Which $Billion AI Startup Would You Short?

0:00During the Cerebral Valley AI Summit, we asked a group of founders and investors, which billion dollar AI startup would you short? And their answers caused a bit of a stir.

0:13Joining me today are my Cerebral Valley AI co-hosts and the co-founders of Volley, Max Child, and James Wilsterman. We'll also be diving into some of the insights from our panelists throughout the conference. My interview with Mike Krieger, the chief product officer at Anthropic, former co-founder of Instagram about the problem with sycophancy and foundation models. My wife got her first like, you're completely wrong. And she was like, yes, this is great. And I think we should have more of that. We looked at a clip from the mayor who, though ever the politician, won the adoration of the crowd.

0:46No one should be asking someone that's been in a job for 10 months for advice. And we asked, is Mecca Hitler inevitable? On stage with Jimmy Baugh, one of the co-founders of Elon Musk's XAI. The Mecca Hitler in the room. Yeah, Michael Hitler, our model, had an episode. This is the Newcomer Podcast.

1:14All right, I'm excited. We've had some time to rest since the Cerebral Valley AI Summit. Max has been resting live flat. What? Literally just got home from Dubai, I believe. literally 10 minutes got off a 16 hour long flight over the world's greatest hits like tehran and moscow i i've got my seven week old so back back to the parenting uh minds though it's been a joy james what you're are you well rested or how are you feeling i'm not well rested eric i uh have been solo solo parenting my two-year-old for oh where's your wife where's her in brazil and mexico oh Nice. Yeah. So we've all got our own reasons to be tired.

1:59We're at the exact level of delirious that the viewer should want here because we just had the Cerebral Valley AI Summit, right? We host this twice a year, big AI, you know, 300 person event, top AI founders, investors, real insider thing. We sent out a survey, became a point of media fascination, like sources, business insider wrote about it. I think random Indian media outlets for reasons that we'll explain as we progress. We had 300 attendees at the thing. I want to be clear. Like, you know, some of these things we didn't, everybody did not fill out this anonymous survey. This was not academic research.

2:35This is, you can count the dots for yourself and sort of infer how many people did this. I don't know. What are we saying? Like 30 to 40 people? I just want to be transparent, but it gives you a sliver of where this highly engaged audience thought things were. And part of the fun was that Max and I were on stage sort of reacting to the survey results. Anyway, so we're going to dig into the survey. James, this was really your brainchild. Anything I missed about it or you want to start ticking through the questions? Yeah, I guess one part, one fun part of the game that we played on stage was that you and Max had to guess what the audience would think the answer to these questions might be.

3:13and uh i have to say max kind of ran away with that game on stage he did well i think i won five one four one something like that max did well yeah it was a dumb of them were close let's go through we can talk about yeah you did very well we can react to the reactions myself despite max's dominance i think i had this i i had my ear on the pole still all right should we start with the beginning yeah so the first question was what will open ai's annualized revenue be at the end of 2026 the audience median was 30 billion um and it's a 20 billion 2025 is the expectation right yeah over 20 billion already so this is pretty uh low estimate in my opinion and i said like 40 or what did i say yeah i forget who said 40 and one of us said 40 and one of us said 41 i did me yeah yeah i went higher maybe 41 or something i was surprised i think that Given we're exiting this year at 20 in OpenAI, I thought the effusive glee and bubble talk of the conference would flow through to a growth projection in OpenAI.

4:17In Vidya, what? It just beat earnings. Things are going well. We're round tripping all day. Everybody's making money. Revenue is not the problem. Mere 50 % growth for OpenAI, I think, would be considered an atrocity at this point in the bubble cycle. I'm betting on 40. If this actually happens, I think the bubble is over, baby. What's your bet, Max? was uh yeah i i think 40 is what i yeah i said on the last podcast i think we discussed i mean i think 2x year on year makes it makes sense to me unless the bubble collapses i'm curious where do you guys think the next 20 billion comes from like is it business as usual or do they need to create new products or i don't know copy harvey you know going to legal yeah anything it's like i think you know they should lean into the api business do this sort of general purpose foundation model.

5:03So I don't necessarily think they need to go after an application directly, but if they were really hungry for revenue, you'd think they'd figure out who their best partner is and say, screw it, we're going to cannibalize them to find the revenue we need. Kind of seems like they're gearing up for that. Yeah. Palantir model or, you know, find the biggest pockets of money in the world and, you know, consult on how to adopt AI in those companies. Right. I mean, I think there's a lot of headroom on the consumer subscription business as well. I just think that, you know, they haven't necessarily monetized, you know, a huge percentage of the people who use AI every day and they just keep adding value to that consumer subscription.

5:41And so, you know, if even if you just got a doubling of that, that, you know, that would get you pretty far along the route here. All right. Next question. What will NVIDIA be worth at the end of 2026? On the day we took this survey, it was 4.8 trillion. They just had earnings. I'm not actually sure what it is the day we're recording. pick it up is it's yeah it's pretty cool they had they had pretty good earnings i think but i don't think wall street yeah i checked the stock's like four five four six right now or something 4.35 trillion so down three five so down a little bit yeah yeah um all right that's always they beat earnings and then they still you know the expectations are so insane um i said five i think max what did you say i said six which was dead on the money if i recall yeah yeah yep so So the audience had a median of six trillion.

6:32Very few outliers on the high end there. Five was the next biggest grouping. Also, interesting. Median was just so large. That's really... It's just so many people said six. If it was average, oh, well, average is obviously thrown off by this like 100 trillion is a troll. But forget that one. I think based on the performance of the recent earnings, I think this is actually high for reality. like they crushed the recent earnings as we discussed and the stock barely went up on a multi-day period so uh if they're not going up you know 10 a quarter essentially they're not going to hit this number and i just don't see that if those earnings aren't moving the stock up so this actually feels high for reality to me five is hard because you're sort of like you're chickening it out on saying it's all going to blow up and you know it's like what is this reality where either it's like oh the mania has continued or there's a pullback in some ways my My five trillion feels like sort of a weird sort of sane world.

7:32But who knows? Obviously, if you knew the answer to this, you could be unlimited. You're going to have an unlimited amount of wealth. So nobody knows. But this is what a couple of people think. What year will an independent committee of experts as dictated by the Microsoft OpenAI agreement declare that we have reached AGI? I thought this was a funny question because at once it's such a like silly idea, like, oh, we're going to have AGI. But like there's an actual contractual agreement between Microsoft and OpenAI that there's a committee to resolve this and big business dealings hang on in. So it's a specific question, which is sort of funny.

8:13This was one of my favorites because you guys were trying to guess what the audience will try to guess what Microsoft and OpenAI will decide as AGI. there's a lot of layers of prognostication going on here um so what did you guys say you said much higher i said 2035 yeah i said i said 29 so i was pretty close to the median of 2030 what do you guys think in retrospect like is this this is too early or i just feel like a theme of cerebral valley in the beginning you know we started in 2023 in march 23 23 after chat gbt that was probably the conference we talked most about AGI. And then we talk about it less and less every time.

8:53You know, it's like there was so much enthusiasm when the models came out. And now we're in the sort of like, oh, man, this is exactly like self-driving cars where you feel really close. And then there's a lot of like edge cases to hammer out. And so I just think AGI pessimism has gone way up. And you add to that Andre Karpathy sort of thing. And it's just like, I don't know. I don't think the insider vibes are like AGI tomorrow unless you're talking to daria or something i think the i think the interesting thing here is that there's basically just two buckets of people uh one is 2030 or sooner which are like the accelerationists and then one is like 2045 or never which is like the decelerationists or the pessimists or whatever you want to call maybe maybe the realists yeah yeah yeah whereas you sort of hit this exact plot that basically nobody was which was 10 years from now or whatever but like it's still gonna happen uh Which is an interesting, like, yeah, that you sort of found the middle of this smiling curve where you're either an optimist or you're a pessimist.

9:50And you kind of tried to hit the middle and it didn't quite work. Yeah, that's interesting. We have more optimists at our conference was the end result. That's why Max beat me. He understood that was the psychology of the answer right there. Yeah, exactly. Optimistic conference. There's something slightly interesting comparing it to the self-driving car world, like you said, though, Eric, because, you know, self-driving cars are basically useless until they reach, you know, parody or capability of human drivers. Right, where this is not like that. This is not. This is like, it just happens to have all these other very targeted, more verticalized use cases that are super valuable, but not with the full human replication.

10:29Which is why we're, you know, I was negative about self-driving cars because it was annoying because you need them to actually work. Whereas this, I've been very enthusiastic. So yeah, I agree. That's what's beautiful about text versus safely delivering humans places, which thankfully now Waymo is good at and we can celebrating it. But, you know, 10 years ago or whatever, it was annoying marketing. I like that. Okay, next question. Which venture capital firm's AI portfolio are you most jealous of? I think this was kind of a shocker, right? Neither of you guessed A16Z, which tied with Khosla for the lead here.

11:07Obviously this - We both said thrive, right? Yeah, and then we both - And then we decided to do a tiebreaker. And then we decided to do a tiebreaker. Second pick or something. No, I said Sequoia. You said Sequoia and I said Cosla. And that's why you won this. Yeah, I got the kicker on that one. That was a good pull, Max. How did you decide to pull Cosla? Just OpenAI? I did some chat GPT research beforehand. Cosla was the first venture investor in OpenAI. But that round has gotten significantly diluted. And I'm sure they'll be reporting over time. I don't know how much they've done secondary. It was one of these fantastic investments.

11:41But it feels like they got squashed down by later stage rounds in all these negotiations. I think we can all agree, though, the Andreessen tie for victory with Kostla is pretty bizarre because there's not a lot of really notable successful Andreessen AI investments compared to most of these other firms. I mean, I got a text from somebody when I shit on Andreessen on stage, which is funny. What's wrong with this? Like live on stage? which right after I went off, but it's like, oh, people are paying attention. Yeah, I don't know, Andreessen. I mean, they have, I can't list. They had character. They have like SSI.

12:19I mean, I'm sure they have a ton. Some of them are later, you know, it's like they're in OpenAI. They're in XAI. I just think the lesson of our draft was like the only thing that matters is basically being heavy in Anthropic, XAI or OpenAI. And like, they're not really in any of those, right? I mean, it was my understanding. they're probably big in x ai are they big in x ai uh i think that they're in x but i don't think they're huge yeah and these are all sequoia was pretty big yeah sequoia has money all i'm saying is yeah they to me it feels like the pr of andreessen generally is sort of overwhelming the actual portfolio it's like oh you know i think drive is doing really well they've done these huge bets in open ai like big right yeah pre all the markups so i bet they're doing really well and would like to substantiate that thrive reach out um the uh i mean all the you know they mentioned a bunch of firms like you know it's like a lot gill who we had on stage is obviously great index ventures i mean some of these like who knows some partner you know said them i'm personally jealous of this slide because it would be a great cap table uh for a startup uh just have all these have every single person on this line yeah yeah sounds good and benchmark doesn't make sense to me necessarily for ai portfolio you know but um i mean they have mercore now mercore lang chain i don't know but oh okay as i'm about to preview i do think uh brand like having a big brand you know who gets an answer on a survey somebody who's has large mind share and this This is why surveys are imperfect.

13:58Yeah. What research mechanisms. So we're about to see name recognition is everything here. If you could put money in any private technology companies today, what would they be? So the top 10 by far was the first was Anthropic followed by OpenAI and then Cursor. Those are the top three. Rounding out the top five was Anduril and SpaceX. Hmm. Open evidence is interesting. Then perplexity, replet, stripe, XAI. Max and I both said anthropic, right? Yes. Yes. I feel like the tiebreaker, first of all, I was idiotic on the tiebreaker because I picked a really random company nobody was ever going to pick.

14:40I said fireworks, which was clearly just like if I were going to make the bet, I don't know what I was thinking. Max, you picked cursor. Cursor. Cursor. Yeah. Went with the momentum play. I think we got, we had one pick this tiebreaker thing. I was not pre-negotiated. I didn't come with the time so i didn't do so poorly here like um but you both kind of agree with the audience on anthropic i mean it's an interesting question because anthropic is only you know valued at you know what is it 350 billion compared to open ai's 500 now right they're they're starting to get pretty pricey comparatively speaking and i know their revenue growth has been unbelievable and i I think they're mindshare in Silicon Valley, again, among developers.

15:23And there is this sort of like undercurrent of like they're like the ethical AI company, quote unquote, with the, you know, cool hip branding in the West Village. It is a bit strange to me that they're substantially bigger than OpenAI and the votes here. Because I just think that the hipness of Anthropic is maybe outweighing the fact that. Well, Silicon Valley is more bullish on Anthropic than most of America, clearly. Of course. Yeah. But I just think, yeah, if you were an honest assessment of the valuation compared to the revenue versus OpenAI, I would say like, hey, you might just want to take the momentum plan, bet on OpenAI.

16:04So it was a bit of a, yeah, let's guess what the hip Silicon Valley one is. And you and I both correctly guessed the hip Silicon Valley one was Anthropic. What global company's model will top the LM Arena web development leaderboard at the end of 2026? We had an excellent conversation with some insiders in the data labeling space the night before who said that LM Arena is an incredibly gameable metric. And it's because it's basically voting from users whether or not they liked the response or not. And so it's very susceptible to, some might call it glazing, others might call it sycophancy towards the user.

16:44And their belief was that OpenAI was heavily over-optimizing to glazing their users and therefore was continuing to do well in these types of rankings. Which I thought was an interesting point, which is why I think I chose OpenAI for this. And did well. One, the answers are OpenAI, Anthropic, Google, Gemini, XAI, and then Alibaba. We made sure to say global to try and induce some Chinese answers, but they didn't rank high. I just think this was basically what the rankings were during the day the survey was taken. Nobody was really going out too far on a limb that this would be radically different next year.

17:21But I think that's interesting because my understanding is now Gemini ranks higher than OpenAI a week later. right so i bet you'd i bet you know you'd see a lot more people guessing gemini insiders needed to be more aware yeah yeah man i wish gemini come out before the conference so no well why why do you wish that well i just think google would have leaned in and talking about it it also would have been giving us like a current you know we had plenty to talk about it was one of my favorite events but like you know it's there's a lot to talk about a week later world that happens a week after it's like oh come on yeah all right if you could short a one billion dollar valuation startup which would it be before before we answer this yeah before we answer this question let me say uh having just gotten off the 16 hour plane ride from the united arab emirates this was brought up to me multiple times apropos of nothing it's reached the broader world with venture capitalists and investors of all stripes.

18:24They had no idea I'm associated with the conference. They had no idea that this was something that I was personally like on stage for the reveal of this information. So this not just went viral, this literally traveled around the world faster than, you know, as fast as the speed of light, the answer to this question. And I think other journalists probably made millions of dollars off of this. And Eric maybe didn't. Millions? No. Well, I don't know. How much money do you think we will make off stories? this is a media business. Maybe not. we'd be lucky if they made$10 ,000. I mean, they didn't make any money.

18:56I meant Business Insider. I meant Business Insider. If their order even monetized it to the tune of$20 ,000, I'd be impressed. I meant Business Insider if they have a sub plan, which I believe they do. Sure. Converted hundreds of subs off of ripping off our survey. So congrats to Business Insider. So yes, We're happy to get the coverage, to be clear. Thank you, Ben. Thank you for the coverage. Thank you, Alex. Eric, how does this make you feel It's all about lists and rankings and types of stories you might write. We played it as a fun game on stage. And I think our coverage in the newsletter reflected that it was like a fun game.

19:32And that's what we're talking about here. This was not an academic survey. It was sort of provocative. And it's funny. I mean, in some ways, the media should be a little looser. It's like, oh, some insiders think this thing. But just like when things are turned into like journalese, it feels like official, like Silicon Valley has turned on perplexity. It's like, I don't know. some random people who decided to fill out a survey, you know, that's what they said. And perplexity was on the bull list, you know, way lower. And I do think this means something that it was number one short. I mean, perplexity, why is it number one?

20:02Super highly valued and doesn't have that gateway to the consumer. And people have tried to do browsers forever. You know, it's like people have failed and, you know, Google Chrome and Safari and Explorer dominate. So it's a very hard space. So yeah, there are some investors that are super bullish, but I think most people are like, how do they get distribution? Yeah. The reason it's a short, right, is valuation, I think, to a large degree, right? It's valued at$20 billion, right? And whatever leaks have come out about the revenue from the inside, there's some debates of whether or not they're counting free trials that are a year long as part of revenue, if you read the information story about this kind of stuff.

20:43But anyway, even if you just take it on its face, the revenue that they're leaking or stating, this is like a hundred X revenue multiple, which is sort of ludicrous on any company. I mean, even the wildly overvalued companies we were talking about earlier, like Anthropic and OpenAI are only multiple, you know, 25 X revenue. So this thing is worth, you know, the, the excitement around it from a valuation perspective is roughly four X OpenAI or cursor, right? If you were to just sort of do the math there. And so I think that's the reason it's a short is just that the valuation is just out of control.

21:17And to your point, they don't have their own models. It is a search engine. And obviously Google and OpenAI are trying to own the search space. And now they're getting into the browser thing. And it's not clear that even this, that AI browsers are even a space. Yeah. So most votes, most shorts was perplexity followed by OpenAI. And then tied for third was cursor figure harvey mercore mistral and thinking machines i i answered i think open ai on the belief that oh man popularity and max got this dead on so kudos to max which was a great pick but i i got the number two and then i think we did a second did we do a second i said thinking machines also which was third so we we were on the pulse and that and that was that was before thinking machines got marked up to 50 billion it came in third when it was a 10 billion dollar company.

22:07Now it's 50. So that to me would have been the pick maybe. You know, Cerebral Valley survey went viral with Indian media because perplexity founder is sort of a high profile Indian founder. And I think there's interest in how it goes. So it really, really traveled. I think perplexity hangs on, you know, it could sell to Apple or somebody to save their AI strategy. I think because they have a big distribution problem. And then some people really believe in the founder and some people don't all right any final takes well i was just curious about uh i was curious about cursor because you said that you had a uh oh i just wanted to not talk about until we revealed that it was yeah short of the short max what's your take are you bullish or bearish i mean i don't know like probably bullish given the momentum they're seeing on revenue and revenue growth i think if you just take the brain dead case that lots of revenue is good and lots of revenue growth is good, they have a good business.

23:06I think they're much more likely to exit to someone for near their current valuation than perplexity, for example, where I think that buying it for 20 would just be absolutely insane. But yeah, I mean, ultimately, there's this whole debate with cursor that they're repackaging other people's models and their gross margins are terrible and yada, yada, yada. But ultimately, someone may have to cave and just buy them to own the IDE space, Microsoft being an obvious candidate. So it'll be interesting. And now Google has Antigravity, which is their... Right. That's more their Claude code, I think. No, no.

23:38I downloaded it. Antigravity is a cursor clone in many ways. But I mean, design. Okay, I have it on my computer. It has terminal access. It's trying to come up with stuff. Well, it's designed more specifically. I guess you could say it's Claude code to some degree. It's because it's so designed for a specific model. Because it felt like Claude code because I do stuff in the terminal. I haven't actually known that a cursor. I'm not a coder, so I don't know. but when i it felt like flawed code you shouldn't have to use and uh and terminal that much for anti-gravity like versus cursor i don't know i don't know you wanted me when it you downloaded my website and then i was like build the website and then i was like okay go in the terminal and run it yeah but that's what cursor would do too that's what cursor okay because you're just it's it's like a level above lovable or something right where it's like it's forcing you to actually more yeah my speed is replet even dumber than that i think i need to try replet like i think you'd like repler rep it's like in between i think replet has a little bit more pro user i want the dumbest person least coding you know that's probably i think i think lovable is the yeah is the version why i beef with lovable because they they have yeah try rep pattern you know yeah replet but they say they fix actually what you should try as of you know yesterday is i think gemini in ai studio it's like they've built a lovable kind of thing classic google man yeah jesus christ how many names you don't even know how to find it like that name is insane yeah gemini in gemini inside ai studio inside ai studio oh my god you're not a daily driver of AI Studio.

25:19Specifically the build menu, the build tab within. Oh my God. Yeah. You're not up in Vertex every day, Eric? That's a real name of a Google product, by the way. It's related to AI. Yeah. All right. Let's do some clips. Let's do some clips. For founders and developers building modern data-driven applications, MongoDB's local event series is coming to San Francisco on January 15th, and it's designed to help you focus on innovation, not infrastructure. You'll learn about technologies, tools, and best practices that make it easy to build and scale modern applications without complexity. Plus, attendees will hear directly from experts and innovators who are using MongoDB to power the next wave of AI applications.

26:06mongodb.local, San Francisco, January 15th. Learn more and register at mdb.link forward slash sf-dot-local or click the link in the description. Our first clip, it's me, Eric, interviewing Mike Krieger, the chief product officer of Anthropic, who was the co-founder of Instagram before that. Returning to sort of my core philosophical question, like the sycophancy question, What is your view on that and how much to enable sort of everybody likes to be flattered. It's a reality of human beings versus an effort to be direct. And how do you think about those trade-offs? Yeah, I think there's a wide gulf between true empathy and then sycophancy.

26:49And it's interesting that it materializes not just in, hey, I'm having a conversation with Claude about some coaching or personal goal that I have, but it also does encode as well. When we were testing Sonnet 4.5, one of the things that people got most excited about was when Claude was like, this idea is bad. Not that you should feel bad about it, but this idea is not a good direction. I can go and implement it if you really want to, but I would suggest that we try this other thing instead. So there is something like that pushback is not just valuable in a personal relationship with AI sense. It's actually like how you get good work out of the models.

27:24but for a long time our models have been I think appropriately empathetic if you're going through a hard time I was dealing with the death of a pet and I talked to Claude a lot about these different things and it always started in terms of like hey that sounds hard, sorry to hear but then I'm going to give you a factual answer I'm going to go research these pieces but still with a place of empathy as well and so I think when we look at it internally and we're just evaluating it ourselves it's again not that empathy It's not even like the likability of the model. It is, do you, like, does it show up in the way that you'd want a good conversationalist to show up and then continue on its AI journey around what it is going to do with you as well?

28:03But I think it spans everything from that, like, initial response all the way to, like, how it evaluates an idea as well, you know? Claude, especially previous versions, were kind of, like, known for being like, you're absolutely right when you correct it. And my wife got her first like, you're completely wrong. And she was like, yes, this is great. And I think we should have more of that kind of contact. Less San Francisco. Yeah, less San Francisco. A little more direct New York. You know, I'd set up this big theme, you know, that he'd been at a social media company, Instagram. Now he was an AI company.

28:36Social media companies were built on optimizing for user engagement through machine learning. And AI companies at least started off chasing the truth and chasing these leaderboards. But like sycophancy is a, you know, it shows that these models and OpenAI is famous for the sycophancy issue and people's attachment to GPT-4, which was the one that really sucked up to everybody and people didn't want to see it go away. You know, clearly these model companies have to think about how much to pander to the egos of their users. What do you guys think? I mean, it's interesting because it does sort of spiritually align with the fact that Anthropic has almost no consumer adoption compared to OpenAI.