1
00:00:02,440 --> 00:00:08,240
Speaker 1: Also media, Hello and welcome to Better Offline. I'm, of

2
00:00:08,280 --> 00:00:21,960
Speaker 1: course your host ed zitron. As ever, go to the

3
00:00:22,000 --> 00:00:24,960
Speaker 1: episode notes and buy some merchandise. We have the Beautiful

4
00:00:25,040 --> 00:00:28,440
Speaker 1: challenge coin available until August twenty four. I want to say,

5
00:00:28,640 --> 00:00:31,000
Speaker 1: subscribe to my newsletter or send me a threatening email.

6
00:00:31,040 --> 00:00:32,879
Speaker 1: I love hearing from you all, and I also love

7
00:00:32,920 --> 00:00:34,840
Speaker 1: to see that our subreddit has gone over twelve thousand

8
00:00:34,920 --> 00:00:38,760
Speaker 1: vagabonds and scallywags. I'm so proud of our community. Anyway,

9
00:00:38,800 --> 00:00:40,440
Speaker 1: this week's going to be a fun one and two

10
00:00:40,440 --> 00:00:42,839
Speaker 1: part of plus a special report on GPT five, and

11
00:00:42,880 --> 00:00:44,720
Speaker 1: it's kind of a culmination of the last year and

12
00:00:44,760 --> 00:00:48,080
Speaker 1: a half of work that I've done covering the AI bubble. Now,

13
00:00:48,120 --> 00:00:51,000
Speaker 1: the reason I return to this subject so often is

14
00:00:51,000 --> 00:00:53,120
Speaker 1: because each week I've become more and more convinced that

15
00:00:53,159 --> 00:00:55,880
Speaker 1: the only way the AI bubble ends is well badly.

16
00:00:56,400 --> 00:00:58,279
Speaker 1: Every week I see more and more evidence that this

17
00:00:58,320 --> 00:01:00,640
Speaker 1: is structured in a way that is doomed when not

18
00:01:00,720 --> 00:01:04,320
Speaker 1: even abruptly, but they were going to be abrupt little farts,

19
00:01:04,400 --> 00:01:07,440
Speaker 1: little burps as the bubble begins letting out air, and

20
00:01:07,480 --> 00:01:09,920
Speaker 1: in a way where I see people in innovation as

21
00:01:09,959 --> 00:01:13,600
Speaker 1: the collateral damage in the global financial crisis. The Federal

22
00:01:13,640 --> 00:01:15,920
Speaker 1: Reserve and the US government, as well as other equivalents

23
00:01:15,920 --> 00:01:18,040
Speaker 1: around the world, work to ensure that when a bank

24
00:01:18,120 --> 00:01:20,559
Speaker 1: or hedge FuMB went under or a big insurance company

25
00:01:20,560 --> 00:01:23,200
Speaker 1: became insolvent, the collapse was managed in a way that

26
00:01:23,240 --> 00:01:26,959
Speaker 1: didn't implode the entire financial system. The difference with AI

27
00:01:27,040 --> 00:01:29,400
Speaker 1: is first that it's nowhere here as essential as financial

28
00:01:29,400 --> 00:01:32,000
Speaker 1: services to the global economy and thus unlikely to be

29
00:01:32,040 --> 00:01:34,600
Speaker 1: saved as AIG was in two thousand and eight, but

30
00:01:34,640 --> 00:01:38,160
Speaker 1: also that there's no real way for this to end gracefully,

31
00:01:38,840 --> 00:01:40,880
Speaker 1: just due to the way all of this is happening,

32
00:01:40,920 --> 00:01:44,120
Speaker 1: how it's built. I'll get into it, and all the

33
00:01:44,160 --> 00:01:46,640
Speaker 1: signs are now there, And I have been hesitant to

34
00:01:46,640 --> 00:01:49,560
Speaker 1: make this call the whole time. By the way, where well,

35
00:01:49,600 --> 00:01:53,000
Speaker 1: there's a contagion here, where one precarious company falls, the

36
00:01:53,000 --> 00:01:56,000
Speaker 1: rest eventually tumble, sparking a chain of events that will

37
00:01:56,000 --> 00:02:00,760
Speaker 1: be well, pretty bad. But how does this end? At

38
00:02:00,760 --> 00:02:02,800
Speaker 1: the start of August, we saw no less than three

39
00:02:02,800 --> 00:02:04,760
Speaker 1: different pieces, two in the Wall Street Journal and one

40
00:02:04,760 --> 00:02:09,560
Speaker 1: in the Financial Times, Financial Nacial Times, Who Cares, published

41
00:02:09,560 --> 00:02:11,680
Speaker 1: within the span of one week, all asking the same

42
00:02:11,760 --> 00:02:14,880
Speaker 1: question whether the massive proliferation of data centers is a

43
00:02:14,880 --> 00:02:18,239
Speaker 1: massive bubble, and honestly, since then there have been several more.

44
00:02:18,840 --> 00:02:21,240
Speaker 1: Although these publications at times seem to have taken the

45
00:02:21,240 --> 00:02:24,640
Speaker 1: default position of AI's inevitable value, they've begun to sour

46
00:02:24,680 --> 00:02:26,760
Speaker 1: on the idea that it's going to happen anytime soon.

47
00:02:27,160 --> 00:02:30,040
Speaker 1: To be clear, these are sensible, well respected business newspapers,

48
00:02:30,080 --> 00:02:34,840
Speaker 1: hardly partisan voices from the AI wars. Separately, CNBC reported

49
00:02:34,840 --> 00:02:36,920
Speaker 1: that quirked up three hundred cores and open ai is

50
00:02:36,960 --> 00:02:39,320
Speaker 1: either raised or is about to raise another eight point

51
00:02:39,400 --> 00:02:42,760
Speaker 1: three billion dollars in cash, less than two months since

52
00:02:42,760 --> 00:02:45,760
Speaker 1: it raised ten billion dollars from SoftBank and a selection

53
00:02:45,840 --> 00:02:50,000
Speaker 1: of other venture catital firms. While this sounds ambiguous, this

54
00:02:50,080 --> 00:02:52,600
Speaker 1: reporting only claims that it has secured that funding, not

55
00:02:52,639 --> 00:02:55,440
Speaker 1: that it's received it, which is an important distinction. I've

56
00:02:55,480 --> 00:02:57,760
Speaker 1: pointed out in the past that secured funding often comes

57
00:02:57,760 --> 00:03:00,560
Speaker 1: with its own conditions attached. On toime top of this,

58
00:03:00,960 --> 00:03:03,160
Speaker 1: I should add that there is currently a rumor that

59
00:03:03,280 --> 00:03:06,240
Speaker 1: open ai insiders, so the people working at the company,

60
00:03:06,400 --> 00:03:08,480
Speaker 1: are selling six billion dollars with stock at a five

61
00:03:08,560 --> 00:03:11,960
Speaker 1: hundred billion dollar valuation that's around the price of Netflix.

62
00:03:12,000 --> 00:03:14,600
Speaker 1: This is fucking stupid. It's so stupid, and every time

63
00:03:14,639 --> 00:03:16,959
Speaker 1: I read about it, I feel a little insane. But

64
00:03:17,960 --> 00:03:20,320
Speaker 1: it's also weird that you've got people like soft Bank

65
00:03:20,360 --> 00:03:23,239
Speaker 1: and Dragonea who invested in the eight point through billion

66
00:03:23,280 --> 00:03:26,400
Speaker 1: dollar round, who are also buying the stock. All this

67
00:03:26,480 --> 00:03:29,120
Speaker 1: sounds stinky and I hate to be crude. I hate

68
00:03:29,120 --> 00:03:30,679
Speaker 1: to be too crude here, But where the fuck is

69
00:03:30,720 --> 00:03:33,920
Speaker 1: all this money going. Is open ai just incinerating capital?

70
00:03:34,160 --> 00:03:37,000
Speaker 1: Is it compute? Is it salaries? Is it more compute?

71
00:03:37,040 --> 00:03:40,480
Speaker 1: Is it data centers? Because soft Bank isn't actually building

72
00:03:40,040 --> 00:03:44,520
Speaker 1: anything for stargate It's all okay, I calm down a

73
00:03:44,560 --> 00:03:47,440
Speaker 1: little bit, but it's all getting a little bit silly now.

74
00:03:47,480 --> 00:03:49,520
Speaker 1: The information, which has some of the best sourcing in

75
00:03:49,560 --> 00:03:52,600
Speaker 1: any publication covering the AI bubble, i'd argue, though I've

76
00:03:52,600 --> 00:03:55,000
Speaker 1: taken issues with their reporting in the past, suggested that

77
00:03:55,080 --> 00:03:57,640
Speaker 1: open ai intends to use this money to build data centers,

78
00:03:57,800 --> 00:04:01,160
Speaker 1: possibly the only worst investment it can make than general ivai,

79
00:04:01,560 --> 00:04:03,960
Speaker 1: and it's the one that open ai can't really avoid

80
00:04:04,000 --> 00:04:07,280
Speaker 1: because they're also somehow running out of compute two and

81
00:04:07,320 --> 00:04:10,440
Speaker 1: amongst this already ridiculous situation. Since the issue of open

82
00:04:10,440 --> 00:04:13,160
Speaker 1: Ai and Anthropic's actual revenues, which I wrote about my

83
00:04:13,200 --> 00:04:16,360
Speaker 1: premium newsletter on the first of August Please give Me Money,

84
00:04:16,640 --> 00:04:19,000
Speaker 1: and have roughly estimated to be five point twenty six

85
00:04:19,040 --> 00:04:21,000
Speaker 1: billion dollars in the case of open Ai and one

86
00:04:21,000 --> 00:04:24,680
Speaker 1: point five billion dollars in Anthropic as of July. Now,

87
00:04:24,800 --> 00:04:26,799
Speaker 1: to be clear, there are very different numbers going around.

88
00:04:27,080 --> 00:04:29,560
Speaker 1: Get to in a second. In any case, these estimates

89
00:04:29,560 --> 00:04:32,080
Speaker 1: were made based on both companies predilection for leaking their

90
00:04:32,120 --> 00:04:37,240
Speaker 1: annualized revenue or month times twelve. Now, this extremely annoying

91
00:04:37,400 --> 00:04:39,800
Speaker 1: term annualized or recurring revenue is one that I keep

92
00:04:39,800 --> 00:04:42,039
Speaker 1: bringing up because it's become the de facto wait for

93
00:04:42,080 --> 00:04:45,360
Speaker 1: GENERATIVEAI companies to express their revenue, and both open ai

94
00:04:45,360 --> 00:04:47,760
Speaker 1: and Anthropic leaking them intentionally and doing so in a

95
00:04:47,760 --> 00:04:50,640
Speaker 1: way that suggests that they're not even using the traditional

96
00:04:50,680 --> 00:04:53,880
Speaker 1: ways of calculating them. Because arr isn't in and of

97
00:04:53,920 --> 00:04:56,920
Speaker 1: itself an evil thing, it's fairly stand up within software

98
00:04:56,920 --> 00:05:00,960
Speaker 1: as a service companies, except they run completely different to

99
00:05:01,040 --> 00:05:05,760
Speaker 1: open ai and Anthropic anyway. Open ai leaked on July thirtieth,

100
00:05:05,760 --> 00:05:07,920
Speaker 1: twenty twenty five, that it was at twelve billion dollars

101
00:05:07,960 --> 00:05:10,560
Speaker 1: of annualized revenue, so it earned around eight hundred and

102
00:05:10,600 --> 00:05:13,880
Speaker 1: thirty three million dollars in some sort of thirty day period,

103
00:05:14,080 --> 00:05:17,040
Speaker 1: and then two days later, on August first, twenty twenty five,

104
00:05:17,400 --> 00:05:19,719
Speaker 1: the New York Times reported it was at thirteen billion

105
00:05:19,760 --> 00:05:23,719
Speaker 1: dollars annualized, at one point eight billion dollars of monthly revenue.

106
00:05:24,600 --> 00:05:27,600
Speaker 1: It's taking the pierce a little bit. It's very clear

107
00:05:27,680 --> 00:05:30,560
Speaker 1: open AI is not talking in actual calendar months, at

108
00:05:30,560 --> 00:05:33,400
Speaker 1: which point we can assume they're using like a trailing

109
00:05:33,480 --> 00:05:35,960
Speaker 1: thirty day window, as in the month is just thirty days,

110
00:05:36,040 --> 00:05:39,280
Speaker 1: rather than a calendar month like May or June or July.

111
00:05:39,880 --> 00:05:43,600
Speaker 1: We can, however, declaratively say that it's not saying the

112
00:05:43,640 --> 00:05:45,719
Speaker 1: month of June or the month of July was twelve

113
00:05:45,880 --> 00:05:48,479
Speaker 1: or thirteen billion dollars annualized, because if it was, they

114
00:05:48,520 --> 00:05:51,080
Speaker 1: wouldn't have given two vastly different goddamn numbers in the

115
00:05:51,080 --> 00:05:54,000
Speaker 1: same two day period. It doesn't make any sense. And

116
00:05:54,040 --> 00:05:55,840
Speaker 1: to be clear, while I can't say for certain, I

117
00:05:55,839 --> 00:05:58,840
Speaker 1: believe these leaks are deliberate. Open AI is timing matches

118
00:05:58,880 --> 00:06:02,720
Speaker 1: exactly with the fundraise. Similarly, on Anthropic side, these revenues

119
00:06:02,720 --> 00:06:05,920
Speaker 1: are beginning to get really, really weird. Anthropic went from

120
00:06:05,920 --> 00:06:08,760
Speaker 1: making around seventy two million dollars eight hundred and seventy

121
00:06:08,760 --> 00:06:11,920
Speaker 1: five million dollars annualized in January to four hundred and

122
00:06:11,960 --> 00:06:14,880
Speaker 1: thirty three million dollars in monthly revenue in July, or

123
00:06:14,880 --> 00:06:17,120
Speaker 1: at least it leaked on July first, twenty twenty five

124
00:06:17,160 --> 00:06:19,520
Speaker 1: that it was at four billion annuallyzed to the Information,

125
00:06:19,800 --> 00:06:21,880
Speaker 1: which is three hundred and thirty three million dollars, and

126
00:06:21,920 --> 00:06:25,160
Speaker 1: then claimed it had reached five billion annualized four hundred

127
00:06:25,160 --> 00:06:28,760
Speaker 1: and sixteen million to Bloomberg on July twenty ninth, twenty

128
00:06:28,760 --> 00:06:33,840
Speaker 1: twenty five. I'm so tired of this. Something stinky is happening.

129
00:06:34,200 --> 00:06:36,919
Speaker 1: The people who will not admit something stinky is happening,

130
00:06:37,080 --> 00:06:40,320
Speaker 1: I'm beginning to question their sanity or whether they're just

131
00:06:40,400 --> 00:06:43,040
Speaker 1: inherently corrupt in some way. And you can be corrupt

132
00:06:43,080 --> 00:06:45,640
Speaker 1: without money, you can just do it for your mates. Anyway.

133
00:06:46,720 --> 00:06:49,560
Speaker 1: How did Anthropic get there, I'm guessing it was from

134
00:06:49,600 --> 00:06:52,599
Speaker 1: cranking up prices on Cursor, a product made by a

135
00:06:52,600 --> 00:06:55,280
Speaker 1: company called any Sphere that lets developers generate code using

136
00:06:55,320 --> 00:06:58,440
Speaker 1: Anthropics models and other models, and we've had the confirmation

137
00:06:58,560 --> 00:07:00,600
Speaker 1: that's the case thanks to the Information reporting the one

138
00:07:00,600 --> 00:07:03,280
Speaker 1: point four billion dollars of the annualized revenue from Anthropic.

139
00:07:03,279 --> 00:07:05,720
Speaker 1: It's from its two top customers, so around one hundred

140
00:07:05,720 --> 00:07:07,880
Speaker 1: and sixteen million dollars in a month, and the biggest

141
00:07:07,880 --> 00:07:11,320
Speaker 1: of which is Cursor. Confusingly, the Information also says that

142
00:07:11,320 --> 00:07:14,240
Speaker 1: Anthropics Clad Code is generating nearly four hundred million dollars

143
00:07:14,240 --> 00:07:17,360
Speaker 1: in annualized revenue, roughly doubling from just a few weeks ago.

144
00:07:17,440 --> 00:07:19,920
Speaker 1: Per their report, it's around thirty three million dollars a

145
00:07:20,000 --> 00:07:24,200
Speaker 1: monthly revenue. They really jacked that product up. I need

146
00:07:24,240 --> 00:07:27,080
Speaker 1: to do a whole thing on glored code if I'm honest,

147
00:07:27,280 --> 00:07:30,040
Speaker 1: because that whole thing has become a complete nightmare. They've

148
00:07:30,080 --> 00:07:33,520
Speaker 1: had to completely the weekly rate limits are coming. God,

149
00:07:33,520 --> 00:07:35,360
Speaker 1: what a fucking mess. Every time I think about this

150
00:07:35,400 --> 00:07:38,040
Speaker 1: company get upset. In any case, I think Cursor is

151
00:07:38,040 --> 00:07:40,360
Speaker 1: a huge indicator of the current fragility of the bubble

152
00:07:40,440 --> 00:07:42,760
Speaker 1: and the fact that for most AI startups there's simply

153
00:07:42,760 --> 00:07:45,880
Speaker 1: no way out because being acquired or going public does

154
00:07:45,920 --> 00:07:49,000
Speaker 1: not appear to be a viable route. But before I explain,

155
00:07:49,080 --> 00:07:52,160
Speaker 1: please listen to one of the many stable, normal advertisers

156
00:07:52,280 --> 00:07:54,880
Speaker 1: that gives money to the show, not to me, though

157
00:07:55,120 --> 00:08:08,920
Speaker 1: I don't get the money from the ads. And we're back.

158
00:08:09,600 --> 00:08:12,440
Speaker 1: So how exactly is one AI powered coding environment such

159
00:08:12,440 --> 00:08:15,080
Speaker 1: a load bearing part of the AI trade. We're going

160
00:08:15,120 --> 00:08:17,400
Speaker 1: to be focusing now on both the fact that Cursor

161
00:08:17,480 --> 00:08:19,960
Speaker 1: is a systemic risk to the AI industry, one where

162
00:08:20,040 --> 00:08:22,320
Speaker 1: if it fails, could make every other player that bit

163
00:08:22,400 --> 00:08:24,920
Speaker 1: less secure, and the fact that Cursor is emblematic of

164
00:08:24,920 --> 00:08:27,840
Speaker 1: what makes general if AI companies so paradoxical, where you

165
00:08:27,840 --> 00:08:30,240
Speaker 1: can have bonkers valuations on one hand, but also no

166
00:08:30,320 --> 00:08:32,920
Speaker 1: way for these companies to go public or sell to

167
00:08:32,960 --> 00:08:37,120
Speaker 1: someone else and justified the actual valuation itself. I know

168
00:08:37,200 --> 00:08:39,480
Speaker 1: it sounds a little insane, but I believe that Cursor

169
00:08:39,559 --> 00:08:41,679
Speaker 1: is the weak point of the entire bubble, and I'm

170
00:08:41,679 --> 00:08:43,959
Speaker 1: going to explain why and how this could go. This

171
00:08:44,000 --> 00:08:46,560
Speaker 1: is by no means inevitable, but I cannot work out

172
00:08:46,679 --> 00:08:50,640
Speaker 1: what Cursor does other than this Cursor makes before at

173
00:08:50,720 --> 00:08:52,959
Speaker 1: least the massive changes to its service. Though Tom the

174
00:08:53,040 --> 00:08:55,680
Speaker 1: town over at Newcomer suggests that the revenue is still

175
00:08:55,720 --> 00:08:58,320
Speaker 1: going up. For now, they make about five hundred million

176
00:08:58,360 --> 00:09:01,720
Speaker 1: dollars in annualized revenue around forty two million dollars a month.

177
00:09:01,840 --> 00:09:04,439
Speaker 1: This makes it the single highest earning general IFAI company

178
00:09:04,520 --> 00:09:07,959
Speaker 1: that isn't called AI or Anthropic, and the highest learning

179
00:09:08,000 --> 00:09:12,080
Speaker 1: company built on top of primarily Anthropics models. Its success

180
00:09:12,120 --> 00:09:14,800
Speaker 1: is symbolic to the greater AI movement and just as

181
00:09:14,840 --> 00:09:17,240
Speaker 1: it hit its peak, Anthropic and open Ai to a

182
00:09:17,280 --> 00:09:21,400
Speaker 1: lesser extent, though they seem more interconnected. During GPT five's launch,

183
00:09:21,600 --> 00:09:25,440
Speaker 1: decided to add priority processing in priority service tiers, demanding

184
00:09:25,440 --> 00:09:28,680
Speaker 1: more money upfront and causing Cursor to have to massively

185
00:09:28,720 --> 00:09:31,680
Speaker 1: degrade its service to keep up. I also have a

186
00:09:31,679 --> 00:09:33,760
Speaker 1: premium piece that explains this from a few weeks go

187
00:09:33,880 --> 00:09:37,120
Speaker 1: within a monologue two and it's over at wezered dot app.

188
00:09:37,160 --> 00:09:40,520
Speaker 1: Please give me money to explain In short, curses AI

189
00:09:40,559 --> 00:09:43,280
Speaker 1: powered coding Editor used to have fairly unrestrained access to

190
00:09:43,320 --> 00:09:46,839
Speaker 1: the various models provided by companies like open Ai and Anthropic.

191
00:09:48,120 --> 00:09:51,000
Speaker 1: In mid June, a few weeks after Anthropic introduced priority

192
00:09:51,000 --> 00:09:54,040
Speaker 1: tiers that required enterprise companies to pay upfront and guarantee

193
00:09:54,040 --> 00:09:57,360
Speaker 1: a certain throughput of tokens and increase costs on prompt cashing,

194
00:09:57,400 --> 00:10:00,000
Speaker 1: which is a big part of AI coding. Curse mass

195
00:10:00,160 --> 00:10:02,400
Speaker 1: the change the amount its users could use its product

196
00:10:02,520 --> 00:10:05,640
Speaker 1: and introduce the two hundred dollars a month subscription. As

197
00:10:05,679 --> 00:10:08,680
Speaker 1: an aside to this, Anthropic also competes with Curser's AI

198
00:10:08,760 --> 00:10:11,720
Speaker 1: coding product with Clawed Code, its own service. I've talked

199
00:10:11,720 --> 00:10:13,120
Speaker 1: about it in the past, and I'll talk about it

200
00:10:13,160 --> 00:10:16,960
Speaker 1: some more in this series too. Cursor, as Anthropic's largest client,

201
00:10:17,040 --> 00:10:20,480
Speaker 1: the second being GitHub Copilot, represents a material part of

202
00:10:20,520 --> 00:10:23,040
Speaker 1: its revenue, and its surging popularity meant that it was

203
00:10:23,080 --> 00:10:26,480
Speaker 1: sending more and more revenue to Anthropic. Anthropic use this

204
00:10:26,480 --> 00:10:29,760
Speaker 1: opportunity to raise prices on accessing its models to continue

205
00:10:29,920 --> 00:10:33,040
Speaker 1: providing service at an acceptable level to Cursor's customers by

206
00:10:33,080 --> 00:10:36,600
Speaker 1: introducing priority toier access on May thirtieth, twenty twenty five.

207
00:10:37,280 --> 00:10:39,600
Speaker 1: This has allowed Anthropic to duce its revenues, and due

208
00:10:39,640 --> 00:10:42,400
Speaker 1: to the upfront nature of these contracts, Cursor is locked

209
00:10:42,400 --> 00:10:45,280
Speaker 1: in regardless of how well it does. The net result

210
00:10:45,280 --> 00:10:47,720
Speaker 1: of these cost increases means that Curser's product is less

211
00:10:47,720 --> 00:10:50,679
Speaker 1: attractive to its customers and will thus eventually make it

212
00:10:50,800 --> 00:10:53,560
Speaker 1: less money. At this point, one has to ask, how

213
00:10:53,600 --> 00:10:57,160
Speaker 1: does Cursor survive. Its product isn't profitable and the means

214
00:10:57,160 --> 00:10:59,920
Speaker 1: it used to make the company so successful have become untenable.

215
00:11:00,200 --> 00:11:02,640
Speaker 1: It is guaranteed a certain throughput of tokens per second

216
00:11:02,640 --> 00:11:05,280
Speaker 1: to the major model developers. Chief of the meanthropic. But

217
00:11:05,320 --> 00:11:08,720
Speaker 1: Cursor itself said in the whole Cursor Ultra two hundred

218
00:11:08,760 --> 00:11:11,320
Speaker 1: dollars a month thing that it signed multi year deals

219
00:11:11,360 --> 00:11:14,680
Speaker 1: with multiple cloud providers like open Ai, Xai and Google.

220
00:11:15,760 --> 00:11:19,000
Speaker 1: Cursor's product is now worse by the way. They also

221
00:11:19,040 --> 00:11:21,200
Speaker 1: have made it so the automodel that used to be

222
00:11:21,240 --> 00:11:24,200
Speaker 1: unlimited will now as of September fifteenth, charge a fee.

223
00:11:24,280 --> 00:11:26,600
Speaker 1: It's so fucking funny. I love this shit. It's so

224
00:11:26,720 --> 00:11:29,640
Speaker 1: funny watching this happen. But really, people are going to

225
00:11:29,640 --> 00:11:33,640
Speaker 1: cancel their subscriptions, their annualized revenue will drop, and their

226
00:11:33,679 --> 00:11:36,160
Speaker 1: ability to raise capital will suffer. As a direct result,

227
00:11:36,600 --> 00:11:39,000
Speaker 1: Cursor will, regardless of this drop in revenue, have to

228
00:11:39,000 --> 00:11:40,840
Speaker 1: pay the cloud companies what it owes them as if

229
00:11:40,880 --> 00:11:43,079
Speaker 1: it had the business it used to because it was

230
00:11:43,120 --> 00:11:46,160
Speaker 1: a guaranteed throughput. I've spoken to a few different people,

231
00:11:46,160 --> 00:11:48,840
Speaker 1: including a company with an enterprise contract, either planning to

232
00:11:48,840 --> 00:11:50,640
Speaker 1: cancel or trying to find a way out of their

233
00:11:50,679 --> 00:11:52,840
Speaker 1: agreements with Cursor, and they were planning to do so

234
00:11:52,920 --> 00:11:56,600
Speaker 1: as recently as April. I'm sorry recently as April. I

235
00:11:56,640 --> 00:11:58,520
Speaker 1: mean they were planning to do back in April, before

236
00:11:58,559 --> 00:12:02,640
Speaker 1: this shit happened. I'm not a here, you do understand anyway.

237
00:12:02,720 --> 00:12:04,760
Speaker 1: If Cursor is allowed to die, it will be unable

238
00:12:04,800 --> 00:12:07,160
Speaker 1: to pay a chunk of Anthropic and open AI's revenue,

239
00:12:07,160 --> 00:12:10,400
Speaker 1: and yes, the revenue of people like Xai in Google

240
00:12:10,440 --> 00:12:13,520
Speaker 1: as well. It also brings into question whether it's possible

241
00:12:13,559 --> 00:12:16,480
Speaker 1: to build, putting aside any questions of profitability, a business

242
00:12:16,480 --> 00:12:18,760
Speaker 1: of any kind offering services built on top of generative

243
00:12:18,800 --> 00:12:21,400
Speaker 1: AI models, and in turn bring into doubt the veracity

244
00:12:21,440 --> 00:12:24,520
Speaker 1: of investing in this sector writ large. It will also

245
00:12:24,559 --> 00:12:27,200
Speaker 1: call into question whether any other generativai company is a

246
00:12:27,240 --> 00:12:29,960
Speaker 1: real business. This will naturally lead to the question of

247
00:12:29,960 --> 00:12:33,679
Speaker 1: why we're building all these goddamn data centers. Cursor at

248
00:12:33,679 --> 00:12:37,000
Speaker 1: this point faces two options, die or get acquiet. This

249
00:12:37,040 --> 00:12:39,079
Speaker 1: is not an attack on anyone who works at the company,

250
00:12:39,160 --> 00:12:42,080
Speaker 1: nor anything personal. The unit economics of this business do

251
00:12:42,200 --> 00:12:44,840
Speaker 1: not make sense, and yet on some level. Its existence

252
00:12:44,880 --> 00:12:48,000
Speaker 1: is deeply important to the valley's future. And a large

253
00:12:48,040 --> 00:12:51,640
Speaker 1: chunk of anthropics revenue I should add, and Tom totan

254
00:12:51,760 --> 00:12:54,280
Speaker 1: over it, and Newcomer added something like ten percent of

255
00:12:54,320 --> 00:12:58,080
Speaker 1: anthropics revenue comes from Cursor. This is so bad, but

256
00:12:58,160 --> 00:13:00,800
Speaker 1: I'm gonna humor this. I'm gonna hue in this big time.

257
00:13:01,400 --> 00:13:06,200
Speaker 1: Who would actually acquire Cursor? Maybe open Ai, but they

258
00:13:06,480 --> 00:13:10,040
Speaker 1: couldn't acquire Windsurf for another AI company because they were

259
00:13:10,080 --> 00:13:12,680
Speaker 1: too worried that Microsoft was going to get the somehow

260
00:13:12,840 --> 00:13:15,360
Speaker 1: essential ip of what appears to be one of one

261
00:13:15,400 --> 00:13:19,200
Speaker 1: hundred different AI powered coding environments. They also already trade

262
00:13:19,200 --> 00:13:22,040
Speaker 1: and failed to buy Cursor. And if I'm honest, I

263
00:13:22,080 --> 00:13:23,880
Speaker 1: would not be surprised if Curser would sell.

264
00:13:23,920 --> 00:13:24,079
Speaker 2: Now.

265
00:13:24,160 --> 00:13:26,120
Speaker 1: I don't know if open ai has the money, but

266
00:13:26,240 --> 00:13:30,760
Speaker 1: you know they probably would sell. Honestly, Cursor fucked up

267
00:13:30,800 --> 00:13:33,400
Speaker 1: bad not selling to open Ai. They could have got

268
00:13:33,400 --> 00:13:35,240
Speaker 1: ten billion dollars in Sam Mortman would have had to

269
00:13:35,280 --> 00:13:38,840
Speaker 1: accelerate the funding clause. It would have been so goddamn sick.

270
00:13:38,880 --> 00:13:41,760
Speaker 1: But now the only sick here is Cursor's fragile, plagued

271
00:13:41,800 --> 00:13:45,480
Speaker 1: business model. But how about Anthropic. I don't know about

272
00:13:45,480 --> 00:13:49,080
Speaker 1: that one either. They've already got their own extremely expensive

273
00:13:49,080 --> 00:13:52,280
Speaker 1: coding environment, which I estimated in my premium newsletter loses

274
00:13:52,320 --> 00:13:54,720
Speaker 1: the company one hundred percent to ten thousand percent of

275
00:13:54,720 --> 00:13:59,000
Speaker 1: a subscript subscription per customer. That's like a monthly subscription

276
00:13:59,679 --> 00:14:01,560
Speaker 1: just from us. And they did that a few weeks ago.

277
00:14:01,640 --> 00:14:04,120
Speaker 1: And now Anthropic, as of August twenty eighth, is adding

278
00:14:04,120 --> 00:14:07,080
Speaker 1: weekly limits on accounts, which I believe creates some of

279
00:14:07,080 --> 00:14:09,880
Speaker 1: the most gnarly churn in SaaS history. It's going to

280
00:14:09,920 --> 00:14:13,400
Speaker 1: be quite nice. Also, does Anthropic really want to acquire

281
00:14:13,400 --> 00:14:18,880
Speaker 1: its largest customer also with more money. They're not raising

282
00:14:18,960 --> 00:14:22,760
Speaker 1: five billion dollars to bail out Cursor. Anthropic needs that

283
00:14:22,800 --> 00:14:25,960
Speaker 1: to feed directly into Andy Jesse's ass pocket to keep

284
00:14:26,000 --> 00:14:29,000
Speaker 1: offering increasingly more complex models that never quite seemed good

285
00:14:29,080 --> 00:14:32,360
Speaker 1: enough to make any money. But how about Google? They

286
00:14:33,200 --> 00:14:36,920
Speaker 1: know they just sort of bought Windsurf, but I'll get

287
00:14:37,000 --> 00:14:39,520
Speaker 1: it's a complete mess, and they can't do that again.

288
00:14:39,920 --> 00:14:43,640
Speaker 1: They've already given out the participation trophy multiple billions of

289
00:14:43,640 --> 00:14:46,000
Speaker 1: dollars to investors and founders, so nobody has to get

290
00:14:46,040 --> 00:14:48,880
Speaker 1: embarrassed about this. And then allowed Cognition to pick up

291
00:14:48,920 --> 00:14:50,880
Speaker 1: the scraps of a business that made six point eighty

292
00:14:50,880 --> 00:14:53,440
Speaker 1: three million dollars a month after burning one hundred and

293
00:14:53,480 --> 00:14:56,640
Speaker 1: forty three million of investor capital, and tech runch reports

294
00:14:56,680 --> 00:14:58,840
Speaker 1: that Windsurf was left for about one hundred million dollars

295
00:14:58,840 --> 00:15:02,400
Speaker 1: in cash post acquisition. Tech Crunch also reports that Cognition

296
00:15:02,440 --> 00:15:04,480
Speaker 1: paid about two hundred and fifty million dollars for what

297
00:15:04,560 --> 00:15:08,600
Speaker 1: remained of the company. Google paid like two point something

298
00:15:08,680 --> 00:15:12,360
Speaker 1: billion dollars to just take the founders, and Cognition picked

299
00:15:12,400 --> 00:15:14,600
Speaker 1: up the rest. And this deal, by the way, didn't

300
00:15:14,600 --> 00:15:18,520
Speaker 1: actually pay out the majority of Windsurf's employees. He I'll

301
00:15:18,520 --> 00:15:20,640
Speaker 1: have the links to the links to all of this

302
00:15:20,720 --> 00:15:23,280
Speaker 1: in the notes. This whole deal is real sickly though,

303
00:15:23,680 --> 00:15:26,440
Speaker 1: that the investors got paid out, the founders got paid out,

304
00:15:26,520 --> 00:15:29,320
Speaker 1: the employees got fugged. It really sucks. I'll get into

305
00:15:29,320 --> 00:15:31,880
Speaker 1: it in a bit, but let's go with meta. Here's

306
00:15:31,920 --> 00:15:34,440
Speaker 1: the thing. If I'm Michael Truell, the CEO of Cursor

307
00:15:34,480 --> 00:15:37,240
Speaker 1: of any Sphere, even I am calling Mark Zuckerberg, and

308
00:15:37,240 --> 00:15:38,800
Speaker 1: I'm pretending that I think the only person in the

309
00:15:38,840 --> 00:15:41,280
Speaker 1: world who can usher in superintelligence is the guy who

310
00:15:41,320 --> 00:15:43,600
Speaker 1: burned more than forty five billion dollars on the metaverse

311
00:15:43,720 --> 00:15:45,920
Speaker 1: and believes that not wearing AI glasses in the future

312
00:15:45,920 --> 00:15:48,480
Speaker 1: will be a disadvantage. I would say all manner of

313
00:15:48,480 --> 00:15:50,440
Speaker 1: shit about the future, and that the only way to

314
00:15:50,480 --> 00:15:52,680
Speaker 1: do this was to buy my AI powered coding startup

315
00:15:52,680 --> 00:15:57,080
Speaker 1: that literally cannot afford to exist. I would tell Zuckerberg anything.

316
00:15:57,080 --> 00:15:59,560
Speaker 1: I would be like, mate, I just watched the actual

317
00:15:59,600 --> 00:16:03,080
Speaker 1: star I think you're the people from Stargate. He's never

318
00:16:03,120 --> 00:16:05,200
Speaker 1: gonna watch Stargate. He's just gonna be like, whoa, yeah,

319
00:16:05,760 --> 00:16:09,400
Speaker 1: I've heard that's a TV show. Wow, A great. And

320
00:16:09,440 --> 00:16:11,640
Speaker 1: then you get however much money you need, just twenty

321
00:16:11,680 --> 00:16:14,040
Speaker 1: billion dollars, He'll just pull that out of his pocket.

322
00:16:14,080 --> 00:16:16,520
Speaker 1: He has it. He has twenty billion. He'll just handed

323
00:16:16,520 --> 00:16:18,600
Speaker 1: to you. He's probably got it in is like in

324
00:16:18,640 --> 00:16:20,600
Speaker 1: his safe. Just go and ask mine, how much money

325
00:16:20,640 --> 00:16:22,800
Speaker 1: do you think Mark Zuckerberg's got his safe? Email me,

326
00:16:23,040 --> 00:16:27,120
Speaker 1: email me with the guests. Anyway, This whole can't afford

327
00:16:27,120 --> 00:16:30,080
Speaker 1: to exist thing that really is the problem. These companies

328
00:16:30,080 --> 00:16:32,200
Speaker 1: are all going through the same motions that every company

329
00:16:32,240 --> 00:16:34,800
Speaker 1: before them did. Raise as much money as possible, get

330
00:16:34,840 --> 00:16:37,000
Speaker 1: as big as possible and eventually scale to the point

331
00:16:37,080 --> 00:16:40,600
Speaker 1: of the e fat with enterprise cash. Except the real

332
00:16:40,640 --> 00:16:43,160
Speaker 1: problem is that, just like the Big text New gluttony

333
00:16:43,200 --> 00:16:46,080
Speaker 1: of physical real estate, it's taken on general ifai company

334
00:16:46,120 --> 00:16:48,200
Speaker 1: is a burden with a constant aggressive form of cloud

335
00:16:48,280 --> 00:16:51,520
Speaker 1: damn the endless punishment of the costs for accessing the

336
00:16:51,560 --> 00:16:54,360
Speaker 1: API for general if ai models that always seem to

337
00:16:54,360 --> 00:16:56,240
Speaker 1: get a little bit better, but never in such a

338
00:16:56,280 --> 00:16:59,360
Speaker 1: way that anything really changes other than how much Anthropic

339
00:16:59,440 --> 00:17:01,280
Speaker 1: and open Ai are going to need at the end

340
00:17:01,280 --> 00:17:03,800
Speaker 1: of the month, or they break your startup's legs. I'm

341
00:17:03,800 --> 00:17:06,600
Speaker 1: not even trying to be funny. Anthropic raised its prices

342
00:17:06,600 --> 00:17:09,800
Speaker 1: on curses so severely it broke its already unprofitable business model.

343
00:17:10,040 --> 00:17:12,400
Speaker 1: These products, while also for the most part not producing

344
00:17:12,400 --> 00:17:14,760
Speaker 1: that much revenue, need to be sold with users being

345
00:17:14,800 --> 00:17:18,080
Speaker 1: aware of and or sensitive to the costs of providing them.

346
00:17:18,400 --> 00:17:20,960
Speaker 1: The Cursor's original product was a twenty dollars a month

347
00:17:21,240 --> 00:17:24,040
Speaker 1: service for five hundred fast requests of different models, in

348
00:17:24,040 --> 00:17:26,480
Speaker 1: the same way that accessing clawed code on any subscription

349
00:17:26,560 --> 00:17:28,840
Speaker 1: is either twenty one hundred or two hundred dollars a month.

350
00:17:29,119 --> 00:17:32,000
Speaker 1: Rather than paying per API cores, because these companies also

351
00:17:32,160 --> 00:17:34,840
Speaker 1: products that shield the customer from the actual costs of

352
00:17:34,880 --> 00:17:38,879
Speaker 1: running the services, and those costs could be vast. One article,

353
00:17:38,880 --> 00:17:40,920
Speaker 1: which I'll link to in the spreadsheet, argues that given

354
00:17:40,960 --> 00:17:43,320
Speaker 1: the trajectory of the development of AI models, it's not

355
00:17:43,400 --> 00:17:46,320
Speaker 1: unreasonable to imagine a full time engineer spending one hundred

356
00:17:46,400 --> 00:17:52,040
Speaker 1: thousand dollars in tokens each year per engineer. Just really

357
00:17:52,080 --> 00:17:54,480
Speaker 1: think about that for a second. I'll have the link

358
00:17:54,560 --> 00:17:57,919
Speaker 1: in there. It's fucking stupid. It's so fucking stupid, I

359
00:17:58,000 --> 00:18:01,200
Speaker 1: want to scream. And the irony is that, despite being

360
00:18:01,200 --> 00:18:03,920
Speaker 1: willing to kill these companies by fundamentally changing the terms

361
00:18:03,960 --> 00:18:06,520
Speaker 1: and put in which they access their models, Anthropic is

362
00:18:06,560 --> 00:18:10,040
Speaker 1: also in some way dependent on Cursor, Repler and other

363
00:18:10,119 --> 00:18:12,800
Speaker 1: similar firms continuing to buy tokens at the same rate

364
00:18:12,840 --> 00:18:15,720
Speaker 1: as before, as that consumption is baked into their own

365
00:18:16,119 --> 00:18:18,880
Speaker 1: arr figures, as well as the forward looking revenue projections

366
00:18:18,920 --> 00:18:22,120
Speaker 1: they've given to investors to convince them to give them

367
00:18:22,160 --> 00:18:25,800
Speaker 1: five billion dollars. It is, in some sense of Kobyashi

368
00:18:25,920 --> 00:18:29,080
Speaker 1: maru Anthropic is an existential need to screw over its

369
00:18:29,080 --> 00:18:32,199
Speaker 1: customers by hiking rates and imposing long term commitments, but

370
00:18:32,240 --> 00:18:35,320
Speaker 1: its existence is also in some way predicated on these

371
00:18:35,320 --> 00:18:38,919
Speaker 1: companies continuing to exist. If Cursor and Replit both die,

372
00:18:39,119 --> 00:18:42,760
Speaker 1: that's a significant chunk of anthropics API business gone, and

373
00:18:42,800 --> 00:18:46,240
Speaker 1: may I remind you that significantly overshadows its subscription business,

374
00:18:46,240 --> 00:18:48,800
Speaker 1: making it almost like an inverse of open AI, where

375
00:18:48,840 --> 00:18:52,240
Speaker 1: subscriptions drive the bulk of revenue, also a shitty business model.

376
00:18:52,800 --> 00:18:56,760
Speaker 1: Anthropic's future is wedded to Cursor, and I just don't

377
00:18:56,800 --> 00:18:59,560
Speaker 1: see how Cursor survives, let alone exists or gets subsumed

378
00:18:59,560 --> 00:19:01,840
Speaker 1: by another company in a way that mirrors our acquisitions

379
00:19:01,840 --> 00:19:05,439
Speaker 1: have worked since Ever, if Curser does not sell for

380
00:19:05,440 --> 00:19:08,359
Speaker 1: a healthy amount, I'm talking ten to twenty billion dollars,

381
00:19:08,400 --> 00:19:10,760
Speaker 1: and I mean actually sell, not the founders are hired

382
00:19:10,760 --> 00:19:13,119
Speaker 1: and a strange contractual agreement that pays out investors and

383
00:19:13,160 --> 00:19:16,320
Speaker 1: its assets are sold to Rick from porn Stars. It's pawn.

384
00:19:16,359 --> 00:19:18,960
Speaker 1: By the way, it will prove that no generative AI

385
00:19:19,040 --> 00:19:23,400
Speaker 1: company to this date has actually been successful in reality.

386
00:19:23,440 --> 00:19:26,119
Speaker 1: I expect a Chumley esque deal that helps CEO Michael

387
00:19:26,160 --> 00:19:29,280
Speaker 1: Truell by a Porsche while his staff makes nothing. It's

388
00:19:29,280 --> 00:19:31,960
Speaker 1: Cursor worth ten billion dollars or more. No, no matter

389
00:19:31,960 --> 00:19:33,840
Speaker 1: how good its product may or may not be, it

390
00:19:33,880 --> 00:19:35,320
Speaker 1: is not good enough to be sold at a price

391
00:19:35,359 --> 00:19:38,119
Speaker 1: that doesn't require Cursor to incinerate hundreds of millions of

392
00:19:38,160 --> 00:19:42,000
Speaker 1: dollars with no end in sight. And this ultimately gives

393
00:19:42,080 --> 00:19:46,480
Speaker 1: us a real conundrum. Why aren't generative AI startups selling

394
00:19:47,280 --> 00:19:49,600
Speaker 1: I'll tell you what, though, other people are selling stuff

395
00:19:49,600 --> 00:19:51,159
Speaker 1: and it's time for you to buy it. Listen to

396
00:19:51,240 --> 00:19:52,719
Speaker 1: these ads and let me know what you think at

397
00:19:52,720 --> 00:20:02,840
Speaker 1: my email address, which is Robert Evans app and we're back. So, really, though,

398
00:20:02,960 --> 00:20:06,399
Speaker 1: why are there so few generative AI acquisitions? I know

399
00:20:06,440 --> 00:20:07,960
Speaker 1: you're probably going to assume that there have been a

400
00:20:07,960 --> 00:20:11,399
Speaker 1: ton of them, right, if you've not been obsessively reading

401
00:20:11,480 --> 00:20:14,080
Speaker 1: every single story about that for over a year, you

402
00:20:14,200 --> 00:20:17,639
Speaker 1: might have missed it. Book before we go any further,

403
00:20:17,680 --> 00:20:21,680
Speaker 1: there have been some, but they've been sparse, and they

404
00:20:21,680 --> 00:20:24,399
Speaker 1: often have a weird structure. That's typically the exception, not

405
00:20:24,520 --> 00:20:27,879
Speaker 1: the rule in tech. So here's a real one. AMD

406
00:20:28,040 --> 00:20:30,919
Speaker 1: bought Silo AI, the largest private AI lab in Europe,

407
00:20:30,960 --> 00:20:33,000
Speaker 1: in August twenty twenty four, for six hundred and sixty

408
00:20:33,040 --> 00:20:35,480
Speaker 1: five million dollars, which appears to be the only real

409
00:20:35,520 --> 00:20:38,200
Speaker 1: acquisition in generative AI history and appears to be partially

410
00:20:38,200 --> 00:20:42,639
Speaker 1: based on Silo's use of AMD's GPUs Elsewhere, n Vidio

411
00:20:42,680 --> 00:20:45,240
Speaker 1: bought Octo AI for an estimated two hundred and fifty

412
00:20:45,240 --> 00:20:48,560
Speaker 1: million dollars in September twenty twenty four, after buying Brev

413
00:20:48,640 --> 00:20:51,639
Speaker 1: dot Dev in July twenty twenty four for an undisclosed sum,

414
00:20:51,720 --> 00:20:54,840
Speaker 1: and then Gretel in March twenty twenty five. Gretel, Brev

415
00:20:55,040 --> 00:20:59,240
Speaker 1: dot Dev, Octa AI. Jesus fucking Christ, these names be normal.

416
00:20:59,320 --> 00:21:03,000
Speaker 1: Be normal for one minute. Yeah. In all three cases,

417
00:21:03,040 --> 00:21:05,639
Speaker 1: these products are used to deploy generative AI and not

418
00:21:05,720 --> 00:21:08,760
Speaker 1: products built on top of generative AI or AI models.

419
00:21:09,000 --> 00:21:12,960
Speaker 1: Canva bought generative AI content and research company Leonardo dot

420
00:21:12,960 --> 00:21:15,720
Speaker 1: Ai in July twenty twenty four from this closed sum.

421
00:21:15,920 --> 00:21:19,520
Speaker 1: But that's small frame. Really. The only significant one I've

422
00:21:19,560 --> 00:21:22,440
Speaker 1: seen was on July twenty ninth, twenty twenty five, publicly

423
00:21:22,520 --> 00:21:27,120
Speaker 1: trained did customer service platform Nice buying AI powered customer

424
00:21:27,240 --> 00:21:31,600
Speaker 1: service company Cognigee in a nine hundred and fifty five

425
00:21:31,840 --> 00:21:34,720
Speaker 1: million dollar deal. I hate these names so much. According

426
00:21:34,720 --> 00:21:38,320
Speaker 1: to c X today, Cognitigi expects about eighty five million

427
00:21:38,359 --> 00:21:40,680
Speaker 1: dollars in revenue this year, though nobody appears to be

428
00:21:40,720 --> 00:21:43,359
Speaker 1: talking about how much it costs to make that revenue. However,

429
00:21:43,480 --> 00:21:46,560
Speaker 1: according to some sources, they judged tens of thousands or

430
00:21:46,640 --> 00:21:49,240
Speaker 1: hundreds of thousands of dollars per contract for their AI

431
00:21:49,359 --> 00:21:52,159
Speaker 1: voice agents that can understand and respond to user input

432
00:21:52,240 --> 00:21:56,120
Speaker 1: in a natural way. Great. Great, we've got one real

433
00:21:56,160 --> 00:21:59,240
Speaker 1: deal company built on models acquisition, and it's a company

434
00:21:59,240 --> 00:22:01,639
Speaker 1: that most people have heard of making around seven million

435
00:22:01,680 --> 00:22:03,879
Speaker 1: dollars a month. But let's take a look at the others.

436
00:22:04,240 --> 00:22:07,080
Speaker 1: So you probably remember Inflection AI to Microsoft, which was

437
00:22:07,119 --> 00:22:09,280
Speaker 1: not an acquisition but a six hundred and fifty million

438
00:22:09,320 --> 00:22:12,159
Speaker 1: dollar licensing deal that, according to Fast Company, may be

439
00:22:12,280 --> 00:22:14,199
Speaker 1: more like a billion dollars when you include things like

440
00:22:14,200 --> 00:22:16,640
Speaker 1: how much it paid to Inflection CEO and former DeepMind

441
00:22:16,640 --> 00:22:19,840
Speaker 1: co founded mister Foss Suleiman, who is most famous for

442
00:22:19,960 --> 00:22:22,360
Speaker 1: just being an abusive and horrible guy to work with.

443
00:22:23,000 --> 00:22:26,399
Speaker 1: Here he's pushing a horrifying nine ninety six culture and

444
00:22:26,520 --> 00:22:30,120
Speaker 1: Microsoft two to do what Muster Phi, You fucking prick. Anyway,

445
00:22:30,320 --> 00:22:32,760
Speaker 1: according to Fast Company, the deal involves a license to

446
00:22:32,800 --> 00:22:36,119
Speaker 1: sell Inflections models, a waiver against any employment claims against

447
00:22:36,160 --> 00:22:39,359
Speaker 1: inflection or Microsoft paying off investors in some sort of

448
00:22:39,400 --> 00:22:43,600
Speaker 1: unnamed compensation for employees. Sounds stinky to me, but let's

449
00:22:43,600 --> 00:22:46,240
Speaker 1: look at Twinsurf to Google and of course Cognition, which

450
00:22:46,280 --> 00:22:49,159
Speaker 1: was also not an acquisition. Windsurf c suite went to

451
00:22:49,160 --> 00:22:51,440
Speaker 1: Google for two point four billion dollars, which paid them

452
00:22:51,480 --> 00:22:53,520
Speaker 1: off along with their investors, and then the rest of

453
00:22:53,600 --> 00:22:55,439
Speaker 1: the staff and the product got acquired by Cognition for

454
00:22:55,440 --> 00:22:59,080
Speaker 1: two hundred and fifteen million dollars. According to tech Crunch,

455
00:22:59,080 --> 00:23:01,240
Speaker 1: investors made one point two billion dollars on the deal,

456
00:23:01,240 --> 00:23:05,359
Speaker 1: with Windsurf co founders Van Mahan and Douglas Chen making

457
00:23:05,400 --> 00:23:07,840
Speaker 1: another one point two billion dollars, and it's staff getting

458
00:23:07,840 --> 00:23:09,800
Speaker 1: to start a new job at a different company building

459
00:23:09,840 --> 00:23:12,639
Speaker 1: something else with According to tech Crunch, a large portion

460
00:23:12,720 --> 00:23:16,280
Speaker 1: of Windsurf's approximately two hundred and fifty employees not benefiting

461
00:23:16,320 --> 00:23:18,800
Speaker 1: from the deal as on the side Mahan and Chen

462
00:23:18,920 --> 00:23:22,879
Speaker 1: fucking suck. I fucking hate these guys, absolute loser psychopaths.

463
00:23:23,000 --> 00:23:25,040
Speaker 1: You make one point two billion dollars and you couldn't

464
00:23:25,040 --> 00:23:29,160
Speaker 1: break off a little bit from that, You fucking ah disgusting.

465
00:23:29,720 --> 00:23:33,400
Speaker 1: You can can't crack that two hundred billion. The point two.

466
00:23:33,560 --> 00:23:35,320
Speaker 1: You can take two hundred million dollars, share it with

467
00:23:35,359 --> 00:23:38,600
Speaker 1: the class. You fucking losers, assholes, pieces of shit. Come

468
00:23:38,640 --> 00:23:40,159
Speaker 1: on the show if you want to hear more words

469
00:23:40,200 --> 00:23:44,080
Speaker 1: like that, you fucking assholes. Anyway, let's talk about Ioproducts

470
00:23:44,080 --> 00:23:46,840
Speaker 1: going to open Ai in an all stock acquisition. This

471
00:23:46,960 --> 00:23:49,560
Speaker 1: dealers a fast and it's unclear of open Ai actually

472
00:23:49,560 --> 00:23:52,760
Speaker 1: bought anything six point four billion dollars in stock for what.

473
00:23:53,280 --> 00:23:55,760
Speaker 1: Johnny Ive's weird face staring at you lovingly as he

474
00:23:55,800 --> 00:23:58,320
Speaker 1: says things like I think we should make it look

475
00:23:58,440 --> 00:24:00,720
Speaker 1: like a circle while making a five million dollars a

476
00:24:00,800 --> 00:24:03,280
Speaker 1: year's salary. Get out of here. It's not real money.

477
00:24:03,840 --> 00:24:06,159
Speaker 1: And then, of course this character dot Ai to Google.

478
00:24:06,240 --> 00:24:09,159
Speaker 1: This was also not an acquisition. Google, to quote the

479
00:24:09,160 --> 00:24:11,720
Speaker 1: Wall Street Journal, paid two point seven billion dollars to

480
00:24:11,760 --> 00:24:14,480
Speaker 1: bring back an AI genius who quit in frustration. I

481
00:24:14,480 --> 00:24:17,639
Speaker 1: would not use the term genius. Well, I don't like

482
00:24:17,760 --> 00:24:20,720
Speaker 1: that term. Just to be clear, Shazir was one of

483
00:24:20,720 --> 00:24:22,840
Speaker 1: the authors of the original Attention is All You Need

484
00:24:22,880 --> 00:24:27,600
Speaker 1: paper that began the transformer based large language model era. Nevertheless,

485
00:24:27,640 --> 00:24:30,200
Speaker 1: much like Inflection, Google paid a licensing fee to character

486
00:24:30,240 --> 00:24:32,320
Speaker 1: dot Ai for its models and hired, according to the

487
00:24:32,320 --> 00:24:35,240
Speaker 1: information it's co founders and many of its engineers, creating

488
00:24:35,240 --> 00:24:37,199
Speaker 1: a fund that would pay out vesting shares as in

489
00:24:37,240 --> 00:24:39,240
Speaker 1: the shares you're given when you join a company that

490
00:24:39,280 --> 00:24:41,320
Speaker 1: you accrue over time as you work there until July

491
00:24:41,359 --> 00:24:45,000
Speaker 1: twenty twenty six. Anything after that you fugged. And now,

492
00:24:45,119 --> 00:24:48,760
Speaker 1: outside of one very industry specific acquisition, there just doesn't

493
00:24:48,760 --> 00:24:51,880
Speaker 1: seem to be an investor with the hunger to buy

494
00:24:51,920 --> 00:24:54,680
Speaker 1: a company like Cursor value that nine point nine billion

495
00:24:54,680 --> 00:24:58,080
Speaker 1: dollars or more if they raise another round, And you

496
00:24:58,160 --> 00:25:01,639
Speaker 1: have to ask why. If ai is has promised the

497
00:25:01,680 --> 00:25:04,240
Speaker 1: thing that will radically change our economy and these companies

498
00:25:04,240 --> 00:25:06,400
Speaker 1: are building the tools that will bring about that change,

499
00:25:06,600 --> 00:25:10,560
Speaker 1: why does nobody want to buy them? And in broader terms,

500
00:25:10,880 --> 00:25:14,040
Speaker 1: what does it mean when these companies, those with ten

501
00:25:14,040 --> 00:25:15,840
Speaker 1: billion dollars are, in the case of open Ai, three

502
00:25:15,960 --> 00:25:20,600
Speaker 1: hundred or even five hundred billion dollar valuations. What does

503
00:25:20,640 --> 00:25:22,959
Speaker 1: it mean when they can't be born and can't go public.

504
00:25:23,000 --> 00:25:26,200
Speaker 1: Where does all this go? What happens next? What's the

505
00:25:26,240 --> 00:25:29,320
Speaker 1: plan here? How will venture firms that plow billions of

506
00:25:29,359 --> 00:25:32,360
Speaker 1: dollars into these businesses bring a return for their LPs.

507
00:25:32,720 --> 00:25:35,159
Speaker 1: That's the limited partners who invest in the venture capitalists.

508
00:25:35,200 --> 00:25:38,200
Speaker 1: If there are no IBOs and no real buyouts, Sure

509
00:25:38,240 --> 00:25:41,000
Speaker 1: investors got paid in these deals, but these deals will

510
00:25:41,000 --> 00:25:44,640
Speaker 1: like pulling teeth. These are like if they do enough

511
00:25:44,680 --> 00:25:47,240
Speaker 1: of these, there will, even in this administration, be an

512
00:25:47,240 --> 00:25:51,920
Speaker 1: antitrust thing. Nevertheless, the economic implications of these questions are

513
00:25:51,960 --> 00:25:55,240
Speaker 1: quite frankly terrifying, especially when you consider the importance that

514
00:25:55,359 --> 00:25:58,440
Speaker 1: VC is historically held in building the US tech ecosystem,

515
00:25:58,680 --> 00:26:00,840
Speaker 1: And they raise further questions about the impact of an

516
00:26:00,840 --> 00:26:03,439
Speaker 1: AI bubble on companies that are promising and do have

517
00:26:03,480 --> 00:26:05,840
Speaker 1: a viable business model and a product with a natural fit,

518
00:26:06,000 --> 00:26:07,879
Speaker 1: but won't be able to raise any cash because everyone's

519
00:26:07,880 --> 00:26:11,760
Speaker 1: putting it into fucking AI. But ed great ed? What

520
00:26:11,840 --> 00:26:15,359
Speaker 1: if Cursor turns profitable? Now great awesome? I would believe

521
00:26:15,400 --> 00:26:18,400
Speaker 1: if it was possible, if it ever ever happened, which

522
00:26:18,440 --> 00:26:21,520
Speaker 1: it's not. I'm not even being sarcastic or rude. It's

523
00:26:21,600 --> 00:26:25,600
Speaker 1: just not happened. No company that actually stakes their entire

524
00:26:25,640 --> 00:26:27,679
Speaker 1: product and genera if AI appears to be able to

525
00:26:27,680 --> 00:26:30,399
Speaker 1: make a profit of any kind, glean a company that

526
00:26:30,440 --> 00:26:32,720
Speaker 1: makes the best eight point three million dollars a month,

527
00:26:32,720 --> 00:26:35,359
Speaker 1: one hundred million dollars an anionized revenue, said that it

528
00:26:35,400 --> 00:26:38,280
Speaker 1: had five hundred and fifty million dollars in cash in

529
00:26:38,320 --> 00:26:40,840
Speaker 1: December of last year, then raised one hundred and fifty

530
00:26:40,840 --> 00:26:43,359
Speaker 1: million dollars in June of this year. Where'd the money go?

531
00:26:44,040 --> 00:26:46,679
Speaker 1: Why does a generative search product with revenues that are

532
00:26:46,760 --> 00:26:49,320
Speaker 1: less than a third of the Cincinnati Reds baseball team

533
00:26:49,640 --> 00:26:52,040
Speaker 1: need a half billion dollars to make eight point three

534
00:26:52,080 --> 00:26:54,440
Speaker 1: billion dollar eight point three million dollars a month. Even

535
00:26:55,000 --> 00:26:57,520
Speaker 1: I'm not even saying these companies are unnecessary so much

536
00:26:57,520 --> 00:26:59,560
Speaker 1: as they may very well be impossible to run as

537
00:26:59,600 --> 00:27:02,760
Speaker 1: real bits businesses. This isn't even a qualitative judgment of

538
00:27:02,760 --> 00:27:05,480
Speaker 1: any one generative AI company. I'm just saying if any

539
00:27:05,520 --> 00:27:08,480
Speaker 1: of these were good businesses, they would be either profitable

540
00:27:08,600 --> 00:27:11,600
Speaker 1: or be acquired in actual deals, and there would also

541
00:27:11,760 --> 00:27:14,399
Speaker 1: be good businesses by now it is not early. That

542
00:27:14,560 --> 00:27:17,880
Speaker 1: argument is stupid. The amount of cash that these businesses

543
00:27:17,880 --> 00:27:20,440
Speaker 1: are burning does not suggest that they're rapidly approaching any

544
00:27:20,560 --> 00:27:23,040
Speaker 1: kind of sane burn rate, or we would have heard

545
00:27:23,480 --> 00:27:26,399
Speaker 1: putting aside any kind of skepticism. I have anything you

546
00:27:26,400 --> 00:27:28,439
Speaker 1: may hold against me for what I say or the

547
00:27:28,440 --> 00:27:30,960
Speaker 1: way I say it. I ask you this, where are

548
00:27:31,040 --> 00:27:34,200
Speaker 1: the profitable companies? Why isn't there one outside of the

549
00:27:34,240 --> 00:27:37,840
Speaker 1: companies creating training data or in Nvidia. We're three years

550
00:27:37,840 --> 00:27:40,960
Speaker 1: in and we haven't had one, one one of them,

551
00:27:41,080 --> 00:27:45,640
Speaker 1: one one. We've also had no real exits and no IPOs.

552
00:27:45,960 --> 00:27:48,520
Speaker 1: There's been no course for celebration, no validation of a

553
00:27:48,520 --> 00:27:51,760
Speaker 1: business model through another company deciding it was necessary to

554
00:27:51,760 --> 00:27:54,520
Speaker 1: continue its dominance by raising funds on the public market

555
00:27:54,680 --> 00:27:58,320
Speaker 1: or allowing actual investors thought they may be, to act

556
00:27:58,320 --> 00:28:01,080
Speaker 1: as a determiner of their value. I also will say

557
00:28:01,119 --> 00:28:04,240
Speaker 1: that very few, if not all, of these companies, they

558
00:28:04,240 --> 00:28:06,880
Speaker 1: would all just turn into a pillar of salt if

559
00:28:06,920 --> 00:28:08,840
Speaker 1: you made them fill out on S one, which is

560
00:28:08,880 --> 00:28:11,800
Speaker 1: the document to go public. It's also unclear what the

561
00:28:11,800 --> 00:28:15,400
Speaker 1: addition of windsurfs intellectual property actually adds the cognition, much

562
00:28:15,440 --> 00:28:18,280
Speaker 1: like it's a little unclear what differentiates cognitions so called

563
00:28:18,280 --> 00:28:22,040
Speaker 1: aipowered software engineered Devin from anything else on the market.

564
00:28:22,600 --> 00:28:24,600
Speaker 1: I had Goldman sachs Is paying for it and said

565
00:28:24,640 --> 00:28:27,440
Speaker 1: the stupidest shit I've ever heard, the CNBC that nevertheless

566
00:28:27,480 --> 00:28:30,120
Speaker 1: shows how little it's actually paying for. This is an

567
00:28:30,119 --> 00:28:34,399
Speaker 1: actual quote from an actual human being. We're going to

568
00:28:34,440 --> 00:28:36,879
Speaker 1: start augmenting our workforce with Devin, which is going to

569
00:28:36,920 --> 00:28:39,000
Speaker 1: be like our new employee, who's going to start doing

570
00:28:39,000 --> 00:28:41,959
Speaker 1: stuff on behalf of our developers are gent told CNBC

571
00:28:42,200 --> 00:28:44,320
Speaker 1: initially we will have hundreds of devons, and that might

572
00:28:44,320 --> 00:28:47,760
Speaker 1: go into the thousands depending on the use cases. Hundreds

573
00:28:47,800 --> 00:28:50,800
Speaker 1: of devons could mean hundreds of seats at are very optimistic,

574
00:28:50,840 --> 00:28:54,120
Speaker 1: five hundred users at the highest end possible at five

575
00:28:54,160 --> 00:28:56,680
Speaker 1: hundred dollars a month, and more than likely it's twenty

576
00:28:56,680 --> 00:29:00,800
Speaker 1: dollars a month. If not less. Let's assume that it

577
00:29:00,880 --> 00:29:03,640
Speaker 1: does discount an enterprise scale too, because that always happens.

578
00:29:03,680 --> 00:29:07,600
Speaker 1: It's about two hundred and fifty thousand dollars a month. Wow, Wow,

579
00:29:07,640 --> 00:29:09,840
Speaker 1: three million dollars a year in revenue on a trial

580
00:29:09,920 --> 00:29:13,040
Speaker 1: basis amazing. To be clear, it's probably far fewer seats

581
00:29:13,040 --> 00:29:15,400
Speaker 1: and far fewer dollars a month. I'm nothing if not

582
00:29:15,480 --> 00:29:18,640
Speaker 1: generous with my adversaries. In fact, I can't find a

583
00:29:18,680 --> 00:29:22,360
Speaker 1: shred of evidence that cognition actually makes money, despite currently

584
00:29:22,440 --> 00:29:25,400
Speaker 1: raising three hundred million dollars at a ten billion dollar valuation.

585
00:29:25,680 --> 00:29:28,400
Speaker 1: I can find no information about their revenues beyond one

586
00:29:28,400 --> 00:29:31,200
Speaker 1: comment from the information from July twenty twenty four, when

587
00:29:31,200 --> 00:29:35,480
Speaker 1: Cognition raised at a two billion dollar valuation. Cognitions fund

588
00:29:35,520 --> 00:29:38,720
Speaker 1: raises the latest example of AI startups raising capital sky

589
00:29:38,800 --> 00:29:42,520
Speaker 1: high valuations despite having little or no revenue. Cool In

590
00:29:42,560 --> 00:29:44,520
Speaker 1: a further move, per the Information that is both a

591
00:29:44,520 --> 00:29:46,960
Speaker 1: pale horse and a deeply scummy thing to do, Cognition

592
00:29:47,040 --> 00:29:49,320
Speaker 1: is now laid off thirty people from the Windsurf team

593
00:29:49,400 --> 00:29:51,760
Speaker 1: and is now offering the remaining two hundred buyouts equal

594
00:29:51,800 --> 00:29:54,239
Speaker 1: to nine months of salary, and I assume the end

595
00:29:54,240 --> 00:29:57,280
Speaker 1: of any chance to recruit further stock and Cognition CEO

596
00:29:57,360 --> 00:29:59,640
Speaker 1: Scott Woo said the following in an email, telling Windsurf

597
00:29:59,680 --> 00:30:02,400
Speaker 1: employer is about layoffs and buyouts. We don't believe in

598
00:30:02,440 --> 00:30:04,840
Speaker 1: work life balance building. The future of software engineering is

599
00:30:04,880 --> 00:30:06,520
Speaker 1: a mission we all care so deeply about that we

600
00:30:06,560 --> 00:30:09,520
Speaker 1: couldn't possibly separate the two. We know that not everyone

601
00:30:09,600 --> 00:30:11,960
Speaker 1: who joined Windsurf is signed up to join Cognition, where

602
00:30:12,000 --> 00:30:14,320
Speaker 1: we spent six days at the office in clock eighty

603
00:30:14,320 --> 00:30:18,640
Speaker 1: plus hour weeks. Fuck you man, fuck you. I'm sorry.

604
00:30:18,720 --> 00:30:20,960
Speaker 1: You shouldn't be proud of that, and all that pissed

605
00:30:21,040 --> 00:30:23,360
Speaker 1: vinegar and burning of the midnight oil does not appear

606
00:30:23,360 --> 00:30:25,880
Speaker 1: to have created a product that actually matters. I realized

607
00:30:25,920 --> 00:30:27,840
Speaker 1: that's a little cold. But if you're braying and smacking

608
00:30:27,880 --> 00:30:30,120
Speaker 1: your chest about your hard charging six days a week

609
00:30:30,200 --> 00:30:32,280
Speaker 1: office culture, you should be able to do better than

610
00:30:32,320 --> 00:30:35,280
Speaker 1: We have one publicly known customer and nobody knows our revenue.

611
00:30:35,800 --> 00:30:38,360
Speaker 1: Maybe it's a little simpler. Cognition may be paid two

612
00:30:38,440 --> 00:30:40,600
Speaker 1: hundred and fifty million dollars to acquire Windsurf so that

613
00:30:40,600 --> 00:30:43,000
Speaker 1: you could, after the transaction say that they had eighty

614
00:30:43,000 --> 00:30:46,000
Speaker 1: two million dollars in anualized revenue. If that's the case,

615
00:30:46,040 --> 00:30:48,480
Speaker 1: this is one of the dodgiest, weirdest acquisitions I've seen

616
00:30:48,480 --> 00:30:51,800
Speaker 1: in my life. Two founders getting what like a few

617
00:30:51,880 --> 00:30:54,080
Speaker 1: hundred million dollars between them and their investors, and a

618
00:30:54,120 --> 00:30:56,880
Speaker 1: few of their colleagues moving with them to Google, leaving

619
00:30:56,920 --> 00:30:59,080
Speaker 1: the rest of the staff effectively jobless or in hell

620
00:30:59,120 --> 00:31:02,120
Speaker 1: with little payoff for them time working at Windsurf. I

621
00:31:02,120 --> 00:31:03,960
Speaker 1: can only imagine how it must have felt to go

622
00:31:04,000 --> 00:31:07,080
Speaker 1: from being supposedly acquired by open Ai to this farcical

623
00:31:07,200 --> 00:31:09,800
Speaker 1: rich get rich of bullshit. It also suggests that the

624
00:31:09,840 --> 00:31:12,640
Speaker 1: actual underlying value of Windsurf's IP was around two hundred

625
00:31:12,640 --> 00:31:16,360
Speaker 1: and fifty million dollars. So I ask why exactly is

626
00:31:16,440 --> 00:31:19,520
Speaker 1: Cognition worth ten billion dollars? And why did it have

627
00:31:19,560 --> 00:31:22,360
Speaker 1: to raise three hundred million dollars after raising hundreds of

628
00:31:22,440 --> 00:31:25,600
Speaker 1: millions of dollars? According to Bloomberg in March, where is

629
00:31:25,640 --> 00:31:31,240
Speaker 1: the money going? Also doesn't seem to have revenue in general,

630
00:31:31,400 --> 00:31:33,920
Speaker 1: and Carl Brown of the Internet Bugs revealed the Cognition

631
00:31:34,080 --> 00:31:36,960
Speaker 1: faked the demo of Devin the AI powered software engineer

632
00:31:37,200 --> 00:31:40,160
Speaker 1: last year. And Devon doesn't even rank on swe benchmark,

633
00:31:40,200 --> 00:31:43,680
Speaker 1: the industry standard for model efficacy at coding tasks. Then, nevertheless,

634
00:31:43,720 --> 00:31:46,680
Speaker 1: why here is based entirely on one coding language. Ah,

635
00:31:46,800 --> 00:31:50,120
Speaker 1: we'll get to that another time. At best, Cognition has

636
00:31:50,120 --> 00:31:53,080
Speaker 1: now acquired their own nonprofitable coding environment and the smidgen

637
00:31:53,120 --> 00:31:56,520
Speaker 1: of revenue associated with them. How would Cognition go public?

638
00:31:56,920 --> 00:31:59,760
Speaker 1: What is the actual exit path for Cognition or any

639
00:31:59,760 --> 00:32:03,160
Speaker 1: other generative AI startup? Because really it comes down to

640
00:32:03,200 --> 00:32:06,800
Speaker 1: three things, get acquired, go public, or die. And that

641
00:32:06,960 --> 00:32:10,640
Speaker 1: right there is silicon Valley's own crisis their own housing crisis,

642
00:32:10,680 --> 00:32:13,960
Speaker 1: except instead of condo houses they can't afford with subprime

643
00:32:14,000 --> 00:32:17,880
Speaker 1: adjustable rate mortgages. Venture capitalists have invested in unprofitable, low

644
00:32:17,920 --> 00:32:20,600
Speaker 1: revenue startups with valuations that they can never sell at,

645
00:32:20,920 --> 00:32:23,080
Speaker 1: and like homeowners in the dismal years of two thousand

646
00:32:23,080 --> 00:32:25,840
Speaker 1: and eight and two thousand and nine, they're almost certainly underwater.

647
00:32:25,960 --> 00:32:29,800
Speaker 1: They just haven't realized it. Yeah, Where customers were unable

648
00:32:29,840 --> 00:32:33,120
Speaker 1: to refinance their mortgages to bring their monthly payments down, Generally,

649
00:32:33,120 --> 00:32:36,440
Speaker 1: of AI startups face pressure to continually raise higher and

650
00:32:36,520 --> 00:32:38,880
Speaker 1: higher evaluations to keep up with their costs, with each

651
00:32:38,920 --> 00:32:41,520
Speaker 1: one making it less likely that their company will sell

652
00:32:41,560 --> 00:32:44,480
Speaker 1: to someone else, by which I mean survive. The other

653
00:32:44,520 --> 00:32:46,880
Speaker 1: difference is that in the case of the housing crisis,

654
00:32:47,040 --> 00:32:49,520
Speaker 1: those who were able to hold onto their properties eventually

655
00:32:49,520 --> 00:32:52,360
Speaker 1: saw their acuity recover to pre crash levels. In part

656
00:32:52,360 --> 00:32:55,480
Speaker 1: because housing is essential and because its price is influenced

657
00:32:55,520 --> 00:32:57,320
Speaker 1: just as much by supply and demand as it is

658
00:32:57,360 --> 00:32:59,880
Speaker 1: the ability for people to finance the purchase of properties,

659
00:33:00,080 --> 00:33:02,719
Speaker 1: and when the population increases, so too does the demand

660
00:33:02,760 --> 00:33:05,479
Speaker 1: for housing. A house is a useful thing that you

661
00:33:05,520 --> 00:33:09,080
Speaker 1: could live in with wars and stuff. None of that

662
00:33:09,240 --> 00:33:12,040
Speaker 1: is true with AI. There is a finite amount of investors,

663
00:33:12,040 --> 00:33:14,720
Speaker 1: a finite number of companies, and the finite amount of capital.

664
00:33:14,880 --> 00:33:17,480
Speaker 1: And those companies are only as valuable as the expectations

665
00:33:17,480 --> 00:33:20,360
Speaker 1: their investors have for them, and as the broader sentiment

666
00:33:20,400 --> 00:33:24,400
Speaker 1: towards AI goes as well. Also, I should add, there

667
00:33:24,440 --> 00:33:27,400
Speaker 1: is only there don't appear to be that many business

668
00:33:27,440 --> 00:33:31,320
Speaker 1: models or different ones. There's so many different AI coding environments,

669
00:33:31,320 --> 00:33:34,480
Speaker 1: including ones made by these companies. Amazon has their own one,

670
00:33:34,560 --> 00:33:38,200
Speaker 1: now open Ai has codex. I mean, at some point

671
00:33:38,720 --> 00:33:42,040
Speaker 1: can all of these things last? And who is going

672
00:33:42,080 --> 00:33:44,640
Speaker 1: to buy Cognition? Because the only other opportunity for the

673
00:33:44,640 --> 00:33:47,040
Speaker 1: investors who put money into this company to make money here,

674
00:33:47,280 --> 00:33:49,920
Speaker 1: let alone recoup their initial investment, is for them to

675
00:33:49,960 --> 00:33:53,840
Speaker 1: go public. Do you think Cognition will go public? How

676
00:33:53,840 --> 00:33:57,600
Speaker 1: about Cursor? It's worth nine point nine billion dollars. There

677
00:33:57,640 --> 00:34:00,000
Speaker 1: was a rumor that was also raising an eighteen bigs

678
00:34:00,000 --> 00:34:02,720
Speaker 1: million and twenty billion dollars back in June. Do you

679
00:34:02,920 --> 00:34:05,880
Speaker 1: do you think that Cursor goes public or gets sold?

680
00:34:05,880 --> 00:34:09,160
Speaker 1: I mean maybe maybe one of these cells? Do you

681
00:34:09,200 --> 00:34:13,960
Speaker 1: see perplexity fucking perplexity at evaluation of eighteen plus billion

682
00:34:14,040 --> 00:34:17,840
Speaker 1: dollars selling to another company or going public. The alternative

683
00:34:17,880 --> 00:34:20,760
Speaker 1: has discussed is that perplexity a company, if with fifteen

684
00:34:20,800 --> 00:34:24,120
Speaker 1: million users and around five hundred and fifty million annualized revenue,

685
00:34:24,160 --> 00:34:26,359
Speaker 1: is still making less than half of the revenue. By

686
00:34:26,400 --> 00:34:30,400
Speaker 1: the way of the Cincinnati Reds baseball team, they're going

687
00:34:30,480 --> 00:34:33,000
Speaker 1: to go public. They're going to go public. Just to

688
00:34:33,000 --> 00:34:34,960
Speaker 1: be clear, the Reds make three hundred and twenty five

689
00:34:35,000 --> 00:34:39,799
Speaker 1: million dollars in a year. They're profitable, real money. How

690
00:34:40,080 --> 00:34:42,560
Speaker 1: is Perplexity, which is an inferior business to the Reds?

691
00:34:43,080 --> 00:34:45,080
Speaker 1: How are they meant to do that? They've at this

692
00:34:45,160 --> 00:34:47,400
Speaker 1: point raised over a billion dollars and they lost sixty

693
00:34:47,440 --> 00:34:49,560
Speaker 1: eight million dollars in twenty twenty four and thirty four

694
00:34:49,560 --> 00:34:53,719
Speaker 1: million dollars of revenue by comparison, By the way, the

695
00:34:53,760 --> 00:34:56,360
Speaker 1: Reds are a great business within their income of twenty

696
00:34:56,440 --> 00:34:58,960
Speaker 1: nine million dollars. Or to provide a service that upsets

697
00:34:58,960 --> 00:35:02,160
Speaker 1: and humiliates millions people from Ohio every year for the

698
00:35:02,200 --> 00:35:05,560
Speaker 1: pleasure of America, that's a great business. We should respect

699
00:35:05,560 --> 00:35:08,640
Speaker 1: the red So the Reds should be interviewed by Bloomberg.

700
00:35:09,040 --> 00:35:12,240
Speaker 1: We should have Elie de la Cruz talking every time

701
00:35:12,560 --> 00:35:14,480
Speaker 1: we hear from Sam Mortman. Just put Eli in there.

702
00:35:14,520 --> 00:35:17,200
Speaker 1: He'd be way more interesting and also works harder and

703
00:35:17,280 --> 00:35:20,920
Speaker 1: provides greater value as a business leader. Anyway, putting aside

704
00:35:20,960 --> 00:35:23,439
Speaker 1: the reds, what exactly is it that Perplexity could offer

705
00:35:23,480 --> 00:35:25,760
Speaker 1: to the public markets as a stock or to an acquirer.

706
00:35:26,080 --> 00:35:28,640
Speaker 1: Apple considered acquiring them back in June. But Apple tends

707
00:35:28,640 --> 00:35:31,240
Speaker 1: to acquire companies that wants to integrate into their core business,

708
00:35:31,239 --> 00:35:33,399
Speaker 1: as was the case with Siri, which makes me think

709
00:35:33,400 --> 00:35:35,480
Speaker 1: that Perplexity leaked that information about a deal that was

710
00:35:35,520 --> 00:35:38,960
Speaker 1: never really serious. Hell, Meta also talked about acquiring them too,

711
00:35:39,239 --> 00:35:42,319
Speaker 1: and again I would assume that Perplexity leaked. That isn't

712
00:35:42,320 --> 00:35:44,680
Speaker 1: it weird by the way that two different companies talked

713
00:35:44,680 --> 00:35:48,520
Speaker 1: about buying Perplexity but neither of them did so. CEO

714
00:35:48,640 --> 00:35:51,200
Speaker 1: Ara vince Rivenas said in July that he wanted to

715
00:35:51,239 --> 00:35:53,520
Speaker 1: remain independent, which is a weird thing to say after

716
00:35:53,680 --> 00:35:56,960
Speaker 1: talking to two giant, multi trillion dollar market cap tech

717
00:35:56,960 --> 00:36:00,360
Speaker 1: films about selling them to them, Like you see, to

718
00:36:00,400 --> 00:36:02,400
Speaker 1: not want to stay independent, why would you have the

719
00:36:02,400 --> 00:36:06,600
Speaker 1: conversation if you're you're just getting intel bullshit. It's almost

720
00:36:06,600 --> 00:36:09,839
Speaker 1: as if nobody wants to buy Perplexity, or that really

721
00:36:09,880 --> 00:36:11,720
Speaker 1: they don't want to buy any of these sham companies,

722
00:36:11,760 --> 00:36:13,960
Speaker 1: which I know sounds mean. But if you're worth tens

723
00:36:14,000 --> 00:36:16,799
Speaker 1: of billions or a billion dollars or anything over like

724
00:36:16,840 --> 00:36:20,239
Speaker 1: five hundred million dollars and you can't make more than

725
00:36:20,280 --> 00:36:23,200
Speaker 1: a bottom tier baseball team in fucking Ohio, you're neither

726
00:36:23,200 --> 00:36:27,640
Speaker 1: innovative nor deserving of said valuation. But really, my pissiness

727
00:36:27,680 --> 00:36:30,279
Speaker 1: and baseball comparisons aside, what exactly is the plan for

728
00:36:30,320 --> 00:36:33,120
Speaker 1: these goddamn companies. They don't make enough money to survive

729
00:36:33,160 --> 00:36:35,600
Speaker 1: without a continuous slow of venture capital, and they don't

730
00:36:35,640 --> 00:36:37,960
Speaker 1: seem to make impressive sums of money even when allowed

731
00:36:37,960 --> 00:36:40,359
Speaker 1: to burn as much as they'd like. These companies are

732
00:36:40,360 --> 00:36:42,600
Speaker 1: not being forced to live frugally, or at least have

733
00:36:42,680 --> 00:36:44,920
Speaker 1: yet to be made to, perhaps because they have all

734
00:36:44,960 --> 00:36:48,120
Speaker 1: actively engaged as at spending as much money as possible

735
00:36:48,160 --> 00:36:50,080
Speaker 1: and perceived of finding an idea that makes more money

736
00:36:50,120 --> 00:36:52,920
Speaker 1: than it loses. This is not a rational nor reasonable

737
00:36:52,920 --> 00:36:56,200
Speaker 1: way to proceed. Yes, there are startups that can justify

738
00:36:56,239 --> 00:36:59,080
Speaker 1: burning capital. Yes there are companies that have burned hundreds

739
00:36:59,080 --> 00:37:01,200
Speaker 1: of millions of dollars to find their business models, or

740
00:37:01,239 --> 00:37:03,680
Speaker 1: billions in the case of Uber. But none of these

741
00:37:03,719 --> 00:37:08,000
Speaker 1: companies are like those companies in the GENERATIVEAI space. GENERALFAI

742
00:37:08,040 --> 00:37:10,839
Speaker 1: businesses don't have the same economics, nor do they have

743
00:37:10,960 --> 00:37:13,600
Speaker 1: the same total addressable markets. Now, if you're gonna say

744
00:37:13,600 --> 00:37:16,800
Speaker 1: Amazon Web Services, there's a three part episode The Hater's

745
00:37:16,800 --> 00:37:19,600
Speaker 1: Guide to the AI Bubble that explains that, in detail,

746
00:37:19,680 --> 00:37:22,440
Speaker 1: I do voices, I have sources. You need to go

747
00:37:22,480 --> 00:37:25,360
Speaker 1: and listen to that. I'm fucking tired of hearing people

748
00:37:25,360 --> 00:37:28,960
Speaker 1: about this one. What about Uber? What about Aws? Yeah?

749
00:37:29,000 --> 00:37:31,600
Speaker 1: You got shit? That's all you've got. You certainly don't

750
00:37:31,600 --> 00:37:34,120
Speaker 1: have a company to point to. These startups are their

751
00:37:34,200 --> 00:37:38,640
Speaker 1: VC firm, subprime mortgages, overstuffed valuations, with no exit route

752
00:37:38,680 --> 00:37:40,400
Speaker 1: and no clear example of how to sell them or

753
00:37:40,440 --> 00:37:43,000
Speaker 1: who to sell them to. The closest they've got is

754
00:37:43,520 --> 00:37:46,960
Speaker 1: using GENERATIVEAI startups as beauty pageants for guys wearing Patagonia,

755
00:37:47,239 --> 00:37:49,160
Speaker 1: finding ways to pretend that the guy who runs an

756
00:37:49,160 --> 00:37:52,960
Speaker 1: AI startup sorry AI lab, is some sort of mysterious

757
00:37:53,000 --> 00:37:55,719
Speaker 1: genius versus just another founder and just another bubble with

758
00:37:55,800 --> 00:38:00,440
Speaker 1: just another overstuffed valuation. The literal only liquidity mechanism outside

759
00:38:00,440 --> 00:38:03,520
Speaker 1: of cognike that Generative AI has had so far is

760
00:38:03,560 --> 00:38:07,160
Speaker 1: selling AI talent to big tech a premium. Nobody has

761
00:38:07,160 --> 00:38:09,720
Speaker 1: gone or is going public, and if they're not going public,

762
00:38:09,760 --> 00:38:12,200
Speaker 1: the only route for these companies is to either become profitable,

763
00:38:12,239 --> 00:38:14,960
Speaker 1: which they haven't, or sell to somebody, which they're not.

764
00:38:15,760 --> 00:38:17,719
Speaker 1: But I'd be dancing around the real reason they won't

765
00:38:17,760 --> 00:38:20,960
Speaker 1: sell because fundamentally, generative AI does not let companies build

766
00:38:21,040 --> 00:38:24,120
Speaker 1: something new. Anyone that builds a generative AI product is

767
00:38:24,200 --> 00:38:27,879
Speaker 1: ultimately just prompting the model, albeit increasingly or complex ways,

768
00:38:27,880 --> 00:38:30,560
Speaker 1: at the scale of something like claud Code or though

769
00:38:30,600 --> 00:38:32,399
Speaker 1: anthropic as the advantage of being one of the main

770
00:38:32,480 --> 00:38:35,399
Speaker 1: veins of their infrastructure, or of course cursor. This means

771
00:38:35,400 --> 00:38:38,160
Speaker 1: that the generative AI company owns very few unique things

772
00:38:38,160 --> 00:38:40,400
Speaker 1: beyond their talent and will forever be at the mercy

773
00:38:40,440 --> 00:38:42,920
Speaker 1: of any and all decisions that their model provider makes,

774
00:38:43,120 --> 00:38:46,759
Speaker 1: such as increasing prices or creating competing products. I know

775
00:38:46,840 --> 00:38:50,120
Speaker 1: it sounds ludicrous, but this is the reality of these companies.

776
00:38:50,280 --> 00:38:52,480
Speaker 1: While there are some companies that have some unique training

777
00:38:52,560 --> 00:38:54,920
Speaker 1: and models and such. None of them seem to be

778
00:38:54,920 --> 00:38:58,719
Speaker 1: building interesting or unique products. Is a result if your

779
00:38:58,800 --> 00:39:02,680
Speaker 1: argument is that these things takes time? How long? No? Really,

780
00:39:03,480 --> 00:39:05,240
Speaker 1: so many of you have said this is what happens.

781
00:39:05,239 --> 00:39:07,000
Speaker 1: They burn a lot of money, they grow, and then

782
00:39:07,680 --> 00:39:10,440
Speaker 1: and then what you stop? Sure because the next thing

783
00:39:10,480 --> 00:39:13,480
Speaker 1: you say is term profitable by getting enterprise customers. Nobody

784
00:39:13,520 --> 00:39:14,759
Speaker 1: can do the first part, and for you can do

785
00:39:14,800 --> 00:39:18,040
Speaker 1: the second part in anything approaching a consistent fashion. But really,

786
00:39:18,040 --> 00:39:21,000
Speaker 1: how long should we give them? Three years? Perplexity has

787
00:39:21,040 --> 00:39:22,879
Speaker 1: had three years and a billion dollars. It doesn't seem

788
00:39:22,920 --> 00:39:25,480
Speaker 1: to be close to profitable. How long does Perplexity deserve

789
00:39:25,520 --> 00:39:30,240
Speaker 1: exactly four years? Five years an eternity. Every single example

790
00:39:30,239 --> 00:39:31,759
Speaker 1: of a company that's burned a lot of money and

791
00:39:31,800 --> 00:39:33,400
Speaker 1: then not done so in the end has been a

792
00:39:33,400 --> 00:39:35,840
Speaker 1: company with a physical thing or connections to the real world,

793
00:39:36,000 --> 00:39:38,400
Speaker 1: with the exception of Facebook, which has never had the

794
00:39:38,480 --> 00:39:41,799
Speaker 1: kind of cash burning monstrusity that General fai Is or

795
00:39:42,120 --> 00:39:45,560
Speaker 1: Meta has now created for no reason. There has never

796
00:39:45,640 --> 00:39:48,240
Speaker 1: been a software company that has just chewed through hundreds

797
00:39:48,280 --> 00:39:50,520
Speaker 1: of millions or billions of dollars and then suddenly become

798
00:39:50,600 --> 00:39:54,560
Speaker 1: profitable magically, mostly because the magical valuations of software have

799
00:39:54,680 --> 00:39:58,800
Speaker 1: been in their ability to transcend infrastructure once unit economics

800
00:39:58,800 --> 00:40:00,840
Speaker 1: in the sales of software like my Microsoft Office or

801
00:40:00,880 --> 00:40:03,920
Speaker 1: providing access to Instagram do not require the most powerful

802
00:40:03,960 --> 00:40:07,400
Speaker 1: graphics processing units run full tilt at all times, and

803
00:40:07,440 --> 00:40:09,200
Speaker 1: those are products that people like and want to use

804
00:40:09,239 --> 00:40:11,200
Speaker 1: every day, even though both of them seem to be

805
00:40:11,239 --> 00:40:14,879
Speaker 1: getting worse every minute. I get people saying that these

806
00:40:14,880 --> 00:40:17,560
Speaker 1: companies are in the growth stage, but when all of

807
00:40:17,560 --> 00:40:20,040
Speaker 1: them are unprofitable, and even the unprofitable ones outside of

808
00:40:20,040 --> 00:40:23,600
Speaker 1: open Ai and Anthropic aren't really making much money. Come on,

809
00:40:24,239 --> 00:40:26,560
Speaker 1: this isn't anything like any boom that leads to something

810
00:40:26,680 --> 00:40:29,400
Speaker 1: good or you saul or profitable, and it's because the

811
00:40:29,440 --> 00:40:32,080
Speaker 1: economics do not make sense. And that's before we get

812
00:40:32,080 --> 00:40:43,440
Speaker 1: into open Ai and Anthropic in the next episode. Thank

813
00:40:43,480 --> 00:40:44,960
Speaker 1: you for listening to Better Offline.

814
00:40:45,080 --> 00:40:47,520
Speaker 2: The editor and composer of the Better Offline theme song

815
00:40:47,560 --> 00:40:50,240
Speaker 2: is Matasowski. You can check out more of his music

816
00:40:50,239 --> 00:40:55,399
Speaker 2: and audio projects at Matasowski dot com, m a ttoso

817
00:40:55,680 --> 00:40:59,440
Speaker 2: w Ski dot com. You can email me an easy

818
00:40:59,520 --> 00:41:01,960
Speaker 2: at better off line dot com or visit better Offline

819
00:41:02,000 --> 00:41:04,080
Speaker 2: dot com to find more podcast links, and of course

820
00:41:04,120 --> 00:41:07,239
Speaker 2: my newsletter. I also really recommend you go to chat

821
00:41:07,239 --> 00:41:09,880
Speaker 2: dot Where's youread dot at to visit the discord, and

822
00:41:09,960 --> 00:41:12,640
Speaker 2: go to our slash Better Offline to check out our reddit.

823
00:41:13,440 --> 00:41:16,680
Speaker 1: Thank you so much for listening. Better Offline is a

824
00:41:16,719 --> 00:41:19,800
Speaker 1: production of cool Zone Media. For more from cool Zone Media,

825
00:41:19,920 --> 00:41:23,080
Speaker 1: visit our website cool Zonemedia dot com, or check us

826
00:41:23,080 --> 00:41:26,040
Speaker 1: out on the iHeartRadio app, Apple Podcasts, or wherever you

827
00:41:26,120 --> 00:41:27,240
Speaker 1: get your podcasts.