1
00:00:02,400 --> 00:00:03,560
Speaker 1: All Media.

2
00:00:04,559 --> 00:00:07,400
Speaker 2: Hello on, Welcome to Better Offline. I am, of course

3
00:00:07,440 --> 00:00:22,880
Speaker 2: your host ed zitron. Now we're in hate a season.

4
00:00:22,880 --> 00:00:25,040
Speaker 2: Hate a season is when I bring people on allowed

5
00:00:25,079 --> 00:00:27,880
Speaker 2: to get rude, because you know, this podcast is usually

6
00:00:27,920 --> 00:00:30,960
Speaker 2: so reserved and nice in the way we refer to people,

7
00:00:31,040 --> 00:00:34,120
Speaker 2: especially tech executives, which we've never referred to as funck

8
00:00:34,159 --> 00:00:37,839
Speaker 2: wits or morons or dumbasses or shittheads or shit heels

9
00:00:37,920 --> 00:00:41,160
Speaker 2: or ass wipes or fuck nuts or nob ends. We've

10
00:00:41,200 --> 00:00:44,240
Speaker 2: never done any of that ever. Nevertheless, joining me today

11
00:00:44,280 --> 00:00:47,080
Speaker 2: is chief cloud economist that Doug Bell Corey. Quinn Corey,

12
00:00:47,159 --> 00:00:47,800
Speaker 2: how are you doing?

13
00:00:48,280 --> 00:00:50,720
Speaker 1: I have delighted to be here because normally in all

14
00:00:50,760 --> 00:00:53,040
Speaker 1: the other places I find myself, I have to be

15
00:00:53,120 --> 00:00:53,840
Speaker 1: so reserved.

16
00:00:54,480 --> 00:00:59,440
Speaker 2: You have to pull your punches. Cory, today's Hatter season.

17
00:00:59,680 --> 00:01:02,720
Speaker 2: Tell me about your feelings on Amazon Web services.

18
00:01:03,360 --> 00:01:07,320
Speaker 1: Oh, dear lord, it's it's like it's weird. I don't

19
00:01:07,360 --> 00:01:10,280
Speaker 1: actually hate Amazon. People think I do because of the

20
00:01:10,319 --> 00:01:12,160
Speaker 1: things I say and the things that I do. But

21
00:01:12,200 --> 00:01:15,000
Speaker 1: if I hate it, then this directions, but ten years

22
00:01:15,040 --> 00:01:17,360
Speaker 1: in I write a newsletter about it. If I actually

23
00:01:17,400 --> 00:01:19,640
Speaker 1: hate the company and want nothing but ill things to

24
00:01:19,680 --> 00:01:22,080
Speaker 1: happen to them. First off, that's a pathology and I

25
00:01:22,120 --> 00:01:25,160
Speaker 1: need a restraining order. And secondly, it always bugs me

26
00:01:25,200 --> 00:01:27,200
Speaker 1: when people think that I have an ax to grind

27
00:01:27,280 --> 00:01:30,679
Speaker 1: against the company, because ten years in, are you seriously

28
00:01:30,760 --> 00:01:33,319
Speaker 1: suggesting this is the best I would be able to

29
00:01:33,400 --> 00:01:35,919
Speaker 1: do if I actually wanted to hit them where it hurts?

30
00:01:36,040 --> 00:01:36,399
Speaker 2: Come on?

31
00:01:37,000 --> 00:01:39,000
Speaker 1: But yeah, they've been annoying the piss out of me lately.

32
00:01:39,480 --> 00:01:42,520
Speaker 2: So AWS is weird because maybe you can explain this

33
00:01:42,600 --> 00:01:46,480
Speaker 2: to me. Over my entire tech career, I've heard this

34
00:01:46,680 --> 00:01:50,000
Speaker 2: constant story that your AWS bill just it just goes up.

35
00:01:50,440 --> 00:01:53,640
Speaker 2: It just arbitrarily increases. It's not really clear whether they're

36
00:01:53,640 --> 00:01:56,920
Speaker 2: doing price increases, though I know those exist too.

37
00:01:56,680 --> 00:01:59,200
Speaker 1: Rarely open but occasionally increasingly these days.

38
00:01:59,320 --> 00:02:02,400
Speaker 2: Yes, but why what is it that makes the bill

39
00:02:02,480 --> 00:02:04,720
Speaker 2: go up? Why? And why does this happen so often?

40
00:02:05,160 --> 00:02:07,240
Speaker 1: I used to think it was a conspiracy, and then

41
00:02:07,280 --> 00:02:09,480
Speaker 1: I realized Amazon is in no way, shape or form

42
00:02:09,560 --> 00:02:12,600
Speaker 1: organizationally competent enough to pull something like that off. It

43
00:02:12,680 --> 00:02:17,519
Speaker 1: is the nature of charging per usage and honestly, assholes

44
00:02:17,800 --> 00:02:20,359
Speaker 1: where Okay, we're going to launch a service and we're

45
00:02:20,360 --> 00:02:22,920
Speaker 1: going to charge you for every gigabyte you put in it.

46
00:02:22,919 --> 00:02:25,880
Speaker 1: That's all we're gonna do. Great, straightforward humans can rationalize

47
00:02:25,919 --> 00:02:28,520
Speaker 1: around that, great eazy psy. They did that when they

48
00:02:28,560 --> 00:02:31,200
Speaker 1: first launched S three and beta back in two thousand

49
00:02:31,200 --> 00:02:33,760
Speaker 1: and three. Is storage right, it's a big object store

50
00:02:33,840 --> 00:02:36,880
Speaker 1: picture it as an infinite sized disc and you're pretty close.

51
00:02:37,200 --> 00:02:40,239
Speaker 1: But they would that Originally they only charged for the

52
00:02:40,320 --> 00:02:43,320
Speaker 1: data you stored there. Great, so people stored a bunch

53
00:02:43,360 --> 00:02:46,720
Speaker 1: of objects that were zero byte in size, so they

54
00:02:46,840 --> 00:02:49,800
Speaker 1: just used the names of those objects, and the ability

55
00:02:49,840 --> 00:02:51,640
Speaker 1: to query with that object existed or not in what

56
00:02:51,680 --> 00:02:55,680
Speaker 1: its name was. As a free database, this at the time, also,

57
00:02:55,800 --> 00:02:58,600
Speaker 1: according to legend, didn't just it wasn't just about taking

58
00:02:58,600 --> 00:03:02,120
Speaker 1: free services. It was also just crippling the performance of

59
00:03:02,360 --> 00:03:04,200
Speaker 1: S three at the time. So what they did is

60
00:03:04,240 --> 00:03:07,959
Speaker 1: they started also charging every request you make with jarj

61
00:03:08,040 --> 00:03:10,200
Speaker 1: a small fee. I think it's something like a penny

62
00:03:10,240 --> 00:03:12,840
Speaker 1: per thousand requests of memory served. Don't book me on that.

63
00:03:12,880 --> 00:03:15,760
Speaker 1: I look it up when it matters, and that solves

64
00:03:15,760 --> 00:03:18,280
Speaker 1: that problem. But then you have things like that that

65
00:03:18,320 --> 00:03:20,680
Speaker 1: are patched on and tied to things, and tied to

66
00:03:20,720 --> 00:03:24,600
Speaker 1: things and tied to more things, and in any complex organization.

67
00:03:25,080 --> 00:03:28,280
Speaker 1: What happens is is now suddenly you have this morass

68
00:03:28,480 --> 00:03:31,400
Speaker 1: of one hundred million dollars a year and spent great,

69
00:03:31,639 --> 00:03:33,720
Speaker 1: what's in there that's costing all the money?

70
00:03:33,840 --> 00:03:34,080
Speaker 2: Well?

71
00:03:34,520 --> 00:03:37,520
Speaker 1: Everything? But if you look at anything individual and you

72
00:03:37,600 --> 00:03:40,400
Speaker 1: don't think it's being used, Okay, am I turn I'm

73
00:03:40,400 --> 00:03:42,839
Speaker 1: going to turn it off. If I'm right, it's going

74
00:03:42,880 --> 00:03:44,880
Speaker 1: to save me a bit of money. If I'm wrong,

75
00:03:44,920 --> 00:03:47,280
Speaker 1: it's going to take down production. And now I have

76
00:03:47,360 --> 00:03:50,520
Speaker 1: a serious problem trying to serve my customers. So the

77
00:03:50,640 --> 00:03:56,120
Speaker 1: safe bias is don't turn things off. Also, you're an engineer. Hypothetically,

78
00:03:56,160 --> 00:03:57,880
Speaker 1: I realize that can be an insult to your part

79
00:03:57,920 --> 00:04:00,920
Speaker 1: of the world. Roll with it for a moment. Exactly,

80
00:04:01,040 --> 00:04:03,680
Speaker 1: you're an engineer. You want something to spin up, and

81
00:04:03,720 --> 00:04:05,880
Speaker 1: no one is letting you spin it up. You will

82
00:04:05,920 --> 00:04:09,040
Speaker 1: take up residents, annoying the hell out of them every

83
00:04:09,080 --> 00:04:11,680
Speaker 1: ten minutes until you get the access that you need.

84
00:04:12,720 --> 00:04:15,560
Speaker 1: You do your job. Things are done. Two things now

85
00:04:15,560 --> 00:04:17,960
Speaker 1: work against you. One, you had to ask and bag

86
00:04:18,000 --> 00:04:19,680
Speaker 1: and plead them to spin that up. You want to

87
00:04:19,680 --> 00:04:21,080
Speaker 1: hang on to it in case you need it again

88
00:04:21,080 --> 00:04:23,040
Speaker 1: because you don't want to go through that. But this

89
00:04:23,200 --> 00:04:27,520
Speaker 1: is an organization exactly exactly. This is not Amazon's fault.

90
00:04:27,560 --> 00:04:29,880
Speaker 1: This is human nature's fault. They haven't patched that yet,

91
00:04:29,920 --> 00:04:32,240
Speaker 1: though they're I think they're trying their best iteration of

92
00:04:32,279 --> 00:04:35,360
Speaker 1: a Laxa plus it. Why not, I'm kidding that thing

93
00:04:35,400 --> 00:04:38,080
Speaker 1: just sells ads. We're good boy way.

94
00:04:38,120 --> 00:04:41,120
Speaker 2: So I'm an engineer and I've advocated for my thing.

95
00:04:41,920 --> 00:04:44,599
Speaker 1: Yes, and now you're done using it. No one is going,

96
00:04:44,640 --> 00:04:47,000
Speaker 1: there's no there's no call to action to get you

97
00:04:47,080 --> 00:04:48,640
Speaker 1: off your ass to go, and no one off or

98
00:04:48,640 --> 00:04:50,520
Speaker 1: no one will come and badger you about it. It

99
00:04:50,640 --> 00:04:54,120
Speaker 1: just it tends to exists. It's why people cynically say

100
00:04:54,120 --> 00:04:55,719
Speaker 1: you're not charged for the things you use in the

101
00:04:55,720 --> 00:04:57,479
Speaker 1: clouds so much as you're charged for the things you

102
00:04:57,520 --> 00:04:58,360
Speaker 1: forget to turn off.

103
00:04:58,880 --> 00:05:01,240
Speaker 2: And Amazon loves this. I'm guessing they love the fact

104
00:05:01,240 --> 00:05:04,200
Speaker 2: that there's just this cloud anchor system.

105
00:05:05,160 --> 00:05:08,000
Speaker 1: I would say that, on some level, contrary to popular belief,

106
00:05:08,160 --> 00:05:10,440
Speaker 1: they don't love it as much as you'd think, because

107
00:05:10,520 --> 00:05:12,880
Speaker 1: all right, I'm trying to run on my blog and

108
00:05:12,920 --> 00:05:15,040
Speaker 1: I've spun up eight attempts at this blog, so it

109
00:05:15,040 --> 00:05:18,279
Speaker 1: now costs eight times what it should. Great and the

110
00:05:18,400 --> 00:05:20,880
Speaker 1: narrative becomes if they're not careful with all that wasted

111
00:05:20,960 --> 00:05:24,760
Speaker 1: usage that's doing nothing is huh, the cloud is pants

112
00:05:24,760 --> 00:05:27,840
Speaker 1: shittingly expensive. We should go back to data centers. There's

113
00:05:27,880 --> 00:05:30,760
Speaker 1: really no one on the other side of this issue.

114
00:05:30,839 --> 00:05:33,640
Speaker 1: So why doesn't Amazon help people turn tune these things

115
00:05:33,680 --> 00:05:35,760
Speaker 1: down and turn it off? They made some a board

116
00:05:35,800 --> 00:05:38,080
Speaker 1: of attempts at part of the organization. But this is

117
00:05:38,120 --> 00:05:41,360
Speaker 1: going to surprise you. Amazon is not the most organizationally

118
00:05:41,520 --> 00:05:43,440
Speaker 1: competent company you're ever going to meet.

119
00:05:43,880 --> 00:05:46,000
Speaker 2: But what does that mean in practice? Is it just

120
00:05:46,080 --> 00:05:48,360
Speaker 2: that they don't know they're ass from the air hole.

121
00:05:48,440 --> 00:05:51,640
Speaker 2: They don't really know how things work. All of the above.

122
00:05:51,800 --> 00:05:54,839
Speaker 1: Silos, it's all silos. They're over two hundred services. And

123
00:05:54,880 --> 00:05:56,960
Speaker 1: we long ago across the point where I can speak

124
00:05:57,040 --> 00:06:01,520
Speaker 1: incredibly convincingly about AWS service is that do not exist

125
00:06:01,800 --> 00:06:05,160
Speaker 1: two Amazon employees and not get called out on it

126
00:06:05,200 --> 00:06:08,200
Speaker 1: because who in the world knows everything that Amazon does?

127
00:06:08,320 --> 00:06:10,720
Speaker 1: AND's just th guy, I sound as plausible as AI

128
00:06:10,839 --> 00:06:11,360
Speaker 1: these days.

129
00:06:11,640 --> 00:06:16,600
Speaker 2: Yeah, So it's just a labyrinth, the Leviathan of different

130
00:06:17,080 --> 00:06:18,960
Speaker 2: SKUs and product categories that.

131
00:06:18,960 --> 00:06:22,880
Speaker 1: Make million SKUs. If you want to be precise, Yes, Jesus.

132
00:06:22,600 --> 00:06:25,360
Speaker 2: And those are all just different kinds of virtual machines

133
00:06:25,400 --> 00:06:25,880
Speaker 2: and the like.

134
00:06:26,680 --> 00:06:29,320
Speaker 1: Okay, we'll start there. Sure, virtual machines. You want to

135
00:06:29,320 --> 00:06:32,839
Speaker 1: spin up a virtual machine in Virginia, US East one? Okay,

136
00:06:32,880 --> 00:06:34,840
Speaker 1: which one are you going to pick? Because there are

137
00:06:34,920 --> 00:06:38,600
Speaker 1: approximately seven hundred different types you can spin up. Are

138
00:06:38,600 --> 00:06:40,440
Speaker 1: you going to pick the right one? Of course you're

139
00:06:40,480 --> 00:06:40,840
Speaker 1: fucking not.

140
00:06:41,720 --> 00:06:44,599
Speaker 2: Oh so how do you pick the right one? This

141
00:06:44,760 --> 00:06:48,200
Speaker 2: is the thing because it seems that the people making

142
00:06:48,240 --> 00:06:51,039
Speaker 2: these decisions of which virtual machines to get or whatever

143
00:06:51,279 --> 00:06:54,960
Speaker 2: spin up on aws or engineers, right, Usually sulfware engineers

144
00:06:55,440 --> 00:06:58,600
Speaker 2: software engineers aren't necessarily infrastructure experts, are they.

145
00:06:58,760 --> 00:07:02,520
Speaker 1: You've noticed very often companies will standardize on one particular

146
00:07:02,560 --> 00:07:05,000
Speaker 1: size of instance across the board. Why well, because in

147
00:07:05,040 --> 00:07:07,360
Speaker 1: development it's the one that the original founding engineer picked,

148
00:07:07,360 --> 00:07:09,760
Speaker 1: and we'll come back and revisit this later ten years ago.

149
00:07:10,080 --> 00:07:11,080
Speaker 1: No one is revisited.

150
00:07:11,520 --> 00:07:12,520
Speaker 2: M take that.

151
00:07:13,000 --> 00:07:15,320
Speaker 1: They do have a tool that I like, it's even free,

152
00:07:15,360 --> 00:07:18,160
Speaker 1: which I'm sure someone loses sleepover called compute optimizer that

153
00:07:18,200 --> 00:07:20,200
Speaker 1: looks at you're running workloads and says, hey, this one

154
00:07:20,200 --> 00:07:22,840
Speaker 1: should be bigger, this one should be smaller. I was

155
00:07:22,880 --> 00:07:26,240
Speaker 1: extraordinarily skeptical when it came out. It is better at

156
00:07:26,240 --> 00:07:28,120
Speaker 1: figuring that out than I am with a tooling that

157
00:07:28,160 --> 00:07:30,480
Speaker 1: I built internally. So once again I'm thrilled to turn

158
00:07:30,480 --> 00:07:33,760
Speaker 1: it off. It's one of the Amazon at its best things.

159
00:07:33,920 --> 00:07:36,600
Speaker 1: I can only assume that the people responsible for that

160
00:07:36,680 --> 00:07:38,240
Speaker 1: are being eyed for layoffs or.

161
00:07:38,160 --> 00:07:43,880
Speaker 2: Pips, because Andy Jesse and several NBA snipers are outside

162
00:07:43,920 --> 00:07:45,360
Speaker 2: at the location right now.

163
00:07:45,960 --> 00:07:51,400
Speaker 1: It's frust stay, but I like AWS truly. Something changed.

164
00:07:51,440 --> 00:07:52,960
Speaker 1: He got a job that no one in their right

165
00:07:53,000 --> 00:07:56,120
Speaker 1: mind could possibly want but also could not refuse. And

166
00:07:56,240 --> 00:07:58,560
Speaker 1: I feel like I sould a soul. Yes, absolutely, but

167
00:07:58,680 --> 00:08:00,480
Speaker 1: he just he was the CEO of a double us.

168
00:08:00,800 --> 00:08:03,640
Speaker 1: And he was he when he gave keynote talks, when

169
00:08:03,680 --> 00:08:07,160
Speaker 1: he gave presentations, when I was fortunate enough to basically

170
00:08:07,200 --> 00:08:09,840
Speaker 1: sneak my way into them, you could see the irrepressible

171
00:08:09,880 --> 00:08:12,800
Speaker 1: humanity leaking out around the edges. He was a person.

172
00:08:13,440 --> 00:08:16,560
Speaker 1: Now he is a figurehead. He doesn't want to customer,

173
00:08:16,640 --> 00:08:17,720
Speaker 1: just to be clear to Congress.

174
00:08:18,440 --> 00:08:21,560
Speaker 2: This shift was when he became CEO of Amazon or

175
00:08:21,640 --> 00:08:22,800
Speaker 2: CEO of a WS.

176
00:08:22,920 --> 00:08:26,080
Speaker 1: CEO of Amazon. AWS was his own Fifdom, and he

177
00:08:26,120 --> 00:08:30,280
Speaker 1: could do whatever he wanted to my understanding within that company,

178
00:08:30,360 --> 00:08:33,080
Speaker 1: it was the only people that mattered, Brother Garrett who

179
00:08:33,080 --> 00:08:35,280
Speaker 1: had an opinion on this were Andy Jassey and God.

180
00:08:35,320 --> 00:08:37,240
Speaker 1: And God was sort of absence, so great, We're just

181
00:08:37,240 --> 00:08:39,480
Speaker 1: gonna leave it to Andy. Andy went deep and he

182
00:08:39,559 --> 00:08:42,320
Speaker 1: understood every aspect of this stuff. Factly, a year ago

183
00:08:42,400 --> 00:08:44,320
Speaker 1: he went to Reinvent in Las Vegas and gave a

184
00:08:44,360 --> 00:08:45,080
Speaker 1: talk on stage.

185
00:08:45,200 --> 00:08:46,160
Speaker 2: I think because he missed it.

186
00:08:46,200 --> 00:08:49,120
Speaker 1: He missed being able to get up there and talk

187
00:08:49,200 --> 00:08:51,920
Speaker 1: about computers with people because now he has to ship

188
00:08:51,960 --> 00:08:53,440
Speaker 1: underpants all over the world instead.

189
00:08:54,080 --> 00:08:57,080
Speaker 2: Right, But is he technical because he's an MBA? Is

190
00:08:57,080 --> 00:08:57,320
Speaker 2: he not?

191
00:08:58,320 --> 00:09:00,400
Speaker 1: This is one of the dangerous things that I have

192
00:09:00,480 --> 00:09:04,960
Speaker 1: found people underestimate Amazon leadership on is it is never

193
00:09:05,000 --> 00:09:07,520
Speaker 1: the smart bet to assume that they don't understand the

194
00:09:07,559 --> 00:09:10,640
Speaker 1: technical nuances of the things that they are working with.

195
00:09:11,120 --> 00:09:14,959
Speaker 1: They go so sarcastically deep in so many different ways

196
00:09:15,320 --> 00:09:17,440
Speaker 1: that it is never a smart move to say that

197
00:09:17,440 --> 00:09:20,800
Speaker 1: they don't understand it is unlike virtually any other tech

198
00:09:20,800 --> 00:09:23,320
Speaker 1: company I've ever spoken with, other than small scale ones

199
00:09:23,320 --> 00:09:27,760
Speaker 1: with technical founders, he very clearly understands the technology and

200
00:09:27,840 --> 00:09:29,040
Speaker 1: used it to build things himself.

201
00:09:29,559 --> 00:09:34,400
Speaker 2: Right, But here's the reason I question that. Please, he's

202
00:09:34,400 --> 00:09:38,600
Speaker 2: talking about general IVAI. I've just pulled up his notice.

203
00:09:38,720 --> 00:09:40,440
Speaker 1: Oh no, one understands how that works under the hood.

204
00:09:40,520 --> 00:09:42,920
Speaker 1: Let's be very clear on that. And yeah, he lost

205
00:09:42,960 --> 00:09:44,880
Speaker 1: the fucking plot when it came to gen AI.

206
00:09:45,040 --> 00:09:47,400
Speaker 2: That's the thing, because he's like, we're also using GENERATIVEAI

207
00:09:47,480 --> 00:09:51,520
Speaker 2: broadly across our internal operations in our fulfillment network. We're

208
00:09:51,600 --> 00:09:54,680
Speaker 2: using AI to improve in ventur replacement. So not generate IVEAI.

209
00:09:54,880 --> 00:09:58,200
Speaker 2: Demand forecasting definitely not generative AI. And the efficiency of

210
00:09:58,200 --> 00:10:00,760
Speaker 2: our robots not generative AI. I don't know, maybe he

211
00:10:00,800 --> 00:10:02,560
Speaker 2: does know what he's talking about. He's just a fucking

212
00:10:02,600 --> 00:10:04,040
Speaker 2: carnival barker now he's.

213
00:10:03,920 --> 00:10:06,040
Speaker 1: Doing to say that it all seems to be. And

214
00:10:06,080 --> 00:10:09,800
Speaker 1: the trick to disprove everything you just said about that,

215
00:10:10,000 --> 00:10:12,000
Speaker 1: rather everything that he said that you relate to us

216
00:10:12,120 --> 00:10:15,760
Speaker 1: is very simple. Go out with an Amazon employee for

217
00:10:15,960 --> 00:10:19,600
Speaker 1: several drinks and then ask them about their experience of

218
00:10:19,760 --> 00:10:23,760
Speaker 1: using AI within the company. Oh, I have two different companies.

219
00:10:24,280 --> 00:10:28,120
Speaker 1: There is no there is no congruity between some of

220
00:10:28,200 --> 00:10:29,239
Speaker 1: those two perspectives.

221
00:10:29,280 --> 00:10:32,360
Speaker 2: So Corey, maybe you can help me with this because you, certainly,

222
00:10:32,400 --> 00:10:35,360
Speaker 2: without saying exactly what we're talking about, you knew about

223
00:10:35,360 --> 00:10:37,800
Speaker 2: a contract that I knew about way before I knew,

224
00:10:37,840 --> 00:10:40,640
Speaker 2: and I knew about it months early. Yes, with a

225
00:10:40,679 --> 00:10:44,320
Speaker 2: large company we discussed, so you're very well connected. It

226
00:10:44,480 --> 00:10:48,480
Speaker 2: seems up until jeneralf Ai, like AWS was at least

227
00:10:48,520 --> 00:10:52,600
Speaker 2: a somewhat sensible operation, that it was something that we had,

228
00:10:53,240 --> 00:10:55,760
Speaker 2: if not focus, at least a focus on scale and

229
00:10:56,040 --> 00:10:59,200
Speaker 2: useful cloud computing. Yes, what the fuck is it about

230
00:10:59,240 --> 00:11:01,080
Speaker 2: GPUs that's having everyone insane?

231
00:11:01,280 --> 00:11:03,800
Speaker 1: Because I want to point something out that I think

232
00:11:03,880 --> 00:11:06,640
Speaker 1: is being lost in the streetwide because I've talked to

233
00:11:06,679 --> 00:11:10,400
Speaker 1: folks at AWS about this, and folks who have left AWS,

234
00:11:10,760 --> 00:11:13,160
Speaker 1: who have I used to dabble working in data centers,

235
00:11:13,160 --> 00:11:14,920
Speaker 1: which means I know just enough to know what I

236
00:11:14,960 --> 00:11:17,960
Speaker 1: don't know, and I've talked to these folks. When Amazon

237
00:11:18,040 --> 00:11:20,959
Speaker 1: commits to being able to bring certain amounts of capacity online,

238
00:11:21,000 --> 00:11:24,400
Speaker 1: it is real capacity. It is each data center they

239
00:11:24,440 --> 00:11:26,240
Speaker 1: build out, each region they build out, is a multi

240
00:11:26,240 --> 00:11:29,320
Speaker 1: billion dollar investment, and it takes years to go from

241
00:11:29,480 --> 00:11:33,319
Speaker 1: sign contract to its serving customer traffic. They are real

242
00:11:33,440 --> 00:11:36,120
Speaker 1: when it comes to this stuff. There are a number

243
00:11:36,120 --> 00:11:38,920
Speaker 1: of other companies building well, yes, and they agree that

244
00:11:39,080 --> 00:11:41,600
Speaker 1: running it. They have operationalized how to run these things

245
00:11:41,640 --> 00:11:43,920
Speaker 1: better than you or I would ever be able to do.

246
00:11:44,000 --> 00:11:47,960
Speaker 1: That is their superpower. Other companies and they're spinning up

247
00:11:47,960 --> 00:11:49,959
Speaker 1: a data center, do not do this. They are, well,

248
00:11:49,960 --> 00:11:52,560
Speaker 1: we've we've rented a warehouse out, we round up, throwing

249
00:11:52,640 --> 00:11:55,319
Speaker 1: some generators in the back and were check. Call Comcast

250
00:11:55,400 --> 00:11:56,720
Speaker 1: and see what they can do to get it hooked

251
00:11:56,800 --> 00:12:00,319
Speaker 1: up to the internet netney to go. Yeah, you say

252
00:12:00,320 --> 00:12:02,880
Speaker 1: other companies, You don't mean Google and Microsoft here, No,

253
00:12:03,080 --> 00:12:06,400
Speaker 1: I'm taught, well, I am talking Oracle, but yes, Microsoft

254
00:12:06,679 --> 00:12:11,240
Speaker 1: kind of Microsoft. Microsoft is a shitty at building data centers.

255
00:12:11,280 --> 00:12:14,880
Speaker 1: Azure has been a disappointment across the board. Capacity shortballs

256
00:12:14,880 --> 00:12:18,640
Speaker 1: plagued them for a long time during COVID. They generally

257
00:12:18,640 --> 00:12:20,960
Speaker 1: tend to view two racks in a data center somewhere

258
00:12:21,000 --> 00:12:24,439
Speaker 1: as a region, which is weird and messed up. Google

259
00:12:24,600 --> 00:12:27,920
Speaker 1: is real, but Google has made other interesting technical trade

260
00:12:27,960 --> 00:12:31,480
Speaker 1: offs that I can see why they did it. But

261
00:12:31,520 --> 00:12:33,959
Speaker 1: I would say when it comes to infrastructure, in my experience,

262
00:12:34,000 --> 00:12:35,200
Speaker 1: AWS is number one.

263
00:12:35,520 --> 00:12:37,720
Speaker 2: Google is a close second. I have heard some crazy

264
00:12:37,760 --> 00:12:40,600
Speaker 2: shit about Google that they're doing these things called stalks

265
00:12:41,360 --> 00:12:47,920
Speaker 2: where they just like like create these pop up data centers.

266
00:12:46,559 --> 00:12:48,079
Speaker 1: For a long time. Every company does that for a

267
00:12:48,120 --> 00:12:49,240
Speaker 1: point centering stuff.

268
00:12:49,320 --> 00:12:51,880
Speaker 2: Yes, anyway, but back to the major question, which is

269
00:12:52,320 --> 00:12:56,199
Speaker 2: AWS seems like a sensible business that does real things.

270
00:12:56,240 --> 00:12:57,160
Speaker 2: I'm not disputing that.

271
00:12:57,640 --> 00:13:01,280
Speaker 1: Yes, it prints money. Let's be clear on despite all, remember,

272
00:13:01,640 --> 00:13:05,719
Speaker 1: I help companies negotiating money cloud contracts and understand what

273
00:13:05,760 --> 00:13:08,120
Speaker 1: they're spending and where the money is going, which means

274
00:13:08,120 --> 00:13:11,240
Speaker 1: that companies can say a lot. People say a lot

275
00:13:11,240 --> 00:13:13,280
Speaker 1: of things. But one of my prime rules about this

276
00:13:13,320 --> 00:13:17,760
Speaker 1: stuff is that customers lie mostly to themselves. They're not

277
00:13:17,880 --> 00:13:20,360
Speaker 1: intentionally trying to mislead me when they tell me what's

278
00:13:20,400 --> 00:13:22,959
Speaker 1: going on. But I find the bills of the ultimate

279
00:13:23,000 --> 00:13:24,920
Speaker 1: source of truth, because if you're not paying for it

280
00:13:24,920 --> 00:13:28,040
Speaker 1: doesn't really exist. And the AI spend in most of

281
00:13:28,080 --> 00:13:31,239
Speaker 1: these companies hovers, in my experience with a few outliers

282
00:13:31,480 --> 00:13:36,280
Speaker 1: between five and seven percent of their total infrastructure spend. Right,

283
00:13:36,440 --> 00:13:39,320
Speaker 1: it is not three times. Companies that are doing a

284
00:13:39,360 --> 00:13:42,559
Speaker 1: three hundred million dollars a year contract with AWS are

285
00:13:42,600 --> 00:13:45,120
Speaker 1: not turning around saying better make it four hundred because

286
00:13:45,120 --> 00:13:47,679
Speaker 1: of all the GENAI that is not happening.

287
00:13:58,520 --> 00:14:01,240
Speaker 2: That's kind of my point though. It's like Amazon Web

288
00:14:01,280 --> 00:14:03,839
Speaker 2: Services doesn't seem like a business that chases fats, or

289
00:14:03,840 --> 00:14:06,880
Speaker 2: if it does chase them in a more sustainable way.

290
00:14:07,000 --> 00:14:07,800
Speaker 2: They now they's credit.

291
00:14:07,840 --> 00:14:10,599
Speaker 1: They docked they Dutch blockchain almost entirely, and I was

292
00:14:10,640 --> 00:14:12,320
Speaker 1: worried they were going to fall down that rabbit hole.

293
00:14:12,760 --> 00:14:15,600
Speaker 2: But I mean, they doing two hundred billion dollars in

294
00:14:15,679 --> 00:14:18,200
Speaker 2: capex this year. What is it they see in AI?

295
00:14:18,240 --> 00:14:20,680
Speaker 2: Because to your point, just now, what like seven to

296
00:14:20,720 --> 00:14:23,560
Speaker 2: eight percent of a three hundred million contract, that's dog

297
00:14:23,640 --> 00:14:26,520
Speaker 2: shit like that's that's not going to pipe out that

298
00:14:26,560 --> 00:14:27,760
Speaker 2: Capex Sandy.

299
00:14:27,960 --> 00:14:31,040
Speaker 1: I believe this is my belief here. I can't base

300
00:14:31,080 --> 00:14:33,200
Speaker 1: it on anything other than what I've seen Amazon do,

301
00:14:33,840 --> 00:14:36,960
Speaker 1: uh I, and what I've seen from customers. The Amazon

302
00:14:37,120 --> 00:14:40,760
Speaker 1: is selling the capacity that they are spinning up. They

303
00:14:40,800 --> 00:14:43,000
Speaker 1: are spinning up as fast as they can spin it.

304
00:14:43,400 --> 00:14:46,800
Speaker 1: The question that I have is twofold one. Okay, you

305
00:14:47,080 --> 00:14:49,320
Speaker 1: invest in all these AI data centers in the AI

306
00:14:49,400 --> 00:14:53,440
Speaker 1: bubble pops to flates whatever, how repurposable are those facilities,

307
00:14:53,480 --> 00:14:56,600
Speaker 1: and that capital expenditure if it's data centers and networking, Yes,

308
00:14:56,800 --> 00:14:59,840
Speaker 1: very repurposable. Yeah, GPUs, well, we're all going to get

309
00:14:59,840 --> 00:15:01,840
Speaker 1: real in online gaming in a few years. I don't

310
00:15:01,840 --> 00:15:02,400
Speaker 1: except they.

311
00:15:02,320 --> 00:15:05,440
Speaker 2: Don't have video out, they don't have DisplayPort.

312
00:15:05,600 --> 00:15:08,000
Speaker 1: Yes, yeah, we still call them graphics, which I love.

313
00:15:08,240 --> 00:15:11,640
Speaker 2: I think I like the people are I've also heard

314
00:15:11,640 --> 00:15:14,400
Speaker 2: people say general purpose units, which is the opposite of

315
00:15:14,400 --> 00:15:16,640
Speaker 2: what a GPU is like. It is very much not

316
00:15:16,760 --> 00:15:18,320
Speaker 2: general purpose. It was.

317
00:15:18,320 --> 00:15:20,280
Speaker 1: It became public through a press release that they have

318
00:15:20,360 --> 00:15:24,080
Speaker 1: signed a thirty eight billion dollar contract with open AI

319
00:15:24,160 --> 00:15:28,560
Speaker 1: with insane. There is zero doubt in my mind that,

320
00:15:28,760 --> 00:15:32,160
Speaker 1: because I did some digging on this, is there line

321
00:15:32,200 --> 00:15:36,080
Speaker 1: of sight for AWS to bring all on line thirty

322
00:15:36,080 --> 00:15:41,120
Speaker 1: eight billion dollars of capacity to provide to open AI. Yes, yes,

323
00:15:41,160 --> 00:15:44,280
Speaker 1: there is where my doubts come in. Is open AI

324
00:15:44,400 --> 00:15:45,280
Speaker 1: good for the money?

325
00:15:45,680 --> 00:15:45,800
Speaker 3: No?

326
00:15:45,880 --> 00:15:48,440
Speaker 1: I have some questions here, mostly from reading your work.

327
00:15:48,800 --> 00:15:53,560
Speaker 2: And that's the thing. It seems like something about GPUs

328
00:15:53,600 --> 00:15:56,560
Speaker 2: and large language models have broken everyone. That's kind of

329
00:15:56,600 --> 00:15:58,720
Speaker 2: the point I'm getting too, because like everything I've read

330
00:15:58,720 --> 00:16:01,120
Speaker 2: about AWS, and I've read because I'm a dickhead, I

331
00:16:01,200 --> 00:16:04,280
Speaker 2: went and I read all of the previous coverage of

332
00:16:04,320 --> 00:16:08,320
Speaker 2: the lead up to AWS becoming profitable. Kevin Rust, by

333
00:16:08,360 --> 00:16:10,400
Speaker 2: the way, literally a month before they did so, wrote

334
00:16:10,400 --> 00:16:12,800
Speaker 2: an article saying it would be a while. What fucking

335
00:16:12,800 --> 00:16:16,960
Speaker 2: three points in that ass wipe? Anyway, everything I read

336
00:16:17,040 --> 00:16:20,360
Speaker 2: was everyone's saying, oh, Amazon's being really spendy. They're spending

337
00:16:20,560 --> 00:16:24,760
Speaker 2: sixty seven billion in kapex a quarter. Oh no, and

338
00:16:24,760 --> 00:16:27,320
Speaker 2: then the total capex of sixty nine billion. But everything

339
00:16:27,320 --> 00:16:29,560
Speaker 2: from Amazon was like, we're building something sustainable, we see

340
00:16:29,560 --> 00:16:33,240
Speaker 2: the revenue potential behind it. It was all very boring. Now

341
00:16:33,280 --> 00:16:35,800
Speaker 2: with AWS, everything they talk about with large language models,

342
00:16:35,800 --> 00:16:38,720
Speaker 2: it sounds like they've all been huffing ayawasco or something.

343
00:16:39,080 --> 00:16:41,360
Speaker 2: They sound insane they.

344
00:16:41,200 --> 00:16:43,760
Speaker 1: Do, which tells me one of two things is true

345
00:16:43,880 --> 00:16:47,600
Speaker 1: or honestly, this is the real world. Unlike on Twitter,

346
00:16:48,080 --> 00:16:50,680
Speaker 1: two complex competing things can in fact be true at

347
00:16:50,680 --> 00:16:53,320
Speaker 1: the same interest time. One of them is they're still

348
00:16:53,320 --> 00:16:55,960
Speaker 1: doing a lot of that boring stuff, but the publicity

349
00:16:56,080 --> 00:16:59,120
Speaker 1: all accrues to the things that are far flung. This

350
00:16:59,200 --> 00:17:01,760
Speaker 1: is also two hundred billion dollars across the entirety of

351
00:17:01,880 --> 00:17:06,560
Speaker 1: Amazon that includes distribution centers, that includes Low Earth Orbits

352
00:17:06,920 --> 00:17:10,119
Speaker 1: and basically Amazon Basic Starlink. I think they're calling it

353
00:17:10,200 --> 00:17:16,320
Speaker 1: Leo now. It includes yes, their AI nonsense, their factory boondoggles,

354
00:17:16,359 --> 00:17:20,160
Speaker 1: their robotic stuff, office buildouts, et cetera, et cetera, et cetera.

355
00:17:20,440 --> 00:17:22,800
Speaker 1: They don't give clear breakdowns on how much of that

356
00:17:22,920 --> 00:17:25,320
Speaker 1: is AI services, and frankly, I wouldn't trust them if

357
00:17:25,359 --> 00:17:27,879
Speaker 1: they did not, since I saw a job ad a

358
00:17:27,880 --> 00:17:32,119
Speaker 1: few years back on Amazon dot Jobs saying experience with

359
00:17:32,160 --> 00:17:38,080
Speaker 1: other AI services like S three storage service, that's not AI.

360
00:17:38,480 --> 00:17:41,400
Speaker 2: You can use it to feed AI. But my hairanthropic

361
00:17:41,760 --> 00:17:44,440
Speaker 2: anthropics one of the largest S three customers I've heard.

362
00:17:44,960 --> 00:17:47,080
Speaker 1: I've heard that. I've heard from seven different companies so

363
00:17:47,119 --> 00:17:49,560
Speaker 1: far that they are the largest customer really three. Yeah,

364
00:17:49,560 --> 00:17:51,240
Speaker 1: that is funny how that tends to work out.

365
00:17:51,320 --> 00:17:54,800
Speaker 2: God, I miss I miss working in cloud software so much,

366
00:17:54,880 --> 00:17:58,280
Speaker 2: where everyone is the everyone's the exclusive partner, everyone's the

367
00:17:58,359 --> 00:17:59,960
Speaker 2: number one partner, and everyone's the larger.

368
00:18:00,600 --> 00:18:03,399
Speaker 4: It's so but the qualifiers get They're like, we're the

369
00:18:03,480 --> 00:18:06,639
Speaker 4: largest S three customers in Australia, great in the Northeast,

370
00:18:06,880 --> 00:18:09,440
Speaker 4: in the Northeast region between the hours of six am

371
00:18:09,480 --> 00:18:11,880
Speaker 4: and twelve am.

372
00:18:11,920 --> 00:18:14,959
Speaker 2: Okay, So here's the thing with the capex. I'm sure

373
00:18:15,000 --> 00:18:18,320
Speaker 2: a lot of it is for is for AI, just

374
00:18:18,359 --> 00:18:20,720
Speaker 2: because if you look at their previously as of Capex,

375
00:18:20,720 --> 00:18:24,840
Speaker 2: it's like four A fifty sixty Yeah, four A sixty

376
00:18:24,880 --> 00:18:28,200
Speaker 2: one sixty three fifty two eighty three one hundred and

377
00:18:28,240 --> 00:18:29,520
Speaker 2: thirty one two hundred billion.

378
00:18:30,560 --> 00:18:32,879
Speaker 1: Gerald Over at Platform and Nomics has been tracking capec

379
00:18:32,920 --> 00:18:35,439
Speaker 1: span to the hyperscalers for a long time now, and

380
00:18:35,800 --> 00:18:37,399
Speaker 1: he was one of the first people to call out

381
00:18:37,520 --> 00:18:41,560
Speaker 1: the cloud pretenders IBM and Oracle specifically, because they're making

382
00:18:41,600 --> 00:18:43,760
Speaker 1: all the cloud noises, but they're not spending anything on

383
00:18:43,880 --> 00:18:45,800
Speaker 1: Capex and you kind of have to build the data

384
00:18:45,800 --> 00:18:48,520
Speaker 1: centers for this to work. Not to worry, Oracle's taking

385
00:18:48,560 --> 00:18:51,040
Speaker 1: care of that. What I want to know, my question is,

386
00:18:51,040 --> 00:18:52,840
Speaker 1: when you start, okay, we're going to spend even assume

387
00:18:52,880 --> 00:18:55,639
Speaker 1: it's all AI whatever, fine, two hundred billion dollars on

388
00:18:55,720 --> 00:18:57,960
Speaker 1: AI spent, how much of that is just in the

389
00:18:57,960 --> 00:18:59,399
Speaker 1: form of a check written to Nvidio.

390
00:19:00,600 --> 00:19:06,320
Speaker 2: And that's the question because from my understanding, my understanding

391
00:19:06,880 --> 00:19:10,280
Speaker 2: the I should bring up the chart. I'm just gonna

392
00:19:10,320 --> 00:19:12,280
Speaker 2: do some live on air. I just bringing up an

393
00:19:12,320 --> 00:19:15,919
Speaker 2: image and of course can't find it. Nevertheless, there is

394
00:19:15,960 --> 00:19:19,320
Speaker 2: an image I have the Amazon doesn't. Well, Amazon builds

395
00:19:19,320 --> 00:19:22,119
Speaker 2: its own the trainium GPUs. And do they still use

396
00:19:22,160 --> 00:19:23,119
Speaker 2: inferentia or.

397
00:19:23,119 --> 00:19:26,000
Speaker 1: They That sounds like that that my grandpa got diagnosed with.

398
00:19:26,720 --> 00:19:29,040
Speaker 2: Yeah, but do they liked inferentia? I thought was there?

399
00:19:29,080 --> 00:19:30,760
Speaker 1: Oh, they talk about it a fair bit. They talk

400
00:19:30,800 --> 00:19:32,920
Speaker 1: about trainium. And the thing is is what you don't

401
00:19:32,920 --> 00:19:35,840
Speaker 1: see is people using it in the wild. I looked

402
00:19:35,880 --> 00:19:38,080
Speaker 1: at a bunch of stuff with OLAMA. That's the run

403
00:19:38,119 --> 00:19:40,200
Speaker 1: your own stuff there. I searched about a year ago

404
00:19:40,240 --> 00:19:42,320
Speaker 1: for the term inferentia. It showed up once in a

405
00:19:42,359 --> 00:19:44,840
Speaker 1: pull request that had been closed after sixty days because

406
00:19:44,840 --> 00:19:47,560
Speaker 1: no one responded to it. No one is using this

407
00:19:47,680 --> 00:19:51,200
Speaker 1: in the real world. A year ago year in the

408
00:19:51,280 --> 00:19:54,480
Speaker 1: month now they had with tranium two I think it was,

409
00:19:54,680 --> 00:20:00,280
Speaker 1: they had speakers from Anthropic on keynote stage talking about it, which, yeah, okay,

410
00:20:00,280 --> 00:20:03,320
Speaker 1: they invested how many tens of billions into Anthropic? Yeah,

411
00:20:03,400 --> 00:20:06,159
Speaker 1: I'm sure there's a contract requirement around to sending it

412
00:20:06,240 --> 00:20:08,600
Speaker 1: executive talk on stage. And the other one was Apple.

413
00:20:09,880 --> 00:20:12,800
Speaker 1: Apple doesn't admit what they're doing to their own employees,

414
00:20:12,960 --> 00:20:16,320
Speaker 1: to each other, let alone publicly. So yeah, that looked

415
00:20:16,320 --> 00:20:18,240
Speaker 1: about as natural as an oral bowel movement.

416
00:20:18,800 --> 00:20:19,920
Speaker 2: Nice, but those are.

417
00:20:19,880 --> 00:20:22,040
Speaker 1: The only two companies I've seen doing anything with it.

418
00:20:22,440 --> 00:20:25,959
Speaker 2: So based on this chart I've got, which involves various

419
00:20:26,000 --> 00:20:29,160
Speaker 2: stars and company names, it looks like Amazon gets their

420
00:20:29,240 --> 00:20:32,760
Speaker 2: GPUs through fox Con also known as On High Precision

421
00:20:33,840 --> 00:20:38,080
Speaker 2: Corporation Limited. I love Chutneyese and Taiwanese names. By theme

422
00:20:38,280 --> 00:20:41,879
Speaker 2: is almost certainly in there. No, TSMC isn't because builds

423
00:20:41,880 --> 00:20:42,320
Speaker 2: the chips.

424
00:20:42,760 --> 00:20:46,320
Speaker 1: These are the training ships themselves, because they have they

425
00:20:46,359 --> 00:20:48,440
Speaker 1: bought and do the design, but they don't do the fab.

426
00:20:48,680 --> 00:20:51,200
Speaker 2: So they ship the I assume that Anna Perna, which

427
00:20:51,200 --> 00:20:54,560
Speaker 2: is the internal place that Amazon makes its chips, ships them.

428
00:20:54,560 --> 00:20:58,119
Speaker 2: It looks like the fox Con, Quantum Computing and Jabil,

429
00:20:59,320 --> 00:21:02,000
Speaker 2: and then they put them in the service. So really

430
00:21:02,000 --> 00:21:04,080
Speaker 2: it's them cutting the checks to them. I'm going to

431
00:21:04,119 --> 00:21:07,040
Speaker 2: be watching the monthly earnings of those companies in Taiwan

432
00:21:07,480 --> 00:21:11,520
Speaker 2: quite viciously, because here's the thing. If it's all two

433
00:21:11,600 --> 00:21:15,120
Speaker 2: hundred billion dollars of capex, if they really think they're

434
00:21:15,160 --> 00:21:17,639
Speaker 2: going to get paid that much, but what happens. Has

435
00:21:17,680 --> 00:21:20,400
Speaker 2: Amazon ever faced the situation where they did and overbuild

436
00:21:20,400 --> 00:21:22,320
Speaker 2: Has that ever happened to them? What did they do.

437
00:21:22,680 --> 00:21:26,120
Speaker 1: Yes, they made a bunch of noises about it. During

438
00:21:26,119 --> 00:21:28,560
Speaker 1: the pandemic. They overbuilt their distribution centers.

439
00:21:29,119 --> 00:21:32,760
Speaker 2: Ah, but that's people buy stuff exactly.

440
00:21:32,800 --> 00:21:34,959
Speaker 1: That was my position on it. It feels like, Okay, they

441
00:21:35,000 --> 00:21:37,000
Speaker 1: slowed it down and they didn't shut them down, but

442
00:21:37,000 --> 00:21:38,920
Speaker 1: they waited for people to grow into it. The thing

443
00:21:39,000 --> 00:21:43,840
Speaker 1: is is a warehouse that you use to ship out underpants. Great,

444
00:21:44,640 --> 00:21:47,400
Speaker 1: that has the same utility more or less five years

445
00:21:47,400 --> 00:21:49,600
Speaker 1: from now as it does. Yes, exactly. A lot of

446
00:21:49,640 --> 00:21:53,280
Speaker 1: these chips are depreciating assets. M like, oh man, I

447
00:21:53,280 --> 00:21:55,159
Speaker 1: can't wait to get that seven year old computer in

448
00:21:55,200 --> 00:21:56,760
Speaker 1: here so I can do some real work, says no

449
00:21:56,800 --> 00:21:57,200
Speaker 1: one ever.

450
00:21:57,720 --> 00:22:00,439
Speaker 2: Well, also training them especially they've got well they are

451
00:22:00,480 --> 00:22:02,400
Speaker 2: on generation two three.

452
00:22:02,280 --> 00:22:05,320
Speaker 1: Now they're now shipping three, right, it's time to pre

453
00:22:05,359 --> 00:22:08,480
Speaker 1: announce four. And then a desperate attempt to Osbourne Computer themselves.

454
00:22:08,800 --> 00:22:10,840
Speaker 2: Yeah, explain that reference.

455
00:22:10,920 --> 00:22:11,080
Speaker 3: Oh.

456
00:22:11,200 --> 00:22:13,840
Speaker 1: Osbourne Computer was a company back in the eighties.

457
00:22:13,960 --> 00:22:14,200
Speaker 2: There.

458
00:22:14,240 --> 00:22:17,439
Speaker 1: They launched the Osbourne one and their CEO went out

459
00:22:17,440 --> 00:22:19,000
Speaker 1: and did a whole press joke and he's like, yeah,

460
00:22:19,080 --> 00:22:21,520
Speaker 1: this is great, but Osbourne two is gonna blow the

461
00:22:21,680 --> 00:22:24,840
Speaker 1: crap out of it. Oh, luck in video, and everyone's like, great,

462
00:22:25,000 --> 00:22:27,240
Speaker 1: we're not gonna buy the Osbourne one. Then the company

463
00:22:27,240 --> 00:22:28,919
Speaker 1: went under and never shipped the Osbourne two.

464
00:22:29,440 --> 00:22:31,720
Speaker 2: Well okay, not like in video. Then this is the

465
00:22:31,800 --> 00:22:35,600
Speaker 2: thing surely that sets Amazon up for a big problem

466
00:22:35,600 --> 00:22:38,760
Speaker 2: with those tradium chips success. Like even if in some

467
00:22:38,840 --> 00:22:42,320
Speaker 2: wid well the Trainium four or five was amazing, I

468
00:22:42,400 --> 00:22:44,760
Speaker 2: don't think it will be. But just saying doesn't that

469
00:22:44,760 --> 00:22:46,720
Speaker 2: mean they're gonna be sitting on a bunch of obsolete

470
00:22:46,760 --> 00:22:50,199
Speaker 2: Trainium It just feels very questionable. You know.

471
00:22:50,320 --> 00:22:52,439
Speaker 1: You can say a lot about open Ai, and you do.

472
00:22:52,720 --> 00:22:54,760
Speaker 1: But that dress release where they announced that thirty eight

473
00:22:54,800 --> 00:22:57,600
Speaker 1: billion dollar contract, they didn't mention Trainium.

474
00:22:57,920 --> 00:22:59,040
Speaker 2: They didn't, did they?

475
00:23:01,000 --> 00:23:02,840
Speaker 1: Which tells me that if it were half as good

476
00:23:02,840 --> 00:23:04,919
Speaker 1: at this at the model training as they say it is,

477
00:23:04,960 --> 00:23:09,040
Speaker 1: open Ai would be all about it. Yeah, I am

478
00:23:09,040 --> 00:23:10,240
Speaker 1: not seeing that to be true.

479
00:23:10,600 --> 00:23:13,399
Speaker 2: Well, Clammy, Sam Moltman, he's he signed a thing with

480
00:23:13,800 --> 00:23:17,280
Speaker 2: Sarah Brusber, Sarah Brush.

481
00:23:17,400 --> 00:23:19,119
Speaker 1: Is there any one that Sam Mortman has not signed

482
00:23:19,160 --> 00:23:19,520
Speaker 1: deals with?

483
00:23:20,080 --> 00:23:23,160
Speaker 2: Ah, let's think Grock. I haven't signed one with Grok yet?

484
00:23:23,240 --> 00:23:23,760
Speaker 1: Okay, good?

485
00:23:23,800 --> 00:23:26,040
Speaker 2: Good? I mean actually sorry, just to be clear, hasn't

486
00:23:26,040 --> 00:23:28,520
Speaker 2: signed with either, well, I guess it would. He wouldn't

487
00:23:28,520 --> 00:23:31,080
Speaker 2: sign one with GROC as technically Grok is a competitor,

488
00:23:31,240 --> 00:23:35,320
Speaker 2: I mean g r o Q. The annoying fast inference

489
00:23:35,960 --> 00:23:39,280
Speaker 2: At this point, my hated dom and people are going

490
00:23:39,320 --> 00:23:41,119
Speaker 2: to say this episode was two technical and the answer

491
00:23:41,160 --> 00:23:45,120
Speaker 2: is calm down, this is my podcast. I'm just I'm

492
00:23:45,160 --> 00:23:49,920
Speaker 2: wondering at what point they they say uncle, because okay,

493
00:23:50,000 --> 00:23:52,840
Speaker 2: two hundred billion dollars and let's say one hundred and

494
00:23:52,840 --> 00:23:55,719
Speaker 2: twenty five billion of that is AI, right, AI chips

495
00:23:56,240 --> 00:23:59,040
Speaker 2: takes four or five years probably to install all of

496
00:23:59,080 --> 00:24:03,679
Speaker 2: that to what's end? What happens in forfit? Are they

497
00:24:03,680 --> 00:24:04,720
Speaker 2: gonna make? Well?

498
00:24:04,760 --> 00:24:06,040
Speaker 1: I mean it would be That's what no one is

499
00:24:06,080 --> 00:24:08,679
Speaker 1: able to explain to me. What use cases today do

500
00:24:08,840 --> 00:24:10,920
Speaker 1: are we looking at? Where? Wow, if we had ten

501
00:24:10,960 --> 00:24:13,600
Speaker 1: times as much inference as we do today, then X

502
00:24:13,640 --> 00:24:16,439
Speaker 1: would be possible solve for X. I'm not hearing it.

503
00:24:16,720 --> 00:24:19,800
Speaker 2: No, I'm not either. I'm not even being my usual

504
00:24:19,800 --> 00:24:21,719
Speaker 2: hate you self, even though this is hate a season.

505
00:24:22,280 --> 00:24:26,160
Speaker 2: It's I genuinely can't get the answer of anyone, even

506
00:24:26,200 --> 00:24:30,320
Speaker 2: the boosters. It's like there's this insatiable demand for compute.

507
00:24:30,680 --> 00:24:33,239
Speaker 1: We're gonna summon God through Jason is step one, and

508
00:24:33,280 --> 00:24:35,199
Speaker 1: step two is we're gonna ask God what to do.

509
00:24:36,160 --> 00:24:38,720
Speaker 2: I mean, that's the thing I do think that they

510
00:24:39,440 --> 00:24:42,119
Speaker 2: I think that have you seen the revenue projections for

511
00:24:42,160 --> 00:24:44,720
Speaker 2: these companies? Have you ever seen like Oracle's cash flow?

512
00:24:45,160 --> 00:24:46,880
Speaker 2: So if you see this that they.

513
00:24:46,800 --> 00:24:49,000
Speaker 1: Put out, also known as there's a reason that every

514
00:24:49,040 --> 00:24:51,600
Speaker 1: time Larry Ellison, co founder of Oracle, goes and gives

515
00:24:51,600 --> 00:24:54,479
Speaker 1: a keynote or talk somewhere, they preface it with a

516
00:24:54,520 --> 00:24:57,719
Speaker 1: disclosure that is the lawyer version of everything you are

517
00:24:57,760 --> 00:25:00,480
Speaker 1: about to hear is a fanciful imagining of reality.

518
00:25:00,720 --> 00:25:03,880
Speaker 2: Don't listen to him. He's he's at a long day.

519
00:25:03,920 --> 00:25:07,600
Speaker 2: He hasn't taken his nap. Fucking if hr Guy Gooz,

520
00:25:07,680 --> 00:25:11,680
Speaker 2: Jerry Stiller. He The thing is, you look at these

521
00:25:12,720 --> 00:25:16,240
Speaker 2: cash flow positive these cash flow diagrams for Oranthropic, open AI,

522
00:25:16,280 --> 00:25:18,800
Speaker 2: and Oracle, they all follow the same thing, which is

523
00:25:19,240 --> 00:25:22,480
Speaker 2: terrible cash flow negative cash flow for years and years

524
00:25:22,520 --> 00:25:24,960
Speaker 2: and years, and then in twenty twenty nine everything changes,

525
00:25:25,040 --> 00:25:28,520
Speaker 2: number go up, and I at this point genuinely think

526
00:25:28,560 --> 00:25:30,680
Speaker 2: that they think they're going to invent Agi in twenty

527
00:25:30,680 --> 00:25:33,480
Speaker 2: twenty nine, that that's the only plan they have.

528
00:25:34,040 --> 00:25:38,280
Speaker 1: Because they're not doing the things that they're not if

529
00:25:38,280 --> 00:25:40,639
Speaker 1: they've genuinely believed that they would not be doing these things.

530
00:25:40,840 --> 00:25:44,000
Speaker 1: We think that we are three years away from launching

531
00:25:44,160 --> 00:25:47,199
Speaker 1: AGI that will change everything. So what are we gonna do.

532
00:25:47,359 --> 00:25:49,800
Speaker 1: We're gonna find a way to put ads into chat jibbity.

533
00:25:50,520 --> 00:25:53,000
Speaker 1: But that's not the play when you are when you

534
00:25:53,000 --> 00:25:54,359
Speaker 1: think you're on the verge of the singularity.

535
00:25:54,920 --> 00:25:59,080
Speaker 2: I mean, nobody was demes habit. I don't know. I

536
00:25:59,119 --> 00:26:01,200
Speaker 2: don't like him to say should extent I refused learn

537
00:26:01,240 --> 00:26:06,560
Speaker 2: his name. The deep mind bloke. I try not to.

538
00:26:06,760 --> 00:26:09,680
Speaker 1: I try not to anthropomorphize either AI or its creators.

539
00:26:09,960 --> 00:26:13,880
Speaker 2: Yeah, deepest mean beepus. He's the one that every few

540
00:26:13,920 --> 00:26:16,600
Speaker 2: weeks does an interview and he's like, and then God

541
00:26:16,720 --> 00:26:19,000
Speaker 2: is going to come out. I'm really scared of what

542
00:26:19,040 --> 00:26:21,280
Speaker 2: the computer will do. He doesn't sound like that he

543
00:26:21,359 --> 00:26:24,919
Speaker 2: wishes And it's just you make it sound exciting, Yeah, exactly.

544
00:26:24,960 --> 00:26:27,400
Speaker 2: It sounds fun and interesting and comical when these people

545
00:26:27,440 --> 00:26:33,040
Speaker 2: are just boring, boring people, boring motherfuckers. Bom. It's just

546
00:26:33,080 --> 00:26:38,320
Speaker 2: these massive overpromises and these weird fucking like these weird

547
00:26:38,359 --> 00:26:40,720
Speaker 2: things where I like, look in let's look at this.

548
00:26:41,200 --> 00:26:44,640
Speaker 2: There are six hundred and fifty billion dollars in CAPEX

549
00:26:44,720 --> 00:26:49,240
Speaker 2: being spent this year just by Alphabet, Amazon, Meta and Microsoft.

550
00:26:49,640 --> 00:26:54,040
Speaker 2: Where is it going? Genuinely? Actually simple question. I actually

551
00:26:54,119 --> 00:26:57,359
Speaker 2: don't know where it's going anymore. It's into a giant

552
00:26:57,359 --> 00:26:59,200
Speaker 2: hole in their balance sheet. At some point.

553
00:27:00,359 --> 00:27:03,080
Speaker 1: You'll notice that all of the hyperscalers, all of them

554
00:27:03,480 --> 00:27:07,399
Speaker 1: that have succeeded, even the pretenders, have other businesses that

555
00:27:07,600 --> 00:27:11,320
Speaker 1: financed this. There has never been a successful hyperscaler that

556
00:27:11,440 --> 00:27:15,119
Speaker 1: launched as a hyperscaler. The closest story you've got to that,

557
00:27:15,160 --> 00:27:17,480
Speaker 1: and I question whether you call them a hyperscaler is

558
00:27:17,480 --> 00:27:20,280
Speaker 1: cloud flair because they start as a CDN and then

559
00:27:20,400 --> 00:27:22,080
Speaker 1: just added the rest of the cloud stuff.

560
00:27:22,560 --> 00:27:24,679
Speaker 2: And let's see, I picked this stock. Is I got

561
00:27:24,760 --> 00:27:28,560
Speaker 2: fifty nine billion dollar market cap? Well, they're a real company. Yeah,

562
00:27:28,840 --> 00:27:30,840
Speaker 2: you wouldn't w I won't want to invest in a

563
00:27:30,880 --> 00:27:35,480
Speaker 2: company that wasn't making egregious promises. But it's just Jesus,

564
00:27:35,680 --> 00:27:40,680
Speaker 2: it's just so strange because even if they think AI

565
00:27:40,800 --> 00:27:42,880
Speaker 2: is going to be big, surely all the money they're

566
00:27:42,880 --> 00:27:45,240
Speaker 2: spending today is not going to do that, Like all

567
00:27:45,280 --> 00:27:47,440
Speaker 2: of these chips aren't going to do it. The models

568
00:27:47,440 --> 00:27:49,360
Speaker 2: they train today aren't going to do it.

569
00:27:49,359 --> 00:27:51,119
Speaker 1: It's like they're trying to bootstrap from thing to thing.

570
00:27:51,200 --> 00:27:54,800
Speaker 1: To think, the story that concerns me is I, okay.

571
00:27:54,840 --> 00:27:57,119
Speaker 1: You take a look at these companies and their subscription plans.

572
00:27:57,280 --> 00:28:00,120
Speaker 1: Top end of it is two hundred bucks. Great. I

573
00:28:00,160 --> 00:28:02,280
Speaker 1: pay it myself just so I can dispatch a winged

574
00:28:02,320 --> 00:28:04,840
Speaker 1: monkey to build me a shit post website, which is awesome,

575
00:28:04,840 --> 00:28:06,800
Speaker 1: and sometimes it works and sometimes I have to prompt

576
00:28:06,840 --> 00:28:10,639
Speaker 1: it to do different things. Great, it's better at shitty

577
00:28:10,640 --> 00:28:12,880
Speaker 1: front end than I am at shitty front end. Terrific.

578
00:28:13,600 --> 00:28:16,160
Speaker 1: I'm not going to spend five grand on that. Yeah,

579
00:28:16,280 --> 00:28:18,840
Speaker 1: but the model of for zooms not. And we're not

580
00:28:18,960 --> 00:28:22,000
Speaker 1: just talking me here, we are talking effectively everyone. It's

581
00:28:22,040 --> 00:28:24,600
Speaker 1: like the unspoken message of a lot of these folks

582
00:28:24,680 --> 00:28:26,320
Speaker 1: is that your boss is going to fire you and

583
00:28:26,320 --> 00:28:28,680
Speaker 1: then take your salary and split it with the AI company.

584
00:28:29,160 --> 00:28:33,200
Speaker 2: Right, and also you will be able to code everything

585
00:28:33,440 --> 00:28:37,240
Speaker 2: with AI versus what people love that I quote the

586
00:28:37,280 --> 00:28:40,400
Speaker 2: same thing Carl Brown's saying, makes the easy things easier

587
00:28:40,400 --> 00:28:41,360
Speaker 2: and the hard things horder.

588
00:28:42,080 --> 00:28:44,360
Speaker 1: Like we are building software at duck Billet, We are

589
00:28:44,440 --> 00:28:47,360
Speaker 1: hiring engineers. Is this because we're stupid?

590
00:28:48,200 --> 00:29:03,640
Speaker 3: I don't believe that to be true.

591
00:29:05,360 --> 00:29:07,600
Speaker 2: Yeah, And that's kind of the thing. It's like, I

592
00:29:07,600 --> 00:29:10,120
Speaker 2: don't know, people still seem to be hiring software engineers.

593
00:29:10,160 --> 00:29:13,040
Speaker 2: The engineers are getting fired and replaced. AI story doesn't

594
00:29:13,080 --> 00:29:15,360
Speaker 2: seem to work. And also, every time I go and

595
00:29:15,400 --> 00:29:18,560
Speaker 2: read someone who claims that clawed code changed their entire life,

596
00:29:18,760 --> 00:29:21,080
Speaker 2: it mostly comes down to, yeah, it took something from

597
00:29:21,080 --> 00:29:25,920
Speaker 2: a few hours to an hour, and it's like, cool, great, okay,

598
00:29:26,400 --> 00:29:26,920
Speaker 2: but in.

599
00:29:27,040 --> 00:29:28,440
Speaker 1: My life, what did it do? It taught me to

600
00:29:28,480 --> 00:29:30,280
Speaker 1: clearly explain what I wanted.

601
00:29:30,400 --> 00:29:34,760
Speaker 2: Yeah, but also I still had to fix things. It's

602
00:29:34,800 --> 00:29:38,520
Speaker 2: like if all of this money just went into creating

603
00:29:39,520 --> 00:29:42,680
Speaker 2: slightly more efficient software engineers, and even then the data

604
00:29:42,720 --> 00:29:47,680
Speaker 2: suggests they're less efficient. I don't know was this worth it?

605
00:29:47,920 --> 00:29:51,120
Speaker 2: And Amazon certainly isn't. Amazon certainly isn't. I guess they're

606
00:29:51,120 --> 00:29:53,800
Speaker 2: getting the chunk out of anthropic from this, like they

607
00:29:54,160 --> 00:29:57,200
Speaker 2: can claim anthropic success. They can have them pay for

608
00:29:57,320 --> 00:29:59,680
Speaker 2: trainium in project.

609
00:29:58,880 --> 00:30:01,840
Speaker 1: Platform company, where when people build interesting things on top

610
00:30:01,880 --> 00:30:04,040
Speaker 1: of you, how do you how do you take credit

611
00:30:04,080 --> 00:30:05,200
Speaker 1: for that without being obnoxious.

612
00:30:05,840 --> 00:30:07,920
Speaker 2: I mean, has that ever been a problem for Amazon?

613
00:30:08,800 --> 00:30:11,200
Speaker 1: Well, not that that's nothing their word about being obnoxiou,

614
00:30:11,200 --> 00:30:12,280
Speaker 1: it's just they're bad at messaging.

615
00:30:12,840 --> 00:30:17,520
Speaker 2: Well, I mean they'll mention Emlierate panos Panee m is

616
00:30:17,800 --> 00:30:20,680
Speaker 2: panas Penee came over from Microsoft from the Surface team

617
00:30:21,160 --> 00:30:27,320
Speaker 2: to lead Amazon's Alexa Plus. I mean, now there's something obnoxious.

618
00:30:27,880 --> 00:30:32,400
Speaker 2: Now there's some real generative AI bullshit. They've lost like

619
00:30:32,520 --> 00:30:38,440
Speaker 2: billions on Alexa now as well? God, what does strange

620
00:30:38,480 --> 00:30:39,720
Speaker 2: company Amazon's become?

621
00:30:41,840 --> 00:30:44,800
Speaker 1: They've always been peculiar. The problem is is at some

622
00:30:44,920 --> 00:30:48,040
Speaker 1: point of scale, like we're a little weird, no longer

623
00:30:48,160 --> 00:30:50,920
Speaker 1: carries water. It's great, you're now one of the lynch

624
00:30:50,920 --> 00:30:54,000
Speaker 1: pins of the global economy. You have to explain what

625
00:30:54,000 --> 00:30:55,320
Speaker 1: you're doing a bit more effectively.

626
00:30:56,440 --> 00:31:01,680
Speaker 2: Yeah, I I what do you think You've known Amazon

627
00:31:01,680 --> 00:31:04,320
Speaker 2: for one hundred years? Now, what do you think happens

628
00:31:04,400 --> 00:31:06,720
Speaker 2: if the AI bubble busts? What do you actually think

629
00:31:06,760 --> 00:31:11,200
Speaker 2: they do? Like, what do you think are they the

630
00:31:11,240 --> 00:31:15,560
Speaker 2: types to actually mothball this stuff? Would they just pretend

631
00:31:15,600 --> 00:31:18,040
Speaker 2: like this never happened aspects of it?

632
00:31:18,080 --> 00:31:20,520
Speaker 1: But there are again, they are making money hand over

633
00:31:20,560 --> 00:31:23,440
Speaker 1: fist on the actual nuts and bolts of cloud computing.

634
00:31:23,480 --> 00:31:27,480
Speaker 1: That's not going away. They will their Stockell tumbled because

635
00:31:27,520 --> 00:31:31,480
Speaker 1: they're not posting massive growth numbers, but they have a

636
00:31:31,520 --> 00:31:36,560
Speaker 1: banging business of providing computers to the world. The question is,

637
00:31:36,560 --> 00:31:38,240
Speaker 1: is what happens when there's a giant hole in their

638
00:31:38,240 --> 00:31:40,840
Speaker 1: balance sheet because well, it turns out that hundreds of

639
00:31:40,880 --> 00:31:43,280
Speaker 1: billions of dollars in contracts everyone's die and they suddenly

640
00:31:43,360 --> 00:31:45,800
Speaker 1: they don't got the money. What do we do? That's

641
00:31:45,840 --> 00:31:47,480
Speaker 1: going to be a systemic problem. It's not going to

642
00:31:47,520 --> 00:31:49,400
Speaker 1: just hit Amazon. It's going to hit a lot of folks.

643
00:31:50,200 --> 00:31:52,400
Speaker 1: And I think that the answer everyone's sort of hoping

644
00:31:52,400 --> 00:31:56,440
Speaker 1: for in that space is government bailout. No but of Amazon,

645
00:31:56,520 --> 00:31:59,040
Speaker 1: though I would be so irritated.

646
00:31:59,240 --> 00:32:01,840
Speaker 2: Well, also what they bailing out? Amazon's not gonna go

647
00:32:01,920 --> 00:32:04,800
Speaker 2: bank or even if Amazon had to take a massive

648
00:32:04,840 --> 00:32:09,120
Speaker 2: impairment on elder Tradeum chips and it was like thirty

649
00:32:09,200 --> 00:32:13,240
Speaker 2: billion dollars the shareholders had the shareholders, they'll be mad,

650
00:32:13,400 --> 00:32:16,600
Speaker 2: don't get me wrong, But it's not gonna destroy Amazon,

651
00:32:17,040 --> 00:32:20,800
Speaker 2: like it's not. I mean, it might destroy Anti Jesse's asshole,

652
00:32:20,960 --> 00:32:23,200
Speaker 2: like it might be might send Andy Jesse and a

653
00:32:23,280 --> 00:32:25,720
Speaker 2: Nye and bail into the sun. But it doesn't feel

654
00:32:25,720 --> 00:32:27,040
Speaker 2: like it will kill them. I just don't know what

655
00:32:27,040 --> 00:32:27,880
Speaker 2: a bailout would be.

656
00:32:30,520 --> 00:32:35,680
Speaker 1: I am so annoyed that I had great hopes when

657
00:32:35,720 --> 00:32:37,760
Speaker 1: Andy took over, but it seems like the company has

658
00:32:37,840 --> 00:32:40,360
Speaker 1: just continued down the in sertification curve. They're not doing

659
00:32:40,400 --> 00:32:42,760
Speaker 1: surprising things. The reason I started the Last Week in

660
00:32:42,800 --> 00:32:46,040
Speaker 1: Aws newsletter in twenty seventeen was that every week they

661
00:32:46,040 --> 00:32:48,800
Speaker 1: were fixing massive customer facing problems and it was hard

662
00:32:48,800 --> 00:32:50,560
Speaker 1: to keep track because they were just as bad that

663
00:32:50,680 --> 00:32:56,640
Speaker 1: as they are now twenty seventeen. Right now, they don't

664
00:32:56,640 --> 00:33:00,200
Speaker 1: do that anymore. It's basically inertia. I would not be

665
00:33:00,200 --> 00:33:02,240
Speaker 1: able to start the newsletter and build in our readership

666
00:33:02,440 --> 00:33:05,560
Speaker 1: today the way that I did back then, just because

667
00:33:05,720 --> 00:33:08,160
Speaker 1: it people do not care nearly so much about the

668
00:33:08,160 --> 00:33:09,280
Speaker 1: platform as they once did.

669
00:33:09,840 --> 00:33:12,000
Speaker 2: So when did you notice the shift happen?

670
00:33:12,880 --> 00:33:14,920
Speaker 1: It was first graduate than all at once, I think

671
00:33:15,040 --> 00:33:18,160
Speaker 1: is probably the way to frame it. When I realized

672
00:33:18,200 --> 00:33:23,040
Speaker 1: that I went from every week having the problem of

673
00:33:23,640 --> 00:33:27,280
Speaker 1: there's so much stuff here, which ones make the cut,

674
00:33:27,440 --> 00:33:30,880
Speaker 1: to there's so much filler here, where's anything actually worth

675
00:33:30,960 --> 00:33:31,520
Speaker 1: writing about.

676
00:33:31,880 --> 00:33:34,600
Speaker 2: And when you say philo, what do you mean they

677
00:33:34,720 --> 00:33:35,680
Speaker 2: used to say?

678
00:33:35,520 --> 00:33:38,000
Speaker 1: They used to talk about things like as lambda, a

679
00:33:38,080 --> 00:33:41,520
Speaker 1: transformative shift in how a compute could be run. They

680
00:33:41,760 --> 00:33:44,920
Speaker 1: put that out with the same enthusiasm corporate voice as

681
00:33:45,120 --> 00:33:48,440
Speaker 1: there's now a third cloud front edge location in Dallas,

682
00:33:48,760 --> 00:33:50,880
Speaker 1: which even people in Dallas do not give.

683
00:33:50,720 --> 00:33:54,040
Speaker 2: It to us about, right, what does LAMB that do?

684
00:33:54,800 --> 00:33:56,200
Speaker 2: In effectively? You give it its code.

685
00:33:56,240 --> 00:33:58,400
Speaker 1: It runs the code when certain trigger events happen, be

686
00:33:58,400 --> 00:34:00,760
Speaker 1: a web request, be it a time being hit, be

687
00:34:00,760 --> 00:34:03,360
Speaker 1: it's something that it's some other event hitting. And it

688
00:34:03,400 --> 00:34:07,480
Speaker 1: only charges you for while it runs. Huh and massively

689
00:34:07,520 --> 00:34:09,640
Speaker 1: scales up and you don't have to worry about any

690
00:34:09,680 --> 00:34:12,040
Speaker 1: of the care and feeding of the infrastructure around it,

691
00:34:12,120 --> 00:34:14,680
Speaker 1: which that's really cool. It is it was cool. There's

692
00:34:14,719 --> 00:34:17,759
Speaker 1: beta in twenty fifteen. But now for the last truly

693
00:34:17,800 --> 00:34:19,640
Speaker 1: great things Amazon put out that change the way I

694
00:34:19,640 --> 00:34:20,680
Speaker 1: thought about how this could work.

695
00:34:21,480 --> 00:34:24,960
Speaker 2: Except now Amazon doesn't really put out fun new updates.

696
00:34:24,960 --> 00:34:28,720
Speaker 1: It's just incremental updates. We have a new thing. We've launched,

697
00:34:28,760 --> 00:34:32,920
Speaker 1: our own large language model called Nova. It's not as

698
00:34:32,960 --> 00:34:37,120
Speaker 1: good as the ones you currently use, but it's less money. Yeah,

699
00:34:37,160 --> 00:34:39,040
Speaker 1: what's start poc stage. They don't care.

700
00:34:39,880 --> 00:34:42,120
Speaker 2: What's great is if you look when Nova is Wow,

701
00:34:42,200 --> 00:34:47,920
Speaker 2: it's in the UH top twenty five below Kat's code

702
00:34:47,920 --> 00:34:51,320
Speaker 2: of pro Kimmy two point five four point one memo

703
00:34:51,480 --> 00:34:54,640
Speaker 2: me and the Max Quinn thinking Kimmy K two basically

704
00:34:54,680 --> 00:34:55,920
Speaker 2: every other model.

705
00:34:56,160 --> 00:34:59,600
Speaker 1: Right, like exactly, it's like they're racing neck and neck

706
00:34:59,640 --> 00:35:04,040
Speaker 1: with the taxidermy in Frontier AI lab that it really is.

707
00:35:03,960 --> 00:35:07,160
Speaker 2: That though, it's like it's something called yeah Ernie from Baidu,

708
00:35:07,440 --> 00:35:10,720
Speaker 2: like every Chinese model appears to cry, that's so funny.

709
00:35:11,160 --> 00:35:14,919
Speaker 2: What a what a horrible situation we found ourselves in.

710
00:35:15,400 --> 00:35:18,000
Speaker 1: Well, Amazon prides itself on being frugal, and it turns

711
00:35:18,000 --> 00:35:20,400
Speaker 1: out that when I have to assume the reason the

712
00:35:20,440 --> 00:35:22,399
Speaker 1: Titan models that were so bad they never really saw

713
00:35:22,480 --> 00:35:26,359
Speaker 1: broad release in light a day was that it's okay,

714
00:35:26,400 --> 00:35:28,680
Speaker 1: it's gonna cost some billion dollars to train this model.

715
00:35:28,800 --> 00:35:31,120
Speaker 1: What can you do for twenty million? Like that? That

716
00:35:31,280 --> 00:35:34,759
Speaker 1: is the Amazon ethos. Fail is the answer.

717
00:35:34,920 --> 00:35:38,120
Speaker 2: Unless it comes to Capex though, Like that's the thing.

718
00:35:38,280 --> 00:35:40,840
Speaker 2: It feels like what you say there goes directly against

719
00:35:40,880 --> 00:35:42,840
Speaker 2: how they're dealing with AI. It feels like AI is

720
00:35:42,920 --> 00:35:44,160
Speaker 2: just poisoning their brains.

721
00:35:44,719 --> 00:35:47,200
Speaker 1: Oh my god. Yes, I wound up doing a so

722
00:35:47,280 --> 00:35:49,160
Speaker 1: I back when these stuff all started coming out, I

723
00:35:49,200 --> 00:35:51,560
Speaker 1: had fun with it. I ran a bunch of custom

724
00:35:51,680 --> 00:35:54,680
Speaker 1: questions against these things, like rank the US presidents by

725
00:35:54,719 --> 00:35:56,480
Speaker 1: absorbency and then see how long long is it? We

726
00:35:56,480 --> 00:35:58,880
Speaker 1: would bully them into doing it or give me some

727
00:35:58,960 --> 00:36:01,560
Speaker 1: criticisms of core Quinn and one of the early chat

728
00:36:01,600 --> 00:36:04,520
Speaker 1: GPT things made me question like aspects of my life

729
00:36:04,560 --> 00:36:07,040
Speaker 1: and go home and drink heavily, Whereas I asked one

730
00:36:07,080 --> 00:36:09,200
Speaker 1: of the early Amazon models that question, and it would

731
00:36:09,200 --> 00:36:11,600
Speaker 1: have been like Exhibit A and the defamation case, except

732
00:36:11,640 --> 00:36:13,759
Speaker 1: I'd have to prove that anyone took anything Amazon said

733
00:36:13,760 --> 00:36:14,800
Speaker 1: even slightly seriously.

734
00:36:15,560 --> 00:36:20,000
Speaker 2: Right. It's it's just weird. There's something that I think

735
00:36:20,040 --> 00:36:22,600
Speaker 2: the I don't know if AI side coast this is it,

736
00:36:22,640 --> 00:36:25,160
Speaker 2: but I mentioned this with David Gerard. It's like people

737
00:36:25,160 --> 00:36:28,160
Speaker 2: getting one shot at by this. It must one show executives.

738
00:36:28,480 --> 00:36:30,200
Speaker 2: I think, oh, it absolutely does.

739
00:36:30,640 --> 00:36:33,880
Speaker 1: I use it myself in the curation of my newsletter

740
00:36:33,920 --> 00:36:37,840
Speaker 1: for example something because this is this is a terrific

741
00:36:38,040 --> 00:36:40,719
Speaker 1: use case, and I think it encapsulates my philosophy on

742
00:36:40,719 --> 00:36:44,600
Speaker 1: this uh where Amazon. I consume all the RSS feeds

743
00:36:44,600 --> 00:36:46,600
Speaker 1: every week and there are roughly one hundred and fifty

744
00:36:46,640 --> 00:36:49,040
Speaker 1: items that come out. So what I do is I

745
00:36:49,080 --> 00:36:51,919
Speaker 1: have a scoring thing that pops up two columns. Here's

746
00:36:52,280 --> 00:36:55,239
Speaker 1: here's the good ones, here's the crap ones. And I

747
00:36:55,320 --> 00:36:57,160
Speaker 1: have a whole rubric that I have put into this.

748
00:36:57,360 --> 00:36:59,200
Speaker 1: It saves me a tremendous amount of time. I go

749
00:36:59,280 --> 00:37:01,480
Speaker 1: through and I put some from, I move some back

750
00:37:01,520 --> 00:37:03,880
Speaker 1: and forth as I read down the list and increasing

751
00:37:04,080 --> 00:37:06,719
Speaker 1: it learns from my decisions. It's getting it mostly right.

752
00:37:07,000 --> 00:37:09,960
Speaker 1: It speeds up my workflow. It is convenient. Is this

753
00:37:10,160 --> 00:37:11,800
Speaker 1: worth changing the entire world?

754
00:37:11,840 --> 00:37:12,640
Speaker 2: For? No?

755
00:37:12,640 --> 00:37:14,680
Speaker 1: No, it is not. But I'll make Hey, will the

756
00:37:14,719 --> 00:37:16,879
Speaker 1: sun shines? Besides, if I'm going to make fun of something,

757
00:37:16,880 --> 00:37:19,319
Speaker 1: I should be using it so I can understand it

758
00:37:19,360 --> 00:37:20,799
Speaker 1: to make fun of it more effectively.

759
00:37:21,280 --> 00:37:24,920
Speaker 2: Right, But would you pay the real rates for it?

760
00:37:24,960 --> 00:37:26,719
Speaker 2: Because that's what I think is going to break this

761
00:37:26,800 --> 00:37:29,279
Speaker 2: as well. It's when these companies start having to charge

762
00:37:29,280 --> 00:37:31,880
Speaker 2: the real costs, when they have to start getting paying

763
00:37:31,920 --> 00:37:32,520
Speaker 2: the margins.

764
00:37:32,719 --> 00:37:34,640
Speaker 1: Right now, this is costing me seven cents a month.

765
00:37:34,680 --> 00:37:37,680
Speaker 1: I am glad to pay that I would probably pay

766
00:37:37,800 --> 00:37:40,439
Speaker 1: as high for that as I don't know, twenty bucks

767
00:37:40,480 --> 00:37:43,960
Speaker 1: a month, which, again that's a significant increase there, sure,

768
00:37:44,480 --> 00:37:46,319
Speaker 1: but too much beyond that.

769
00:37:46,640 --> 00:37:46,799
Speaker 2: No.

770
00:37:46,880 --> 00:37:49,640
Speaker 1: I spent seven years going and doing this by hand.

771
00:37:49,760 --> 00:37:52,440
Speaker 1: I don't feel the need to stop. I did use

772
00:37:52,480 --> 00:37:55,280
Speaker 1: claud code for several weeks to rebuild the entire interface

773
00:37:55,360 --> 00:37:58,520
Speaker 1: because I was I had this janky, horrible newsletter publication

774
00:37:58,719 --> 00:38:01,440
Speaker 1: thing that I kept meaning to. I looked into what

775
00:38:01,440 --> 00:38:03,360
Speaker 1: it would take to have an engineer do it between

776
00:38:03,360 --> 00:38:05,960
Speaker 1: twenty to fifty thousand dollars. This was just three weeks

777
00:38:06,000 --> 00:38:08,120
Speaker 1: over the holiday break of prompting the thing, and it

778
00:38:08,200 --> 00:38:11,360
Speaker 1: finally got it dialed in and correct. Now, in practice,

779
00:38:11,400 --> 00:38:13,560
Speaker 1: if neither one of those two things is a viable option,

780
00:38:13,840 --> 00:38:15,959
Speaker 1: there's a bunch of stuff off the shelf these days.

781
00:38:15,960 --> 00:38:18,759
Speaker 1: It didn't exist back then that I could have worked

782
00:38:18,800 --> 00:38:21,400
Speaker 1: with and had a much better base. But this was fun.

783
00:38:21,480 --> 00:38:24,440
Speaker 1: This was I could use some software in this small place.

784
00:38:24,719 --> 00:38:28,600
Speaker 1: The cost of generating this software now from my time perspective,

785
00:38:28,920 --> 00:38:32,080
Speaker 1: is minimal, and it fills in a gap. And these

786
00:38:32,120 --> 00:38:34,520
Speaker 1: things are good at solving a building tools to solve

787
00:38:34,520 --> 00:38:37,760
Speaker 1: one very specific problem. Where these things fall down. Software

788
00:38:37,760 --> 00:38:40,040
Speaker 1: has always fallen down on this is as soon as

789
00:38:40,160 --> 00:38:43,280
Speaker 1: the requirements change, I said on a link list every Monday.

790
00:38:43,320 --> 00:38:46,000
Speaker 1: I built software myself to handle this. When I started

791
00:38:46,000 --> 00:38:47,960
Speaker 1: sending out blog posts with it as well, I had

792
00:38:48,000 --> 00:38:49,839
Speaker 1: to go back to the drawing board and redo an

793
00:38:49,880 --> 00:38:53,160
Speaker 1: awful lot because oh, it's a different model. The entire

794
00:38:53,280 --> 00:38:58,080
Speaker 1: flow changes when I introduced the requirement downstream. When software

795
00:38:58,160 --> 00:39:01,080
Speaker 1: becomes free, is longer the bottleneck more or less. You

796
00:39:01,080 --> 00:39:04,520
Speaker 1: can build custom tools for any given situation far more effectively.

797
00:39:05,280 --> 00:39:07,640
Speaker 1: I started writing a Bash script to give someone access

798
00:39:07,680 --> 00:39:09,560
Speaker 1: to just run run this one command in your term,

799
00:39:09,600 --> 00:39:12,160
Speaker 1: It'll give you access to this thing. And I started

800
00:39:12,160 --> 00:39:14,279
Speaker 1: writing it myself, threw it into one of these things.

801
00:39:14,280 --> 00:39:16,440
Speaker 1: It came out with a much better script as a

802
00:39:16,480 --> 00:39:18,400
Speaker 1: one off that I don't need again. I could I

803
00:39:18,440 --> 00:39:20,839
Speaker 1: have done it myself. Yeah, about four hours, But would

804
00:39:20,840 --> 00:39:21,120
Speaker 1: I have.

805
00:39:21,239 --> 00:39:24,400
Speaker 2: No It's soldings. It's also those.

806
00:39:24,200 --> 00:39:27,280
Speaker 1: Small problems or point solutions. But I'm not firing people

807
00:39:27,280 --> 00:39:29,719
Speaker 1: for this. I'm not replacing segments of the economy with

808
00:39:29,760 --> 00:39:32,600
Speaker 1: this I have. My brother is one of the few

809
00:39:32,600 --> 00:39:35,000
Speaker 1: people I have heard about being definitely impacted by the

810
00:39:35,040 --> 00:39:38,520
Speaker 1: rise of AI. Until last year he was a freelance translator.

811
00:39:39,000 --> 00:39:41,200
Speaker 1: I believe that that feels like it's something that is

812
00:39:41,239 --> 00:39:45,160
Speaker 1: easily prone to disruption. I buy less stock photography when

813
00:39:45,160 --> 00:39:47,360
Speaker 1: I can have shit post images of giraffes on fire

814
00:39:47,400 --> 00:39:48,120
Speaker 1: and data centers.

815
00:39:48,160 --> 00:39:52,000
Speaker 2: Now that they're not They're not gonna love that. The

816
00:39:52,080 --> 00:39:53,959
Speaker 2: list is not gonna love that. Oh no, I don't

817
00:39:53,960 --> 00:39:57,480
Speaker 2: even I think Ai is just genuinely ugly, you know,

818
00:39:57,560 --> 00:39:59,759
Speaker 2: even that it would suck the life out of the

819
00:40:00,440 --> 00:40:01,399
Speaker 2: out of the ship post.

820
00:40:01,440 --> 00:40:03,360
Speaker 1: And I paid for artists to do things correctly.

821
00:40:03,440 --> 00:40:04,760
Speaker 2: Okay, good by have stickers.

822
00:40:04,840 --> 00:40:07,799
Speaker 1: And I gave out a reinvent for ship posting dot

823
00:40:07,840 --> 00:40:11,240
Speaker 1: Ai and it was bad Ai art in a nineteen

824
00:40:11,320 --> 00:40:14,760
Speaker 1: fifty style, like a dog with an extra leg, people

825
00:40:14,800 --> 00:40:18,160
Speaker 1: with extra hands and whatnot. That we hired an illustrator

826
00:40:18,480 --> 00:40:21,480
Speaker 1: as a human to draw in the style of bad Ai.

827
00:40:21,760 --> 00:40:23,960
Speaker 1: I thought that was a lot of fun for something

828
00:40:24,040 --> 00:40:26,399
Speaker 1: like this. Yeah, I am paying artists and I'm making

829
00:40:26,440 --> 00:40:29,160
Speaker 1: it work. I'm building a ship post meme to throw

830
00:40:29,239 --> 00:40:31,239
Speaker 1: on Twitter and never think about it against the thing.

831
00:40:31,360 --> 00:40:34,279
Speaker 2: Though posting is an art, posting is an art. It

832
00:40:34,360 --> 00:40:38,800
Speaker 2: is bringing this. There is no bushido in in generating images,

833
00:40:38,920 --> 00:40:41,920
Speaker 2: but let's move on from this subject because I have

834
00:40:42,000 --> 00:40:45,560
Speaker 2: one final question, Please hit me with it. So you've

835
00:40:45,600 --> 00:40:48,839
Speaker 2: seen this thing about how everyone's freaking out about, and

836
00:40:48,880 --> 00:40:52,120
Speaker 2: I know you're going to love this. Everyone's like, oh, hey,

837
00:40:52,760 --> 00:40:57,040
Speaker 2: clawed code is replacing software everybody. How funny is it?

838
00:40:57,200 --> 00:41:00,480
Speaker 2: How ridiculous is it to you? I'm leading question. I

839
00:41:00,520 --> 00:41:03,880
Speaker 2: realized the idea that instead of buying on a percy basis,

840
00:41:03,920 --> 00:41:06,040
Speaker 2: people are going to just replace SaaS with their own

841
00:41:06,080 --> 00:41:08,799
Speaker 2: internal share. I think it's one of the funniest and

842
00:41:08,840 --> 00:41:10,040
Speaker 2: stupidest things I've ever heard.

843
00:41:10,440 --> 00:41:10,600
Speaker 4: Oh.

844
00:41:10,640 --> 00:41:13,240
Speaker 1: I think it's a terrific thing that is totally happening

845
00:41:13,760 --> 00:41:18,560
Speaker 1: because obviously, again I am building software for enterprises. Well,

846
00:41:18,760 --> 00:41:21,479
Speaker 1: they can just write this code overnight to do it. Yes,

847
00:41:21,560 --> 00:41:25,360
Speaker 1: they can easily generate a website that purports to do something.

848
00:41:25,719 --> 00:41:27,680
Speaker 1: But what if they need that thing to be correct?

849
00:41:27,719 --> 00:41:30,280
Speaker 1: What if they're, oh, I don't know, signing a billion

850
00:41:30,320 --> 00:41:33,240
Speaker 1: dollar contract on the result of what that software spits

851
00:41:33,280 --> 00:41:36,920
Speaker 1: out exactly? Maybe Yolo slamming it isn't the right answer.

852
00:41:37,400 --> 00:41:39,840
Speaker 1: We're going to just replace all the software with custom

853
00:41:39,880 --> 00:41:40,760
Speaker 1: bespoke stuff.

854
00:41:40,960 --> 00:41:41,240
Speaker 2: Great.

855
00:41:41,480 --> 00:41:45,279
Speaker 1: A terrific example I saw is that for payroll anthropic.

856
00:41:45,360 --> 00:41:48,399
Speaker 1: The company uses ADP, Well why would they do that?

857
00:41:48,400 --> 00:41:53,000
Speaker 1: They can clearly build their own eternally because you're not

858
00:41:53,080 --> 00:41:56,319
Speaker 1: buying the software, you're buying the compliance, the keeping up

859
00:41:56,320 --> 00:41:58,919
Speaker 1: with all the different jurisdictions in which you operate. You're

860
00:41:59,000 --> 00:42:01,920
Speaker 1: someone who is going to certify with their reputation that

861
00:42:01,960 --> 00:42:06,560
Speaker 1: this is going to solve a problem. There are a

862
00:42:06,760 --> 00:42:09,040
Speaker 1: bunch of things out there, like there's some software that

863
00:42:09,080 --> 00:42:11,200
Speaker 1: will be disrupted by this picture, a bunch of the

864
00:42:11,480 --> 00:42:13,760
Speaker 1: obnoxious stuff like I just want to convert this from

865
00:42:13,880 --> 00:42:16,799
Speaker 1: one from one video type to a different video type,

866
00:42:16,800 --> 00:42:19,719
Speaker 1: and you'll find some like random thing for ten bucks. Yeah, well,

867
00:42:19,920 --> 00:42:22,319
Speaker 1: now cloud code or whatnot is going to fix that

868
00:42:22,360 --> 00:42:25,160
Speaker 1: because I don't know how FFmpeg command line flags are

869
00:42:25,160 --> 00:42:28,440
Speaker 1: supposed to work. It does great better, it's easier than

870
00:42:28,440 --> 00:42:30,840
Speaker 1: looking it up. Am I going to go and replace

871
00:42:30,960 --> 00:42:33,800
Speaker 1: actual business? Line of business? So line is a software

872
00:42:34,960 --> 00:42:35,440
Speaker 1: I wish?

873
00:42:35,480 --> 00:42:38,120
Speaker 2: But no, Well that's the thing as well. People think

874
00:42:38,200 --> 00:42:43,000
Speaker 2: the people pay for like Salesforce what have you, because

875
00:42:43,040 --> 00:42:46,279
Speaker 2: they're like, oh, because I have to pay for it,

876
00:42:46,320 --> 00:42:48,800
Speaker 2: and also just because it's a simple piece of software. No,

877
00:42:48,960 --> 00:42:53,400
Speaker 2: they're these Intrica API ratking things that roll around in

878
00:42:53,440 --> 00:42:56,520
Speaker 2: their own filth, and then you they get they maintain

879
00:42:56,600 --> 00:42:58,960
Speaker 2: them on the back end. Even if even the shittiest

880
00:42:58,960 --> 00:43:01,840
Speaker 2: cloud software you use is has some weird back end stuff.

881
00:43:01,880 --> 00:43:04,360
Speaker 2: They have to constantly maintain because things break, and the

882
00:43:04,400 --> 00:43:06,080
Speaker 2: bigger the company, the more things that break.

883
00:43:06,400 --> 00:43:09,560
Speaker 1: Yeah, and it's just the gen one of all of

884
00:43:09,560 --> 00:43:12,680
Speaker 1: Salesforce is just an Excel spreadsheet. It is who am

885
00:43:12,719 --> 00:43:14,600
Speaker 1: I contacting? What is the last thing I did in

886
00:43:14,640 --> 00:43:17,440
Speaker 1: the rest? And it gets complex quickly, and where the

887
00:43:17,560 --> 00:43:19,759
Speaker 1: leap happens is Okay, Now it's not just you. Now

888
00:43:19,760 --> 00:43:21,560
Speaker 1: it's an entire sales team. Now you're going to like

889
00:43:21,680 --> 00:43:23,759
Speaker 1: you have some turnover in that sales team over the

890
00:43:23,800 --> 00:43:26,120
Speaker 1: course of doing business. How do you wind up having

891
00:43:26,160 --> 00:43:29,319
Speaker 1: a provable chain of issue on this? Someone pops up

892
00:43:29,360 --> 00:43:31,000
Speaker 1: for app for three years from now they said they

893
00:43:31,000 --> 00:43:33,120
Speaker 1: talked to someone here. We have no record of that

894
00:43:33,719 --> 00:43:37,040
Speaker 1: because Claude rebuilt it five times this week. Where does

895
00:43:37,080 --> 00:43:39,400
Speaker 1: that going to live? How is that going to exist?

896
00:43:39,600 --> 00:43:44,400
Speaker 2: And even then it's like audibility, auditability, even Jesus Christ,

897
00:43:44,520 --> 00:43:44,960
Speaker 2: but this is.

898
00:43:44,920 --> 00:43:48,040
Speaker 1: Not an AI phenomenon. When Dropbox launched, the number one

899
00:43:48,120 --> 00:43:51,440
Speaker 1: comment on hacker News was this just is just our

900
00:43:51,520 --> 00:43:53,040
Speaker 1: sync with some extra bells and whistles.

901
00:43:53,040 --> 00:43:58,080
Speaker 2: This isn't a product. Yeah, yeah, well that's about yeah.

902
00:43:58,320 --> 00:44:01,160
Speaker 2: I mean even then drop Boxes like, it's not just storage,

903
00:44:01,440 --> 00:44:03,040
Speaker 2: is the support services around it.

904
00:44:03,360 --> 00:44:05,120
Speaker 1: I don't know believe a dropbox doesn't want to do

905
00:44:05,160 --> 00:44:07,200
Speaker 1: what Dropbox does anymore every time I use it. A

906
00:44:07,239 --> 00:44:10,160
Speaker 1: folder that sinks everywhere is what I loved about drops

907
00:44:10,480 --> 00:44:13,280
Speaker 1: Now it's trying to basically compete with zoom in Google

908
00:44:13,360 --> 00:44:15,000
Speaker 1: Docs and all the other stuff.

909
00:44:16,600 --> 00:44:21,040
Speaker 2: Stiles, well, gory. Let's wrap it up there. We've done

910
00:44:21,080 --> 00:44:24,520
Speaker 2: some good hating we can find. I'll have some good links.

911
00:44:24,600 --> 00:44:27,919
Speaker 2: View will link to duc Bill, link to last week

912
00:44:27,920 --> 00:44:28,640
Speaker 2: and Amazon was it.

913
00:44:28,760 --> 00:44:33,040
Speaker 1: Last week at aws dot com with an obnoxious, sarcastic platypus.

914
00:44:33,280 --> 00:44:35,319
Speaker 1: That's one my use of AI before we go that

915
00:44:35,400 --> 00:44:39,040
Speaker 1: I think you'll appreciate. If people email me Corey dot

916
00:44:39,160 --> 00:44:43,160
Speaker 1: Quinn at DUCKBILLHQ dot com and I don't like what

917
00:44:43,200 --> 00:44:45,360
Speaker 1: you're doing and you're trying to sell me something. I

918
00:44:45,440 --> 00:44:48,680
Speaker 1: now have an AI powered executive assistant in the form

919
00:44:48,719 --> 00:44:51,200
Speaker 1: of Billy the platypus, who will tell you in corporate

920
00:44:51,239 --> 00:44:55,160
Speaker 1: appropriate ways to go fuck yourself, which is just chef's kiss,

921
00:44:55,160 --> 00:44:57,799
Speaker 1: because I can't say that to people. I have a reputation.

922
00:44:58,040 --> 00:45:01,759
Speaker 1: The obnoxious platypus responding to so salespeople does not have

923
00:45:01,800 --> 00:45:04,640
Speaker 1: this constraint. This is the one liability shift I have

924
00:45:04,680 --> 00:45:07,800
Speaker 1: found that actually works. And you can blame AI for

925
00:45:07,920 --> 00:45:09,240
Speaker 1: things you personally approved.

926
00:45:09,600 --> 00:45:12,279
Speaker 2: Well, I don't endorse the AI platypus. You should get

927
00:45:12,280 --> 00:45:14,160
Speaker 2: a real one. You should get a real one.

928
00:45:14,280 --> 00:45:16,920
Speaker 1: Danger that customs officials had many questions I can not

929
00:45:17,000 --> 00:45:18,000
Speaker 1: wish to answer.

930
00:45:17,880 --> 00:45:20,719
Speaker 2: Train the platypus Cory. All right, you've been listening to

931
00:45:20,760 --> 00:45:23,040
Speaker 2: Better Offline. I'm at Zidra and you have a monologue

932
00:45:23,040 --> 00:45:25,800
Speaker 2: this week. I guess. Thank you so much for listening.

933
00:45:25,840 --> 00:45:37,240
Speaker 5: Everyone, thank you for listening to Better Offline. The editor

934
00:45:37,280 --> 00:45:40,160
Speaker 5: and composer of the Better Offline theme song is Matasowski.

935
00:45:40,480 --> 00:45:42,280
Speaker 5: You can check out more of his music and audio

936
00:45:42,320 --> 00:45:46,160
Speaker 5: projects at Matasowski dot com, M A T T O

937
00:45:46,360 --> 00:45:50,520
Speaker 5: S O W s ki dot com. You can email

938
00:45:50,560 --> 00:45:53,000
Speaker 5: me at easy at Better offline dot com, or visit

939
00:45:53,040 --> 00:45:55,360
Speaker 5: Better Offline dot com to find more podcast links and

940
00:45:55,400 --> 00:45:56,440
Speaker 5: of course, my newsletter.

941
00:45:56,840 --> 00:45:59,319
Speaker 2: I also really recommend you go to chat dot Where's

942
00:45:59,360 --> 00:46:02,040
Speaker 2: youreaed dot to visit the discord and go to our

943
00:46:02,120 --> 00:46:05,400
Speaker 2: slash Better Offline to check out I'll Reddit. Thank you

944
00:46:05,480 --> 00:46:06,440
Speaker 2: so much for listening.

945
00:46:07,280 --> 00:46:09,760
Speaker 1: Better Offline is a production of cool Zone Media.

946
00:46:09,880 --> 00:46:13,280
Speaker 2: For more from cool Zone Media, visit our website coolzonmedia

947
00:46:13,320 --> 00:46:16,120
Speaker 2: dot com, or check us out on the iHeartRadio app,

948
00:46:16,200 --> 00:46:18,600
Speaker 2: Apple Podcasts, or wherever you get your podcasts.