unquote native app it just feels so much better in your hand it's also it's

2127
02:27:43,960 --> 02:27:47,200
like what most of the nostril developers miss or many of the nostril developers

2128
02:27:47,200 --> 02:27:51,000
miss like those that are building the web apps it's like 98% of people are on

2129
02:27:51,000 --> 02:27:54,940
phones yeah it's just like your web app is completely uninteresting if there

2130
02:27:54,940 --> 02:27:57,640
isn't an app that people can install those people are not gonna use it but I

2131
02:27:57,640 --> 02:28:01,120
mean you can install it as a PWA and like I think mutiny showed that you can get

2132
02:28:01,120 --> 02:28:05,560
a pretty good PWA yeah you can do this but it's not it's not it's not people

2133
02:28:05,560 --> 02:28:08,800
don't know what it is and how do you do it's kind of jamming a square into a

2134
02:28:08,800 --> 02:28:13,280
round hole a little bit right but it's like I'm I'm kind of you know I'm of two

2135
02:28:13,280 --> 02:28:16,200
minds with all that stuff because it's also I think that the app stores are

2136
02:28:16,200 --> 02:28:20,040
gonna be a massive yeah issue going forward this is the other thing that

2137
02:28:20,040 --> 02:28:24,760
and that makes Android is fantastic just because of Sep Store alone yeah no that

2138
02:28:24,760 --> 02:28:26,720
that's the other thing that makes Android so much more appealing is like

2139
02:28:26,720 --> 02:28:29,880
let's say I did this on iOS like literally the only way I would do this

2140
02:28:29,880 --> 02:28:33,280
actually so the idea that one person could do it alone is false cuz I could

2141
02:28:33,280 --> 02:28:37,040
write all the if you could sidelone on app stores yeah on iOS maybe I could do

2142
02:28:37,040 --> 02:28:40,600
this yeah but without that like I would literally kill myself if I had to like

2143
02:28:40,600 --> 02:28:46,360
literally if I had to manage 10 different app like that is the code

2144
02:28:46,360 --> 02:28:50,520
signing along I mean it's funny the we'll know that the code sign I could do all

2145
02:28:50,520 --> 02:28:55,640
in next but I can automate all that the fucking clicking around the the generating

2146
02:28:55,640 --> 02:29:00,680
the screen Joe oh and and then like the retarded things the reviewers send back

2147
02:29:00,680 --> 02:29:04,920
like it's more of that like doing with that person that's different every time

2148
02:29:04,920 --> 02:29:07,480
and is an idiot and doesn't understand what you're doing yeah and you are

2149
02:29:07,480 --> 02:29:10,880
probably you know going against what they want you to do to some extent yeah

2150
02:29:10,880 --> 02:29:15,760
sure like that was so I would have to hire somebody to do that yeah like there's

2151
02:29:15,760 --> 02:29:19,480
no way I would do that myself I would I would just crash and burn if I tried to

2152
02:29:19,480 --> 02:29:22,960
do that yeah but that's that's the thing that's so appealing with zap store is

2153
02:29:22,960 --> 02:29:27,240
like you know you could get it where it's like if CI is green 10 minutes later

2154
02:29:27,240 --> 02:29:31,120
there's a new yes absolutely that's how it works right now yeah and and that's

2155
02:29:31,120 --> 02:29:35,320
that's really appealing yeah and yourself sign it so like distribution

2156
02:29:35,320 --> 02:29:39,880
distribution is like in if I kept doing this like in principle I think I

2157
02:29:39,880 --> 02:29:45,160
could get the distribution story for zaps for Android to be as good as it is

2158
02:29:45,160 --> 02:29:48,040
on the web because that's the thing that's actually the thing right now that the

2159
02:29:48,040 --> 02:29:52,840
web is really good at is the distribution is so easy yeah like you can just set

2160
02:29:52,840 --> 02:29:57,080
up get hooks to resell and your website shows up yeah when it's done building

2161
02:29:57,080 --> 02:30:02,880
yeah like no like you push it to github yeah as long as it's on yeah you know

2162
02:30:02,880 --> 02:30:07,280
like but that sort enables this for AP AP case but and that's one thing that

2163
02:30:07,280 --> 02:30:10,120
I you know like I don't think any of us understand how important that is like

2164
02:30:10,120 --> 02:30:14,440
we don't oh I maybe most people don't most people don't understand yeah I

2165
02:30:14,440 --> 02:30:17,600
don't think I appreciate yeah I don't think I appreciate yeah like every week

2166
02:30:17,600 --> 02:30:22,800
I'm like more realizing super important how important that is yeah so I don't

2167
02:30:22,800 --> 02:30:25,880
think I fully understand it because I did I'm more bullish on one than I was

2168
02:30:25,880 --> 02:30:29,640
last week yeah and last week I was more bullish on that was the previous week

2169
02:30:29,640 --> 02:30:33,960
so I just look back forward like I don't think I understand it quite yet and also

2170
02:30:33,960 --> 02:30:36,600
with enough tooling this stuff will be easier and easier and easier and it's

2171
02:30:36,600 --> 02:30:39,920
still so underutilized because Step Store is mostly an Android thing now but

2172
02:30:39,920 --> 02:30:44,680
also if I if I'm downloading Sparrow Wallet or like Bitcoin core client or

2173
02:30:44,680 --> 02:30:48,200
whatever it's like I want to make sure that it's signed by the right people and

2174
02:30:48,200 --> 02:30:50,760
they are part of my web of trust yeah I would just want to see it immediately I

2175
02:30:50,760 --> 02:30:54,280
want to I want to have these checks happen automatically yeah and and we

2176
02:30:54,280 --> 02:30:58,360
don't do all that yet we have like primitive CLI tools that implement that

2177
02:30:58,360 --> 02:31:02,360
a little bit that's on like like friends that built up those and so on but but

2178
02:31:02,360 --> 02:31:09,080
we don't really like and no one stops us from like improving up up on home

2179
02:31:09,080 --> 02:31:12,680
room for example by connects yeah like we can't do that also all kinds of

2180
02:31:12,680 --> 02:31:16,520
package managers and so on like we can improve just by adding nostril to to this

2181
02:31:16,520 --> 02:31:20,920
layer and also also saps and micro payments yeah it's like it's it's they

2182
02:31:20,920 --> 02:31:24,280
even have it in NPM right where it's like these these projects are looking for

2183
02:31:24,280 --> 02:31:28,520
funding and those kind of things like like if you if if that's how I use

2184
02:31:28,520 --> 02:31:32,280
not right now if I find something amazing I'm gonna set it one if it's like an

2185
02:31:32,280 --> 02:31:35,640
amazing tool I'm I'm happy to set the package straight up thing I'm really

2186
02:31:35,640 --> 02:31:38,760
interested in trying to do is integrate Zapp's row with Nix so like the whole

2187
02:31:38,760 --> 02:31:45,880
idea of Nix is that you you declare what inputs are to a build process so a

2188
02:31:45,880 --> 02:31:48,560
build process is like you have code and you want to turn it into something you

2189
02:31:48,560 --> 02:31:52,720
can run right this is for the you know listeners right so you have code like

2190
02:31:52,800 --> 02:31:56,400
text in a file and you want to turn it into a thing you can run on Linux or

2191
02:31:56,400 --> 02:32:00,320
an APK or something and you know somebody has to build that right and so like for

2192
02:32:00,320 --> 02:32:04,960
example you know Nix is very similar to geeks geeks it's the same idea but a

2193
02:32:04,960 --> 02:32:09,280
different implementation geeks is like more hardcore Fos type people that don't

2194
02:32:09,280 --> 02:32:13,200
want to be stuck switch to geeks like before you use geeks right so the thing

2195
02:32:13,200 --> 02:32:18,560
is it allows like a very very reproducible build where with almost zero

2196
02:32:19,560 --> 02:32:23,360
dependencies right like and for context right like you know the old way they

2197
02:32:23,360 --> 02:32:28,040
did it was Gideon and so the environment you use to build your new version of

2198
02:32:28,040 --> 02:32:33,000
Bitcoin core was Ubuntu Ubuntu contains like 400 megabytes of trusted

2199
02:32:33,000 --> 02:32:36,560
binaries yeah so like you have 400 megabytes of stuff that could be an

2200
02:32:36,560 --> 02:32:40,040
attack surface that could be an inflation bug yeah and you don't know and like

2201
02:32:40,040 --> 02:32:46,520
it's reproducible meaning like you put code in you code plus Ubuntu equals

2202
02:32:46,520 --> 02:32:50,200
binary you get the same one every time but since the inflation bug could be

2203
02:32:50,200 --> 02:32:53,160
in the you want to and not the code yeah you might have an inflation bug every

2204
02:32:53,160 --> 02:32:58,320
time yeah so get geeks was an attempt yeah the issue is that you can you can

2205
02:32:58,320 --> 02:33:02,520
build a hostile compiler that injects stuff yeah do your things and you will

2206
02:33:02,520 --> 02:33:06,800
it will be almost impossible to detect it so you have to bootstrap the whole

2207
02:33:06,800 --> 02:33:11,240
thing yeah out of nothing yeah knowing every step along the way what it does

2208
02:33:11,240 --> 02:33:14,640
without adding any external dependency the craziest thing the geeks goes called

2209
02:33:14,640 --> 02:33:21,020
stage zero so this is where they they start with 200 bytes I think 200

2210
02:33:21,020 --> 02:33:24,320
bytes of trusted binary so it's literally like assembly code that's

2211
02:33:24,320 --> 02:33:28,440
commented saying what each instruction does yeah and that is bootstrapped into

2212
02:33:28,440 --> 02:33:31,440
something that can like interpret a little bit more and so there's like

2213
02:33:31,440 --> 02:33:34,760
how paranoid those people are there's three layers of reading text and then

2214
02:33:34,760 --> 02:33:39,360
they write like three layers of scheme compilers like the simplest list dialect

2215
02:33:39,360 --> 02:33:43,040
and then they build like TCC it's called it's like a really primitive basic

2216
02:33:43,040 --> 02:33:46,400
compiler and then they build another one and then they build GCC which can

2217
02:33:46,400 --> 02:33:52,040
build everything see like nine stages to get to see a real C compiler and so

2218
02:33:52,040 --> 02:33:58,560
nix is the same thing that's not quite as you can't bootstrap it quite as well

2219
02:33:58,560 --> 02:34:03,160
but it's it's very good in the sense that you declare what do I need to build this

2220
02:34:03,160 --> 02:34:07,840
software and then it can just build it without any globally installed stuff

2221
02:34:07,840 --> 02:34:12,000
which is where you get all the you know problems both for a developer experience

2222
02:34:12,000 --> 02:34:16,120
but also for like building a trusted binary so what you could do is you could

2223
02:34:16,120 --> 02:34:20,840
make it so like every time I run a project that contains some of this stuff

2224
02:34:20,840 --> 02:34:25,400
I could have like one little command to like build and publish event of what

2225
02:34:25,400 --> 02:34:29,360
hash the binary was you know stuff like that right so you could you could have

2226
02:34:29,360 --> 02:34:33,400
like an extension to nix where every developer who uses it in their flow

2227
02:34:33,400 --> 02:34:37,240
whenever they publish a binary or whenever they run a binary for anything

2228
02:34:37,320 --> 02:34:42,280
it can just publish it and something and as an ostrich events and so you could

2229
02:34:42,280 --> 02:34:49,360
get sort of like a real web of trust on binaries for free yeah you know and like

2230
02:34:49,360 --> 02:34:53,400
not necessarily for free like someone like Zabstor would have to be you know

2231
02:34:53,400 --> 02:34:56,680
aggregating all this and trying to notice if two people disagreed but you

2232
02:34:56,680 --> 02:35:01,160
could just you could see automatically like well did Gigi and Justin ever build

2233
02:35:01,160 --> 02:35:05,520
amethyst and get different hashes for a given version yeah right so like that

2234
02:35:05,520 --> 02:35:08,640
would be really cool where you could get almost automatically just a normal

2235
02:35:08,640 --> 02:35:12,480
developer workflow or CI pipe yeah now we have dedicated you know like I can

2236
02:35:12,480 --> 02:35:15,640
be able to spot with the let's say I don't know I am first but let's say they

2237
02:35:15,640 --> 02:35:21,160
publish it via their CI yeah right like I could take their code and I could run

2238
02:35:21,160 --> 02:35:26,600
into my own CI server and and you could see whether like my setup in there setup

2239
02:35:26,600 --> 02:35:30,240
disagreed or not so like I think there could be stuff like that where you could

2240
02:35:30,480 --> 02:35:36,120
I think this Zabstor idea like binaries plus kind of attestation I think that

2241
02:35:36,120 --> 02:35:39,120
the same the meta idea is like you want to you want to have the trust

2242
02:35:39,120 --> 02:35:42,720
distributed for multiple sources as much as possible yeah you don't want to

2243
02:35:42,720 --> 02:35:46,360
have a single point of failure a single source of trust yeah and what you are

2244
02:35:46,360 --> 02:35:49,080
saying is basically you'll have that automatically just because every single

2245
02:35:49,080 --> 02:35:52,920
developer so I could have a I could have a check I have I could have a next

2246
02:35:52,920 --> 02:35:56,880
server that just watches 50 github repos yeah if they if they you care about

2247
02:35:56,880 --> 02:36:01,320
that's the thing yeah if they use if they use Nix or if I think my social

2248
02:36:01,320 --> 02:36:03,440
network here's there could be something out that I don't ever want to

2249
02:36:03,440 --> 02:36:08,920
yeah but you still want to check it you know Jesus is his his Jesus in the

2250
02:36:08,920 --> 02:36:14,360
sandwiches Tamagotchi idea yeah maybe I never use that yeah but maybe I think

2251
02:36:14,360 --> 02:36:18,280
my followers might be interested so I could set up a basically a CI server it's

2252
02:36:18,280 --> 02:36:22,560
not it's kind of like CI but like a server that watches 49 githubs that I

2253
02:36:22,560 --> 02:36:26,280
care about plus that one whenever there's a new thing whenever a new release

2254
02:36:26,280 --> 02:36:30,640
branch whenever a new release is published I download the code I it just

2255
02:36:30,640 --> 02:36:36,520
does a Nix build it's all automated with no real coding effort on my part and it

2256
02:36:36,520 --> 02:36:41,160
just publishes what about what hash I got yeah and so you could get you know like

2257
02:36:41,160 --> 02:36:45,440
well how Nix currently works is there's like you know there's a big package

2258
02:36:45,440 --> 02:36:49,560
registry that's on github and I think they basically have one server like this

2259
02:36:49,560 --> 02:36:53,120
Hydra server that builds stuff and publishes it and says those are what the

2260
02:36:53,120 --> 02:36:56,240
real hashes are right yeah I don't understand it very well but I think

2261
02:36:56,240 --> 02:36:59,600
that's basically how it works so it's like you do in the Nix ecosystem you are

2262
02:36:59,600 --> 02:37:02,280
kind of trusting I'm sure there are other people checking each other but it's

2263
02:37:02,280 --> 02:37:07,720
not it's it's very it's very like centralized in the sense that there's not

2264
02:37:07,720 --> 02:37:10,520
more than a hundred of them yeah I mean that's that's that's the name of the

2265
02:37:10,520 --> 02:37:13,760
game we want to we want to we don't want to remove trust entirely we just want to

2266
02:37:13,760 --> 02:37:17,600
remove trusted third parties as much as possible yeah that's why you know like

2267
02:37:17,600 --> 02:37:21,880
I'm I'm still trusting my web of trust yeah that they're not malicious and not

2268
02:37:21,880 --> 02:37:25,040
pushing malicious binaries and all those kind of things but it's just it's

2269
02:37:25,040 --> 02:37:29,080
always better to have I believe to have multiple sources and kind of like crowd

2270
02:37:29,080 --> 02:37:33,960
source this stuff and just kind of yeah like have a balance that keeps everyone

2271
02:37:33,960 --> 02:37:36,640
in check right now also with the with the app stores for example there's one

2272
02:37:36,640 --> 02:37:40,440
instance if they want to remove you from the app store you're done yeah that's

2273
02:37:40,440 --> 02:37:45,040
it yeah if you don't have any checks and balances or anything like that and should

2274
02:37:45,040 --> 02:37:48,760
we do one more like up and down and then grab a coffee I'm really I'll need some

2275
02:37:48,760 --> 02:37:52,960
I'll need some drugs very soon yeah the heat stroke yeah hits and I'm still

2276
02:37:52,960 --> 02:37:56,000
half sick and limping

2277
02:37:56,240 --> 02:38:00,480
that's a nice day it's not raining it hasn't rained yet oh yeah that's that was

2278
02:38:00,480 --> 02:38:03,640
something I never factored into the you know oh let's just go on and walk and

2279
02:38:03,640 --> 02:38:07,200
have a conversation kind of thing it's like what if it starts raining what if

2280
02:38:07,200 --> 02:38:10,000
there's like you know like I really should have brought like a raincoat I

2281
02:38:10,000 --> 02:38:14,200
did zero research and I get here and we're going on these walks and stuff I'm

2282
02:38:14,200 --> 02:38:20,680
like oh shit yeah it's it's unusually it's unusually rainy it's off Morocco

2283
02:38:20,680 --> 02:38:25,080
this is gonna be like a beach town I brought the flip-flops you know it's gonna be

2284
02:38:25,080 --> 02:38:29,240
like even look at the weather like 80 degrees every day but it was winter

2285
02:38:29,240 --> 02:38:34,120
right now basically so it's it's it's really like unusually cold and rainy and

2286
02:38:34,120 --> 02:38:38,120
oh man it was so windy last time with with Pablo like a bus stop flew away and

2287
02:38:38,120 --> 02:38:44,480
like it was you it was mayhem it was like hurricane left away yeah so very

2288
02:38:44,480 --> 02:38:52,400
very unusual weather but yeah no I I to me it's super exciting to to watch

2289
02:38:52,400 --> 02:38:58,960
kind of the culmination of so many things and strains of thought and it's

2290
02:38:58,960 --> 02:39:04,040
all kind of culminating now in the not their e-cash world in some sense it's

2291
02:39:04,040 --> 02:39:07,400
like because a lot of the I talked with Kelly about this briefly a lot of the

2292
02:39:07,440 --> 02:39:11,360
cares computer club hacker scene Berlin type people they really love e-cash they

2293
02:39:11,360 --> 02:39:15,120
hate Bitcoin they love e-cash they know it's the right idea lots of the AI

2294
02:39:15,120 --> 02:39:19,480
people they know they need micro payments yeah some of the serial knowledge

2295
02:39:19,480 --> 02:39:23,360
compute people and so on they figure out now noster is the perfect substrate to

2296
02:39:23,360 --> 02:39:27,760
do stuff not a new shit coin not a new whatever other thing and for communication

2297
02:39:27,760 --> 02:39:31,040
and like like simple X and all those kind of things as well it's like you you

2298
02:39:31,040 --> 02:39:34,600
need a substrate that does the the the social lookup that initiates the

2299
02:39:34,600 --> 02:39:37,960
handshake for example you know it's just like you said with your phone and

2300
02:39:37,960 --> 02:39:42,080
NPAB idea you need to whole punch to something yeah and it might as well be an

2301
02:39:42,080 --> 02:39:45,120
NPAB you know you don't want to build up a new layer of well the whole the whole

2302
02:39:45,120 --> 02:39:48,520
point of building a phone like the phone a building a phone calling out it's like

2303
02:39:48,520 --> 02:39:51,680
the worst idea possible normally it's like how the fuck am I gonna get another

2304
02:39:51,680 --> 02:39:55,680
phone number but you know that's not I can't you know phone numbers are

2305
02:39:55,680 --> 02:39:58,800
massive shit but like whatever the analog is what you're doing it's like

2306
02:39:58,800 --> 02:40:02,520
you're you're looking at such a monumental effort yeah to develop a

2307
02:40:02,520 --> 02:40:06,400
contact book of people you could interact with yeah but with noster you

2308
02:40:06,400 --> 02:40:09,640
get it for free yeah you know so it's it's kind of funny that I got interested

2309
02:40:09,640 --> 02:40:14,440
in this because it's like it's like the worst idea normally right yeah yeah

2310
02:40:14,440 --> 02:40:19,240
yeah yeah it provides a lot of things for free and I think also like like this

2311
02:40:19,240 --> 02:40:22,080
calling thing isn't like one of the apps that the Bitcoin world has produced it

2312
02:40:22,080 --> 02:40:26,280
as I think the one of the best user experiences is Keith yeah Keith has a

2313
02:40:26,280 --> 02:40:30,920
it is no one really knows how it works for for some for some things like you get

2314
02:40:30,920 --> 02:40:34,360
on a call there I've been on a few called video calls where it's like this

2315
02:40:34,360 --> 02:40:38,600
has better latency than zoom yeah right it's like the call is higher quality

2316
02:40:38,600 --> 02:40:42,440
than zoom yeah what's going on here right because it's direct that you both

2317
02:40:42,440 --> 02:40:45,800
have good connections exactly so but then there's there's some downsides

2318
02:40:45,800 --> 02:40:50,320
right like no one knows how it works it's not open source which is to me

2319
02:40:50,320 --> 02:40:55,320
extraordinarily suspect security focused app from a company with a

2320
02:40:55,640 --> 02:41:01,200
money printer yeah like a literal money printer yeah like very suspect yeah and

2321
02:41:01,200 --> 02:41:05,600
you know architect like I went and looked at the the binary to try to understand

2322
02:41:05,600 --> 02:41:08,840
how it works and it's like they do all this obfuscation to see to hide where

2323
02:41:08,840 --> 02:41:12,880
the code is right like the actual source code is in a platform that mp3 file

2324
02:41:12,880 --> 02:41:20,760
yeah mp3 it's an mp3 file so I had I had to I had to well the crazy thing is

2325
02:41:20,760 --> 02:41:23,800
one of the I was asking one of the alums about stuff it actually knew about the

2326
02:41:23,800 --> 02:41:29,400
platform that mp3 file from a Chinese forum there's a Chinese I wouldn't go

2327
02:41:29,400 --> 02:41:32,440
I'm for I have like an hour conversation I was trying to get the LLM to tell me

2328
02:41:32,440 --> 02:41:37,040
how it knew that because it's it was O1 Pro and like O1 Pro is hard to get it

2329
02:41:37,040 --> 02:41:41,800
to answer direct questions so you know because it has bad social skills yeah he

2330
02:41:41,800 --> 02:41:46,600
has bad social skills yeah as I mentioned but someone knew this so I was able to

2331
02:41:46,600 --> 02:41:51,040
get in there and it's like say the single hyperdrive which is a way of like

2332
02:41:51,040 --> 02:41:55,600
having like it's like a file it's like a file system mounted on the DHT and

2333
02:41:55,600 --> 02:41:59,480
this is where like the UI components are saved so the UI components are in

2334
02:41:59,480 --> 02:42:04,720
like a binary format in their DHT file system that's how it is that's

2335
02:42:04,720 --> 02:42:09,600
interesting which is like what's going on here you know and then and then the

2336
02:42:09,600 --> 02:42:12,320
identity is something I don't even know how you have to just do it all from

2337
02:42:12,320 --> 02:42:15,320
scratch that's where nozer comes in it's like well if you could just add even if

2338
02:42:15,320 --> 02:42:19,080
you kept all the and it's where we're to see if you just kept all the technical

2339
02:42:19,080 --> 02:42:22,840
stuff and just added nostril identities I think it would be at least for my

2340
02:42:22,840 --> 02:42:26,120
social group it would be better yeah because then I could call you yeah I

2341
02:42:26,120 --> 02:42:29,400
can't call you on keep because I don't I doubt we'd have to do some damn thing

2342
02:42:29,400 --> 02:42:34,920
yeah but if it was if you had nostril is I could call you directly yeah and but

2343
02:42:34,920 --> 02:42:37,800
yeah and that's in the end always what you want to do right like you want to

2344
02:42:37,800 --> 02:42:40,480
send money to someone you want to call someone you want to send a message to

2345
02:42:40,480 --> 02:42:45,840
someone you want to tag someone you want to you want to I often just want to go

2346
02:42:45,840 --> 02:42:49,280
like hey I think this article would be interesting for you I just want to

2347
02:42:49,280 --> 02:42:53,600
tag people in articles it's just like hey like that's I I suggest you read it I

2348
02:42:53,600 --> 02:42:56,800
don't necessarily want to DM you the link that's kind of you know what I mean

2349
02:42:56,800 --> 02:43:01,040
yeah yeah we we are evolving all these these different levels and layers now

2350
02:43:01,040 --> 02:43:04,360
it's like I definitely don't want to send you an email with like hey dear

2351
02:43:04,360 --> 02:43:06,880
Justin I mean that's another one that would be like a Unix thing that would

2352
02:43:06,880 --> 02:43:09,840
be interesting it's like they're just the notifications that you want to get in

2353
02:43:09,840 --> 02:43:12,760
their notifications tab yeah but like you don't necessarily I don't

2354
02:43:12,760 --> 02:43:18,040
necessarily want my yeah it's like my Twitter feed time so take it or leave

2355
02:43:18,040 --> 02:43:21,200
it yeah well I'm not mad if you don't but I don't I don't necessarily want to

2356
02:43:21,200 --> 02:43:25,440
see like long-form stuff like to me the one form should be like a separate world

2357
02:43:25,440 --> 02:43:29,080
I agree I'm in a totally different mind space when I'm using like an art

2358
02:43:29,080 --> 02:43:33,720
theatre I still hate kind of the fact that first of all Twitter removed the

2359
02:43:33,720 --> 02:43:37,200
character limit second of all no sir didn't reintroduce the character limit

2360
02:43:37,200 --> 02:43:39,960
for kind of ones yeah I think kind ones should be tweets and they should be

2361
02:43:39,960 --> 02:43:44,160
short yeah and a long-form is something completely different but that's a good

2362
02:43:44,160 --> 02:43:49,380
example in the in the kind of the monolithization of nostril right like

2363
02:43:49,380 --> 02:43:52,280
it's like the clients are becoming everything and everything is in these

2364
02:43:52,280 --> 02:43:55,880
clients and it's like it'd be but I'm already using DVM some clients that hide

2365
02:43:55,880 --> 02:44:02,880
long-form stuff it's like I use I use the you know I did buy this one I know

2366
02:44:02,880 --> 02:44:06,880
I'm left side of the of the welfare kind of you know I want to use the stupidest

2367
02:44:06,880 --> 02:44:12,160
tools imaginable yeah and and so so I vibe coded the DVM into existence with

2368
02:44:12,160 --> 02:44:16,840
like let's someone build a DVM that only shows tweets that are shorted in

2369
02:44:16,840 --> 02:44:20,440
210 characters basically yeah and I use it all the time because I don't want to

2370
02:44:20,440 --> 02:44:23,960
read walls of text when I scroll through my feet yeah it's like it's it's

2371
02:44:23,960 --> 02:44:28,080
limitation is what fastest creativity and limitation is what made Twitter

2372
02:44:28,080 --> 02:44:32,760
great I think yeah and but I think we will have this specialization it's like

2373
02:44:32,760 --> 02:44:37,320
you know a live stream and like a comment in a chat of a live stream is

2374
02:44:37,320 --> 02:44:40,680
different than a tweet yeah you know it's a different thing it's a different

2375
02:44:40,680 --> 02:44:45,880
interaction yeah it's like a highlight of a long-form article or margin notes in

2376
02:44:45,880 --> 02:44:50,520
a book are a different thing than a tweet yeah it's it's the kind of this

2377
02:44:50,520 --> 02:44:53,840
conversation slash war is currently going on because people obviously have

2378
02:44:53,840 --> 02:44:56,600
different opinions on it like what should be a kind one and what shouldn't

2379
02:44:56,600 --> 02:45:00,360
and so yeah and then and how should we render it because Pablo has a great

2380
02:45:00,360 --> 02:45:04,080
point that you should make it at least discoverable that this stuff exists on

2381
02:45:04,080 --> 02:45:08,280
us so it should pop up in your kind one feet from time to time in a sense like

2382
02:45:08,280 --> 02:45:13,080
you know like this is this is a highlight and but your client right now

2383
02:45:13,080 --> 02:45:16,440
cannot really run the highlights for example and all those kind of stuff but

2384
02:45:16,440 --> 02:45:22,320
yeah we'll figure it out all right coffee yeah any more sick thoughts no

2385
02:45:22,320 --> 02:45:30,120
I don't know it's a distant story you're back it's but a nightmare it's sick

2386
02:45:30,120 --> 02:45:35,040
developer who is building rust multi-platform yeah I'm gonna go build some

2387
02:45:35,040 --> 02:45:40,080
web apps now let's do it it's that's that's that's that's the new after all

2388
02:45:40,080 --> 02:45:45,160
this gig of rain thing I'm gonna go build up you're gonna just speak

2389
02:45:45,160 --> 02:45:54,160
into my yeah awesome no let's let's close it here okay I think that was

2390
02:45:54,160 --> 02:45:58,120
great I'm a little bit dizzy but it's just me being sick let me give it let me

2391
02:45:58,120 --> 02:46:02,800
give you a hug and I'll I forgot to do the snap in the beginning so I'll have to

2392
02:46:02,800 --> 02:46:07,840
do that now because otherwise I have no way to sync it up if we if we fucked up

2393
02:46:07,840 --> 02:46:11,800
and it might be the first recording that we have actually has an mp3

2394
02:46:11,800 --> 02:46:16,760
file from the get-go okay I think it's the first time after eight tries or so or

2395
02:46:16,760 --> 02:46:22,200
nine tries now that I got it right but we'll find out I always press to I

2396
02:46:22,200 --> 02:46:26,680
forget to press record on the phone or the cable breaks or it like it always

2397
02:46:26,680 --> 02:46:30,240
fucked it up where's the cable that okay that thing oh so it's a good receiver

2398
02:46:30,240 --> 02:46:34,640
yeah so it's still working and it's which they put sore sometimes it might be a

2399
02:46:34,640 --> 02:46:39,400
graphene thing I don't know and then it just recorded my pocket so it's a all

2400
02:46:39,400 --> 02:46:44,320
right that's a good one fantastic