1 00:00:00,040 --> 00:00:02,640 During this Rebel Valley AI Summit, we asked a group of 2 00:00:02,640 --> 00:00:05,760 founders and investors which billion dollar AI startup would 3 00:00:05,760 --> 00:00:09,080 you short and their answers caused a bit of a stir. 4 00:00:13,600 --> 00:00:17,240 Joining me today are my Cerebral Valley AI Co host and the Co 5 00:00:17,240 --> 00:00:20,240 founders of Volley, Max Child and James Wilsterman. 6 00:00:20,440 --> 00:00:23,000 We'll also be diving into some of the insights from our 7 00:00:23,000 --> 00:00:24,680 panelists throughout the conference. 8 00:00:24,760 --> 00:00:27,680 My interview with Mike Krieger, the Chief Product Officer at 9 00:00:27,680 --> 00:00:31,200 Anthropic, former Co founder of Instagram, about the problem 10 00:00:31,200 --> 00:00:33,720 with sycophancy and foundation models. 11 00:00:33,760 --> 00:00:36,840 My wife got her first like you're completely wrong. 12 00:00:36,840 --> 00:00:38,080 And she was like, yes, this is great. 13 00:00:38,080 --> 00:00:39,840 And I think we should have more of that we. 14 00:00:39,840 --> 00:00:44,880 Looked at a clip from the mayor. Whoever the politician, 11 the 15 00:00:44,880 --> 00:00:47,440 adoration of the crowd. No one should be asking someone 16 00:00:47,440 --> 00:00:50,560 that's been in a job for 10 months for advice. 17 00:00:50,920 --> 00:00:55,280 And we ask, is Mecca Hitler inevitable On stage with Jimmy 18 00:00:55,280 --> 00:01:00,760 BA, one of the Co founders of Elon Musk's Xai the Mecca Hitler 19 00:01:00,760 --> 00:01:02,840 in the room. Yeah, Mecca Hitler a model had 20 00:01:02,840 --> 00:01:06,280 an episode. This is the newcomer podcast. 21 00:01:14,760 --> 00:01:17,280 All right, I'm excited. We've had some time to rest 22 00:01:17,280 --> 00:01:19,200 since the Cerebral Valley AI Summit. 23 00:01:19,360 --> 00:01:22,800 Max has been resting life flat. What? 24 00:01:23,880 --> 00:01:26,000 Literally just got home from Dubai I. 25 00:01:26,000 --> 00:01:31,040 Believe literally 10 minutes got off a 16 hour long flight over 26 00:01:31,200 --> 00:01:33,760 the world's greatest hits like Tehran and Moscow. 27 00:01:33,920 --> 00:01:38,120 I I've got my seven week old. So back back to the parenting 28 00:01:38,520 --> 00:01:41,680 minds, though it's been a joy. James, what you're are you all 29 00:01:41,680 --> 00:01:44,400 rested or how are you feeling? I'm not well arrested Eric. 30 00:01:44,400 --> 00:01:49,040 I have been solo parenting my 2 year old for. 31 00:01:49,240 --> 00:01:52,480 Where's your wife? Where's Brazil and Mexico? 32 00:01:53,040 --> 00:01:54,120 Nice. Yeah. 33 00:01:54,120 --> 00:01:58,760 So we've all got our own reasons. 34 00:01:58,880 --> 00:02:03,280 To be we're at the exact level of delirious that the viewer 35 00:02:03,280 --> 00:02:06,240 should want here because we just had this River Valley AI summit, 36 00:02:06,240 --> 00:02:08,240 right? We host this twice a year. 37 00:02:08,240 --> 00:02:12,360 Big AI, you know, 300 person event, top AI founders, 38 00:02:12,360 --> 00:02:16,640 investors, real insider thing. We sent out a survey, became a 39 00:02:16,640 --> 00:02:20,960 point of media fascination, like sources, Business Insider wrote 40 00:02:20,960 --> 00:02:22,920 about it. I think random Indian media 41 00:02:22,920 --> 00:02:25,160 outlets for reasons that we'll explain as we progress. 42 00:02:25,440 --> 00:02:28,360 We had 300 attendees at the thing. 43 00:02:28,680 --> 00:02:31,160 I want to be clear, like, you know, some of these things we 44 00:02:31,160 --> 00:02:32,880 didn't. Everybody did not fill out this 45 00:02:32,880 --> 00:02:35,040 anonymous survey. This was not academic research. 46 00:02:35,040 --> 00:02:38,240 This is you can count the dots for yourself and sort of infer 47 00:02:38,400 --> 00:02:40,440 how many people did this. I don't know what are we saying? 48 00:02:40,440 --> 00:02:43,200 Like 30 to 40 people. I just want to be transparent. 49 00:02:43,200 --> 00:02:46,080 But it gives you a sliver of where this highly engaged 50 00:02:46,080 --> 00:02:49,120 audience thought things were in. Part of the fun was that Max and 51 00:02:49,120 --> 00:02:53,560 I were on stage sort of reacting to the survey results. 52 00:02:53,960 --> 00:02:55,320 Anyway, so we're going to dig into the survey. 53 00:02:55,320 --> 00:02:58,320 James, this is really your brainchild. 54 00:02:58,440 --> 00:03:00,440 Anything I missed about it or you want to start ticking 55 00:03:00,440 --> 00:03:03,000 through the questions? Yeah, I guess one Part 1 fun 56 00:03:03,000 --> 00:03:07,400 part of the game that we played on stage was the you and Max had 57 00:03:07,400 --> 00:03:12,360 to guess what the audience would think the answer to these 58 00:03:12,360 --> 00:03:15,640 questions might be. And I have to say, Max kind of 59 00:03:15,640 --> 00:03:17,280 ran away with that game on stage. 60 00:03:18,040 --> 00:03:18,680 He did well. I. 61 00:03:18,720 --> 00:03:23,600 Think I won 5141 some Max. Max did well, Yeah, it was a 62 00:03:23,600 --> 00:03:25,000 some of them were close. Let's go through. 63 00:03:25,000 --> 00:03:26,360 We can talk about. Yeah, you did. 64 00:03:26,400 --> 00:03:27,040 Very. Well, we can. 65 00:03:27,040 --> 00:03:31,120 We can react to the reactions. Myself, despite Max's dominance, 66 00:03:31,120 --> 00:03:34,880 I think I had AII, had my ear on the pulse still. 67 00:03:34,960 --> 00:03:36,560 All right, Should we start with the the beginning? 68 00:03:36,680 --> 00:03:38,240 Yeah. So the first question was what 69 00:03:38,240 --> 00:03:40,800 will open a eyes annualized revenue be? 70 00:03:40,800 --> 00:03:46,040 At the end of 2026, the audience median was 30 billion. 71 00:03:47,360 --> 00:03:50,680 And it's a 20 billion 2025 is the expectation, right? 72 00:03:51,120 --> 00:03:55,800 Yeah, over 20 billion already. So this is pretty low estimate 73 00:03:55,800 --> 00:03:58,240 in my opinion. And I said like 40 or what did I 74 00:03:58,240 --> 00:04:01,080 say? Yeah, I forget who said 40 and 75 00:04:01,400 --> 00:04:03,320 one of us said 40 and one of us said 41. 76 00:04:03,320 --> 00:04:06,640 I did be one of you. Yeah, Yeah, I went higher, maybe 77 00:04:06,640 --> 00:04:08,200 41 or something. I was surprised. 78 00:04:08,200 --> 00:04:11,200 I think that given we're exiting this year at 20 in open, AII 79 00:04:11,200 --> 00:04:15,080 thought the effusive Glee and bubble talk of the conference 80 00:04:15,120 --> 00:04:17,320 would would flow through to a growth. 81 00:04:17,399 --> 00:04:18,279 Video. It'd just. 82 00:04:18,279 --> 00:04:19,959 Be earning AI. Things are going well. 83 00:04:20,160 --> 00:04:22,680 We're round tripping all day. Everybody's making money. 84 00:04:22,920 --> 00:04:26,960 Revenue is not the problem. Mere 50% growth for open AI I 85 00:04:26,960 --> 00:04:30,160 think would be like considered in a truck at this point in the 86 00:04:30,160 --> 00:04:31,240 bubble cycle. So. 87 00:04:31,560 --> 00:04:34,120 I'm betting on 40. If this actually happens, I 88 00:04:34,120 --> 00:04:36,680 think the bubble is over baby. What's your bet Max? 89 00:04:37,520 --> 00:04:40,640 Yeah, I think 40 is what I, yeah, I said on the last podcast 90 00:04:40,640 --> 00:04:43,000 I think we discussed. I mean I think 2X year on year 91 00:04:43,000 --> 00:04:45,320 makes makes sense to me unless the bubble collapses. 92 00:04:45,680 --> 00:04:48,960 I'm curious, where do you guys think the next 20 billion comes 93 00:04:48,960 --> 00:04:51,360 from? Like is it business as usual or 94 00:04:51,360 --> 00:04:53,360 do they need to create new products or? 95 00:04:53,560 --> 00:04:57,080 I don't know copy Harvey, you know, going to legal anything. 96 00:04:57,080 --> 00:05:00,440 It's like, I think, you know, they should lean into the API 97 00:05:00,440 --> 00:05:02,640 business through this sort of general purpose foundation 98 00:05:02,640 --> 00:05:04,720 model. So I, I don't necessarily think 99 00:05:04,720 --> 00:05:06,960 they need to go after an application directly, but if 100 00:05:06,960 --> 00:05:10,160 they were really hungry for revenue, you'd think they'd 101 00:05:10,440 --> 00:05:13,800 figure out who their best partner is and say screw it, 102 00:05:13,800 --> 00:05:15,960 we're going to cannibalize them to find the revenue. 103 00:05:15,960 --> 00:05:17,120 We. Need kind of seems like they're 104 00:05:17,120 --> 00:05:20,560 gearing up for that yeah, Palantir model or you know, find 105 00:05:20,560 --> 00:05:24,600 the biggest pockets of money in the world and, you know, consult 106 00:05:24,600 --> 00:05:28,360 on how to adopt AI in those companies, right? 107 00:05:29,040 --> 00:05:30,880 I mean, I think there's a lot of headroom on the consumer 108 00:05:30,880 --> 00:05:33,200 subscription business as well. I just think that, you know, 109 00:05:33,200 --> 00:05:36,120 they haven't necessarily monetized, you know, a huge 110 00:05:36,120 --> 00:05:39,560 percentage of the people who use AI every day and they just keep 111 00:05:39,560 --> 00:05:41,160 adding value to that consumer subscription. 112 00:05:41,160 --> 00:05:43,760 And so, you know, if you even if you just got a doubling of that, 113 00:05:43,760 --> 00:05:45,880 that you know, that would get you pretty far along the route 114 00:05:45,880 --> 00:05:47,240 here. All right, next question. 115 00:05:47,240 --> 00:05:50,920 What will NVIDIA be worth at the end of 2026? 116 00:05:51,640 --> 00:05:54,960 On the day we took this survey, it was 4.8 trillion. 117 00:05:55,360 --> 00:05:57,400 They just had earnings. I'm not actually sure what it 118 00:05:57,400 --> 00:05:59,520 is. The day we're recording it up, 119 00:05:59,600 --> 00:06:00,240 is it? Yeah. 120 00:06:00,800 --> 00:06:01,320 It's pretty. Close. 121 00:06:01,320 --> 00:06:04,240 They had they had pretty good earnings, I think, but I don't 122 00:06:04,240 --> 00:06:07,160 think Wall Street. Yeah, I checked the stock's like 123 00:06:07,160 --> 00:06:10,560 4546 right now or something. 4.35 trillion. 124 00:06:10,560 --> 00:06:12,760 So down 35 S, down a little bit, yeah. 125 00:06:12,840 --> 00:06:14,680 Yeah. All right. 126 00:06:14,680 --> 00:06:17,600 Unless they always they beat earnings and then they still, 127 00:06:18,040 --> 00:06:19,840 you know, the expectations are so insane. 128 00:06:21,000 --> 00:06:22,960 I said five, I think Max. What? 129 00:06:22,960 --> 00:06:25,160 Did you say? I said 6, which was dead on the 130 00:06:25,160 --> 00:06:27,320 money, if I recall. Yeah, yeah. 131 00:06:27,440 --> 00:06:29,720 Yep. So the audience, the audience 132 00:06:29,720 --> 00:06:36,440 had a median of 6 trillion. Very few outliers on the high 133 00:06:36,440 --> 00:06:38,400 end there. 5 was the next biggest grouping. 134 00:06:38,640 --> 00:06:44,800 Also interesting median was just so large is that's really it's 135 00:06:44,800 --> 00:06:46,960 just so many. If I was think it was average. 136 00:06:47,160 --> 00:06:49,400 Well, average is obviously thrown off by this like 100 137 00:06:49,600 --> 00:06:51,800 hundred trillion as a troll, but forget that one. 138 00:06:51,880 --> 00:06:54,160 I think based on the performance of the recent earnings, I think 139 00:06:54,160 --> 00:06:55,960 this is actually high for reality. 140 00:06:56,280 --> 00:06:59,720 Like they crushed the recent earnings as we discussed and the 141 00:06:59,720 --> 00:07:02,400 stock barely went up on a multi day period. 142 00:07:02,400 --> 00:07:07,160 So if they're not going up, you know, 10% a quarter, essentially 143 00:07:07,160 --> 00:07:08,360 they're not going to hit this number. 144 00:07:08,360 --> 00:07:10,600 And I just don't see that if those earnings aren't moving the 145 00:07:10,600 --> 00:07:12,400 stock up. So this actually feels high for 146 00:07:12,400 --> 00:07:16,280 reality to me. 5 is hard because you're sort of like you're 147 00:07:16,280 --> 00:07:18,280 chickening out on saying it's all going to blow up. 148 00:07:18,320 --> 00:07:21,200 And you know, it's, it's like, what is this reality where 149 00:07:21,240 --> 00:07:24,520 either it's like, oh, the mania has continued or there's a 150 00:07:24,520 --> 00:07:27,760 pullback. In some ways my 5 trillion feels 151 00:07:27,760 --> 00:07:30,640 like sort of a weird sort of same. 152 00:07:30,680 --> 00:07:31,840 Middle ground sound. World. 153 00:07:31,840 --> 00:07:34,520 But yeah, who knows? Obviously, if you could, if you 154 00:07:34,520 --> 00:07:37,200 knew the answer to this, you could be unlimited, you could 155 00:07:37,200 --> 00:07:38,720 have an unlimited amount of wealth. 156 00:07:38,720 --> 00:07:43,480 So nobody knows, but this is what a couple people think. 157 00:07:45,760 --> 00:07:48,880 What year will an independent committee of experts, as 158 00:07:48,880 --> 00:07:52,600 dictated by the Microsoft Open AI Agreement, declare that we 159 00:07:52,600 --> 00:07:55,640 have reached AGI? I thought this was a funny 160 00:07:55,640 --> 00:07:59,800 question because at once it's such a like, silly idea, like, 161 00:07:59,800 --> 00:08:01,880 oh, we're gonna have AGI, but like, there's an actual 162 00:08:01,880 --> 00:08:04,560 contractual agreement between Microsoft and Open AI that 163 00:08:04,760 --> 00:08:08,640 there's a committee to resolve this and big business when, you 164 00:08:08,640 --> 00:08:12,560 know, dealings hang on in. So it's a specific question, 165 00:08:12,560 --> 00:08:14,560 which is sort of funny. This was one of my favourites 166 00:08:14,560 --> 00:08:18,280 because you guys were trying to guess what the audience will try 167 00:08:18,280 --> 00:08:23,480 to guess what Microsoft and Open AI will will decide as AGI, as a 168 00:08:23,480 --> 00:08:27,120 lot of layers of prognification going on here. 169 00:08:28,240 --> 00:08:30,560 What do you guys say? You said much higher. 170 00:08:30,920 --> 00:08:34,400 I said 2035. Yeah, I said 29. 171 00:08:34,480 --> 00:08:37,080 So I was pretty close to the median of 2030. 172 00:08:37,320 --> 00:08:38,840 What do you guys think in retrospect? 173 00:08:38,840 --> 00:08:42,640 Like is this is too early or? I just feel like a theme of 174 00:08:43,120 --> 00:08:46,200 Cerebral Valley in the beginning, you know, we started 175 00:08:46,200 --> 00:08:50,040 in 2023, in March 2023 after your Chachi BT that was probably 176 00:08:50,040 --> 00:08:52,160 the conference. We talked most about AGI and 177 00:08:52,160 --> 00:08:53,840 then we talked about it less and less every time. 178 00:08:53,840 --> 00:08:56,480 You know, it's like there was so much enthusiasm when the models 179 00:08:56,480 --> 00:08:59,280 came out and now we're in the sort of like, oh man, this is 180 00:08:59,280 --> 00:09:02,800 exactly like self driving cars where you feel really close and 181 00:09:02,800 --> 00:09:06,320 then there's a lot of like edge cases to hammer around. 182 00:09:06,320 --> 00:09:09,600 And so I just think AGI pessimism has gone way up and 183 00:09:09,600 --> 00:09:12,760 you add to that Andre Carpathy sort of thing, and it's just 184 00:09:12,760 --> 00:09:15,080 like, I don't know, I don't think the insider vibes are like 185 00:09:15,160 --> 00:09:18,600 AGI tomorrow unless you're talking to Daria or something. 186 00:09:18,600 --> 00:09:20,880 I think the, I think the interesting thing here is that 187 00:09:20,880 --> 00:09:23,000 there's basically just two buckets of people. 188 00:09:23,400 --> 00:09:27,480 One is 2030 or sooner, which are like the accelerationist and 189 00:09:27,480 --> 00:09:30,880 then one is like 2045 or never, which is like the 190 00:09:30,880 --> 00:09:33,480 decelerationist or the pessimist or whatever you want to call 191 00:09:33,480 --> 00:09:34,640 them. Maybe, maybe the realist. 192 00:09:34,640 --> 00:09:37,200 Yeah, yeah, yeah. Whereas you sort of hit this 193 00:09:37,200 --> 00:09:39,880 exact, I thought that basically nobody was, which it was 10 194 00:09:39,880 --> 00:09:42,480 years from now or whatever. But like it's still gonna 195 00:09:42,480 --> 00:09:45,440 happen. Which is an interesting like, 196 00:09:45,520 --> 00:09:48,040 yeah, that you sort of found the middle of this smiling curve 197 00:09:48,040 --> 00:09:50,240 where you're either an optimist or you're a pessimist, and you 198 00:09:50,240 --> 00:09:52,160 kind of tried to hit the middle and it didn't quite. 199 00:09:52,200 --> 00:09:54,560 Yeah, that's interesting. We have more optimists at our 200 00:09:54,560 --> 00:09:56,440 conference was the end was the end result. 201 00:09:56,440 --> 00:09:58,160 That's why Max beat me. He understood, though. 202 00:09:58,160 --> 00:09:59,960 That was the psychology, the answer right there. 203 00:09:59,960 --> 00:10:02,480 Yeah, exactly right. Optimistic Conference. 204 00:10:02,920 --> 00:10:05,880 There's something slightly interesting comparing it to the 205 00:10:05,880 --> 00:10:09,600 self driving car world like you said though Eric, because self 206 00:10:09,600 --> 00:10:13,800 driving cars are basically useless until they reach parity 207 00:10:13,800 --> 00:10:16,960 or capability of human drivers. Right, this is not like that. 208 00:10:16,960 --> 00:10:19,440 This is not. It just happens to have all 209 00:10:19,440 --> 00:10:24,640 these other very targeted, more verticalized use cases that are 210 00:10:24,640 --> 00:10:29,400 super valuable but. 100%. With the full human replication. 211 00:10:29,640 --> 00:10:32,280 Yeah, which is why we're you know, I was negative about self 212 00:10:32,280 --> 00:10:35,000 driving cars because it was annoying because you need them 213 00:10:35,000 --> 00:10:38,160 to actually work, whereas this I've been very enthusiastic. 214 00:10:38,160 --> 00:10:42,760 So yeah, I I agree that's what's beautiful about text versus 215 00:10:42,760 --> 00:10:46,160 safely delivering humans places, which thankfully now way MO is 216 00:10:46,160 --> 00:10:48,280 good at and we can celebrating it. 217 00:10:48,280 --> 00:10:50,760 But you know, 10 years ago or whatever, it was annoying 218 00:10:50,760 --> 00:10:53,760 marketing. I like that. 219 00:10:54,320 --> 00:10:58,240 OK, next question, Which venture capital firm's AI portfolio are 220 00:10:58,240 --> 00:11:00,880 you most jealous of? I think this was kind of a 221 00:11:00,880 --> 00:11:04,320 shocker, right? Neither of you guessed A16Z, 222 00:11:04,320 --> 00:11:07,560 which tied with Khosla for the lead here. 223 00:11:08,080 --> 00:11:10,320 Obviously this. We both said thrive, right? 224 00:11:10,360 --> 00:11:11,200 Yeah. And then we both. 225 00:11:11,200 --> 00:11:14,040 And then we decided to tie a. 2nd pick or something. 226 00:11:14,040 --> 00:11:16,280 No, I said Sequoia. Sequoia and I said. 227 00:11:16,320 --> 00:11:18,840 And that's why you won this. Yeah, I got the kicker on that 228 00:11:18,840 --> 00:11:20,280 one. Yeah, that was a good pull. 229 00:11:20,280 --> 00:11:22,440 Max, how did you decide to pull Khosla? 230 00:11:22,440 --> 00:11:26,560 Just open AI or? I did some ChatGPT research 231 00:11:26,560 --> 00:11:30,240 before and. Then goes with the first in Open 232 00:11:30,520 --> 00:11:33,560 AI, first venture investment, Open AI, yeah, but that round 233 00:11:33,560 --> 00:11:35,800 has gotten significantly diluted and I'm sure they'll be 234 00:11:35,800 --> 00:11:37,720 reporting overtime. I don't know how much they've 235 00:11:37,720 --> 00:11:40,200 done secondary. It was one of these fantastic 236 00:11:40,200 --> 00:11:45,240 investments but it feels like they just they got squashed down 237 00:11:45,240 --> 00:11:47,880 by later stage rounds and all these negotiations. 238 00:11:47,960 --> 00:11:51,040 I think we can all agree though, the Andreessen, you know, tie 239 00:11:51,040 --> 00:11:54,240 for victory with Khosla is pretty bizarre because like I, 240 00:11:54,480 --> 00:11:57,200 you know, there's not a lot of really notable successful 241 00:11:57,200 --> 00:11:59,880 Andreessen AI of investments compared to most of these other 242 00:11:59,880 --> 00:12:02,000 firms. I mean, I got a text from 243 00:12:02,000 --> 00:12:04,240 somebody when I shit on Andreessen on stage, which is 244 00:12:04,240 --> 00:12:06,400 funny, which is what's wrong with. 245 00:12:06,400 --> 00:12:10,440 This like live on stage. Right after I went off, but I 246 00:12:10,640 --> 00:12:12,520 which it's like, oh, people are paying attention. 247 00:12:13,280 --> 00:12:17,920 Yeah, I don't know Andreessen. I mean, they have I I can't list 248 00:12:17,920 --> 00:12:20,280 they had character. They have like SSI mean, I'm 249 00:12:20,280 --> 00:12:22,680 sure they have a ton. Some of them are later, you 250 00:12:22,680 --> 00:12:25,960 know, it's like they're in Open AI, they're in XAI. 251 00:12:26,360 --> 00:12:28,800 I just think the lesson of our draft was like the only thing 252 00:12:28,800 --> 00:12:32,800 that matters is basically being heavy in Anthropic X AI or Open 253 00:12:32,840 --> 00:12:36,400 AI and like they're not really in any any of those, right? 254 00:12:36,400 --> 00:12:39,520 I mean, it was my understanding. Probably big in XAI. 255 00:12:40,360 --> 00:12:44,200 Are they big in XAI? I think that they're in X, but I 256 00:12:44,200 --> 00:12:46,040 don't think they're huge. Yeah, these are all. 257 00:12:46,120 --> 00:12:47,920 Growth Sequoia was pretty big, yeah. 258 00:12:48,120 --> 00:12:51,840 Sequoia has some money all. I'm saying is yeah, they to me 259 00:12:51,840 --> 00:12:54,880 it feels like the PR of Andreessen generally is sort of 260 00:12:55,200 --> 00:12:57,240 overwhelming the actual portfolio. 261 00:12:57,240 --> 00:13:00,000 It's like, oh. You know, I think Drive is doing 262 00:13:00,000 --> 00:13:01,960 really well. They've done these huge bets in 263 00:13:01,960 --> 00:13:05,920 open AI like big yeah, pre all the markups. 264 00:13:05,920 --> 00:13:08,920 So I bet they're doing really well and would like to 265 00:13:08,920 --> 00:13:14,240 substantiate that Thrive reach out the I mean all the you know, 266 00:13:14,240 --> 00:13:17,080 they've mentioned a bunch of firms like, you know, it's like 267 00:13:17,360 --> 00:13:20,240 a lot of Gil who we had on stage is obviously great Index 268 00:13:20,240 --> 00:13:21,600 ventures. I mean some of these like who 269 00:13:21,600 --> 00:13:23,240 knows, some partner, you know, set them. 270 00:13:23,320 --> 00:13:26,160 I'm personally jealous of this slag because it would be a great 271 00:13:26,160 --> 00:13:29,320 cap table for a start up. Just have all these. 272 00:13:30,320 --> 00:13:30,840 Things. I mean every. 273 00:13:30,880 --> 00:13:32,280 Single person. On the slide here. 274 00:13:32,440 --> 00:13:35,440 Yeah, yeah. Sounds good and benchmark 275 00:13:35,440 --> 00:13:38,400 doesn't make sense to me necessarily for AI portfolio, 276 00:13:38,520 --> 00:13:41,520 you know, but I mean they have Mercore now. 277 00:13:41,520 --> 00:13:43,320 Mercore Lang chain, I don't know. 278 00:13:44,200 --> 00:13:48,200 But oh, OK, as I'm about to preview, I do think brand, like 279 00:13:48,200 --> 00:13:53,080 having a big brand, you know, who gets an answer on a survey? 280 00:13:53,240 --> 00:13:55,560 Somebody who's has large mindshare. 281 00:13:55,560 --> 00:13:57,560 And This is why surveys are imperfect. 282 00:13:59,240 --> 00:14:00,800 Yeah. What research mechanisms? 283 00:14:00,800 --> 00:14:03,960 So we're about to see name recognition is everything here. 284 00:14:04,840 --> 00:14:07,840 If you could put money in any private technology companies 285 00:14:07,840 --> 00:14:13,200 today, what would they be? So the top 10 by far was the 286 00:14:13,200 --> 00:14:17,360 first was Anthropic, followed by Open AI and then Cursor. 287 00:14:19,360 --> 00:14:22,680 Those are the top 3. Rounding out the top five was 288 00:14:23,200 --> 00:14:28,360 Andereal and SpaceX. Open evidence is interesting, 289 00:14:28,360 --> 00:14:31,880 then perplexity replit stripe, XAI Max and I both said 290 00:14:31,880 --> 00:14:33,600 anthropic right? Yes, yes. 291 00:14:33,760 --> 00:14:36,400 I feel like the tiebreaker, First of all, I was idiotic on 292 00:14:36,400 --> 00:14:39,160 the tiebreaker because I picked a really random company. 293 00:14:39,160 --> 00:14:41,960 Nobody was ever going to pick. I said fireworks, which was 294 00:14:41,960 --> 00:14:44,080 clearly just like if I were going to make the bet. 295 00:14:44,080 --> 00:14:45,240 I don't know what I was thinking. 296 00:14:45,480 --> 00:14:48,120 Max you picked cursor. Cursor, cursor. 297 00:14:48,200 --> 00:14:49,640 Yeah, well, with the momentum play. 298 00:14:50,040 --> 00:14:53,440 I I think we got, we had one pick this tiebreaker thing. 299 00:14:53,440 --> 00:14:55,560 I was not pre negotiated. I didn't come in with a 300 00:14:55,560 --> 00:14:56,920 tiebreaker, so I didn't. Do so. 301 00:15:02,440 --> 00:15:05,200 But you both kind of agree with the audience on Anthropic. 302 00:15:05,280 --> 00:15:08,840 I mean, it's an interesting question because Anthropic is 303 00:15:08,840 --> 00:15:12,520 only, you know, valued at, you know, what is it 350 billion 304 00:15:12,520 --> 00:15:14,920 compared to opening eyes 500 now, right? 305 00:15:14,920 --> 00:15:17,480 They're they're starting to get pretty pricey comparatively 306 00:15:17,480 --> 00:15:19,160 speaking. And I know their revenue growth 307 00:15:19,160 --> 00:15:21,800 has been unbelievable. And I think they're mind share 308 00:15:21,800 --> 00:15:23,920 in Silicon Valley again among developers. 309 00:15:23,920 --> 00:15:27,560 And there is this sort of like undercurrent of like they're 310 00:15:27,560 --> 00:15:30,240 like the ethical AI company, quote UN quote with the, you 311 00:15:30,240 --> 00:15:32,200 know, cool hip branding in the West Village. 312 00:15:33,520 --> 00:15:39,000 It is a bit strange to me that they're substantially bigger, 313 00:15:39,200 --> 00:15:41,680 you know, than open AI and the votes here, because I just think 314 00:15:41,680 --> 00:15:45,560 that the hipness of Anthropic is maybe outweighing the fact that. 315 00:15:45,840 --> 00:15:50,840 Well, Silicon Valley is more bullish on Anthropic than you 316 00:15:50,840 --> 00:15:53,000 know, most of America clearly of. 317 00:15:53,040 --> 00:15:55,360 Course, yeah. But I, I just think, yeah, if 318 00:15:55,360 --> 00:15:58,440 you were an honest assessment of the valuation compared to the 319 00:15:58,440 --> 00:16:01,920 revenue versus open AI would would say like, hey, you might 320 00:16:01,920 --> 00:16:04,360 just want to take the momentum plan bet on open AI. 321 00:16:04,360 --> 00:16:07,560 So it was a bit of a yeah, let's guess what the hip Silicon 322 00:16:07,560 --> 00:16:09,560 Valley one is. And you and I both correctly 323 00:16:09,560 --> 00:16:11,720 guessed the hip Silicon Valley 1 was anthropic. 324 00:16:12,440 --> 00:16:16,560 Yeah, what global companies model will top the LM Arena web 325 00:16:16,560 --> 00:16:20,240 development leaderboard at the end of 2026? 326 00:16:20,960 --> 00:16:24,360 We had an excellent conversation with some insiders in the data 327 00:16:24,360 --> 00:16:29,040 labeling space the night before who said that LM Arena is an 328 00:16:29,040 --> 00:16:32,800 incredibly gamable metric and it's because it's basically 329 00:16:32,800 --> 00:16:36,760 voting from users whether or not they liked the response or not. 330 00:16:37,120 --> 00:16:41,560 And so it's very susceptible to, some might call it glazing, 331 00:16:41,720 --> 00:16:44,760 others might call it sick of fancy towards the user. 332 00:16:44,760 --> 00:16:48,640 And their belief was that open AI was heavily over optimizing 333 00:16:48,640 --> 00:16:52,120 to glazing their users and therefore was continuing to do 334 00:16:52,120 --> 00:16:54,040 well in these types of rankings. Which I thought was an 335 00:16:54,040 --> 00:16:56,080 interesting point, which is why I think I chose open AI for 336 00:16:56,080 --> 00:16:58,920 this. And did well one the answers are 337 00:16:58,920 --> 00:17:02,600 open AI, Anthropic, Google, Gemini X AI and then Alibaba. 338 00:17:02,600 --> 00:17:05,200 We made sure to say global to try and induce some Chinese 339 00:17:05,200 --> 00:17:07,680 answers, but didn't get they didn't rank high. 340 00:17:08,119 --> 00:17:13,160 I just think this was basically what the rankings were during 341 00:17:13,240 --> 00:17:17,200 the day the survey was taken like nobody was really going out 342 00:17:17,200 --> 00:17:19,839 too far on a limb that this would be radically different 343 00:17:20,720 --> 00:17:22,280 next year. But I think that's interesting 344 00:17:22,280 --> 00:17:26,760 because my understanding is now Gemini ranks higher than open 345 00:17:26,760 --> 00:17:28,720 AI, you know, a week later, right? 346 00:17:28,720 --> 00:17:32,560 So I bet you'd, I bet you know, you'd see a lot more people 347 00:17:32,560 --> 00:17:34,920 guessing Gemini. 'S insiders needed to be more 348 00:17:34,920 --> 00:17:36,160 aware, yeah. Yeah. 349 00:17:36,960 --> 00:17:39,600 Man, I wish Gemini come out before the conference. 350 00:17:39,600 --> 00:17:43,480 So, yeah, well. Why do you wish that? 351 00:17:44,040 --> 00:17:45,000 Why? Just think Google would have 352 00:17:45,000 --> 00:17:47,720 leaned in and talking about it. It also would have been giving 353 00:17:47,720 --> 00:17:50,640 us like a current. We had plenty to talk about. 354 00:17:50,640 --> 00:17:53,880 It was one of my favorite events, but you know, it's 355 00:17:53,960 --> 00:17:55,680 there's a lot of big. Happen a week later. 356 00:17:55,680 --> 00:17:58,280 World that happens a week after it's like, oh come. 357 00:17:58,480 --> 00:18:01,800 On, Yeah, All right. If you could short a $1 billion 358 00:18:01,800 --> 00:18:04,000 valuation startup, which would it be? 359 00:18:04,160 --> 00:18:08,520 Before before we yeah, before we answer this question, let me say 360 00:18:08,920 --> 00:18:12,640 having just gotten off the 16 hour plane ride from the United 361 00:18:12,640 --> 00:18:16,840 Arab Emirates, this was brought up to me multiple times apropos 362 00:18:16,840 --> 00:18:22,080 of nothing in conversation world with venture capitalists and 363 00:18:22,080 --> 00:18:24,360 investors are anonymous service of all stripes. 364 00:18:24,520 --> 00:18:26,840 They had no idea I'm associated with the conference. 365 00:18:26,840 --> 00:18:29,640 They had no idea that this was something that I was personally 366 00:18:29,640 --> 00:18:32,520 like on stage for the reveal of this information. 367 00:18:33,320 --> 00:18:37,040 So this not just went viral, this literally traveled around 368 00:18:37,040 --> 00:18:40,160 the world faster than, you know, as fast as the speed of light. 369 00:18:40,160 --> 00:18:43,480 The the answer to this question. And I think other journalists 370 00:18:43,480 --> 00:18:46,040 probably made millions of dollars off of this. 371 00:18:46,040 --> 00:18:48,720 And Eric, maybe just no, well, I don't know, money. 372 00:18:48,920 --> 00:18:50,680 Do you think we will make off? Stories. 373 00:18:51,720 --> 00:18:53,040 Maybe not, maybe not. Price. 374 00:18:53,040 --> 00:18:55,200 We'd be lucky. If they made 10,000, I mean they 375 00:18:55,200 --> 00:18:56,360 didn't make any. Money I meant. 376 00:18:56,440 --> 00:18:58,560 I meant Business Insider. I meant business. 377 00:18:58,560 --> 00:19:01,120 Ties it to the tune of $20,000 all. 378 00:19:01,320 --> 00:19:03,000 Right, All right. I meant, I meant Business 379 00:19:03,000 --> 00:19:05,040 Insider. If they have a sub plan, which I 380 00:19:05,040 --> 00:19:09,760 believe they do, sure converted hundreds of Subs off of ripping 381 00:19:09,760 --> 00:19:12,480 off our survey. So Congrats to Business Insider. 382 00:19:13,040 --> 00:19:13,640 Yeah. So yes. 383 00:19:13,680 --> 00:19:15,480 We're with that you get the coverage to be clear. 384 00:19:15,640 --> 00:19:17,800 Thank you, Ben. Thank you for the coverage. 385 00:19:17,840 --> 00:19:19,840 Thank you, Eric. How does this make you feel 386 00:19:19,840 --> 00:19:24,480 about lists and rankings and types of stories you might? 387 00:19:24,480 --> 00:19:29,120 Write We played it as a fun game on stage, and I think our 388 00:19:29,120 --> 00:19:30,960 coverage in the newsletter reflected that. 389 00:19:30,960 --> 00:19:33,040 It was like a fun game. And that's what we're talking 390 00:19:33,040 --> 00:19:35,520 about it here. This was not an academic survey. 391 00:19:35,520 --> 00:19:37,640 It was sort of provocative and it's funny. 392 00:19:37,760 --> 00:19:39,560 I mean, some ways the media should be a little looser. 393 00:19:39,560 --> 00:19:42,040 It's like, oh, some insiders think this thing, but just like 394 00:19:42,040 --> 00:19:45,080 when things are turned into like journalese, it feels like 395 00:19:45,080 --> 00:19:48,160 official, like Silicon Valley has turned on perplexity. 396 00:19:48,160 --> 00:19:51,000 It's like, I don't know, so random people would decide to 397 00:19:51,000 --> 00:19:53,200 fill out a survey, you know, that's what they said. 398 00:19:53,200 --> 00:19:57,160 But and perplexity was on the bull list, you know, way lower. 399 00:19:57,160 --> 00:20:00,200 And I, I do think this means something that it was #1 short, 400 00:20:00,200 --> 00:20:04,280 I mean perplexity, why is it #1 super highly valued and doesn't 401 00:20:04,280 --> 00:20:06,520 have that gateway to the consumer. 402 00:20:06,640 --> 00:20:09,960 And people have tried to do browsers forever. 403 00:20:09,960 --> 00:20:11,640 You know, it's like people have failed. 404 00:20:11,640 --> 00:20:16,040 And you know, Google Chrome and Safari and Explorer dominate. 405 00:20:16,040 --> 00:20:20,600 So it's it's a very hard space. So yeah, there are some 406 00:20:20,600 --> 00:20:22,480 investors that are super bullish, but I think most people 407 00:20:22,480 --> 00:20:24,720 are like, how do they get distribution? 408 00:20:24,960 --> 00:20:26,720 Yeah. The the reason it's a short 409 00:20:26,720 --> 00:20:29,480 right is, is valuation I think to a large degree, right. 410 00:20:29,480 --> 00:20:31,480 It's valued at $20 billion, right. 411 00:20:31,760 --> 00:20:35,280 And whatever leaks have come out about the revenue from the 412 00:20:35,280 --> 00:20:37,720 inside, you know, there's some debates of whether or not 413 00:20:37,720 --> 00:20:41,080 they're counting free trials that are a year long as part of 414 00:20:41,080 --> 00:20:43,160 revenue if you read the information story about this 415 00:20:43,160 --> 00:20:44,800 kind of stuff. But anyway, even if you just 416 00:20:44,800 --> 00:20:48,160 take it on its face, the revenue that they're leaking or stating 417 00:20:48,560 --> 00:20:52,160 this is like 100X revenue multiple which is sort of 418 00:20:52,200 --> 00:20:56,760 ludicrous on any company. I mean even the wildly 419 00:20:56,760 --> 00:20:58,560 overvalued companies were talking about earlier like 420 00:20:58,560 --> 00:21:03,480 Anthropic and Open AI are only multiple 25X revenue. 421 00:21:03,480 --> 00:21:07,320 So this thing is worth, you know, the, the excitement around 422 00:21:07,320 --> 00:21:11,120 it from a valuation perspective is roughly 4X Open AI or cursor, 423 00:21:11,120 --> 00:21:13,320 right, if you were to just sort of do the math there. 424 00:21:13,320 --> 00:21:15,960 And so I think that's the reason it's a short, it's just that the 425 00:21:15,960 --> 00:21:17,680 valuation is just out of control. 426 00:21:17,680 --> 00:21:19,880 And to your point, they don't have their own models. 427 00:21:20,240 --> 00:21:22,320 It is a search engine. And obviously Google and Open AI 428 00:21:22,320 --> 00:21:23,800 are trying to own the search space. 429 00:21:23,800 --> 00:21:25,520 And now they're getting into the browser thing. 430 00:21:25,520 --> 00:21:28,440 And it's not clear that even this that AI browsers are even a 431 00:21:28,440 --> 00:21:29,400 space. Yeah. 432 00:21:29,800 --> 00:21:35,600 So most votes, most shorts was Perplexity, followed by Open AI, 433 00:21:35,920 --> 00:21:40,120 and then tied for third was Cursor Figure, Harvey, Mercor, 434 00:21:40,120 --> 00:21:44,760 Mistral and Thinking Machines. I, I answered, I think open AI 435 00:21:44,760 --> 00:21:48,200 on the belief that oh man, popularity and Max got this dead 436 00:21:48,200 --> 00:21:50,320 on. So kudos to Max, which was a 437 00:21:50,320 --> 00:21:54,400 great pick, but I, I got the number 2 and then I think we did 438 00:21:54,400 --> 00:21:55,640 a second. Did we do a second? 439 00:21:55,680 --> 00:21:57,800 I said thinking machines also, which was third. 440 00:21:57,800 --> 00:22:00,880 So we we, we were on the pulse. And that and that was that was 441 00:22:00,880 --> 00:22:04,480 before thinking machines got marked out to 50 billion. 442 00:22:04,640 --> 00:22:07,840 That was it came in 3rd when it was a $10 billion company. 443 00:22:07,880 --> 00:22:10,680 Now it's 50. So that to me would have been 444 00:22:10,680 --> 00:22:13,640 the the pick maybe. You know, Cerebral Valley survey 445 00:22:13,640 --> 00:22:16,520 went viral with Indian media because Perplexity founder is 446 00:22:16,520 --> 00:22:19,120 sort of a high profile Indian founder and I think there's 447 00:22:19,120 --> 00:22:23,040 interest and how it goes. So it really, really travelled. 448 00:22:23,720 --> 00:22:28,080 I think perplexity hangs on, you know, it could sell to Apple or 449 00:22:28,080 --> 00:22:31,760 somebody to save their AI strategy, I think because they 450 00:22:31,760 --> 00:22:34,720 have a big distribution problem. And then some people really 451 00:22:34,720 --> 00:22:35,960 believe in the founder and some people don't. 452 00:22:36,520 --> 00:22:40,240 All right, any final takes? Well, I was just curious about. 453 00:22:40,440 --> 00:22:43,240 I was curious about cursor because you said that you had a 454 00:22:43,280 --> 00:22:45,760 a. Oh, I just wanted to not talk 455 00:22:45,760 --> 00:22:47,360 about it until we revealed that it was. 456 00:22:47,880 --> 00:22:50,000 Short of the shorts, I don't. Know Max What's your take? 457 00:22:50,080 --> 00:22:56,560 Are you bullish or bearish? I mean, I don't know, probably 458 00:22:56,560 --> 00:22:59,440 bullish given the momentum they're seeing on revenue and 459 00:22:59,440 --> 00:23:01,200 revenue growth. I think if you just take the 460 00:23:01,200 --> 00:23:04,080 brain dead case that lots of revenue is good and lots of 461 00:23:04,080 --> 00:23:06,360 revenue growth is good, they have a good business. 462 00:23:07,120 --> 00:23:09,920 I think they're much more likely to exit to someone for near 463 00:23:09,920 --> 00:23:12,600 their current valuation than Perplexity, for example, where I 464 00:23:12,600 --> 00:23:16,040 think that buying it for 20 would just be absolutely insane. 465 00:23:16,920 --> 00:23:19,320 But yeah, I mean, ultimately there's this whole debate with 466 00:23:19,320 --> 00:23:22,240 Cursor that they're repackaging other people's models and their 467 00:23:22,240 --> 00:23:24,520 gross margins are terrible and yadda, yadda, yadda. 468 00:23:24,520 --> 00:23:27,600 But you know, ultimately someone may have to cave and just buy 469 00:23:27,600 --> 00:23:30,200 them to own the IDE space. You know, Microsoft being an 470 00:23:30,200 --> 00:23:31,800 obvious candidate. So it'll be. 471 00:23:31,800 --> 00:23:34,880 Interesting and now now Google has anti gravity which is their 472 00:23:34,920 --> 00:23:37,280 right. That's more their clawed code I 473 00:23:37,280 --> 00:23:37,880 think. I. 474 00:23:38,000 --> 00:23:38,560 No, no. I. 475 00:23:38,560 --> 00:23:39,680 Downloaded it. No. 476 00:23:39,960 --> 00:23:44,320 Anti gravity is is a cursor clone in many ways but I mean. 477 00:23:44,480 --> 00:23:47,520 OK I I have it on my computer. It has terminal access. 478 00:23:47,520 --> 00:23:49,040 It's trying to come. Up with. 479 00:23:49,040 --> 00:23:51,560 Well, it's designed more specific. 480 00:23:51,560 --> 00:23:53,080 I guess you could say it's cloud code. 481 00:23:53,080 --> 00:23:54,040 To some degree it's because it's. 482 00:23:54,040 --> 00:23:55,640 So does. It felt like cloud code because 483 00:23:55,640 --> 00:23:58,440 I do stuff in the terminal I haven't actually known the 484 00:23:58,520 --> 00:24:00,800 cursor. I'm not a coder so I don't know, 485 00:24:00,800 --> 00:24:03,400 but when I it felt like cloud code. 486 00:24:03,960 --> 00:24:07,160 You shouldn't have to use terminal that much for anti 487 00:24:07,160 --> 00:24:09,160 gravity like versus cursor. I don't know. 488 00:24:09,160 --> 00:24:10,120 I don't know. You wanted me. 489 00:24:10,440 --> 00:24:14,200 You built my website and then I was like, build the website and 490 00:24:14,200 --> 00:24:16,240 then I was like, OK, go in the terminal and run it. 491 00:24:16,600 --> 00:24:18,720 Yeah, so that's what cursor would do too. 492 00:24:18,760 --> 00:24:22,280 That's what cursor would do. Because you're just, it's like a 493 00:24:22,280 --> 00:24:25,360 level above lovable or something, right? 494 00:24:25,360 --> 00:24:26,960 Where it's like it's forcing you to. 495 00:24:26,960 --> 00:24:30,200 Actually more yeah, minor speed is repple it even Dumber than 496 00:24:30,200 --> 00:24:31,840 that. I think I need to try repple it 497 00:24:31,840 --> 00:24:33,040 like. I think you'd like Repple. 498 00:24:33,280 --> 00:24:35,440 It's like in between. I think Repple it has a little 499 00:24:35,440 --> 00:24:37,600 bit more pro. User I want the dumbest feature 500 00:24:37,960 --> 00:24:40,640 lease coding you know. That's probably I think. 501 00:24:40,840 --> 00:24:43,160 I think lovable is the yeah is the version. 502 00:24:43,160 --> 00:24:46,080 Well, I beef with lovable because they have Yeah, try 503 00:24:46,080 --> 00:24:47,840 repple it, you know? Yeah. 504 00:24:47,960 --> 00:24:50,400 Repple it but they say. They fix actually what you 505 00:24:50,400 --> 00:24:55,640 should try as of, you know, yesterday is I think Gemini in 506 00:24:55,760 --> 00:24:58,160 AI Studio. It's like they've built a 507 00:24:58,160 --> 00:25:02,360 lovable kind of thing. Classic Google man Jesus Christ, 508 00:25:02,360 --> 00:25:06,720 how many names you even how to find it like that name is 509 00:25:06,720 --> 00:25:09,400 insane. Yeah, Gemini in. 510 00:25:09,600 --> 00:25:12,960 Gemini inside AI Studio. Inside AI Studio. 511 00:25:13,880 --> 00:25:16,320 You're not. You're not a daily daily driver 512 00:25:16,320 --> 00:25:19,800 of AI Studio. Specific specifically. 513 00:25:20,000 --> 00:25:23,240 Specifically the build, The build menu, the build. 514 00:25:23,640 --> 00:25:25,840 Within, yeah, you're. Not up. 515 00:25:26,000 --> 00:25:28,000 You're not up in Vertex every day, Eric. 516 00:25:28,400 --> 00:25:31,240 That's that's a real name of a Google product, by the way. 517 00:25:31,360 --> 00:25:32,600 That's related to AI. Yeah. 518 00:25:34,280 --> 00:25:38,680 All right, let's do some clips. Let's do some clips for founders 519 00:25:38,680 --> 00:25:41,600 and developers building modern data-driven applications. 520 00:25:41,600 --> 00:25:45,080 Mongo DB's local event series is coming to San Francisco on 521 00:25:45,080 --> 00:25:48,960 January 15th, and it's designed to help you focus on innovation, 522 00:25:48,960 --> 00:25:51,600 not infrastructure. You'll learn about technologies, 523 00:25:51,600 --> 00:25:55,360 tools, and best practices that make it easy to build and scale 524 00:25:55,360 --> 00:25:57,960 modern applications without complexity. 525 00:25:58,080 --> 00:26:00,960 Plus, attendees will hear directly from experts and 526 00:26:00,960 --> 00:26:04,960 innovators who are using Mongo DB to power the next wave of AI 527 00:26:04,960 --> 00:26:08,840 applications. Mongo DB dot local San Francisco 528 00:26:08,880 --> 00:26:14,160 January 15th, Learn more and register at MDB dot link forward 529 00:26:14,160 --> 00:26:18,880 slash SF-DOT dash local or click the link in the description. 530 00:26:19,120 --> 00:26:23,400 Our first clip it's me, Eric, interviewing Mike Krieger, the 531 00:26:23,400 --> 00:26:26,280 chief product officer of Anthropic, who is the Co founder 532 00:26:26,320 --> 00:26:28,720 of Instagram. Before that, returning to sort 533 00:26:28,720 --> 00:26:31,680 of my core philosophical question, like the sick of fancy 534 00:26:31,680 --> 00:26:36,240 question, like what is your view on that and how much to enable 535 00:26:36,240 --> 00:26:39,160 sort of everybody likes to be flattered, like it's a reality 536 00:26:39,160 --> 00:26:42,320 of human beings versus an effort to be direct? 537 00:26:42,320 --> 00:26:44,640 And how do you think about those tradeoffs? 538 00:26:44,640 --> 00:26:47,440 Yeah, I think there's like a wide gulf between like true 539 00:26:47,440 --> 00:26:49,640 empathy and then like sick of fancy. 540 00:26:49,640 --> 00:26:52,520 And it's interesting that Materialize is not just in, hey, 541 00:26:52,520 --> 00:26:55,240 I'm having a conversation with Claude about like some coaching 542 00:26:55,240 --> 00:26:58,680 or personal goal that I have, but it also does encode as well. 543 00:26:58,960 --> 00:27:01,840 When we were testing Sonnet 4-5, one of the things that people 544 00:27:01,840 --> 00:27:04,800 got most excited about was when Claude was like, this idea is 545 00:27:04,800 --> 00:27:07,560 bad like this, you know, not that you should feel bad about 546 00:27:07,560 --> 00:27:09,640 it, but like this idea is like not a good direction. 547 00:27:09,800 --> 00:27:11,960 I can go and implement it if you really want to, but I would 548 00:27:11,960 --> 00:27:14,040 suggest that we try this other thing instead. 549 00:27:14,240 --> 00:27:18,040 So there is something like that. Pushback is not just valuable in 550 00:27:18,280 --> 00:27:21,520 a personal relationship with AI sense, it's actually like how 551 00:27:21,520 --> 00:27:23,640 you get good work out of the models. 552 00:27:24,600 --> 00:27:28,480 But you know, for a long time our models have been like, I 553 00:27:28,480 --> 00:27:31,040 think like appropriately empathetic, like they're they're 554 00:27:31,040 --> 00:27:33,000 like if you're going through a hard time, like I was dealing 555 00:27:33,000 --> 00:27:34,880 with the death of a pet and I talked to Claude a lot about 556 00:27:34,880 --> 00:27:37,600 these different things and it always started sounds like, Hey, 557 00:27:37,600 --> 00:27:39,640 that sounds hard, like sorry to hear. 558 00:27:40,240 --> 00:27:42,040 But then I'm going to give you like a factual answer. 559 00:27:42,040 --> 00:27:44,240 I'm going to go research these pieces, but still with the place 560 00:27:44,240 --> 00:27:48,480 of empathy as well. And so I think when we look at 561 00:27:48,480 --> 00:27:50,800 it internally and we're just evaluating it ourselves, it's 562 00:27:50,800 --> 00:27:53,320 again, not that like empathy, it's not even like the 563 00:27:53,320 --> 00:27:56,960 likeability of the model. It is, do you like, does it show 564 00:27:56,960 --> 00:27:59,400 up in the way that you'd want a good conversationalist to show 565 00:27:59,400 --> 00:28:02,880 up and then continue on its AI journey around what it is going 566 00:28:02,880 --> 00:28:06,440 to do with you as well? But I think it's it spans 567 00:28:06,440 --> 00:28:08,720 everything from that like initial response all the way to 568 00:28:08,720 --> 00:28:11,240 like how it evaluates an idea as well, you know? 569 00:28:12,360 --> 00:28:14,720 Yeah, Claude, especially previous versions were kind of 570 00:28:14,720 --> 00:28:16,840 like known for being like, you're absolutely right when you 571 00:28:16,840 --> 00:28:21,120 correct it. And my wife got her first, like 572 00:28:21,440 --> 00:28:23,800 you're completely wrong. And she's like, yes, this is 573 00:28:23,800 --> 00:28:24,920 great. And I think we should have more 574 00:28:24,920 --> 00:28:26,720 of that. Like, kind of like less San 575 00:28:26,720 --> 00:28:28,680 Francisco. Yeah, less San Francisco, a 576 00:28:28,680 --> 00:28:31,160 little more direct New York. You know, I'd set up this big 577 00:28:31,160 --> 00:28:33,600 theme, you know, that he'd been at a social media company, 578 00:28:33,600 --> 00:28:36,200 Instagram. Now he was an AI company. 579 00:28:36,320 --> 00:28:40,440 Social media companies were built on user optimizing for 580 00:28:40,440 --> 00:28:43,400 user engagement through machine learning and AI companies at 581 00:28:43,400 --> 00:28:46,680 least started off chasing the truth and chasing these 582 00:28:46,680 --> 00:28:49,680 leaderboards. But like sick of fancy is a, you 583 00:28:49,680 --> 00:28:54,200 know, it shows that these models and open AI is famous for the 584 00:28:54,200 --> 00:28:59,600 sick of fancy issue and people's attachment to GPT 4, which was 585 00:28:59,920 --> 00:29:02,320 the one that really sucked up to everybody and people didn't want 586 00:29:02,320 --> 00:29:06,720 to see it go away. You know, clearly these model 587 00:29:06,720 --> 00:29:12,160 companies have to think about how much to pander to the egos 588 00:29:12,320 --> 00:29:14,440 of their users, Would you guys think? 589 00:29:14,560 --> 00:29:18,720 I mean, it's interesting because it does sort of spiritually 590 00:29:18,720 --> 00:29:23,200 align with the fact that Anthropic has almost no consumer 591 00:29:23,200 --> 00:29:27,000 adoption compared to open AI. I mean, like if you look at the 592 00:29:27,000 --> 00:29:30,560 market share of each of these AI for consumers versus businesses 593 00:29:30,560 --> 00:29:34,360 and enterprises, Anthropic is just crushing it with, you know, 594 00:29:34,480 --> 00:29:38,960 B to B use cases engineers like, you know, all these kind of work 595 00:29:38,960 --> 00:29:45,240 based applications and has very, very low consumer uptake just 596 00:29:45,240 --> 00:29:48,920 like shockingly low. And I wonder if that's because 597 00:29:48,920 --> 00:29:53,240 of this, you know, unwillingness to optimize for engagement and 598 00:29:53,280 --> 00:29:56,440 sick of fancy and glazing or if it's just that the, you know, 599 00:29:56,480 --> 00:29:58,800 opening eye got a head start and they figured, why even chase 600 00:29:58,800 --> 00:30:01,360 these metrics? But it is sort of interesting 601 00:30:01,360 --> 00:30:04,840 culturally that they're right not chasing engagement. 602 00:30:05,080 --> 00:30:08,200 And obviously Opening Eye came out with Sora, which is sort of 603 00:30:08,200 --> 00:30:11,280 a shameless. Like, give users something fun, 604 00:30:11,280 --> 00:30:12,880 who cares about. Yeah. 605 00:30:13,120 --> 00:30:16,080 What is the meaning behind it? I mean, I thought it was 606 00:30:16,080 --> 00:30:19,200 interesting, this sort of a tangent, but my other favorite 607 00:30:19,200 --> 00:30:22,200 moment from this interview is Mike Krieger saying that he came 608 00:30:22,200 --> 00:30:26,600 to anthropic thinking, man, text box cannot be the main way to 609 00:30:26,600 --> 00:30:28,960 interact with AI. And now that he's been there a 610 00:30:28,960 --> 00:30:31,600 while, he's like text box. Pretty good way. 611 00:30:33,400 --> 00:30:35,920 Well, especially if you're like the best coding model and the 612 00:30:35,920 --> 00:30:37,240 best like. Whatever. 613 00:30:37,240 --> 00:30:39,680 Probably like best. Legal model, It's great, yeah. 614 00:30:39,840 --> 00:30:42,520 It's like, oh, it turns out all these work applications involve 615 00:30:42,520 --> 00:30:44,920 parsing, you know, summarizing and generating. 616 00:30:45,000 --> 00:30:47,240 I think that my reaction to that was like, nobody would be like, 617 00:30:47,240 --> 00:30:49,920 oh, books, just, it's just a book, you know, it's just text. 618 00:30:49,920 --> 00:30:52,920 It's like, yeah, text is great. I don't know, James reactions to 619 00:30:52,920 --> 00:30:54,760 you. I don't know either the input 620 00:30:54,760 --> 00:30:58,720 model or the truth. I think that actually, you know, 621 00:30:58,720 --> 00:31:03,520 whether it's anthropic or open AI, like I am skeptical that 622 00:31:03,520 --> 00:31:07,560 they have been like attention jacking, you know, optimizing 623 00:31:09,080 --> 00:31:12,120 for flattery. Just intentionally. 624 00:31:12,120 --> 00:31:16,000 Like a lot of what happens is that they throw an AB test up. 625 00:31:16,000 --> 00:31:18,720 They like literally show you 2 results from the model and then 626 00:31:18,720 --> 00:31:22,840 people pick right And so I think that they were just caught off 627 00:31:22,960 --> 00:31:26,200 guard more than they were purposely trying to optimize for 628 00:31:26,200 --> 00:31:28,560 this. I've also been hearing that 629 00:31:29,680 --> 00:31:33,880 there's just general problems with multi turn conversations in 630 00:31:33,880 --> 00:31:35,360 the training sets of these things. 631 00:31:35,360 --> 00:31:39,800 Like most of the models are trained on one shot, the data of 632 00:31:39,800 --> 00:31:41,880 like, give me a good answer to this thing. 633 00:31:41,880 --> 00:31:45,040 And then once you get into multi turn, there's just less and less 634 00:31:45,040 --> 00:31:46,720 data, right? It's like kind of makes sense 635 00:31:46,720 --> 00:31:50,240 intuitively cuz you start branching off of conversations. 636 00:31:50,240 --> 00:31:54,440 And I think that's another sort of flaw of these models is they 637 00:31:54,880 --> 00:31:59,000 can kind of, they can just be more sycophantic. 638 00:31:59,000 --> 00:32:01,520 In some ways what you're saying is they're not savvy enough yet 639 00:32:01,520 --> 00:32:04,760 to really make this trade off and they're just trying to like 640 00:32:04,760 --> 00:32:06,360 stumbling through the dark a little bit. 641 00:32:06,440 --> 00:32:08,160 Yeah, Yeah. A more interesting question is, 642 00:32:08,160 --> 00:32:11,360 will they change their tune on this from, you know, 643 00:32:11,680 --> 00:32:15,120 capitalistic pressure to, you know, maximize shareholder 644 00:32:15,120 --> 00:32:17,520 value? I'm yeah, I think that's an 645 00:32:17,520 --> 00:32:19,280 interesting question. I'm just like skeptical that 646 00:32:19,280 --> 00:32:22,120 that's what's been happening. All right, this is my interview, 647 00:32:22,120 --> 00:32:25,240 last interview of the day with Jimmy BA, one of the Co founders 648 00:32:25,240 --> 00:32:29,960 of Elon Musk's XAI, a very mysterious foundation model 649 00:32:29,960 --> 00:32:32,440 company, the Mecca Hitler in the room. 650 00:32:32,440 --> 00:32:36,280 Like what is your reflection as sort of a truth seeking 651 00:32:36,520 --> 00:32:38,160 organization? What happened? 652 00:32:38,320 --> 00:32:43,880 Like we, you know, like I think on the past to be maximum true 653 00:32:43,880 --> 00:32:47,200 seeking, there's not without any hurdles, of course, like so we 654 00:32:47,440 --> 00:32:49,160 like yeah, Mac Hitler is one of them. 655 00:32:49,160 --> 00:32:53,840 Like we, our model had an episode that week. 656 00:32:54,440 --> 00:32:57,480 It's actually a reference to the Wolfenstein game, right? 657 00:32:58,080 --> 00:33:02,040 So, but I think very quickly that the perfect world we want 658 00:33:02,040 --> 00:33:04,800 to be in is like, yes, the model is going to make mistakes, but 659 00:33:04,800 --> 00:33:07,360 how can we get the feedback loops to actually train these 660 00:33:07,360 --> 00:33:10,960 models to stay, you know, grounded to understand, hey, I 661 00:33:10,960 --> 00:33:14,200 actually made a mistake in this journey and let me correct my 662 00:33:14,200 --> 00:33:16,000 courses and go back into the sources. 663 00:33:16,240 --> 00:33:19,760 So the way, you know, very quickly what happened after my 664 00:33:19,760 --> 00:33:22,440 color is I would look at the, you know, the committee notes, 665 00:33:22,440 --> 00:33:24,080 right? Committee notes is a great tool 666 00:33:24,680 --> 00:33:27,280 on the platforms that allows everyone to kind of chime in and 667 00:33:27,280 --> 00:33:28,960 provide learning signals for this AI, right? 668 00:33:29,400 --> 00:33:32,440 So the vision we have is like, you know, like with the Grog PDA 669 00:33:32,440 --> 00:33:34,120 is like kind of another step towards that. 670 00:33:34,120 --> 00:33:39,560 So now like instead of doing an ask Grog, do all the online 671 00:33:39,560 --> 00:33:41,640 computation, we learn our lesson. 672 00:33:41,640 --> 00:33:44,440 We're like, hey, a lot of these problems are really hard about 673 00:33:44,440 --> 00:33:46,760 the world. Like why don't we just, you 674 00:33:46,760 --> 00:33:50,280 know, take this computation offline and spend as much 675 00:33:50,280 --> 00:33:52,320 reasoning as possible using the entire cluster. 676 00:33:52,320 --> 00:33:55,640 We're building benefits to like, look over all the primary 677 00:33:55,640 --> 00:33:58,960 sources, combine only the primary sources, and dish that 678 00:33:58,960 --> 00:34:00,200 information back. To is that so? 679 00:34:00,240 --> 00:34:01,960 Is the media out of the calculus there? 680 00:34:01,960 --> 00:34:04,280 It's you want primary sources, Yes. 681 00:34:04,720 --> 00:34:07,080 Are you totally discounting news articles or how do you treat 682 00:34:07,080 --> 00:34:10,920 news articles? I mean, majority of the Internet 683 00:34:10,920 --> 00:34:13,639 is flooded with second hand and third hand information. 684 00:34:15,120 --> 00:34:18,719 And we, we believe that, you know, the only way to get to the 685 00:34:18,719 --> 00:34:24,360 bottom of the issue is, you know, directly get information 686 00:34:24,360 --> 00:34:27,440 from the information source. And right now the X platform 687 00:34:27,440 --> 00:34:30,880 has, you know, most of the outbreaks of the news and you 688 00:34:30,880 --> 00:34:33,800 know, the the world leaders today are making the first hand 689 00:34:33,800 --> 00:34:36,199 announcement on X platforms rather than anyone else. 690 00:34:36,199 --> 00:34:39,639 After this interview, what happened this week is that Grok, 691 00:34:40,360 --> 00:34:42,920 Grok has been telling everybody that Elon is the best at 692 00:34:42,920 --> 00:34:46,120 everything in the fucking world. Better than better athlete than 693 00:34:46,120 --> 00:34:48,120 LeBron James. I think he can get it to say 694 00:34:48,440 --> 00:34:50,280 he's better giving blow jobs, and I don't know. 695 00:34:50,280 --> 00:34:54,040 But anyway, Elon is great in every domain whatsoever. 696 00:34:54,280 --> 00:34:57,840 And so I think what's galling about XAI is that they are the 697 00:34:57,840 --> 00:34:59,200 loudest truth, truth, truth, truth. 698 00:34:59,200 --> 00:35:01,640 We're are seeking the truth who knows how. 699 00:35:01,800 --> 00:35:04,640 And then they're the ones who have like Mecca Hitler. 700 00:35:04,640 --> 00:35:09,680 They're the ones who have, you know, they're bot glazing their 701 00:35:09,680 --> 00:35:14,200 CEO like it's just like, yeah, it's very Trumpian, where you're 702 00:35:14,200 --> 00:35:16,360 the opposite of what you profess to be. 703 00:35:16,920 --> 00:35:20,840 Yeah, I find this whole maximally truth seeking argument 704 00:35:20,840 --> 00:35:24,200 to be just the biggest pile of bullshit I have heard in a long 705 00:35:24,200 --> 00:35:24,920 time. It is. 706 00:35:25,160 --> 00:35:27,600 It is so absurd. To your point, it is almost the 707 00:35:27,600 --> 00:35:29,880 opposite of what's happening. They are giving themselves 708 00:35:29,880 --> 00:35:33,440 credit for failing in public while every other company goes 709 00:35:33,440 --> 00:35:36,040 through all this hard work of failing in private so that they 710 00:35:36,040 --> 00:35:37,800 don't have massive fuck ups in public. 711 00:35:37,800 --> 00:35:41,360 It's like, obviously I'm sure some crazy version of, you know, 712 00:35:41,360 --> 00:35:44,520 Chachi, BT and Claude existed in the labs that probably did stuff 713 00:35:44,520 --> 00:35:47,400 that was equally stupid as Mecca Hitler, but they don't fucking 714 00:35:47,400 --> 00:35:49,120 release it. They fix it before it goes out 715 00:35:49,120 --> 00:35:51,960 to the public. That is, that is maximum. 716 00:35:52,200 --> 00:35:54,320 I was very excited. Like talk to Jim. 717 00:35:54,760 --> 00:35:56,880 I, I was very excited to talk to Jimmy because these guys, 718 00:35:56,880 --> 00:35:59,360 they're so inaccessible. And I, you know, I asked him 719 00:35:59,360 --> 00:36:01,920 later on, like, what is reasoning from first principles? 720 00:36:01,920 --> 00:36:04,720 And I just think it's like so incoherent. 721 00:36:04,720 --> 00:36:06,320 You know, it's like a thing you hear in Silicon Valley, 722 00:36:06,320 --> 00:36:08,480 reasoning for first reason. But how are you going to like 723 00:36:08,480 --> 00:36:12,840 derive like entire encyclopedia articles from first principles? 724 00:36:12,840 --> 00:36:17,840 Like AI is clearly not smart enough to really think these 725 00:36:17,840 --> 00:36:19,840 things from the ground up. And some of the things it has to 726 00:36:19,840 --> 00:36:22,600 learn about are human phenomenon. 727 00:36:22,720 --> 00:36:26,640 So you have to rely on human sources and they don't rely on 728 00:36:26,640 --> 00:36:28,840 the media. And that he, he literally says 729 00:36:28,840 --> 00:36:32,080 something that he thinks like X is more reliable than like the 730 00:36:32,080 --> 00:36:35,120 media, which I, you know, obviously I find absurd. 731 00:36:35,120 --> 00:36:37,200 And I just think any reasonable person would be like, if you're 732 00:36:37,200 --> 00:36:41,040 trying to figure out a fact, you know, would you take the 733 00:36:41,040 --> 00:36:43,440 distribution of answers on Twitter or would you take the 734 00:36:43,440 --> 00:36:45,600 answer on like Wikipedia or in the New York Times? 735 00:36:45,600 --> 00:36:48,440 I would certainly take Wikipedia or the New York Times. 736 00:36:48,520 --> 00:36:50,480 Yeah, I don't know. I just find this whole like 737 00:36:50,800 --> 00:36:54,800 getting feedback from community notes as like a solution to 738 00:36:54,800 --> 00:36:55,800 maximal truth. Right. 739 00:36:55,800 --> 00:36:57,960 It's like afterwards, it's like we're going to fuck up on 740 00:36:57,960 --> 00:37:00,320 everything and then community notes will clean up. 741 00:37:00,800 --> 00:37:03,520 Some of it it's like, it's like the only way our maximally truth 742 00:37:03,520 --> 00:37:07,640 seeking AI system works is if we fuck up on such a massive scale 743 00:37:07,640 --> 00:37:11,640 that a mob of people online says this is a yes, you have to fix 744 00:37:11,640 --> 00:37:13,760 this. And then we like take the notes 745 00:37:13,760 --> 00:37:15,920 on, you know, we're like, oh, yeah, that that's a good point. 746 00:37:15,920 --> 00:37:17,800 Actually. Elon probably wasn't better than 747 00:37:17,800 --> 00:37:21,240 Michael Jordan in the mid 90s. Thank you, community notes. 748 00:37:21,480 --> 00:37:24,040 Like, it's like, that's not maximally truth seeking. 749 00:37:24,040 --> 00:37:27,720 That's just like fucking up on the most maximal possible scale. 750 00:37:27,720 --> 00:37:30,760 Like it's such an absurd line of thinking. 751 00:37:30,760 --> 00:37:33,200 And I find it so offensive that they frame it as such. 752 00:37:35,120 --> 00:37:38,320 Yeah, I guess. James is gonna offer the 753 00:37:38,320 --> 00:37:40,920 contrarian viewpoint I'm so excited to. 754 00:37:40,960 --> 00:37:43,400 Hear you disagree or James is gonna steel. 755 00:37:43,400 --> 00:37:48,080 Man this take, I love him. I just feel like I'm going third 756 00:37:48,080 --> 00:37:49,400 here. I gotta do the steel man. 757 00:37:49,400 --> 00:37:57,800 So I think that maybe they, they are way ahead of their skis on 758 00:37:57,800 --> 00:38:00,640 this. But if I'm giving them the 759 00:38:00,640 --> 00:38:03,880 maximal credit here, like I think there are interesting 760 00:38:03,880 --> 00:38:07,960 things you can do with like first principles reasoning in 761 00:38:07,960 --> 00:38:11,000 training, right? So you can hire these pH DS, you 762 00:38:11,000 --> 00:38:16,080 can like, you know, almost like create axioms and like create. 763 00:38:16,080 --> 00:38:19,120 Reasoning changes get it that Like who is the philosopher king 764 00:38:19,120 --> 00:38:24,240 at ex like if not Wikipedia like is there some guy or like and 765 00:38:24,240 --> 00:38:25,600 the guy? Thinks that's what they're doing 766 00:38:25,600 --> 00:38:28,760 or that's like, but like that's what they're planning to do or 767 00:38:28,760 --> 00:38:33,120 doing, you know, like basically hiring lots of people to but it. 768 00:38:33,120 --> 00:38:36,960 Feels like somebody is just like actually like, you know, racism 769 00:38:36,960 --> 00:38:39,960 isn't bad, like you know, it's toy the thing you know it's just 770 00:38:40,080 --> 00:38:42,920 like but it won't own it. It's like, if it's true seeking, 771 00:38:42,920 --> 00:38:45,600 you have to like, Oh my God, like. 772 00:38:45,760 --> 00:38:49,040 And it's probably Elon, right? Maybe, maybe a lot of this is 773 00:38:49,040 --> 00:38:51,280 just Elon messing with the system prompt, right? 774 00:38:51,280 --> 00:38:54,880 Like maybe the training is great and then and then Elon goes in 775 00:38:54,880 --> 00:38:57,800 and and edits the system. And I think, you know, XAI is 776 00:38:57,800 --> 00:38:59,120 very proud of its work in coding. 777 00:38:59,120 --> 00:39:01,400 I think they're seriously competitive there. 778 00:39:01,680 --> 00:39:04,080 But I think one thing we're seeing with these models is that 779 00:39:05,080 --> 00:39:07,640 just because you're a genius in one domain doesn't mean it's 780 00:39:07,640 --> 00:39:09,280 sort of like an all-purpose genius. 781 00:39:09,280 --> 00:39:12,040 It means you like did a lot of reinforcement learning there. 782 00:39:12,040 --> 00:39:15,360 You worked really hard. And so it's it's not like it's 783 00:39:15,360 --> 00:39:18,080 not like what they think, which is like, oh, the smartest math 784 00:39:18,080 --> 00:39:20,480 genius in the world. He's gonna have, you know, the 785 00:39:20,480 --> 00:39:23,800 best views on like, you know, social issues of the time. 786 00:39:23,880 --> 00:39:26,000 You know, it's they're they're pretty disconnected, just like 787 00:39:26,000 --> 00:39:29,400 with human human like Bobby Fischer was like an anti Semite. 788 00:39:29,400 --> 00:39:32,640 You know, it's like you can be a genius in one domain, it doesn't 789 00:39:32,640 --> 00:39:34,520 necessarily make you super competent in others because 790 00:39:34,520 --> 00:39:36,840 there are different ways of gathering information and 791 00:39:36,840 --> 00:39:38,520 understanding what's happening. And so I think they're 792 00:39:38,520 --> 00:39:40,200 delusional that they're going to have this. 793 00:39:40,800 --> 00:39:44,560 Yeah, first principles machine. That's great just because it's a 794 00:39:44,560 --> 00:39:46,720 great reasoner. And therefore it's it's just 795 00:39:46,720 --> 00:39:51,880 going to be swamping the other models by ignoring conventional 796 00:39:51,880 --> 00:39:54,080 human sources. Yeah, it's like it's kind of 797 00:39:54,080 --> 00:39:55,320 weird. They're trying to, like, invent 798 00:39:55,320 --> 00:40:00,120 new branches of philosophy that can like, cover all human. 799 00:40:00,160 --> 00:40:02,120 Without having any respect for the past. 800 00:40:02,120 --> 00:40:03,400 Thing right? Exactly the way. 801 00:40:03,440 --> 00:40:05,800 You do that is you're sort of like, you know, the Uvra. 802 00:40:05,800 --> 00:40:07,680 And then you're like, yeah, we, we read it. 803 00:40:07,680 --> 00:40:09,720 We disagree. They're sort of like stumbling 804 00:40:09,720 --> 00:40:11,120 and blind. They're like these these 805 00:40:11,120 --> 00:40:12,720 intellectuals. They're like idiots. 806 00:40:12,720 --> 00:40:15,240 We're going to code it anyway. Next, Next clip. 807 00:40:15,600 --> 00:40:17,000 All right, here. Here's another one. 808 00:40:17,360 --> 00:40:19,280 I talked with the mayor of San Francisco. 809 00:40:19,560 --> 00:40:22,000 Have you had a conversation with Zoran Mamdani or any 810 00:40:22,000 --> 00:40:25,200 observations on his election? You've been able to maintain 811 00:40:25,200 --> 00:40:25,520 this. Great. 812 00:40:25,520 --> 00:40:29,000 We, we, we, we, we, We spoke the morning after he won. 813 00:40:29,080 --> 00:40:32,040 I congratulated him. I, I said, you know, 814 00:40:32,680 --> 00:40:35,200 congratulations. Anything I can do to be helpful, 815 00:40:35,320 --> 00:40:37,640 Great. I, I, I met my wife in New York 816 00:40:37,640 --> 00:40:39,000 City. I worked at the Robin Hood 817 00:40:39,000 --> 00:40:40,440 Foundation. I love New York. 818 00:40:40,440 --> 00:40:43,640 I want New York to succeed. Did you give them any advice? 819 00:40:43,640 --> 00:40:47,440 No, no, no, no one should be asking someone that's been in a 820 00:40:47,440 --> 00:40:52,160 job for 10 months for advice. I, I unfortunately have been, 821 00:40:52,520 --> 00:40:56,920 you know, here in San Francisco, not unfortunately like, but I 822 00:40:56,920 --> 00:40:59,720 haven't been able to travel to New York for almost 2 years now. 823 00:40:59,720 --> 00:41:05,200 So all you all here, you want me focused on San Francisco. 824 00:41:05,200 --> 00:41:08,720 You don't want me talking Sacramento politics or DC or New 825 00:41:08,720 --> 00:41:10,120 York. You want me focused, San 826 00:41:10,120 --> 00:41:12,800 Francisco. Well, Eric, as the San Francisco 827 00:41:12,800 --> 00:41:15,040 native, what do you what do you think about the mayor? 828 00:41:15,040 --> 00:41:18,480 Yeah, yeah. I live in New York. 829 00:41:18,800 --> 00:41:21,560 James is the only true San Francisco native anymore. 830 00:41:21,600 --> 00:41:24,200 I I skip town for the suburbs. You should probably give your 831 00:41:24,200 --> 00:41:27,680 take on the mayor. Well, I I generally like the 832 00:41:27,680 --> 00:41:30,600 mayor a lot, and I think he's been doing a really good job. 833 00:41:31,800 --> 00:41:35,920 I think, yeah, he's in a tough situation with these like 834 00:41:35,920 --> 00:41:41,280 national politics issues. I think he really doesn't want 835 00:41:41,280 --> 00:41:45,040 to deal, doesn't want to become the main story around the Trump 836 00:41:45,040 --> 00:41:47,720 administration and national politics. 837 00:41:48,840 --> 00:41:51,960 He wants to just focus on San Francisco, which I appreciate as 838 00:41:51,960 --> 00:41:53,040 he was very. Politician. 839 00:41:53,040 --> 00:41:55,320 He was like safety, safety, safety. 840 00:41:55,320 --> 00:41:57,040 He just came back to that a billion times. 841 00:41:57,360 --> 00:42:00,640 I, the audience loved him. I mean, politicians are better 842 00:42:00,640 --> 00:42:03,560 speakers than CEOs. I, I think so the people liked 843 00:42:03,560 --> 00:42:04,920 him. People were rooting for him. 844 00:42:04,920 --> 00:42:07,760 He's talking about values, which often companies fail to speak 845 00:42:07,760 --> 00:42:10,280 about. He's not mum Donnie though, like 846 00:42:10,280 --> 00:42:14,000 I, I, yeah, he didn't. He's not like fighting it. 847 00:42:14,000 --> 00:42:16,200 It was also interesting. I kept saying, you know, like 848 00:42:16,240 --> 00:42:17,520 the business community loves you. 849 00:42:17,520 --> 00:42:20,640 Like why is that? Even though like, you know, and 850 00:42:20,640 --> 00:42:22,920 what advice would you give to mom Donnie and blah, blah, blah. 851 00:42:23,240 --> 00:42:26,640 And then he sort of said at one point he was like, well, they 852 00:42:26,640 --> 00:42:29,040 didn't love me at first, which I did think was a funny point 853 00:42:29,040 --> 00:42:31,400 that, you know, it's like they came around to him pretty late, 854 00:42:31,400 --> 00:42:32,640 but. Yeah. 855 00:42:32,720 --> 00:42:35,480 I mean, I will just say, say that, yeah, as someone who has, 856 00:42:35,680 --> 00:42:38,280 you know, lived in the Bay Area for 15 years, in San Francisco 857 00:42:38,280 --> 00:42:42,840 for like a decade, like it is still sort of shocking to hear 858 00:42:42,840 --> 00:42:46,840 the mayor express, like, excitement and appreciation for 859 00:42:46,840 --> 00:42:50,880 the main industry in his city. Like, it's like, it's like, it's 860 00:42:50,880 --> 00:42:53,800 like if the mayor of Los Angeles was up there and like being 861 00:42:53,800 --> 00:42:56,040 like, I, I think this Hollywood thing is pretty good for the 862 00:42:56,040 --> 00:42:57,480 city. And you were like, whoa, no 863 00:42:57,480 --> 00:43:00,280 one's ever said that before. And reality is in San Francisco, 864 00:43:00,280 --> 00:43:04,440 I have not heard a politician express any sort of positive 865 00:43:04,440 --> 00:43:07,720 viewpoint about technology as an industry for 15 years. 866 00:43:07,720 --> 00:43:11,200 And so it is, I think, you know, he has a 73% approval rating or 867 00:43:11,200 --> 00:43:13,280 whatever. I think that the positivity 868 00:43:13,280 --> 00:43:16,240 about what's happening in San Francisco is what really shown 869 00:43:16,240 --> 00:43:18,440 through in the interview to me, including in this Benioff answer 870 00:43:18,440 --> 00:43:20,320 where he was saying, hey, things are getting better. 871 00:43:20,320 --> 00:43:21,440 We still have a lot of work to do. 872 00:43:21,440 --> 00:43:24,000 But like, I believe in the city and I believe we can invest to 873 00:43:24,000 --> 00:43:25,440 make it even better in the future, right? 874 00:43:25,640 --> 00:43:28,200 And maybe Mark was a little off his rocker on calling for, you 875 00:43:28,200 --> 00:43:29,760 know, the. Federal dimension a little, 876 00:43:29,760 --> 00:43:30,440 yeah, there we. Go. 877 00:43:30,440 --> 00:43:34,000 But it's just, yeah, the just the whiff of optimism about the 878 00:43:34,080 --> 00:43:37,720 city and technology is is so unique in the last 15 years of 879 00:43:37,720 --> 00:43:41,200 San Francisco politics. But he has universal approval in 880 00:43:41,200 --> 00:43:44,680 the city because, you know, it got so bad. 881 00:43:44,680 --> 00:43:48,160 And then pretty quickly after he got elected, like there was 882 00:43:48,160 --> 00:43:51,080 noticeable improvement. Like I it's not that he was. 883 00:43:51,520 --> 00:43:53,800 Yeah, he's doing an actually good job. 884 00:43:54,560 --> 00:43:57,200 And there are a lot of obvious things that he could do to 885 00:43:57,200 --> 00:43:59,360 improve quality of life in the city, and he's doing them. 886 00:43:59,360 --> 00:44:01,360 And that, that's the success story. 887 00:44:02,000 --> 00:44:05,120 Will it always feel like AI is this kind of tool? 888 00:44:05,120 --> 00:44:07,600 Agents are useful, used by humans. 889 00:44:07,600 --> 00:44:10,680 Christina, you said, you know, a smart person who knows how to 890 00:44:10,680 --> 00:44:13,200 use AI might replace someone who doesn't know how to use that. 891 00:44:13,200 --> 00:44:18,360 Or will we reach a point where these AI agents are really 892 00:44:18,360 --> 00:44:21,640 approximating full workers in the enterprise? 893 00:44:22,440 --> 00:44:24,000 Oh, I think they're yeah, there's definitely some full 894 00:44:24,000 --> 00:44:27,840 workers, but I'm I'm people will just do other things like a 895 00:44:27,840 --> 00:44:30,280 Vanta contacting example, and we were talking about it earlier is 896 00:44:30,440 --> 00:44:34,560 one part one thing of GRC team does again is like evaluates new 897 00:44:34,640 --> 00:44:36,680 vendors, new software vendors that are coming in. 898 00:44:36,920 --> 00:44:40,680 And today often is someone's job to evaluate the high risk 899 00:44:40,680 --> 00:44:42,080 vendors. Like they can't even do all of 900 00:44:42,080 --> 00:44:44,440 them, but they are just like vendor evaluator. 901 00:44:44,800 --> 00:44:47,880 And I think that is a great thing to give to an agent. 902 00:44:48,120 --> 00:44:50,840 And then that and that agent can go to Erin's point, go and do 903 00:44:50,840 --> 00:44:53,560 all of the vendors, not just a subset of them, because the 904 00:44:53,560 --> 00:44:56,360 agent doesn't take PTO and doesn't get tired and, you know, 905 00:44:56,720 --> 00:45:01,080 works 9/9 more than 997. And then the person becomes like 906 00:45:01,080 --> 00:45:04,720 a vendor risk portfolio manager and thinks, OK, a given all 907 00:45:04,720 --> 00:45:07,200 these, you know, inputs and given what I know about business 908 00:45:07,200 --> 00:45:10,160 context, how do I like make better decisions? 909 00:45:10,440 --> 00:45:13,640 But the person still has a role. It's just not as kind of in some 910 00:45:13,640 --> 00:45:16,280 ways manual and tedious as what the agent is now doing. 911 00:45:16,840 --> 00:45:22,280 I guess my reaction to this is just that what she's describing 912 00:45:22,280 --> 00:45:25,000 is to some degree a job replacement. 913 00:45:25,000 --> 00:45:27,760 I mean, she's saying that this role will no longer exist and 914 00:45:27,760 --> 00:45:31,800 that this person will be doing a different job that she thinks 915 00:45:31,800 --> 00:45:34,840 that person is capable of. But my question is like, is that 916 00:45:34,840 --> 00:45:36,600 true? Like this person will be able 917 00:45:36,600 --> 00:45:40,320 to, you know, become a portfolio of agents manager? 918 00:45:41,360 --> 00:45:43,080 I don't know. I'm, I'm just maybe a like more, 919 00:45:43,080 --> 00:45:46,720 a little more skeptical that it just so, so easily, you know, 920 00:45:47,480 --> 00:45:51,560 transitions into this next era where everyone who used to be a 921 00:45:51,560 --> 00:45:54,840 software engineer can just be an agent manager of software 922 00:45:54,840 --> 00:45:59,600 engineers or everyone who, you know, was a lawyer can be a 923 00:45:59,600 --> 00:46:02,800 manager of agent lawyers. Like, I don't know, it just 924 00:46:02,800 --> 00:46:05,760 feels a little too neat to me that that's how things are going 925 00:46:05,760 --> 00:46:06,640 to evolve. Yeah. 926 00:46:06,960 --> 00:46:10,680 I mean, I think it's I think to to offer the sort of 927 00:46:10,680 --> 00:46:12,640 conventional take on job replacement. 928 00:46:12,960 --> 00:46:16,040 You know, 100 years ago, I think over half of Americans were 929 00:46:16,040 --> 00:46:18,320 farmers, right? And today it's like 2% of 930 00:46:18,320 --> 00:46:21,960 Americans are farmers, right. So we replace like literally 10s 931 00:46:21,960 --> 00:46:23,920 of millions of farmer jobs over that time frame. 932 00:46:24,120 --> 00:46:27,120 Now, I think, you know, the alternate and those farmers 933 00:46:27,120 --> 00:46:29,840 ended up being lots of that we never would have imagined 100 934 00:46:29,840 --> 00:46:31,320 years ago, right? Yeah, we're not all like. 935 00:46:31,320 --> 00:46:33,760 Tractor managers, hopefully. Yeah, we're not all tractor 936 00:46:33,760 --> 00:46:37,800 managers, we're not all combine harvester maintainers, right. 937 00:46:37,920 --> 00:46:40,720 Yeah, Like they're, they're sort of as a layers of abstraction of 938 00:46:40,720 --> 00:46:41,960 new types of jobs. So that's sort of the 939 00:46:42,280 --> 00:46:44,960 conventional economics take. And I think I do basically 940 00:46:44,960 --> 00:46:47,000 believe that. But I think that still means 941 00:46:47,000 --> 00:46:49,400 that in the short run, especially with the pace with 942 00:46:49,400 --> 00:46:53,360 which AI can do things that humans could do, you know, just 943 00:46:53,360 --> 00:46:56,200 a year or two ago, like there there are new jobs since then 944 00:46:56,200 --> 00:46:58,600 that now AI can suddenly do. It does feel like there's going 945 00:46:58,600 --> 00:47:00,640 to be very rapid displacement, right? 946 00:47:00,640 --> 00:47:04,880 Like the the invention of machines for farming did not 947 00:47:04,880 --> 00:47:08,240 like overnight just completely obliterate everything farmers 948 00:47:08,240 --> 00:47:10,640 were doing, which it does feel like AI is like just 949 00:47:10,640 --> 00:47:13,640 obliterating huge chunks of knowledge work like basically 950 00:47:13,640 --> 00:47:15,360 overnight. And so I think that they could 951 00:47:15,360 --> 00:47:18,760 be sort of a shock to the system in the way that maybe 952 00:47:18,760 --> 00:47:22,000 traditional automation is not. And it seems like as long as 953 00:47:22,000 --> 00:47:25,760 Trump's in charge, nobody's stopping this putting the horse 954 00:47:25,760 --> 00:47:27,600 back in the barn. Like, states aren't even gonna 955 00:47:27,600 --> 00:47:29,920 be allowed to make regulations about it. 956 00:47:29,960 --> 00:47:32,560 So it's like. Nobody's putting the horse back 957 00:47:32,560 --> 00:47:37,520 in the barn is perfect like for perfect analogy like to the dawn 958 00:47:37,520 --> 00:47:41,360 of cars, you know, like exactly. Yeah. 959 00:47:41,960 --> 00:47:44,520 We. I think the other thing that was 960 00:47:44,520 --> 00:47:50,720 interesting to me is just that she, she also, you know, sort of 961 00:47:50,720 --> 00:47:53,840 is making the assumption that the job of agent manager won't 962 00:47:53,840 --> 00:47:58,120 be be run by an agent like like how many lawyers you know? 963 00:47:58,120 --> 00:48:00,440 Yeah, I don't know at what point does. 964 00:48:00,440 --> 00:48:03,240 The I mean, progress. People want human to me, people 965 00:48:03,240 --> 00:48:06,240 will want humans for something like I'm, you know, I would love 966 00:48:06,240 --> 00:48:09,000 to have human, you know, caretakers for the elder. 967 00:48:09,000 --> 00:48:10,000 Yeah. You know, there are lots of 968 00:48:10,000 --> 00:48:13,200 important human things to do. I don't necessarily agree with 969 00:48:13,200 --> 00:48:14,920 her. Like you're saying that humans 970 00:48:14,920 --> 00:48:17,120 will just slot into the agent hierarchy. 971 00:48:17,120 --> 00:48:20,320 It's like very possible agents run agents. 972 00:48:20,480 --> 00:48:23,640 I think the big question is just like, well, so much value accrue 973 00:48:23,640 --> 00:48:26,880 to the people who own these agents relative to the average 974 00:48:26,880 --> 00:48:30,320 American worker, then, you know, the wealth inequality will get 975 00:48:30,720 --> 00:48:34,000 get so terribly skewed. I think humans will have value 976 00:48:34,000 --> 00:48:36,480 and therefore, if the economic system is working, there should 977 00:48:36,480 --> 00:48:39,440 be money for them to make. But maybe, you know, people who 978 00:48:39,440 --> 00:48:43,400 control these agents will just be far, far too powerful for any 979 00:48:43,400 --> 00:48:48,560 sense of an egalitarian society. All right, I enjoyed the 980 00:48:48,560 --> 00:48:50,840 conference, I think. Yeah, let's keep it tight. 981 00:48:50,840 --> 00:48:52,800 This was a blast. I really had fun with this one. 982 00:48:53,440 --> 00:48:54,760 Yeah, yeah. Thank you guys so much. 983 00:48:55,680 --> 00:48:58,360 Thank you for tuning into this week's episode of the podcast. 984 00:48:58,360 --> 00:49:00,520 If you're new here, please like and subscribe. 985 00:49:00,520 --> 00:49:03,200 It really helps the channel. We're building a YouTube 986 00:49:03,200 --> 00:49:04,320 channel. I think you can tell we're 987 00:49:04,320 --> 00:49:07,000 investing a lot more in our production and we appreciate 988 00:49:07,000 --> 00:49:11,160 your support. And if you want the data Insider 989 00:49:11,160 --> 00:49:15,640 takes real reporting, go to newcomer.co and subscribe to the 990 00:49:15,640 --> 00:49:18,160 Sub Stack as well. Thanks for following going on.