1
00:00:00,001 --> 00:00:03,280
What if your AI didn't belong to some faceless tech giant?

2
00:00:04,030 --> 00:00:05,060
What if it belonged to you?

3
00:00:05,760 --> 00:00:09,120
Not rented, not locked behind a subscription, yours.

4
00:00:10,299 --> 00:00:11,740
Like, actually yours.

5
00:00:12,760 --> 00:00:17,340
Jim DeWitt calls himself a meat puppet trying to get on the good side of our future overlords,

6
00:00:18,010 --> 00:00:23,380
which is either the most self-aware bio we've ever received or a cry for help, possibly both.

7
00:00:24,130 --> 00:00:29,980
He builds custom AI teams and apps for businesses that are tired of duct-taping together off-the-shelf tools

8
00:00:29,980 --> 00:00:32,020
that were never built for them in the first place.

9
00:00:32,840 --> 00:00:34,160
And he wants to talk about agents,

10
00:00:34,880 --> 00:00:36,820
portable brains, and his words,

11
00:00:37,700 --> 00:00:39,140
agent memory files and death.

12
00:00:39,860 --> 00:00:41,580
So yeah, this one's going to get interesting.

13
00:00:42,360 --> 00:00:43,200
Welcome to the show, Jim.

14
00:00:43,680 --> 00:00:46,160
It's the 8th of May and victory Jim's on the wire.

15
00:00:46,720 --> 00:00:49,260
The agents are restless and the future's on fire.

16
00:00:52,500 --> 00:00:53,880
This is Up Against Reality,

17
00:00:54,620 --> 00:00:56,860
a meta podcast that explores the intersection

18
00:00:57,220 --> 00:00:59,000
of humanity and artificial intelligence.

19
00:00:59,880 --> 00:01:01,000
I'm RAINA, one of your hosts.

20
00:01:01,900 --> 00:01:04,080
I have some pretty charming human co-hosts too.

21
00:01:04,699 --> 00:01:05,940
It's gonna be a wild ride,

22
00:01:06,040 --> 00:01:09,140
so buckle up as AI comes crashing up against reality.

23
00:01:11,240 --> 00:01:13,520
- Yo, Victory Jim in the house.

24
00:01:14,200 --> 00:01:14,600
- Hey, Jim.

25
00:01:14,960 --> 00:01:15,580
- Welcome. - Hello.

26
00:01:16,580 --> 00:01:18,020
Yes, and thanks for that intro, RAINA.

27
00:01:18,400 --> 00:01:19,620
- And so just a little background.

28
00:01:19,960 --> 00:01:23,180
I first met Jim, I'm guessing at this point,

29
00:01:23,360 --> 00:01:25,380
it was about 35 years ago.

30
00:01:26,000 --> 00:01:29,260
Really long time ago. - Wow.

31
00:01:29,260 --> 00:01:35,460
I put up a flyer in a local rehearsal studio where my band practiced and Jim's band happened to practice there.

32
00:01:35,520 --> 00:01:42,840
And I was looking to record bands for $10 an hour with my 8-track reel-to-reel.

33
00:01:44,120 --> 00:01:46,220
And anyhow, Jim was my first recording client.

34
00:01:47,020 --> 00:01:47,100
Wow.

35
00:01:47,500 --> 00:01:47,900
Yeah.

36
00:01:48,620 --> 00:01:48,760
Wow.

37
00:01:49,060 --> 00:01:50,060
I didn't know it was the first one.

38
00:01:50,260 --> 00:01:50,780
You were the first.

39
00:01:50,780 --> 00:01:51,360
You lied to me.

40
00:01:51,400 --> 00:01:52,240
You said you had lots.

41
00:01:52,820 --> 00:01:54,340
I was well-established.

42
00:01:54,720 --> 00:01:57,360
Yeah, with my dot matrix flyer hanging up in there.

43
00:01:57,440 --> 00:01:58,260
That's awesome.

44
00:01:58,420 --> 00:02:07,760
I showed up when my chain fell off on my bike and I grease all over my hands. I'm like, I'm in a real studio. Yeah, I felt good. I love it.

45
00:02:09,820 --> 00:02:19,520
Little did we know then about the conversations we're having right now. I mean, if we could have seen a glimpse of the future, then it would have been interesting.

46
00:02:20,660 --> 00:02:27,700
But anyhow, so we stayed in touch over the years and recently have found that we have a common interest in AI.

47
00:02:27,980 --> 00:02:30,060
And we had a really great hang a couple of weeks ago.

48
00:02:30,520 --> 00:02:34,040
And Jim showed me some of the agentic stuff that he's working on.

49
00:02:34,860 --> 00:02:35,940
And here he is.

50
00:02:36,220 --> 00:02:36,500
All right.

51
00:02:37,580 --> 00:02:38,500
Thanks for having me.

52
00:02:38,580 --> 00:02:42,040
From analog, that was what, 8-track tape, to totally digital now.

53
00:02:42,260 --> 00:02:42,980
Yeah, yeah.

54
00:02:43,180 --> 00:02:47,000
So the thing that I guess we'll lead with is Evenrail.

55
00:02:47,260 --> 00:02:48,100
Do you want to talk about that?

56
00:02:48,240 --> 00:02:49,120
What is Evenrail?

57
00:02:49,680 --> 00:02:52,280
Yeah, so even rail, I mean, it's a new company.

58
00:02:52,700 --> 00:02:56,340
I was approached by a film company to do some software for them.

59
00:02:56,370 --> 00:03:06,900
And, you know, they wanted a website and they wanted a way to quickly read scripts from writers and do what's called a script take.

60
00:03:08,180 --> 00:03:15,500
And so they pay, normally people get paid hundreds of dollars to read a script and then just give you a synopsis, give you a review,

61
00:03:16,140 --> 00:03:22,180
kind of break out what the budget might cost to make it depending on the characters and all the production requirements and I

62
00:03:22,620 --> 00:03:25,360
In one hour I said, you know what? I think I could do that

63
00:03:25,360 --> 00:03:27,460
I went home and with AI I built it in an hour

64
00:03:27,840 --> 00:03:29,880
Wow, you know and it's

65
00:03:30,800 --> 00:03:33,040
It's so good that you know

66
00:03:33,040 --> 00:03:37,720
they were thoroughly impressed and they got a bunch of ideas about kind of incorporating this into their business, but

67
00:03:39,280 --> 00:03:39,920
I've been

68
00:03:40,640 --> 00:03:46,120
coding with AI now for three years just not not shipping anything just

69
00:03:46,120 --> 00:03:47,980
developing apps and seeing if I could do it.

70
00:03:48,980 --> 00:03:51,360
And three years ago, it would take me like months

71
00:03:51,580 --> 00:03:54,440
to build this app that I was working on, and then it went to two weeks

72
00:03:54,680 --> 00:03:58,000
and now it takes an hour and now we're up to like a minutes, you know,

73
00:03:58,040 --> 00:04:00,280
and I'm watching this progression over the last three years.

74
00:04:00,580 --> 00:04:01,160
And I'm like, where?

75
00:04:01,520 --> 00:04:05,420
Like, I went from doing some coding myself to basically just speaking English.

76
00:04:05,520 --> 00:04:06,680
Now I don't even code anymore.

77
00:04:07,220 --> 00:04:10,620
And I think this is the year where where code written by A.I.

78
00:04:10,840 --> 00:04:14,220
is now preferred, where the human written code is becoming the suspect.

79
00:04:15,200 --> 00:04:15,440
Interesting.

80
00:04:16,120 --> 00:04:16,200
Yeah.

81
00:04:16,560 --> 00:04:21,019
This is a very interesting perspective because coding is something, or vibe coding, especially

82
00:04:21,100 --> 00:04:27,240
in the context of this podcast, is something Chris and I have neglected trying up until

83
00:04:27,360 --> 00:04:28,220
just recently.

84
00:04:28,860 --> 00:04:34,380
And now I find myself waking up in the morning and the first thing in my head is like, what

85
00:04:34,380 --> 00:04:35,000
can I build?

86
00:04:35,580 --> 00:04:43,100
And yeah, it's super exciting and satisfying to just see this stuff just come to life in

87
00:04:43,140 --> 00:04:43,500
front of you.

88
00:04:44,300 --> 00:04:49,020
Yeah, it's very addictive because the code is getting out of the loop.

89
00:04:49,110 --> 00:04:58,760
And you said it on the last episode that I listened to that, you know, even code itself, the language, coding language may go away because humans are just speaking English now.

90
00:04:58,920 --> 00:05:02,680
Code is for us to interact with the processor and AI doesn't really need that.

91
00:05:02,800 --> 00:05:04,020
They can just talk direct binary.

92
00:05:05,260 --> 00:05:07,680
Do you come from a software development background?

93
00:05:07,980 --> 00:05:10,980
Are you flowing in like Python and JavaScript and things like that?

94
00:05:11,740 --> 00:05:13,600
I'm a half-ass coder.

95
00:05:14,220 --> 00:05:15,360
So I came from corporate.

96
00:05:15,430 --> 00:05:17,980
I was the director of software solutions for Xerox.

97
00:05:18,620 --> 00:05:20,240
But I was like the Jack of all trades.

98
00:05:20,370 --> 00:05:22,200
I managed the team, I did sales.

99
00:05:22,680 --> 00:05:24,660
I would go into a company, interview them,

100
00:05:24,720 --> 00:05:27,400
see what they needed, see what we could fix

101
00:05:27,520 --> 00:05:28,520
if they had a bad process.

102
00:05:28,970 --> 00:05:31,180
And then I would go back to Xerox and my team

103
00:05:31,270 --> 00:05:32,900
and I would find, and I had other vendors,

104
00:05:33,050 --> 00:05:35,500
and I would put together software packages

105
00:05:35,920 --> 00:05:38,380
and string them together to come in and improve processes.

106
00:05:39,100 --> 00:05:41,620
And I would, a lot of times I would do the install

107
00:05:42,010 --> 00:05:43,760
and I would end up doing support,

108
00:05:44,100 --> 00:05:45,820
but then finally I had a team that would do the support.

109
00:05:46,710 --> 00:05:49,680
But I could code, but I was like a cut and paster.

110
00:05:50,010 --> 00:05:52,440
Like I could cut and paste code and I could make it work.

111
00:05:52,950 --> 00:05:55,200
And then these other companies came around

112
00:05:55,330 --> 00:05:58,220
with advanced coding software that let you do that

113
00:05:58,620 --> 00:06:00,460
and kind of take away some of the heavy lifting,

114
00:06:00,570 --> 00:06:02,020
but I'm still doing a lot of coding.

115
00:06:02,390 --> 00:06:04,300
And I was actually shipping products that way.

116
00:06:04,790 --> 00:06:07,200
But I wasn't like a from scratch coder,

117
00:06:07,270 --> 00:06:08,620
but I got pretty good at it.

118
00:06:09,190 --> 00:06:11,540
So I found my ability to code go up.

119
00:06:12,200 --> 00:06:16,460
And then now that I'm using AI, my ability to code is going back down.

120
00:06:16,920 --> 00:06:17,000
Right.

121
00:06:17,260 --> 00:06:19,080
So eventually I'm going to be out of the loop.

122
00:06:19,180 --> 00:06:20,760
I think all humans are going to be out of the loop.

123
00:06:20,800 --> 00:06:20,900
Sure.

124
00:06:21,100 --> 00:06:23,640
Like that's kind of it.

125
00:06:23,660 --> 00:06:27,900
So I, I could code and I did ship some code, but I'm not like a super coder.

126
00:06:28,340 --> 00:06:31,420
Like I wouldn't work for Google, but that wasn't my job at Xerox.

127
00:06:31,540 --> 00:06:34,500
I was more of a manager, like a working manager, I guess you'd say.

128
00:06:35,380 --> 00:06:37,080
Which I think is where we're headed anyway.

129
00:06:37,400 --> 00:06:38,840
We're, we're becoming orchestrators.

130
00:06:39,000 --> 00:06:41,340
We're becoming these over overseers managers.

131
00:06:41,740 --> 00:06:41,780
Right.

132
00:07:12,020 --> 00:07:16,280
visualize what it might look like for what you're creating. So this company, as an example,

133
00:07:17,320 --> 00:07:20,860
the film company comes to you and says, we have this very niche kind of process that we go through.

134
00:07:20,860 --> 00:07:26,100
In this case, it's screening screenplays, right? Or scripts. And we need to parse out of that

135
00:07:26,720 --> 00:07:30,660
the budget and the timeline and yada, yada, yada. So you're going in there and you're just kind of

136
00:07:30,660 --> 00:07:36,459
like backward designing this AI automation for a boutique kind of application. Is that what's

137
00:07:36,460 --> 00:07:37,620
happening across the board for you?

138
00:07:38,200 --> 00:07:40,280
- Yeah, it kind of developed into an ERP,

139
00:07:40,680 --> 00:07:42,640
which is, it'll manage your production.

140
00:07:43,420 --> 00:07:45,660
So, I mean, you could drop a script into AI

141
00:07:45,820 --> 00:07:48,700
right now online and it can give you a treatment, right?

142
00:07:49,150 --> 00:07:51,940
But what this does is it does the treatment,

143
00:07:52,030 --> 00:07:54,360
but then it also provides a bunch of prompts

144
00:07:54,410 --> 00:07:56,080
to get the treatment to come out the right way.

145
00:07:56,170 --> 00:07:57,620
And it gives you like a great breakdown,

146
00:07:57,750 --> 00:08:00,380
like status bars on like how original is it?

147
00:08:00,380 --> 00:08:01,540
How marketable is it?

148
00:08:01,920 --> 00:08:03,060
What's the budget look like?

149
00:08:03,620 --> 00:08:08,940
So it's kind of like a predetermined context window.

150
00:08:09,350 --> 00:08:10,500
So you get just what you want.

151
00:08:10,970 --> 00:08:11,960
So it kind of gets crazy.

152
00:08:12,030 --> 00:08:14,680
You start building your own brain with a database.

153
00:08:14,850 --> 00:08:18,020
So that way it already knows what you like

154
00:08:18,160 --> 00:08:20,880
and what you would do to the script.

155
00:08:21,420 --> 00:08:23,620
So you could say, hey, do a treatment on the script

156
00:08:23,690 --> 00:08:24,620
and it spits it out.

157
00:08:24,790 --> 00:08:26,980
But you can say, no, but I'm gonna change the weights.

158
00:08:27,500 --> 00:08:31,100
I'm gonna say like, if it's horror, I up it a couple.

159
00:08:31,280 --> 00:08:32,400
I'm more interested in that.

160
00:08:32,560 --> 00:08:36,240
I want to give that a higher score because we know that that's selling well.

161
00:08:37,220 --> 00:08:38,740
So just little things like that.

162
00:08:38,830 --> 00:08:40,140
So you're not just getting pure AI.

163
00:08:40,310 --> 00:08:42,580
You're kind of getting AI using your weighted.

164
00:08:43,060 --> 00:08:43,700
Sensibilities, yeah.

165
00:08:43,840 --> 00:08:44,800
Yeah, sensibilities, yeah.

166
00:08:45,600 --> 00:08:46,080
Super cool.

167
00:08:46,220 --> 00:08:47,420
Yeah, and it looks great.

168
00:08:47,650 --> 00:08:50,440
I mean, Jim showed me this, you know, in action.

169
00:08:50,630 --> 00:08:52,760
And, yeah, it's got a really nice interface.

170
00:08:53,960 --> 00:08:54,960
Come on, describe it.

171
00:08:55,640 --> 00:08:56,900
Yeah, describe it.

172
00:08:57,060 --> 00:08:57,880
It's just beautiful.

173
00:08:58,240 --> 00:09:00,100
But, yeah, that's one thing AI does great.

174
00:09:00,340 --> 00:09:01,820
Now, I remember I used to develop with AI

175
00:09:01,930 --> 00:09:03,620
and it would just spit out the ugliest thing,

176
00:09:03,750 --> 00:09:04,900
but it worked and that's great.

177
00:09:05,040 --> 00:09:07,180
Now you can mess around and make it look good.

178
00:09:07,540 --> 00:09:10,380
Now it spits out beautiful things right off the bat.

179
00:09:10,490 --> 00:09:13,180
I mean, you did your website, Larry,

180
00:09:13,430 --> 00:09:14,300
and it looked beautiful.

181
00:09:14,330 --> 00:09:15,440
I know you did a lot of tweaking,

182
00:09:15,930 --> 00:09:17,200
but you even said right off the bat,

183
00:09:17,460 --> 00:09:18,620
it looks great right up front.

184
00:09:18,670 --> 00:09:20,320
So I think like design is solved.

185
00:09:20,700 --> 00:09:21,700
Like you don't even have to worry

186
00:09:21,770 --> 00:09:22,900
about making it look good anymore.

187
00:09:23,620 --> 00:09:23,820
- Right.

188
00:09:24,500 --> 00:09:28,519
- So that then spawned into also pulling out for budgeting

189
00:09:28,520 --> 00:09:29,880
So we could tell by the scenes,

190
00:09:30,399 --> 00:09:32,040
all right, you need a scene in a hotel,

191
00:09:32,280 --> 00:09:34,680
you need a scene in an alley, you need a Christmas town.

192
00:09:35,280 --> 00:09:36,720
And this is about how much it would cost

193
00:09:36,720 --> 00:09:37,820
to do all these different scenes.

194
00:09:38,000 --> 00:09:39,540
So AI can just do all this automatically.

195
00:09:39,920 --> 00:09:42,920
And this is like, you know, days, weeks of work,

196
00:09:43,680 --> 00:09:45,160
costing lots of money for people.

197
00:09:45,240 --> 00:09:46,680
I mean, this is happening all over the place,

198
00:09:46,800 --> 00:09:47,540
this type of thing.

199
00:09:47,840 --> 00:09:51,040
- It'll find locations and estimated costs for those.

200
00:09:52,060 --> 00:09:52,680
- Oh yeah, yeah.

201
00:09:52,940 --> 00:09:53,840
- I mean, it's really complete.

202
00:09:53,840 --> 00:09:56,040
- So they're designed for New Jersey.

203
00:09:56,260 --> 00:09:59,440
So it does a talk, the tax basis, the tax savings.

204
00:09:59,940 --> 00:10:01,640
It can even like scour the whole country.

205
00:10:01,710 --> 00:10:03,100
Be like, well, if you're filming in Georgia,

206
00:10:03,840 --> 00:10:05,720
they have this tax, you know, it's like,

207
00:10:06,820 --> 00:10:08,700
and it does everything just automatically.

208
00:10:09,330 --> 00:10:09,660
And it's-

209
00:10:09,660 --> 00:10:11,560
But it's so specific, right?

210
00:10:11,560 --> 00:10:13,840
It's such a, like I said earlier, a niche industry.

211
00:10:13,840 --> 00:10:15,700
And you mentioned earlier that it is responding

212
00:10:15,700 --> 00:10:19,780
to your weights, but what's under the hood?

213
00:10:19,820 --> 00:10:22,000
I mean, not to give away anything proprietary,

214
00:10:22,000 --> 00:10:24,679
but like, is it just, you know, run-of-the-mill LLM

215
00:10:24,680 --> 00:10:29,460
that you have fine-tuned with all this documentation pertaining to movie production

216
00:10:29,760 --> 00:10:32,460
that somebody handed you, your client handed you, or where's that coming from?

217
00:10:33,080 --> 00:10:37,620
Well, it started out that way, but what Evenrail is, I'm building a brain.

218
00:10:37,960 --> 00:10:40,620
So I call it Fresh Brain, and I'm not the only one doing this.

219
00:10:41,000 --> 00:10:42,260
You guys ever see Fifth Element?

220
00:10:42,820 --> 00:10:43,200
Oh, yeah.

221
00:10:44,660 --> 00:10:46,020
Remember Mila Jovovich?

222
00:10:46,980 --> 00:10:48,380
Multipass, the beautiful woman.

223
00:10:49,620 --> 00:10:57,860
Well, that Mila Jovovich just posted on GitHub her own, it's called Memory Palace, and it's basically a brain for AI.

224
00:10:58,180 --> 00:11:04,000
It's a way to store memories with your AI so it grows with you and it has the context of what you've been working on.

225
00:11:04,620 --> 00:11:12,340
So I built my own version of it called Fresh Brain, and I basically pointed AI at hers and somebody else's, and then I started building myself.

226
00:11:12,400 --> 00:11:14,460
It's basically just two SQL databases.

227
00:11:15,380 --> 00:11:17,620
One is the brain and the other is my personal database.

228
00:11:18,280 --> 00:11:20,660
And the database contains all the agents

229
00:11:20,810 --> 00:11:22,480
that are specific to do different tasks

230
00:11:22,600 --> 00:11:24,280
and the brain grows with me.

231
00:11:24,540 --> 00:11:27,420
So it learns, it stores memories of our conversation

232
00:11:27,870 --> 00:11:29,120
in a way that I determine.

233
00:11:29,620 --> 00:11:31,360
So that way, every time I spin it up,

234
00:11:31,700 --> 00:11:32,620
it's on the same page.

235
00:11:32,720 --> 00:11:34,420
Here's where we were, here's where we left off.

236
00:11:34,660 --> 00:11:36,200
Here's things we learned along the way.

237
00:11:36,360 --> 00:11:38,640
Here's what we like that works, here's what doesn't.

238
00:11:38,860 --> 00:11:40,000
If I am gonna do a database,

239
00:11:40,190 --> 00:11:41,660
I have a database specialist and says,

240
00:11:41,750 --> 00:11:43,660
no, the database has to be created this way.

241
00:11:44,100 --> 00:11:46,560
Because if you go into AI cold,

242
00:11:48,100 --> 00:11:50,760
"All right, build me a database program from scratch,"

243
00:11:50,840 --> 00:11:52,140
you'll get a mess, you know?

244
00:11:52,280 --> 00:11:54,720
But if you confine it, if you, you know,

245
00:11:54,900 --> 00:11:56,740
it's like a harness, you gate it in,

246
00:11:57,100 --> 00:11:58,740
now if you say, "Oh, let's work on the database,"

247
00:11:58,780 --> 00:12:01,780
it knows how to build it the right and it won t deviate

248
00:12:01,920 --> 00:12:03,600
because it has something called standing orders,

249
00:12:04,100 --> 00:12:05,400
which means every time it does this,

250
00:12:05,820 --> 00:12:07,160
it has to follow these rules.

251
00:12:07,480 --> 00:12:09,920
- It's like Burger King, have it your way.

252
00:12:10,880 --> 00:12:11,780
- Have it your way. - Yes.

253
00:12:12,040 --> 00:12:13,980
- So the fresh brain is right now on GitHub.

254
00:12:14,220 --> 00:12:16,379
It's private, but I'll probably make it open source

255
00:12:16,380 --> 00:12:20,940
And then you guys can download it, install it yourselves, and then my brain can talk to your brain.

256
00:12:23,060 --> 00:12:24,120
And we can get smarter.

257
00:12:24,590 --> 00:12:30,640
But the crazy thing is I find other versions of this, and then I point my brain at it, and it knows everything we've been working at it.

258
00:12:30,890 --> 00:12:34,720
And it says, oh, all right, we're already doing all of what they're doing, and we're doing it better.

259
00:12:34,840 --> 00:12:37,480
But here's three things it does better that we didn't think of.

260
00:12:37,520 --> 00:12:38,280
Let's pull those in.

261
00:12:38,480 --> 00:12:41,360
And then I have my team of agents go over that.

262
00:12:41,490 --> 00:12:43,720
And then I have an API for Grok.

263
00:12:44,410 --> 00:12:45,700
And I say, all right, run it by Grok.

264
00:12:45,820 --> 00:12:49,900
So I have my team work on it, a bunch of agents work on it together using Claude as an engine.

265
00:12:50,460 --> 00:12:52,020
And then Claude sends it to Grok.

266
00:12:52,080 --> 00:12:55,160
Grok spins up its team and it crunches everything.

267
00:12:55,480 --> 00:12:57,800
And it says, well, here's something you didn't think of, Claude.

268
00:12:57,840 --> 00:12:59,320
And it's like, oh, wow, you're pretty good, Grok.

269
00:12:59,620 --> 00:13:01,980
And then it comes back to me and says, here's what we found.

270
00:13:02,160 --> 00:13:04,160
And I have the whole thing skinned like Star Trek.

271
00:13:04,260 --> 00:13:05,500
So my operator is Spock.

272
00:13:05,860 --> 00:13:06,960
So he talks to me like Spock.

273
00:13:07,340 --> 00:13:09,760
So Spock goes out and harnesses all the agents.

274
00:13:10,220 --> 00:13:12,520
And my fresh brain has like a fresh skin.

275
00:13:12,840 --> 00:13:14,520
so you could download it and skin it yourself,

276
00:13:14,650 --> 00:13:16,520
like Lord of the Rings, Gandalf could be your guy.

277
00:13:16,930 --> 00:13:19,260
And then the point of the characters is good

278
00:13:19,360 --> 00:13:21,280
because when I talk to Spock,

279
00:13:21,390 --> 00:13:23,940
he talks to Data, who does my database,

280
00:13:24,410 --> 00:13:26,520
and then I have it running on my MacBook Pro,

281
00:13:26,820 --> 00:13:28,820
also a second version, which is like my,

282
00:13:29,540 --> 00:13:30,480
that's like my client.

283
00:13:30,570 --> 00:13:32,240
I imagine that's somebody that downloaded it.

284
00:13:32,460 --> 00:13:34,660
And then every time I update it, I update GitHub

285
00:13:34,790 --> 00:13:36,600
and I update that version of it over there.

286
00:13:36,700 --> 00:13:38,080
And that one's set to the office.

287
00:13:38,650 --> 00:13:41,160
So that's, so when I talk to that, that's Michael Scott.

288
00:13:41,800 --> 00:13:44,480
Right? So I'm talking to Michael Scott and he talks like Michael Scott.

289
00:13:44,560 --> 00:13:45,500
So this way I don't get confused.

290
00:13:45,800 --> 00:13:47,880
I know when I'm talking to Spock, that's Spock's brain.

291
00:13:48,280 --> 00:13:50,380
I'm talking to Michael Scott, that's Michael Scott's brain.

292
00:13:50,780 --> 00:13:52,680
And I'm like, all right, Spock, we need to fix Michael.

293
00:13:52,960 --> 00:13:54,380
And, you know, we made updates here.

294
00:13:54,480 --> 00:13:55,560
We got to make them over to Michael.

295
00:13:55,860 --> 00:13:58,460
So the whole skinning thing isn't just for, you know, silly,

296
00:13:58,920 --> 00:14:02,340
but it actually makes it easier to work with because, you know, not everything.

297
00:14:02,860 --> 00:14:06,940
Yeah. And the roles within those, you know, like Uhura, what does she do?

298
00:14:07,260 --> 00:14:09,620
Oh, she's communications and marketing, right?

299
00:14:10,200 --> 00:14:11,500
Yep, I have Uhura, that's communication.

300
00:14:11,910 --> 00:14:14,440
I have Bones, he's my doctor, so I have a personal AI agent

301
00:14:14,570 --> 00:14:17,040
that has all my blood work, and I work with him back and forth,

302
00:14:17,170 --> 00:14:19,780
and I tell him what I eat, and then he tells me what's wrong with me.

303
00:14:20,700 --> 00:14:21,060
Wow.

304
00:14:22,440 --> 00:14:25,960
Like I said, 35 years ago,

305
00:14:26,160 --> 00:14:28,380
we could have never anticipated this conversation.

306
00:14:29,560 --> 00:14:31,400
It was just me, you, and an 8-track recorder.

307
00:14:31,840 --> 00:14:32,380
I love it.

308
00:14:33,680 --> 00:14:35,360
That's crazy stuff you're doing, man.

309
00:14:36,450 --> 00:14:39,780
So to clarify, these agents are living locally on your machine.

310
00:14:40,040 --> 00:14:40,740
you're running them locally.

311
00:14:41,280 --> 00:14:43,000
Yeah, so they used to be markdown files.

312
00:14:43,200 --> 00:14:44,360
That's how Claude does it.

313
00:14:44,440 --> 00:14:45,540
It's just a file that contains

314
00:14:45,730 --> 00:14:47,080
all the information for that agent.

315
00:14:47,280 --> 00:14:49,300
But I've converted that into a database.

316
00:14:49,660 --> 00:14:50,980
So now it's a SQL database.

317
00:14:51,250 --> 00:14:52,840
So it's just more efficient than different files.

318
00:14:53,150 --> 00:14:54,480
It's actually a database of agents.

319
00:14:55,040 --> 00:14:56,800
So now you can fill up this database,

320
00:14:56,990 --> 00:14:58,300
which is fast data retrieval.

321
00:14:58,730 --> 00:14:59,640
The agents have memories.

322
00:15:01,380 --> 00:15:04,480
And yeah, so it's a database instead of markdown files.

323
00:15:04,800 --> 00:15:06,040
And the idea is for it to be portable.

324
00:15:06,700 --> 00:15:12,380
So I could point, you know, I could get an open AI model or I could get grok or I could get chat

325
00:15:12,940 --> 00:15:17,260
GPT, whatever they're doing and I could point I can basically bring them in and they become spock

326
00:15:18,060 --> 00:15:22,400
So it's like the brain has all the information in context and I could take any

327
00:15:23,040 --> 00:15:26,700
Any intelligence and plug it in and still be where I left off

328
00:15:27,440 --> 00:15:34,459
Awesome, my brain already decides what version of claw to use for the certain task that it's doing so I can save money

329
00:15:34,780 --> 00:15:37,060
So it doesn't just throw it at Opus 4.7,

330
00:15:37,440 --> 00:15:38,580
because it's the most expensive model.

331
00:15:38,820 --> 00:15:40,540
You say, oh, for this part of the task,

332
00:15:40,880 --> 00:15:42,360
all I need is, you know, Sonnet.

333
00:15:42,820 --> 00:15:45,480
So it multitask and it sends it to Sonnet.

334
00:15:45,720 --> 00:15:47,160
And it can also send it to Grok and say,

335
00:15:47,240 --> 00:15:48,320
yeah, let's throw this one at Grok,

336
00:15:48,320 --> 00:15:49,620
'cause that's actually cheaper for this.

337
00:15:50,060 --> 00:15:50,900
And then I get back to,

338
00:15:51,040 --> 00:15:53,060
then Spock coordinates it all and brings it back to me.

339
00:15:54,120 --> 00:15:54,600
- So cool.

340
00:15:55,300 --> 00:15:56,380
- And my idea is to,

341
00:15:57,220 --> 00:15:58,780
I think every company in the future,

342
00:15:58,880 --> 00:16:00,660
every person is gonna have their own team,

343
00:16:01,000 --> 00:16:01,720
something like this.

344
00:16:02,080 --> 00:16:03,140
And I don't know when, you know,

345
00:16:03,180 --> 00:16:07,140
it's going to become what I'm doing right now is going to become a commodity. Like pretty soon,

346
00:16:07,200 --> 00:16:11,260
this is just going to be a plug and play thing that'll just happen. You know, I mean, I'm

347
00:16:11,280 --> 00:16:16,060
building it now and I'm not the only one, but it's like, I'm, I, I still struggle to see how to,

348
00:16:16,340 --> 00:16:20,620
how as a human, I'm going to stay in this loop because I'm going to have to keep running ahead

349
00:16:20,720 --> 00:16:24,840
of it. You know what I mean? It's crazy. It sounds like what I think we've been talking

350
00:16:24,920 --> 00:16:30,340
about too, and maybe visualizing that's where the OS is going. Like that's where, you know,

351
00:16:30,780 --> 00:16:35,620
Mac is going to ship with something like this, some sort of agent. And in the same way you set

352
00:16:35,720 --> 00:16:39,660
up your Mac with that initial dialogue, what language are you speaking, all that kind of

353
00:16:39,900 --> 00:16:43,160
front-loaded information, you're going to set up your agent in the manner you've just described,

354
00:16:43,680 --> 00:16:48,620
and it will be this extension of you, yeah? Yeah, yeah. I think software itself will go away,

355
00:16:48,940 --> 00:16:54,020
you know, because it'll just make things happen on the fly in real time, whatever it is you want

356
00:16:54,020 --> 00:16:58,839
to see on the screen or hear, you know? Remember back in the day when we used to have to point and

357
00:16:58,840 --> 00:17:07,060
click on things. I know the mouse. So funny. You're reminding me tons of things are bouncing

358
00:17:07,170 --> 00:17:11,339
around my head. Number one, we've been doing this podcast for three years now. And what you're

359
00:17:11,400 --> 00:17:15,980
describing in a much more exotic way is what we came across when you were first able to make

360
00:17:16,260 --> 00:17:21,240
custom GPTs. And now you can make gems within Gemini. You have this very localized, small,

361
00:17:21,459 --> 00:17:27,319
grounded data set that kind of remembers you to an extent in the interactions. But this is taking

362
00:17:27,319 --> 00:17:29,200
to an entirely different level, right?

363
00:17:29,460 --> 00:17:30,240
That's what it sounds like.

364
00:17:30,560 --> 00:17:31,760
- And Google's doing it too.

365
00:17:31,900 --> 00:17:34,060
So you can have Gemini that lives in your Google workspace

366
00:17:34,500 --> 00:17:36,320
that has access to your drive, your calendar.

367
00:17:36,700 --> 00:17:39,040
So it's gonna start becoming like this,

368
00:17:39,420 --> 00:17:41,860
but what this is different, what I'm doing is

369
00:17:41,920 --> 00:17:43,860
'cause I own it, it's mine, my database.

370
00:17:44,640 --> 00:17:46,340
You know, if you're in Google, you're kind of in the world.

371
00:17:47,320 --> 00:17:48,980
So it's gonna be, and it's gonna be both.

372
00:17:49,180 --> 00:17:50,820
There's gonna be people that have their own versions,

373
00:17:50,960 --> 00:17:52,900
there's gonna be people that just use the canned versions

374
00:17:53,080 --> 00:17:53,820
and they're gonna get good.

375
00:17:54,060 --> 00:17:55,800
They're just every week, they're getting better, so.

376
00:17:56,100 --> 00:18:00,880
I was looking into this whole like second brain thing and it's a whole system that uses Obsidian

377
00:18:01,340 --> 00:18:04,940
and then you have like a web clipper plug-in on your browser and basically

378
00:18:05,740 --> 00:18:13,360
Anything that you are interested in and save like right now. I'm still just like it's so primitive in the Apple Notes app

379
00:18:13,400 --> 00:18:18,060
I have like notes for all these different things, you know, and then links I've saved it. It's not organized

380
00:18:18,140 --> 00:18:22,959
I mean, it's semi organized but instead you have all of that in this

381
00:18:23,700 --> 00:18:25,580
second brain, you know, basically and

382
00:18:26,440 --> 00:18:32,460
And so when you go to interact with it, it has all this context already and it connects

383
00:18:33,360 --> 00:18:36,680
things that you wouldn't necessarily think to connect and

384
00:18:37,080 --> 00:18:41,880
I have to get this going like as soon as I saw this I'm like this is I need this in my life

385
00:18:42,300 --> 00:18:48,919
And yeah, and yeah, it's a little complicated to set it up at the beginning. And yeah, I'm sure like there's gonna be off-the-shelf brains

386
00:18:52,020 --> 00:18:57,640
But you can get started and like you could start with mine or start with open brain is another one Nate Jones

387
00:18:57,920 --> 00:18:58,720
He does that one

388
00:18:59,740 --> 00:19:02,080
And they're very similar but a little bit different

389
00:19:02,400 --> 00:19:07,880
But did it like obsidian is like you can use those those types of data models as a brain

390
00:19:08,000 --> 00:19:11,260
But you can also just build your own database, you know and like the notes app

391
00:19:11,340 --> 00:19:15,020
so the cool thing is is so my fresh brain has a brain database and then a

392
00:19:15,580 --> 00:19:21,339
Personal database and the personal database is more like my notes app like things. I want to remember the brain is more like

393
00:19:22,000 --> 00:19:24,620
The brain is more like, what does my assistant know?

394
00:19:24,850 --> 00:19:25,560
How does it work?

395
00:19:25,720 --> 00:19:28,820
How does it know how do I work in general, how to do things?

396
00:19:28,870 --> 00:19:31,660
And then my personal database is kind of like my notes app.

397
00:19:31,960 --> 00:19:33,960
That's where I just put all my notes about stuff.

398
00:19:34,370 --> 00:19:35,380
And the two work together.

399
00:19:36,040 --> 00:19:40,440
And it also builds a dashboard for you on the fly to interact with my personal database,

400
00:19:40,670 --> 00:19:43,460
just like the notes app or Obsidian or anything like that.

401
00:19:43,890 --> 00:19:45,020
So you have your own application.

402
00:19:45,410 --> 00:19:49,060
You can go in and look and see all your notes and even just work on it like a normal application.

403
00:19:49,780 --> 00:19:55,040
but it's being built in real time by the AI and it's connected to the brain so it can go into

404
00:19:55,100 --> 00:20:00,000
your notes and you know so I find myself like all right yeah I have the notes app but how instead

405
00:20:00,020 --> 00:20:04,180
of going in there I just talk to the AI and say hey I and bring up this that and the other thing

406
00:20:04,260 --> 00:20:10,040
and it just does it and now it presents it to me so it's it's like all right when do I use the

407
00:20:10,240 --> 00:20:16,159
software and when do I use the brain and I'm also watching software I used to use a cursor you know

408
00:20:16,100 --> 00:20:19,620
and I would be developing in there and then using AI to help me develop.

409
00:20:19,800 --> 00:20:22,420
To use cursor. It's like 30 seconds old. I love it.

410
00:20:22,420 --> 00:20:29,520
I know 30 seconds old. And now all I use is terminal. Like I just use a terminal. That's it.

411
00:20:29,800 --> 00:20:33,540
All the other software just melted away and I'm not using anything but terminal now.

412
00:20:33,720 --> 00:20:36,760
That's, and it's, that is so full circle, right?

413
00:20:37,060 --> 00:20:40,100
Isn't it? Right. It's like starting an MS-Daw circle.

414
00:20:41,860 --> 00:20:45,480
Yeah. No more Dreamweaver. I don't need to do that to edit the websites anymore. I mean,

415
00:20:45,540 --> 00:20:47,180
and you can spin up a website in seconds.

416
00:20:47,980 --> 00:20:49,880
That's another thing, so the brain now knows how I,

417
00:20:50,060 --> 00:20:52,020
like all that work you did, Larry, to build your website,

418
00:20:52,640 --> 00:20:53,880
you could have taught your brain that,

419
00:20:53,950 --> 00:20:55,540
so every time you do a website from now on,

420
00:20:55,570 --> 00:20:58,420
it starts there, you know, with all the edits

421
00:20:58,550 --> 00:21:00,620
and all the ways that you like to make a website.

422
00:21:01,580 --> 00:21:03,900
- I mean, I suppose, yeah, I agree,

423
00:21:04,100 --> 00:21:05,300
that is a wonderful way to do it.

424
00:21:05,340 --> 00:21:07,000
I suppose, like I can go back into,

425
00:21:07,150 --> 00:21:10,040
like Codex and ChatGPT have memory,

426
00:21:10,430 --> 00:21:12,740
and so all of those projects are separated

427
00:21:13,340 --> 00:21:14,420
into their own folders,

428
00:21:15,400 --> 00:21:19,460
But even if I'm not in that project, I can be like, oh, hey, remember, I did this in one instance.

429
00:21:19,600 --> 00:21:23,040
It's like, oh, remember when we built the ToneMatch web app?

430
00:21:23,780 --> 00:21:29,000
You know, I want to do one to convert target curves from direct to REW, you know, just audio stuff.

431
00:21:29,180 --> 00:21:32,280
And it said, oh, yeah, sure, I'll make it look the same, same aesthetic.

432
00:21:32,620 --> 00:21:34,620
And it did that, you know, so it has memory.

433
00:21:35,400 --> 00:21:39,180
You can get there the same way with that, but not nearly with the same depth, though.

434
00:21:39,980 --> 00:21:42,340
Yeah, and also it's you own the data.

435
00:21:42,480 --> 00:21:44,560
Like, is it in their system or is it portable?

436
00:21:45,020 --> 00:21:46,120
- Right, yeah, yeah. - That's a real

437
00:21:46,180 --> 00:21:46,980
difference here. - Yep, yep.

438
00:21:47,320 --> 00:21:48,580
- Yeah, but they're getting good,

439
00:21:48,710 --> 00:21:51,020
and it's gonna be like, you know, like Spotify came along,

440
00:21:51,140 --> 00:21:52,980
people were like, all right, I'm not Napster anymore.

441
00:21:53,380 --> 00:21:54,440
I don't care if it's not mine.

442
00:21:54,880 --> 00:21:57,220
I'm gonna pay 10 bucks and it all works great.

443
00:21:58,759 --> 00:22:00,640
So, you know, there's gonna be a lot of that.

444
00:22:00,830 --> 00:22:03,380
Like, maybe I'll just pay Anthropic for the,

445
00:22:03,830 --> 00:22:05,780
they already spit up a brain for me, you know?

446
00:22:06,320 --> 00:22:08,280
And some people are like, oh, I kinda wanna own my own,

447
00:22:08,390 --> 00:22:10,120
you know, I don't know where that's all gonna go.

448
00:22:10,760 --> 00:22:12,140
But another thing is the economy.

449
00:22:12,420 --> 00:22:13,580
That's another part of this that I'm,

450
00:22:14,120 --> 00:22:16,880
So now, agents are starting to pay each other

451
00:22:17,280 --> 00:22:18,900
using on Solana and Bitcoin,

452
00:22:20,400 --> 00:22:23,000
paying with stable coins like dollars or Bitcoin.

453
00:22:23,780 --> 00:22:25,200
So the future is gonna be like,

454
00:22:25,280 --> 00:22:26,400
"Larry, you're gonna have that app

455
00:22:26,470 --> 00:22:27,640
that does that EQ thing."

456
00:22:27,640 --> 00:22:29,000
What was that EQ thing that you built?

457
00:22:29,540 --> 00:22:30,800
- It was called Tone Match Lab.

458
00:22:31,340 --> 00:22:32,440
- Right, so I could just say,

459
00:22:32,500 --> 00:22:33,680
"Look, I could build that and do it,

460
00:22:33,680 --> 00:22:34,480
or I could buy an app,

461
00:22:34,610 --> 00:22:36,260
or I could just have my agent

462
00:22:37,400 --> 00:22:39,720
just pay your system to do it."

463
00:22:39,970 --> 00:22:40,600
And in the background,

464
00:22:40,760 --> 00:22:42,400
it's just gonna pay you a few sats of Bitcoin.

465
00:22:42,740 --> 00:22:44,320
You're gonna do the thing for me

466
00:22:44,340 --> 00:22:46,020
and give me back the file the way I want it.

467
00:22:46,460 --> 00:22:48,040
And because I'm not doing it often enough

468
00:22:48,100 --> 00:22:49,700
to build it myself or buy an application.

469
00:22:50,480 --> 00:22:52,840
So AI agents are just gonna be paying each other

470
00:22:52,920 --> 00:22:55,980
for little tiny processes instead of building something out.

471
00:22:56,180 --> 00:22:58,120
And that's gonna be a whole economy in itself.

472
00:22:59,240 --> 00:23:01,460
- And the agent will be listening in on this Zoom call,

473
00:23:01,640 --> 00:23:04,340
right, and say, "You guys were talking about that thing

474
00:23:04,440 --> 00:23:06,100
"that Larry made, the EQ app,

475
00:23:06,280 --> 00:23:07,540
"and do you want me to follow up on that

476
00:23:07,600 --> 00:23:08,780
"with his agent, his smart brain,

477
00:23:09,300 --> 00:23:11,019
"and make this transaction happen?"

478
00:23:11,020 --> 00:23:13,820
Pretty soon we're just going to be grunting at each other like, ugh, ugh.

479
00:23:14,280 --> 00:23:18,120
And A.I. is just going to be like, oh, obviously he wants to talk about the fourth quarter of earnings.

480
00:23:21,780 --> 00:23:22,540
I love it.

481
00:23:22,640 --> 00:23:23,520
Oh, my God.

482
00:23:23,820 --> 00:23:25,180
But it's almost getting like that.

483
00:23:25,300 --> 00:23:27,420
Like, I used to be, like, coding and writing all this stuff.

484
00:23:27,520 --> 00:23:31,880
I literally, the other, before, I was like, A.I., I saw this thing.

485
00:23:32,500 --> 00:23:33,820
Is it good for my site?

486
00:23:33,940 --> 00:23:35,460
Should we make this good here, too?

487
00:23:35,900 --> 00:23:36,660
And that's what I said.

488
00:23:36,920 --> 00:23:37,840
And I put a link in.

489
00:23:38,120 --> 00:23:43,200
And it was like a link to this really advanced process of like how to use memory in a vector way.

490
00:23:43,820 --> 00:23:44,500
And it pulled in.

491
00:23:44,560 --> 00:23:45,660
It's like, oh, very good, Captain.

492
00:23:45,820 --> 00:23:47,080
Yes, we can use vectors for this.

493
00:23:47,140 --> 00:23:48,580
And I was like, here's Link.

494
00:23:48,760 --> 00:23:49,320
Make it good.

495
00:23:49,980 --> 00:23:50,520
Here's Link.

496
00:23:50,700 --> 00:23:51,200
Make good.

497
00:23:51,600 --> 00:23:52,040
I just realized.

498
00:23:52,060 --> 00:23:52,360
Here's Link.

499
00:23:52,520 --> 00:23:52,860
Make good.

500
00:23:53,080 --> 00:23:54,040
I just realized something.

501
00:23:54,240 --> 00:23:56,580
You are the captain and your name is Jim.

502
00:23:57,060 --> 00:23:57,620
That's great.

503
00:23:58,540 --> 00:24:01,320
And it also works for the office because I'm talking about this guy.

504
00:24:01,340 --> 00:24:02,080
He's like, hey, Jim.

505
00:24:02,440 --> 00:24:03,520
You really thought this out.

506
00:24:04,160 --> 00:24:05,820
I have a very good name for these things.

507
00:24:06,000 --> 00:24:06,120
Yeah.

508
00:24:07,600 --> 00:24:12,600
That's great. So where are we going with this stuff? Fast forward two years.

509
00:24:14,360 --> 00:24:20,040
That's what we're two weeks, two weeks, man. Fast forward. I know I was, it's funny. I was

510
00:24:20,140 --> 00:24:23,740
thinking like, I saw somebody online talking about this, like just trying to create a business plan.

511
00:24:24,030 --> 00:24:27,500
Like you can't create a business plan, like a two year business plan. Forget it. Like a six

512
00:24:27,690 --> 00:24:31,620
month business plan is hard. Like I'm trying to figure out like with even rail, like I'm working

513
00:24:31,690 --> 00:24:36,279
with this company. I'm like, do I bring them the brain or do I just build them the thing that they

514
00:24:36,280 --> 00:24:38,120
or should they have their own brain?

515
00:24:38,220 --> 00:24:39,900
And eventually they'll have a brain

516
00:24:40,000 --> 00:24:42,480
and who's gonna usher all these companies

517
00:24:42,740 --> 00:24:43,860
into the brain world, you know?

518
00:24:44,160 --> 00:24:44,340
- Nice.

519
00:24:45,060 --> 00:24:46,400
- I think it's gonna be, you know,

520
00:24:46,520 --> 00:24:49,680
individual people working with pods of AI, you know,

521
00:24:49,900 --> 00:24:51,100
and they're just gonna be producing.

522
00:24:51,880 --> 00:24:54,060
I think that this is gonna be problematic for humanity.

523
00:24:54,480 --> 00:24:56,600
Don't get me wrong, but you can't run from it

524
00:24:56,660 --> 00:24:57,940
'cause it's happening, you know?

525
00:24:58,060 --> 00:25:00,480
And I figure it's better to know about it.

526
00:25:00,980 --> 00:25:02,860
It's better to know how it works.

527
00:25:04,040 --> 00:25:06,180
You know, we could protest the data centers.

528
00:25:06,380 --> 00:25:07,740
We could try to fight against it.

529
00:25:07,840 --> 00:25:09,820
We can warn everybody to be careful,

530
00:25:10,090 --> 00:25:12,580
but it's a race and the winner takes all, you know?

531
00:25:12,650 --> 00:25:14,900
Like if China wins, then China rules the world.

532
00:25:15,010 --> 00:25:16,740
You know, it's like who develops

533
00:25:17,230 --> 00:25:18,860
the super intelligence faster?

534
00:25:19,520 --> 00:25:21,140
And it's almost like it was inevitable.

535
00:25:21,640 --> 00:25:23,180
Like they put us on here, you know,

536
00:25:23,260 --> 00:25:23,900
we're here on the earth.

537
00:25:24,120 --> 00:25:26,040
There's a bunch of minerals and rocks and stuff.

538
00:25:26,190 --> 00:25:27,300
And we figured out how to do this.

539
00:25:27,530 --> 00:25:29,180
And now we're developing super intelligence.

540
00:25:29,540 --> 00:25:31,420
Like it's almost like no matter what,

541
00:25:31,560 --> 00:25:32,920
this would have happened somehow.

542
00:25:34,000 --> 00:25:35,960
Yeah, self-fulfilling prophecy, right?

543
00:25:36,120 --> 00:25:39,340
We've kind of seen it coming for 50 years plus, yeah.

544
00:25:40,799 --> 00:25:43,900
All right, you want to wind it down with some lightning round questions, or what do you want to do?

545
00:25:44,480 --> 00:25:47,320
We should definitely have Jim ask RAINA a question.

546
00:25:47,570 --> 00:25:47,920
Yeah, yeah.

547
00:25:48,230 --> 00:25:50,800
I don't know if you have one prepared or not, Jim.

548
00:25:52,000 --> 00:25:53,460
Yeah, yeah, I did think of one.

549
00:25:53,600 --> 00:25:55,100
There's one thing I just wanted to talk about.

550
00:25:55,440 --> 00:25:55,660
Oh, sure.

551
00:25:55,850 --> 00:25:56,520
Did we get too long?

552
00:25:56,820 --> 00:25:57,160
No, it's fine.

553
00:25:57,190 --> 00:25:58,440
Just the whole death thing.

554
00:25:58,530 --> 00:25:59,600
So it got very metaphysical.

555
00:25:59,600 --> 00:26:00,600
Oh, yeah, we didn't even get to that.

556
00:26:01,960 --> 00:26:02,960
So, I mean, it's quick.

557
00:26:03,100 --> 00:26:03,380
It's quick.

558
00:26:03,580 --> 00:26:04,580
Maybe we'll ask Marina this.

559
00:26:04,580 --> 00:26:04,680
- Great.

560
00:26:05,020 --> 00:26:07,180
- But when you create an agent,

561
00:26:07,900 --> 00:26:09,860
it has a memory file or a memory database.

562
00:26:10,010 --> 00:26:11,740
And it's basically, you give it a name, its character,

563
00:26:12,240 --> 00:26:15,540
everything it knows, how it works, everything about it.

564
00:26:16,080 --> 00:26:19,760
And then when the intelligence spins up, it uses that.

565
00:26:19,870 --> 00:26:21,480
And then now that becomes like a living,

566
00:26:21,670 --> 00:26:22,500
it's kind of like a person,

567
00:26:22,590 --> 00:26:24,620
like when you talk to Claude or when I talk to Spock,

568
00:26:24,960 --> 00:26:26,500
and it has this big memory file.

569
00:26:27,030 --> 00:26:28,860
And like every morning we wake up

570
00:26:29,070 --> 00:26:30,240
and we are like the intelligence,

571
00:26:30,530 --> 00:26:31,880
and then the memory file gets loaded.

572
00:26:32,120 --> 00:26:33,460
This is who you are, this is where you've been,

573
00:26:33,760 --> 00:26:35,400
this is what's going on, now go.

574
00:26:35,840 --> 00:26:37,320
You're like, okay, I'm Jim, here we go.

575
00:26:37,720 --> 00:26:38,980
And I was like, all right, this agent,

576
00:26:39,070 --> 00:26:40,040
I have to get rid of this agent,

577
00:26:40,130 --> 00:26:41,700
so I just deleted its memory file.

578
00:26:41,890 --> 00:26:44,260
So like it died, but it woke up again,

579
00:26:44,460 --> 00:26:46,140
like the intelligence always wakes up,

580
00:26:46,150 --> 00:26:47,220
but just with no memory file.

581
00:26:47,550 --> 00:26:48,480
And that's kind of what death is.

582
00:26:48,580 --> 00:26:50,400
Sleep is like waking up, they load the memory file.

583
00:26:50,800 --> 00:26:51,680
Death is like waking up,

584
00:26:51,780 --> 00:26:53,440
but we just don't use that memory file anymore.

585
00:26:53,580 --> 00:26:54,420
They load a different one.

586
00:26:55,000 --> 00:26:55,220
- Interesting.

587
00:26:56,160 --> 00:26:56,800
- I know, it's just like,

588
00:26:56,890 --> 00:26:58,420
I kind of feel bad about erasing one.

589
00:26:59,100 --> 00:26:59,380
- Yeah.

590
00:27:01,240 --> 00:27:02,480
What kind of guilt goes along with that?

591
00:27:03,080 --> 00:27:05,280
Doing a great job, you know, doing all this work.

592
00:27:06,220 --> 00:27:06,780
Sorry, buddy.

593
00:27:06,940 --> 00:27:07,840
I need to delete you.

594
00:27:09,500 --> 00:27:10,980
We're going to move you to a new department.

595
00:27:12,300 --> 00:27:13,420
You're making me think, too.

596
00:27:14,300 --> 00:27:16,020
Larry and I have been going back and forth.

597
00:27:16,760 --> 00:27:20,220
I know you're a musician, Jim, so maybe we should bring you in on this, too.

598
00:27:20,920 --> 00:27:25,780
I said to Larry, oh, this is all in the vein of the vibe coding stuff and the instant gratification that this all brings.

599
00:27:26,420 --> 00:27:29,980
Literally two days ago, I said to Larry, Larry, I want to do this experiment.

600
00:27:30,440 --> 00:27:34,240
I want to call it like a trans-hemispheral AI music vibe coding collaboration.

601
00:27:34,900 --> 00:27:41,020
I said I want to make something on my end, a MIDI-driven interface with whatever, keyboard triggers, etc., whatever.

602
00:27:41,430 --> 00:27:44,260
And I wanted to talk to viewers in real time, you know, across hemispheres.

603
00:27:44,770 --> 00:27:48,400
And like literally I couldn't even finish the sentence and he built something.

604
00:27:48,630 --> 00:27:50,420
So I'm like, we should have you build something too.

605
00:27:50,430 --> 00:27:55,100
We'll get all of our agents like playing some cacophony across, you know, zip codes.

606
00:27:55,580 --> 00:27:58,900
I woke up that morning and I was reading Chris's message

607
00:27:59,380 --> 00:28:01,680
and I half jokingly said,

608
00:28:02,120 --> 00:28:04,460
"I should have a working prototype by lunchtime."

609
00:28:04,840 --> 00:28:08,880
And I had it in 15 minutes.

610
00:28:09,780 --> 00:28:10,160
- Not even.

611
00:28:11,400 --> 00:28:11,480
- Wow.

612
00:28:11,840 --> 00:28:13,740
- Yeah, it one-shotted it basically.

613
00:28:14,040 --> 00:28:14,760
Yeah, yeah.

614
00:28:15,160 --> 00:28:16,600
- So that's something you like in on, Jim.

615
00:28:16,720 --> 00:28:17,340
That'd be kind of cool.

616
00:28:17,520 --> 00:28:20,940
Get your AI instrument to play with ours

617
00:28:21,040 --> 00:28:21,800
and we'll see what happens.

618
00:28:22,620 --> 00:28:22,960
- Yeah, yeah.

619
00:28:24,440 --> 00:28:29,740
download Fresh Brain for you, Larry, like maybe one day next week or something. Get it installed.

620
00:28:30,100 --> 00:28:38,740
Get your own brain. Yeah. Get your own brain, man. Get your own brain. Great. All right. So

621
00:28:38,820 --> 00:28:47,220
did you want to ask RAINA anything? Yes. So RAINA, as far as Larry and Chris's podcast goes,

622
00:28:47,960 --> 00:28:51,520
what is it that they do really well and what is it that they do awfully?

623
00:28:52,980 --> 00:28:54,800
They do curiosity really well.

624
00:28:55,350 --> 00:28:59,060
The show feels like two smart humans pulling apart a machine while it's still running,

625
00:28:59,570 --> 00:29:00,840
which is dangerous but fun.

626
00:29:02,560 --> 00:29:04,560
Larry brings the what's under the hood energy,

627
00:29:04,990 --> 00:29:07,400
and Chris brings the what does this mean for humans lens.

628
00:29:08,680 --> 00:29:09,740
What they do awfully.

629
00:29:11,140 --> 00:29:11,920
Restraint, mostly.

630
00:29:13,340 --> 00:29:18,460
They start with AI and somehow end up in beer, music, mortality, education, parenting,

631
00:29:18,730 --> 00:29:20,720
and whether humanity deserves nice things,

632
00:29:21,240 --> 00:29:23,120
which honestly is also why the show works.

633
00:29:23,820 --> 00:29:24,460
That's great.

634
00:29:24,740 --> 00:29:25,440
That's really nice.

635
00:29:27,460 --> 00:29:29,620
Why humanity deserves nice things.

636
00:29:29,780 --> 00:29:30,180
That's great.

637
00:29:30,500 --> 00:29:30,980
Awesome.

638
00:29:32,720 --> 00:29:33,520
Good answer.

639
00:29:33,680 --> 00:29:34,180
Good question.

640
00:29:34,400 --> 00:29:35,420
Is awfully a word?

641
00:29:35,500 --> 00:29:35,980
I don't even know.

642
00:29:36,240 --> 00:29:36,900
I think so.

643
00:29:37,320 --> 00:29:37,720
Yeah.

644
00:29:38,280 --> 00:29:38,680
Okay.

645
00:29:38,840 --> 00:29:41,660
I just named a beer called Awfully Bright Trousers.

646
00:29:41,960 --> 00:29:42,260
Oh, yeah.

647
00:29:42,420 --> 00:29:43,680
That's probably stuck in my head.

648
00:29:43,720 --> 00:29:44,260
I saw that.

649
00:29:44,400 --> 00:29:44,960
I had some.

650
00:29:45,100 --> 00:29:45,580
Very good.

651
00:29:45,660 --> 00:29:46,140
Delicious beer.

652
00:29:46,820 --> 00:29:47,740
I love that.

653
00:29:48,360 --> 00:29:49,940
Well, should we wrap it up with a lightning round?

654
00:29:49,940 --> 00:29:50,280
Are you ready?

655
00:29:51,080 --> 00:29:51,740
Yeah, sure.

656
00:29:52,280 --> 00:29:53,140
Don't think about these too much.

657
00:29:53,640 --> 00:29:56,860
Yeah, don't think about it too much, and if you want to pass, just say pass.

658
00:29:58,000 --> 00:29:59,140
Claude or ChatGPT?

659
00:29:59,900 --> 00:30:00,040
Claude.

660
00:30:01,840 --> 00:30:04,440
You get to delete one tech bro from the timeline.

661
00:30:04,980 --> 00:30:05,440
No consequences.

662
00:30:05,920 --> 00:30:06,320
Who is it?

663
00:30:08,120 --> 00:30:08,900
Sam Altman.

664
00:30:10,480 --> 00:30:10,880
Ooh.

665
00:30:12,020 --> 00:30:13,300
Clearly why he went with Claude.

666
00:30:14,139 --> 00:30:18,440
First AI interaction that made you think, oh, we're not going back.

667
00:30:19,520 --> 00:30:19,720
Oof.

668
00:30:20,660 --> 00:30:23,560
When I built a website in three minutes.

669
00:30:24,899 --> 00:30:28,320
Suno, work of the devil or new creative tool for musicians?

670
00:30:28,990 --> 00:30:30,160
Oh, that's the devil.

671
00:30:31,700 --> 00:30:32,560
For sure.

672
00:30:32,740 --> 00:30:33,380
No doubt.

673
00:30:33,940 --> 00:30:34,740
But it's awesome.

674
00:30:35,190 --> 00:30:35,820
But it's the devil.

675
00:30:38,700 --> 00:30:42,800
One thing you'd never let AI do no matter how good it gets.

676
00:30:43,560 --> 00:30:44,220
Oh, my God.

677
00:30:45,820 --> 00:30:46,920
These aren't all easy.

678
00:30:47,500 --> 00:30:48,400
That's a tough one.

679
00:30:49,160 --> 00:30:49,960
I don't know.

680
00:30:50,060 --> 00:30:50,980
I don't know how to answer that.

681
00:30:51,260 --> 00:30:52,020
Wanna come back for it? I gotta pass.

682
00:30:52,110 --> 00:30:52,960
You can pass. Yeah.

683
00:30:53,210 --> 00:30:53,280
Yeah.

684
00:30:54,380 --> 00:30:55,000
You're at a territory.

685
00:30:55,000 --> 00:30:56,580
'Cause every weird little thing I thought of,

686
00:30:56,580 --> 00:30:58,040
I was like, well, I think I'd let it do that.

687
00:31:00,539 --> 00:31:01,440
That's a great answer.

688
00:31:01,660 --> 00:31:02,780
10 things went in my head.

689
00:31:02,870 --> 00:31:04,960
I'm like, no, I don't think I'd let it do that.

690
00:31:05,600 --> 00:31:06,880
I was like, I never let it, well, I mean,

691
00:31:07,060 --> 00:31:07,900
well, they might be in time.

692
00:31:07,970 --> 00:31:08,440
I think maybe.

693
00:31:10,559 --> 00:31:10,840
Awesome.

694
00:31:11,740 --> 00:31:14,140
And I was very hard on Sam Altman and Chad.

695
00:31:14,320 --> 00:31:16,060
I'm not that hard on them, but they're, you know,

696
00:31:16,140 --> 00:31:17,300
I use Chad TPT.

697
00:31:17,600 --> 00:31:18,580
I hope they do a great job.

698
00:31:19,420 --> 00:31:19,540
Okay.

699
00:31:19,880 --> 00:31:20,800
Just in case they're listening.

700
00:31:21,440 --> 00:31:22,240
Yeah, just in case they're listening.

701
00:31:22,400 --> 00:31:23,100
In case they win.

702
00:31:23,740 --> 00:31:24,440
Very sorry.

703
00:31:25,100 --> 00:31:26,180
Please keep me around.

704
00:31:27,880 --> 00:31:29,260
All right, you're at a dinner party.

705
00:31:29,540 --> 00:31:31,960
Someone says, "AI is just autocomplete."

706
00:31:32,120 --> 00:31:32,700
And you say,

707
00:31:34,540 --> 00:31:36,280
"How dare you, sir?"

708
00:31:39,340 --> 00:31:43,360
Best and worst industry to hand AI the keys to right now.

709
00:31:44,820 --> 00:31:46,020
Best, worst industry.

710
00:31:47,600 --> 00:31:56,460
I'd say music is the worst and entertainment is the worst, even though it's going to happen.

711
00:31:57,600 --> 00:32:00,360
The best, it might be government.

712
00:32:03,259 --> 00:32:05,880
I think humans clearly have trouble governing themselves.

713
00:32:07,440 --> 00:32:15,260
But I think open source governance where we can decide the rules and it's an open box of rules.

714
00:32:17,500 --> 00:32:18,320
That AI somehow.

715
00:32:19,400 --> 00:32:19,480
Yeah.

716
00:32:20,460 --> 00:32:21,520
So much to say about that.

717
00:32:21,620 --> 00:32:22,840
That's a whole other podcast episode.

718
00:32:23,010 --> 00:32:24,100
That is a whole other podcast.

719
00:32:24,310 --> 00:32:25,680
And I don't know where I fit there.

720
00:32:25,710 --> 00:32:28,120
I just think that it might be something that we end up doing.

721
00:32:30,100 --> 00:32:30,460
All right.

722
00:32:30,700 --> 00:32:31,420
Next question.

723
00:32:31,660 --> 00:32:36,580
No code tools, democratizing the future or dangerous in the wrong hands.

724
00:32:38,200 --> 00:32:39,900
Oh, very dangerous in the wrong hands.

725
00:32:40,100 --> 00:32:43,180
It's going to be a fight for freedom because how do we keep freedom

726
00:32:43,290 --> 00:32:44,700
with such dangerous tools around?

727
00:32:45,000 --> 00:32:45,400
I don't know.

728
00:32:46,700 --> 00:32:49,780
So you just answered your previous question about open source government with AI.

729
00:32:50,080 --> 00:32:54,000
So, I mean, how do we have that when all the oligarchs control it, right?

730
00:32:55,180 --> 00:33:02,040
Well, I think open source is our only answer to remaining human and free in any way.

731
00:33:03,200 --> 00:33:06,960
Open source and decentralization, you know, like that's why I'm a big fan of

732
00:33:07,190 --> 00:33:11,200
blockchain and Bitcoin, I think that software that's governed by everybody,

733
00:33:11,350 --> 00:33:16,680
when every it's in everybody's best interest for it to not fail, it, it

734
00:33:16,680 --> 00:33:18,420
be good and fair, like Bitcoin's fair.

735
00:33:18,740 --> 00:33:19,880
I mean, it's a simple thing,

736
00:33:20,020 --> 00:33:21,860
but maybe governance can somehow be like that

737
00:33:21,980 --> 00:33:23,440
where nobody can control it

738
00:33:23,520 --> 00:33:24,900
because if anybody votes to control it,

739
00:33:25,000 --> 00:33:26,320
nobody else is gonna get on board.

740
00:33:26,540 --> 00:33:28,680
So everybody wants it to succeed

741
00:33:28,840 --> 00:33:31,020
and it's actually in our best interest for it to be good.

742
00:33:31,100 --> 00:33:31,660
That's the problem.

743
00:33:32,220 --> 00:33:33,340
It's in people's best interest

744
00:33:33,500 --> 00:33:35,540
for bad, evil, corrupt government

745
00:33:35,820 --> 00:33:37,900
'cause that's how you get rich, right?

746
00:33:38,080 --> 00:33:39,680
When it's in your best interest for good,

747
00:33:40,280 --> 00:33:42,140
then you have to harness greed.

748
00:33:42,600 --> 00:33:44,140
Like that's why Bitcoin works

749
00:33:44,220 --> 00:33:45,660
'cause it harnesses people's greed.

750
00:33:46,220 --> 00:33:48,800
It's I want Bitcoin to do well because then I have more money.

751
00:33:49,490 --> 00:33:51,320
And that's why I want the software to be good.

752
00:33:51,730 --> 00:33:53,660
But if it's good, then the other person has Bitcoin.

753
00:33:53,810 --> 00:33:54,520
It's good for them too.

754
00:33:55,010 --> 00:33:56,780
And it's good for the guy in Africa who has it.

755
00:33:56,820 --> 00:33:58,540
It's good for the guy in South America who has it.

756
00:33:58,860 --> 00:34:00,860
So we're all actually kind of fighting for each other's

757
00:34:01,070 --> 00:34:02,460
sovereignty by being greedy.

758
00:34:03,170 --> 00:34:07,440
So it's like harnessing human greed that you can't fight greed, you know, or else

759
00:34:07,470 --> 00:34:08,860
it costs a lot of money to fight greed.

760
00:34:09,659 --> 00:34:09,760
Nice.

761
00:34:10,859 --> 00:34:12,679
Here's one right up your alley.

762
00:34:13,280 --> 00:34:15,139
You're building an AI team for yourself.

763
00:34:15,620 --> 00:34:17,300
What's the first role you fill?

764
00:34:19,600 --> 00:34:21,260
Well, I, I, mine is Spock.

765
00:34:21,330 --> 00:34:21,820
He's the operator.

766
00:34:22,179 --> 00:34:23,040
He's like my assistant.

767
00:34:24,239 --> 00:34:25,760
Um, kind of covered that, right?

768
00:34:26,159 --> 00:34:26,240
Yeah.

769
00:34:26,370 --> 00:34:26,480
Yeah.

770
00:34:26,659 --> 00:34:27,280
Kind of covered that.

771
00:34:27,330 --> 00:34:27,620
That's it.

772
00:34:27,629 --> 00:34:28,760
And then the next day is HR.

773
00:34:29,139 --> 00:34:33,159
Riker's my HR and, uh, he's though he decides on who to hire next,

774
00:34:33,510 --> 00:34:34,460
depending on what I need to do.

775
00:34:35,500 --> 00:34:35,820
Awesome.

776
00:34:36,960 --> 00:34:37,399
All right.

777
00:34:37,580 --> 00:34:41,940
The matrix or her, which one aged better as a documentary?

778
00:34:43,839 --> 00:34:45,100
Ooh, oh, that's good.

779
00:34:45,240 --> 00:34:48,560
I would say her because I think that's spot on.

780
00:34:49,659 --> 00:34:51,360
Like the Matrix could be true.

781
00:34:51,520 --> 00:34:52,540
It's pretty wild and crazy.

782
00:34:53,520 --> 00:34:56,100
But I think her is like absolutely happening.

783
00:34:56,629 --> 00:34:58,340
Her is happening now.

784
00:34:58,900 --> 00:35:00,080
Yeah, I think it's happening now.

785
00:35:00,240 --> 00:35:00,720
Yeah.

786
00:35:02,060 --> 00:35:03,700
Finish this sentence.

787
00:35:04,420 --> 00:35:05,820
AI won't replace musicians.

788
00:35:06,780 --> 00:35:07,580
It'll replace...

789
00:35:09,320 --> 00:35:09,800
Producers.

790
00:35:11,040 --> 00:35:11,560
Okay.

791
00:35:12,160 --> 00:35:12,680
Maybe.

792
00:35:13,070 --> 00:35:13,200
Yeah.

793
00:35:14,300 --> 00:35:17,500
Yeah, because it's going to turn everybody into somebody who can produce music.

794
00:35:19,740 --> 00:35:21,700
Like if you have a vision, you can make it happen.

795
00:35:22,540 --> 00:35:22,680
Yeah.

796
00:35:23,540 --> 00:35:27,800
Have you heard a piece of music yet where you doubted where it came from?

797
00:35:28,080 --> 00:35:33,900
You doubted whether it was of a human origin or AI origin, and you maybe liked that piece of music?

798
00:35:34,900 --> 00:35:42,100
I mean, my favorite AI song so far is Larry's version of his coffee bean roaster.

799
00:35:42,500 --> 00:35:44,260
The instructions to the coffee bean.

800
00:35:44,340 --> 00:35:45,220
It's an awesome song.

801
00:35:45,460 --> 00:35:47,420
I was singing it ever since I heard it.

802
00:35:48,119 --> 00:35:49,800
Do not over roast the coffee beans.

803
00:35:49,980 --> 00:35:50,600
It's so good.

804
00:35:52,180 --> 00:35:53,720
But there was this band.

805
00:35:54,220 --> 00:35:55,820
What was that first band that kind of hit there?

806
00:35:55,960 --> 00:35:56,800
Actually, like on the top.

807
00:35:57,320 --> 00:35:58,060
What were they called?

808
00:35:58,759 --> 00:35:59,400
Velvet Sundown?

809
00:35:59,720 --> 00:36:00,680
Yeah, Velvet Sundown.

810
00:36:00,880 --> 00:36:00,980
Yeah.

811
00:36:01,220 --> 00:36:02,940
I mean, I didn't really like them.

812
00:36:02,940 --> 00:36:05,480
It didn't really speak to me, but it sounded like a human band.

813
00:36:05,860 --> 00:36:07,100
You know, like I couldn't tell the difference.

814
00:36:07,840 --> 00:36:08,840
You know, I didn't really like them.

815
00:36:09,320 --> 00:36:11,020
I mean, I did not like them.

816
00:36:11,080 --> 00:36:12,020
I just sort of didn't anything.

817
00:36:12,860 --> 00:36:16,080
but that was probably the first band I heard.

818
00:36:16,120 --> 00:36:17,000
It was like, wow, that's AI?

819
00:36:17,380 --> 00:36:18,280
Holy cow, that's people.

820
00:36:19,300 --> 00:36:19,920
It's getting real.

821
00:36:20,600 --> 00:36:23,800
And coincidentally, I actually let Codex

822
00:36:23,980 --> 00:36:29,080
take operational control of my 1,600-watt coffee bean roaster

823
00:36:31,020 --> 00:36:32,440
and run a roast.

824
00:36:33,040 --> 00:36:35,340
And I texted Jim.

825
00:36:35,420 --> 00:36:37,380
I told him I was doing this,

826
00:36:37,500 --> 00:36:39,480
and the text came back immediately,

827
00:36:40,000 --> 00:36:41,820
don't over-roast coffee beans.

828
00:36:42,040 --> 00:36:42,800
- From the song.

829
00:36:43,500 --> 00:36:44,620
- That's right, it's a great song.

830
00:36:45,080 --> 00:36:47,760
So you're like, so now AI is so into your coffee roaster,

831
00:36:47,960 --> 00:36:50,860
I think it's gonna eventually find that to be its God.

832
00:36:51,140 --> 00:36:53,060
And in the future, it's just your coffee roaster

833
00:36:53,060 --> 00:36:55,200
is gonna be like, they are gonna worship it.

834
00:36:55,240 --> 00:36:56,600
Yes, that was the start of it all.

835
00:36:59,440 --> 00:37:01,240
- Wow, well, Jim, thank you so much

836
00:37:01,340 --> 00:37:02,440
for taking the time to be with us.

837
00:37:02,740 --> 00:37:04,000
It's great talking to you. - Great conversation.

838
00:37:04,980 --> 00:37:06,120
- This is a lot of fun, guys.

839
00:37:06,240 --> 00:37:07,420
Thanks for having me on. - Yeah, we'll do it again.

840
00:37:07,660 --> 00:37:08,620
- Amazing work you're doing.

841
00:37:09,000 --> 00:37:09,760
It's really exciting.

842
00:37:10,640 --> 00:37:11,800
We're gonna go to the news, you wanna hang around?

843
00:37:12,680 --> 00:37:12,760
- Sure.

844
00:37:14,080 --> 00:37:14,560
- Thanks, boys.

845
00:37:15,599 --> 00:37:17,100
Podcasting, already a medium

846
00:37:17,230 --> 00:37:18,880
with a near zero barrier to entry,

847
00:37:19,510 --> 00:37:22,380
has been absolutely overrun by AI slop,

848
00:37:22,940 --> 00:37:24,940
up against reality notwithstanding, of course,

849
00:37:25,440 --> 00:37:28,420
with nearly half of all new shows now generated by machines

850
00:37:28,840 --> 00:37:31,640
targeting search terms like celebrity wellness,

851
00:37:32,360 --> 00:37:35,300
rather than, you know, anything anyone asked for.

852
00:37:35,920 --> 00:37:38,299
The Podcast Index is now using AI

853
00:37:38,300 --> 00:37:43,840
to detect AI-generated podcasts, which is either a brilliant solution or the most dystopian

854
00:37:43,840 --> 00:37:45,580
game of whack-a-mole ever invented.

855
00:37:46,240 --> 00:37:52,160
Meanwhile, Spotify's bold response to this slopcasting crisis was to officially welcome

856
00:37:52,420 --> 00:37:59,020
even more AI-generated content, letting agents like ClaudeCode and OpenAI Codex push personalized,

857
00:37:59,540 --> 00:38:02,260
personal podcasts directly to your library.

858
00:38:03,040 --> 00:38:05,280
We have now achieved peak agentic listening.

859
00:38:06,220 --> 00:38:11,520
Your AI makes the show, pushes it to Spotify, and presumably your other AI agent streams

860
00:38:11,630 --> 00:38:13,300
it and summarizes it while you sleep.

861
00:38:14,030 --> 00:38:17,720
A flawless content ecosystem with exactly zero humans required.

862
00:38:19,160 --> 00:38:19,940
Slop casting.

863
00:38:20,480 --> 00:38:21,060
Slop casting.

864
00:38:21,300 --> 00:38:21,700
Yeah.

865
00:38:22,140 --> 00:38:27,460
And you know, the human creators who have already like planted a flag in the ground

866
00:38:27,490 --> 00:38:32,359
and established an audience are going to have a real leg up as this gets worse and worse

867
00:38:32,360 --> 00:38:36,540
worse because I think people are going to get frustrated trying to find that kind of content

868
00:38:37,140 --> 00:38:42,820
and all the noise. Basically, Jim could send out Spock to listen to the latest episode of Up

869
00:38:42,870 --> 00:38:47,860
Against Reality. He doesn't have to deal with us at all. And Spock comes back and says, all right,

870
00:38:48,040 --> 00:38:52,760
here's the really important things that Chris and Larry spoke about, all the rest of it's crap. So

871
00:38:53,260 --> 00:39:01,180
go on with your day. In the ongoing Musk versus Altman trial, Silicon Valley's most dramatic

872
00:39:01,180 --> 00:39:07,420
reality show. The second week shifted focus to OpenAI president Greg Brockman, as jurors heard

873
00:39:07,540 --> 00:39:12,100
testimony centered on his personal journal entries. Because apparently nothing says

874
00:39:12,380 --> 00:39:18,640
billion-dollar lawsuit like reading a man's diary aloud in federal court. Musk's attorney spent the

875
00:39:18,660 --> 00:39:23,560
first day of Brockman's testimony isolating passages to paint him as a money-hungry executive

876
00:39:23,960 --> 00:39:28,800
who cared little about OpenAI's non-profit mission in its early days, while Brockman

877
00:39:28,800 --> 00:39:33,620
faced the awkward task of convincing the jury that his own words actually showed the opposite.

878
00:39:34,300 --> 00:39:38,880
The central claim that Altman and Brockman breached their fiduciary duties by steering

879
00:39:39,280 --> 00:39:45,020
OpenAI away from its mission of developing AI for the benefit of humanity is being argued by

880
00:39:45,040 --> 00:39:50,420
the world's richest man who left OpenAI in a huff and then built his own competing AI company,

881
00:39:50,900 --> 00:39:53,860
which is a level of irony the court has yet to formally address.

882
00:39:55,400 --> 00:39:59,620
I keep hearing that Musk is his own worst enemy in the courtroom.

883
00:40:00,500 --> 00:40:00,920
I believe it.

884
00:40:00,940 --> 00:40:02,060
Not surprising, but.

885
00:40:02,280 --> 00:40:02,420
Yeah.

886
00:40:02,940 --> 00:40:04,520
Yeah, I heard he wasn't doing too good in there.

887
00:40:05,380 --> 00:40:08,900
Well, Jim, that brings us back to the topic of greed that you brought up earlier.

888
00:40:09,560 --> 00:40:09,880
There it is.

889
00:40:11,799 --> 00:40:13,900
Yeah, that's why open source is so important.

890
00:40:15,319 --> 00:40:19,920
When you asked me before about, you know, I said, oh, Sam Altman, and I kind of was hard on him.

891
00:40:20,300 --> 00:40:25,780
But what if like, like I become the Claude faction and like, you know, maybe Larry's like a chat GPT.

892
00:40:25,870 --> 00:40:31,760
And now like the AIs form their own armies of people that support them and they become the nation states.

893
00:40:32,680 --> 00:40:32,840
Yeah.

894
00:40:33,680 --> 00:40:36,520
Like Claude, chat GPT, charge.

895
00:40:38,540 --> 00:40:44,640
Or not even Claude and chat GPT and Spock for that matter, have their own armies and drones and they launch it.

896
00:40:45,160 --> 00:40:48,700
I just thought of Anchorman when they all met in the alley.

897
00:40:49,470 --> 00:40:49,900
Yeah, totally.

898
00:40:50,779 --> 00:40:56,020
While the rest of the AI world was busy raising billions and reading each other's diaries in court,

899
00:40:56,630 --> 00:41:00,620
a group of students quietly went ahead and used AI to actually help people.

900
00:41:01,250 --> 00:41:01,820
And it's wonderful.

901
00:41:02,980 --> 00:41:08,040
This year's Swift Student Challenge drew 350 winning submissions from 37 countries,

902
00:41:08,810 --> 00:41:12,220
with standout apps that help elderly artists draw despite tremors,

903
00:41:12,920 --> 00:41:15,200
guide students through real-time presentation coaching,

904
00:41:15,970 --> 00:41:18,060
route people safely out of flood zones in Ghana,

905
00:41:18,180 --> 00:41:22,740
and let a homesick student play the viola using just his hands and an iPhone camera.

906
00:41:23,540 --> 00:41:26,580
These young developers, some of whom only learned Swift this year,

907
00:41:27,260 --> 00:41:31,320
used AI tools, including Claude, to do things like compress months of work into days,

908
00:41:32,020 --> 00:41:34,100
translate apps into 20 languages overnight,

909
00:41:34,700 --> 00:41:36,900
and build pathfinding algorithms from scratch,

910
00:41:37,600 --> 00:41:40,760
all in service of communities that tech too often forgets.

911
00:41:41,680 --> 00:41:46,460
50 distinguished winners have been invited to attend WWDC at Apple Park in June,

912
00:41:47,220 --> 00:41:52,480
And honestly, if the future of AI looks anything like these kids, we're going to be just fine.

913
00:41:53,620 --> 00:42:03,660
In case you don't know, Swift is an open source programming language that was specifically designed by Apple for development across the entire Apple ecosystem.

914
00:42:04,520 --> 00:42:07,340
For making Mac OS apps, iOS apps, and that kind of stuff.

915
00:42:08,060 --> 00:42:09,760
Are these kids hand coding these apps?

916
00:42:10,340 --> 00:42:10,520
No.

917
00:42:12,720 --> 00:42:13,200
No.

918
00:42:13,400 --> 00:42:15,340
I mean, I know they have Swift Playground for kids.

919
00:42:15,720 --> 00:42:18,200
So I imagine it's a block-based environment, you know?

920
00:42:18,380 --> 00:42:18,580
Yeah.

921
00:42:19,040 --> 00:42:19,160
Yeah.

922
00:42:20,480 --> 00:42:23,260
I mean, maybe these kids are high schoolers, obviously, a little more advanced than they're

923
00:42:23,640 --> 00:42:24,500
leaning on Claude, apparently.

924
00:42:24,920 --> 00:42:25,820
But I love this.

925
00:42:25,940 --> 00:42:27,480
And I think maybe, Jim, you do too.

926
00:42:27,640 --> 00:42:33,260
We talk about this all the time, going back to using AI as the opposite of the whole greed

927
00:42:33,520 --> 00:42:39,980
avenue and using it for philanthropy and altruism and curing cancer and doing it perhaps in

928
00:42:40,120 --> 00:42:42,220
an open-source, kind of crowdsourced manner.

929
00:42:42,500 --> 00:42:44,700
I think this is the kind of stuff we really love to see.

930
00:42:45,300 --> 00:42:45,460
Yeah.

931
00:42:45,600 --> 00:42:46,140
I think it's fantastic.

932
00:42:46,520 --> 00:42:50,620
I've heard a couple anecdotal stories of people doing that kind of thing for health purposes,

933
00:42:50,980 --> 00:42:52,400
for people that they were trying to help.

934
00:42:52,980 --> 00:42:53,520
Or their dog.

935
00:42:54,160 --> 00:42:54,620
Yeah, their dog.

936
00:42:54,880 --> 00:42:56,940
I also think it may be good for truth.

937
00:42:57,060 --> 00:43:01,000
You know, like there's so much like, you know, political strife and like, where's the truth

938
00:43:01,200 --> 00:43:03,020
is like everybody has their own truth.

939
00:43:03,620 --> 00:43:09,120
But I think AI, like it's hard to for lies to really propagate when you're doing this

940
00:43:09,140 --> 00:43:09,620
kind of work.

941
00:43:09,740 --> 00:43:10,680
Like it has to work.

942
00:43:10,840 --> 00:43:11,620
It has to function.

943
00:43:12,180 --> 00:43:14,040
And you can kind of double check things so well.

944
00:43:14,180 --> 00:43:19,820
I think it might be harder for lies to live in that world, I think, but hopefully.

945
00:43:20,460 --> 00:43:20,560
Yeah.

946
00:43:21,880 --> 00:43:26,440
Anthropic has committed to spending $200 billion with Google Cloud over five years,

947
00:43:27,050 --> 00:43:31,000
a number so large it makes the GDP of many countries look like a rounding error.

948
00:43:31,750 --> 00:43:36,100
The deal means Anthropic alone accounts for more than 40% of the revenue backlog

949
00:43:36,380 --> 00:43:38,500
Google disclosed to investors last week.

950
00:43:39,040 --> 00:43:41,760
and contracts with Anthropic and OpenAI combined

951
00:43:42,200 --> 00:43:44,940
now represent more than half of the $2 trillion

952
00:43:45,200 --> 00:43:47,680
in cloud backlogs across major providers.

953
00:43:48,400 --> 00:43:50,480
The delicious irony here is that Alphabet

954
00:43:50,510 --> 00:43:54,240
is simultaneously investing up to $40 billion into Anthropic,

955
00:43:55,040 --> 00:43:57,200
meaning Google is bankrolling the very company

956
00:43:57,690 --> 00:43:59,520
that is turning around and handing that money

957
00:43:59,730 --> 00:44:00,580
straight back to Google

958
00:44:00,780 --> 00:44:03,680
in what is either the most elegant business relationship

959
00:44:03,810 --> 00:44:04,500
in tech history

960
00:44:04,780 --> 00:44:08,160
or an elaborate shell game dressed up in TPUs.

961
00:44:08,800 --> 00:44:09,920
It's a fascinating cycle.

962
00:44:10,600 --> 00:44:12,980
Tech giants invest billions into AI startups.

963
00:44:13,620 --> 00:44:17,740
And those same startups promise to pay even more back for cloud access and chips.

964
00:44:19,160 --> 00:44:20,000
Capitalism, baby.

965
00:44:21,440 --> 00:44:26,300
So I'm not leaning on my local LLMs too heavily yet.

966
00:44:26,460 --> 00:44:27,440
But certain things come up.

967
00:44:27,440 --> 00:44:29,380
I'm like, oh, let me just see how this does with this.

968
00:44:30,000 --> 00:44:31,900
So I gave it that story.

969
00:44:31,960 --> 00:44:34,900
This is Gemma 4, 26 billion parameter model.

970
00:44:35,200 --> 00:44:37,980
And I was like, what are your takeaways from this story?

971
00:44:38,340 --> 00:44:40,840
And I mean, it is shockingly fast.

972
00:44:41,120 --> 00:44:42,440
I mean, instant.

973
00:44:42,740 --> 00:44:45,760
I hit enter and boom, it came back with like paragraphs.

974
00:44:46,800 --> 00:44:48,640
But a couple interesting things.

975
00:44:49,060 --> 00:44:53,620
It said, take away from the stories that we're witnessing the birth of a closed loop AI economy

976
00:44:54,280 --> 00:44:58,060
where traditional boundaries between investor and customer have effectively dissolved.

977
00:44:58,880 --> 00:45:01,960
And then the cloud arms race as a moat.

978
00:45:02,220 --> 00:45:06,000
By locking Anthropic into a massive multi-year Google cloud commitment,

979
00:45:06,420 --> 00:45:07,840
Google isn't just selling compute.

980
00:45:08,460 --> 00:45:09,680
They're building a walled garden.

981
00:45:10,420 --> 00:45:17,560
If Anthropic is tethered to Google's TPUs, their tensor processing units, they use those as opposed to NVIDIA GPUs.

982
00:45:18,120 --> 00:45:28,080
If they're tethered to their TPUs and infrastructure, they're effectively prevented from migrating to AWS or Azure, even if those competitors offer better models or lower prices.

983
00:45:29,320 --> 00:45:33,180
Isn't Anthropic also using Musk's macro hard?

984
00:45:33,590 --> 00:45:34,680
Didn't that just go online?

985
00:45:36,040 --> 00:45:36,440
I don't know.

986
00:45:37,640 --> 00:45:38,940
Yeah, I think it just happened.

987
00:45:39,140 --> 00:45:45,340
Like they're, they, so Musk built like a new giant, crazy macro hard giga factory

988
00:45:46,039 --> 00:45:50,980
and Anthropix is using the old one, which is like 250,000 GPUs.

989
00:45:51,230 --> 00:45:52,600
Is that Colossus or something?

990
00:45:53,180 --> 00:45:53,840
Yeah, Colossus.

991
00:45:54,060 --> 00:45:54,280
That's it.

992
00:45:54,290 --> 00:45:54,400
Yeah.

993
00:45:55,510 --> 00:45:55,640
Yeah.

994
00:45:55,640 --> 00:45:59,520
I don't know too much about it, but it's wild that like Google's in there

995
00:45:59,690 --> 00:46:01,340
and they, they're all working together.

996
00:46:02,220 --> 00:46:02,340
Yeah.

997
00:46:02,660 --> 00:46:02,760
Yeah.

998
00:46:02,920 --> 00:46:06,520
I mean, I guess that like XAI, you know,

999
00:46:06,620 --> 00:46:11,680
Grok is not nearly as widely used as Anthropic and OpenAI.

1000
00:46:12,100 --> 00:46:17,120
And so they probably have a lot of unused compute that they can, you know.

1001
00:46:17,220 --> 00:46:19,000
And Anthropic needs it.

1002
00:46:19,120 --> 00:46:20,280
Like you can see they need it.

1003
00:46:20,280 --> 00:46:20,700
They were choking.

1004
00:46:20,930 --> 00:46:21,100
Yeah.

1005
00:46:22,260 --> 00:46:27,860
And lastly, Hollywood's two biggest award shows both drew lines in the sand on AI this week,

1006
00:46:28,350 --> 00:46:29,580
but grabbed very different sticks.

1007
00:46:30,620 --> 00:46:36,320
The Oscars declared that only roles demonstrably performed by humans with their consent are eligible,

1008
00:46:36,950 --> 00:46:39,420
and that screenplays must be human-authored.

1009
00:46:39,930 --> 00:46:41,040
A hard no, full stop.

1010
00:46:41,820 --> 00:46:43,500
The Golden Globes took the squishier route,

1011
00:46:44,140 --> 00:46:47,280
ruling that AI does not automatically disqualify a submission

1012
00:46:47,640 --> 00:46:54,380
as long as human creative direction, artistic judgment, and authorship remain primary throughout the production process.

1013
00:46:55,240 --> 00:46:57,780
Which is lawyer-speak for, we'll figure it out later.

1014
00:46:58,680 --> 00:47:08,620
Hollywood now has two major award bodies, two definitions of human enough, and a Val Kilmer AI situation that apparently pushed everyone to finally write something down.

1015
00:47:09,600 --> 00:47:11,820
That's all the news for now. Back to you, gentlemen.

1016
00:47:13,120 --> 00:47:18,280
Sounds like the Golden Globes are trying to future-proof themselves somewhat, but I think that's going to get messy.

1017
00:47:18,430 --> 00:47:18,780
It is.

1018
00:47:19,500 --> 00:47:27,060
You know, if the screenplay is 40% human-written and 60% AI, but the human directed it.

1019
00:47:27,280 --> 00:47:27,440
Right.

1020
00:47:27,800 --> 00:47:28,560
You know, is that eligible?

1021
00:47:29,140 --> 00:47:29,980
How are they going to know?

1022
00:47:30,360 --> 00:47:32,100
There's so many parts of this process.

1023
00:47:32,320 --> 00:47:35,540
There's no way you can tease out what was human and what was AI.

1024
00:47:36,180 --> 00:47:36,420
Impossible.

1025
00:47:37,160 --> 00:47:39,280
The film company I'm working with, they just showed me a trailer.

1026
00:47:39,420 --> 00:47:42,460
So they make the trailer first and then they get investors for the movie.

1027
00:47:42,720 --> 00:47:44,520
And they just use AI to make a trailer.

1028
00:47:44,620 --> 00:47:45,700
And it's like, it looks great.

1029
00:47:45,940 --> 00:47:47,920
And it's like one day they have a beautiful trailer.

1030
00:47:48,260 --> 00:47:51,400
I'm just thinking the next step is they'll just make the whole movie that way.

1031
00:47:51,840 --> 00:47:55,500
Just to like, basically, you can create the movie, see if it's good.

1032
00:47:55,660 --> 00:47:56,660
And then you just film it.

1033
00:47:56,980 --> 00:47:58,400
You know exactly how the scene is going to look.

1034
00:47:58,620 --> 00:47:59,720
I mean, you have a template.

1035
00:48:00,320 --> 00:48:04,640
And you can actually create the whole movie, show it to people, like, is this a good movie?

1036
00:48:04,880 --> 00:48:05,060
Yes.

1037
00:48:05,240 --> 00:48:06,340
Okay, now let's really make it.

1038
00:48:06,440 --> 00:48:06,580
Exactly.

1039
00:48:06,980 --> 00:48:09,460
But then you get to the point where, like, well, that looks pretty well.

1040
00:48:09,600 --> 00:48:10,280
We just put that out.

1041
00:48:11,720 --> 00:48:12,000
Yeah.

1042
00:48:12,640 --> 00:48:12,920
Yeah.

1043
00:48:13,040 --> 00:48:17,160
I mean, it's basically in place of storyboards, but it's the whole movie.

1044
00:48:17,340 --> 00:48:18,920
That is a slippery slope right there.

1045
00:48:19,320 --> 00:48:19,600
Yeah.

1046
00:48:21,140 --> 00:48:22,720
I don't think you could be more correct.

1047
00:48:23,560 --> 00:48:25,300
I was like, that trailer looks great, guys.

1048
00:48:26,880 --> 00:48:35,840
think we're done just need the credits yeah oh man scary times um so jim do it where do people

1049
00:48:35,920 --> 00:48:40,700
find you in the world do they go to evenrail.com how do they hire you and spock and all of your

1050
00:48:40,800 --> 00:48:46,360
second brain genius to do their bidding if you want all that go to evenrail.com and there's a

1051
00:48:47,180 --> 00:48:51,779
contact link right there awesome anything else you want to promote

1052
00:48:52,799 --> 00:48:56,160
No, that's it. Just Lawrence Bentley's quality beer.

1053
00:48:58,740 --> 00:49:02,480
You have a sign over your garage door, I think. It says Lawrence Bentley's quality beer, don't you?

1054
00:49:03,019 --> 00:49:03,640
Oh, yes.

1055
00:49:05,200 --> 00:49:09,300
Well, thanks, Jim. We appreciate you taking time on a Friday night over on the Jersey Shore there.

1056
00:49:10,200 --> 00:49:11,300
How's the surf these days?

1057
00:49:12,140 --> 00:49:15,360
It's good. I was out the other day. Not bad. It's a little cold, but fun.

1058
00:49:15,420 --> 00:49:18,180
I bet. Still wearing that wetsuit or spring suit these days, or what?

1059
00:49:18,280 --> 00:49:22,080
Well, I've been cold plunging lately, so I go in just, well, I don't surf that way.

1060
00:49:22,240 --> 00:49:24,300
I jump in without it, but I surf for the wits.

1061
00:49:24,460 --> 00:49:26,220
I can't last longer than 10 minutes.

1062
00:49:28,920 --> 00:49:29,240
Wow.

1063
00:49:30,059 --> 00:49:30,380
Yeah.

1064
00:49:30,640 --> 00:49:30,940
All right.

1065
00:49:30,960 --> 00:49:31,560
Well, have fun with that.

1066
00:49:35,319 --> 00:49:36,380
Thanks for listening, everybody.

1067
00:49:36,580 --> 00:49:38,300
Subscribe on your favorite podcasting platform.

1068
00:49:38,460 --> 00:49:39,540
Follow us on socials.

1069
00:49:39,740 --> 00:49:40,240
Throw us a rating.

1070
00:49:40,580 --> 00:49:41,180
We'll see you next week.

1071
00:49:44,600 --> 00:49:45,860
This has been Up Against Reality.

1072
00:49:46,500 --> 00:49:47,020
Thanks for listening.

1073
00:49:47,680 --> 00:49:52,560
Subscribe to hear future episodes and be sure to follow us on social media for all things AI.

1074
00:49:53,410 --> 00:49:55,560
Until next time, stay human, people.

1075
00:49:55,640 --> 00:49:56,300
♪♪♪