Are LLMs Hitting a Wall? VC Shifts Focus

[00:00] The antitrust environment is just so harsh. [00:03] that... [00:05] You know, I think it'll keep going. [00:07] I almost wonder, will investors... [00:10] put in terms to try and, [00:12] Make it difficult. Like in some – it seems like with inflection and adept, you know, investors are basically – [00:18] getting their money back and maybe, [00:21] Given how those companies were doing, getting your money back is a pretty good deal. [00:26] Maybe investors are happy about it, but... [00:28] It's definitely... [00:30] sort of a dangerous situation [00:33] for startup world. [00:43] Welcome to Sorcery. I'm Molly O'Shea, the founder of Sorcery, a weekly top VC deals newsletter hiving the top GPs, founders, and deal announcements leading the next innovation cycle. Today, we have Eric Newcomer, founder of Newcomer, a deeply reported newsletter on the inner workings of the startup and venture capital industry. We discuss the current state of the AI markets. [01:05] the outlook for 2024 and beyond, as well as covering all of his Cerebral Valley Summits to date. [01:12] I hope you enjoy. [01:13] Thank you. [01:14] Hi, Eric. It's great to have you on. Hey, Molly. Thanks for having me. Well, also thank you to you, because I just got back from the New York's Re-World Valley Summit. I think this is the third one you've hosted, and this is the second one I've gone to. And... [01:28] Not too bad. You keep on killing it. The other one, you went to the November one or the March one before that? I went to the November one. Right.

[01:35] Yeah. [01:35] Yeah, we've done-- [01:37] Three in like a year and a half. You know, we did one in... [01:40] March, 2023, super early, open office. [01:44] Then by November, we were like grownups and we had it, you know, in a theater at SF Jazz. [01:50] Yeah, we just brought it to March in June. Not bad. So what were the biggest takeaways from the New York one? Obviously, it's been a complete evolution since... Yeah. [02:00] The first one. [02:02] Yeah, I mean, I feel like we really got to watch some of the themes emerge. [02:09] At this one... [02:12] I'm not going to say disillusionment. [02:14] disillusionment of foundation models, but I'd say there's acceptance that foundation models are plateauing. [02:21] And that, [02:22] People want to see the value accrue. [02:25] at the application layer. So, [02:27] You know, I think a sub point of that is that definitely in March and November last year, OpenAI was like the elephant in the room. You know, some ways the conference represents all the people trying to like find an answer to OpenAI, like all the startups. [02:42] gunning for it and the investors looking for the next one. And so in March and November, it really felt like [02:48] OpenAI loomed large, and certainly I was a voice for, like, is... [02:52] OpenAI just gonna eat your lunch and everything you wanna do. And like being a chat GPD rapper, [02:58] was sort of the ultimate insult last year. I think this year there's much more excitement about, find a product, you know, the models are pretty smart as they are, you know, Pack Radiant Sequoia, I think,

[03:12] articulated a lot of the big themes. And I think he really presented this as sort of like, [03:17] a harvest moment where people should make, [03:20] the best of already pretty intelligent, [03:23] AI. So yeah, I think the core theme was sort of build applications out of, [03:29] The foundation models exist today and don't wait for open AI to pull sort of another – [03:35] rabbit out of a hat. What was the main impetus behind all of that? [03:39] Like why the shift to applications? I think one just sort of the belief that the technology is there. I think literally like yesterday or today Google put out a paper talking about how they want to have – [03:50] You need to really build like frameworks around large language models when you're judging them because they're quite capable if you sort of like put them in the correct settings. So, yeah, I think one is the belief that foundation models exist. [04:02] are useful and can be put to work and so real [04:05] Energy needs to be spent. [04:07] thinking about the applications. [04:11] You know, two, I just – I mean, I just think there's a little bit of backwards thinking where – [04:16] I like thinking from if AI is going to be huge and create the next Facebook page, [04:22] And foundation models are really competitive and sort of, [04:27] old models seem totally outdated as somebody comes up with a smarter one. And as Facebook's llama is like making it. So, uh, foundation models are a commodity, uh, [04:37] then therefore all this money is going to need to be made probably in applications and not at the sort of core foundation models or the next hyperscaler.

[04:51] theme. I mean, we saw, I think, both Pat Grady and [04:54] OLLI, the data dog CEO, sort of compared foundation models to MongoDB and database companies. So like good, but not great. [05:03] businesses and [05:05] And so, yeah, I think there's a lot of hope that somebody is going to. [05:07] sort of invent [05:09] the unexpected at the application layer. I think Pat also said it was interesting to see OpenAI go from an enterprise to a consumer product and ChatGPT just take off. [05:24] Right. Yeah, I think he was saying that, like, his belief in the Sequoia belief behind OpenAI is rooted in... [05:31] them having chat GPT and having this direct customer more than, [05:36] sort of the foundation model and providing it to sort of enterprises or being, [05:41] sort of the smartest person [05:43] model in the room forever. [05:45] So definitely a large shift from last year when it was like, you're not building your own model. What are you, crazy? And then now it's like, okay, let's utilize these tools. Everybody was... [05:57] Building their own model. And there was also more of a question mark around open source and whether open source could catch up. And then I think so this year it's like, yeah, I mean, Lama especially. And like, you know, we had Databricks there. And so there was certainly, yeah, open source is super relevant. No worries. Whereas there was a total open question, I think, at the first two, how relevant. [06:20] open source would be.

[06:22] Yeah. [06:23] Were there any other like buckets of themes that you saw throughout the talks of this recent summit? [06:31] I mean, we had... [06:33] you know, the sort of creativity theme, you know, I surveyed on the first, you [06:38] panel with three VCs. [06:41] What? [06:41] companies they were most bullish about outside of their portfolio. And I think two of them said, [06:47] Mikey Schulman's company, Suno, who is speaking later. And we did not tee that up. But as a conference organizer, nothing you want more than the people in the first panel saying, oh, the most important company is coming on stage later. So I do think [07:01] Suno, you know, really feels amazing. You're creating songs. It feels like... [07:06] something new. It's like, I mean, Mikey was sort of making the pitch [07:11] on stage that... [07:14] you know, [07:15] you're going to get much more sort of, [07:17] consumption of music, you know, you're going to level up sort of the Spotify type experience by allowing people to really [07:25] make it more of a video game and actually build songs with, [07:28] themselves. So I think [07:30] That was super interesting in, like, you know, creativity overall and sort of where the value accrues. And, you know, the generative piece of generative AI is going to shape all these creative professions. So I'd say that was a big bucket. Yeah. [07:50] And then, I mean, I think there's always –

[07:53] and still sort of, [07:56] Nvidia and like how, how much sort of like the metal, uh, [08:00] sort of decides everything here and like the scramble to get, [08:05] GPUs is super important still. [08:09] on people's minds. I mean, we had like Lambda and Modal trying to optimize people's use of these GPUs. [08:19] So, yeah, I mean, there's so much going on, but I would say – [08:24] applications sort of [08:27] figuring out creativity and sort of how it, [08:29] how these companies sort of coexist and, [08:32] with creatives today and then... [08:34] Uh... [08:35] Yeah, I don't know, making it all work on the metal. [08:37] Do you anything you're you think I'm missing or? [08:40] Was there a theme that you sort of took away? [08:43] I mean, I think there's always the debate of open source. And I think maybe this was even crazier at the last summit, like the one in November, because we had Vinod talking and then we had Reid Hoffman talking. And they were all kind of debating the open source of AI and whether or not these models should be accessible. [09:05] And that was a little bit more muted at this one, which was interesting. [09:13] having even more [09:15] pragmatic, [09:16] Like I feel generally, you know, it's an event for startup founders and investors. [09:22] Primarily so there isn't as much soul searching of like should we build this or whatever you know it's it's very like practical I think people aren't worried about AGI particularly because it's like I don't know there are a lot of small problems to solve before we get there but yeah I agree.

[09:38] In the November event, we had that sort of contrast between Reid Hoffman and Vinod Khosla and then Mustafa. [09:45] Um, [09:46] who's now at Microsoft, [09:48] uh, [09:49] You know, Vinod's position... [09:52] is anti-open source because of the national security threat. [09:56] He's a big investor in OpenAI, the first venture investor. [10:01] I don't know. I don't think his position represents a lot of people. [10:05] particularly where [10:08] to be sort of totally... [10:11] closed source, and it doesn't seem... [10:14] likely to win out to me. It's just like [10:17] It's kind of... [10:18] Interesting to see the undercurrent of that, like reflecting back over time, even in the [10:24] one and a half year timeframe of like, [10:27] how much hype there was and anxiety and like pure passion debates. And now to your point, we are in like the pragmatic cycle of, okay, now we're going to the application layer and we know how to utilize these tools. And we're going to start to really enhance our businesses with it. So like, let's actually like, [10:45] use our money and our resources in like the most efficient way possible and make sure that people are trained, they're up to date, that we're fully like emphasizing everything that's there. [10:56] Well, I think whenever there's a... [10:58] big leapfrog in sort of the intelligence. People are like, oh my God, like what's going to happen? You extrapolate out a ton of growth. [11:05] And then now that we're in this period where it hasn't been sort of as insane in terms of like –

[11:10] technological improvements. I think this sort of perceived, you know, like exponential curve that we're riding has sort of flattened a little bit. So people are more interested in sort of practical things. [11:23] today. I have one more theme I sort of forgot about, but I do think, especially because it was in New York, [11:30] and you have a lot of sort of established businesses, [11:33] I mean, we had... [11:34] the CIO, Marco Argenti of Goldman Sachs speak, and then... [11:39] another very established business, but technology business. We had, uh, [11:43] Sridhar, the CEO of Snowflake, close out the day. [11:47] And so I think – and, you know, we had the CEOs of Ramp and Intercom. I think we had a lot of these businesses that – [11:53] We're not... [11:54] born AI companies, but really are trying to... [11:59] sort of figure out [12:00] AI in their businesses and see it [12:04] is really important for [12:07] You know, their story... [12:08] going forward. And so I think, you know, sort of real established businesses trying to get [12:14] uh, [12:15] their hands around how they're going to do AI was another big theme. I would say that... [12:20] That's a really good point. And [12:23] I'd say even in November, right, I feel like most of it was marketing, but now there's tangible outcomes. Like they do have products that are, let's say, like integrated with AI or infused with AI and are really, really running more intelligently. Yeah. [12:37] Again, AI has been around for forever, but like, [12:40] this is different. This is, like, completely different because we can actually, like, automate our workflows and our processes. Right. But, like, in fact...

[12:51] We put own an intercom on stage because, you know, I do think customer support is an area where we're actually going to see. [13:01] you know, chatbots talking to people. And so I thought that was an interesting experience. [13:05] company [13:06] you know, where you-- [13:08] You know, you really needed to bet the whole company on... [13:11] on AI. His point on that the whole workforce is going to be eradicated was quite [13:18] Funny and spicy. Yeah, I don't know. Yeah, he struck both extremes to me in that it's like, oh, we're all going to be ruined, but also it's like, you know, very flawed at the moment. So, yeah, maybe in a long arc of history we'll all be replaced, but... [13:37] I don't know. It feels very far away to me. [13:40] Hey, we'll get right back to the conversation after a word from our sponsor. [13:45] Sorcery is brought to you by Archer. I'm genuinely amazed at what Archer has been able to accomplish. Archer's goal is to transform urban travel, replacing 60 to 90 minute car commutes with estimated 10 to 20 minute electric air taxi flights. They are safe, sustainable, low noise, and cost competitive with ground transportation. Archer's Midnight is a piloted four passenger aircraft designed to perform rapid back-to-back flights with minimal charge time between flights. [14:15] for passengers by providing safe and efficient access to people, places, and events across the communities they live. Visit Archer.com.

[14:24] You're certainly aware of that, which was nice. But yeah, I would say like in November, you had the Salesforce CEO of AI. [14:34] attend and you had writer.ai there and all these companies that were really like [14:40] going forward, they were putting their [14:42] you know, first foot out and trying this new way to, [14:46] market their businesses and [14:48] it kind of demonstrated [14:51] What? [14:52] Taking that step early on really does have an impact on your business. And then you do become known as a market leader, even if it's early on in marketing. But, like, you do have to put effort behind the actual business. [15:05] you know, [15:06] I don't know, let's say like propaganda and stuff, but I don't know if you've, [15:10] You're reporting on this every day, every week. So have you started to see real... [15:17] kind of outcomes with [15:20] what's been marketed over the last year and who's actually leading, like who has, uh, [15:25] you know, like a real pulse on creating capture of market. Yeah. [15:31] Well, I mean, I think one thing I'd say from the event is it was interesting with Marco at Goldman Sachs. [15:37] that, you know, they just launched like this platform for foundation models so that they can use the best one for every case. You know, we're going to have small models here, big models there. So, one, I mean, it's interesting that, you know, they're clearly not letting any one company sort of have a stranglehold with them. It's like the whole sort of poly cloud thing.

[15:58] with AWS, Microsoft. So I thought that was interesting. [16:04] The other point was that even though Goldman... [16:07] is so clearly leaning super hard into AI and, [16:10] So really saying, okay, we're... [16:12] We're sophisticated about this. Marco did say that they hadn't really made any better trades because of... [16:19] LLMs and current technology like they were trying and that maybe it was close to parity, but it didn't sound there There wasn't sort of the bravado of like [16:28] foundation models are leading us to whole new [16:31] investment strategies. [16:33] So that felt a ways away. I mean, I do think, [16:37] even in the Goldman case, like many other people, [16:40] It seems like the use case people are most bullish about is, [16:43] like code assistants and coders. [16:47] getting sort of, you know, support from [16:50] from these autopilot and assistant. [16:54] programs and there's certainly a lot of competition there. So I definitely took away overall the [17:00] just from talking to everybody. [17:03] I'm most bullish on coding as an actual application. [17:08] not necessarily most bullish on, [17:11] sort of the market size of that, but just as that being something that's, like, useful to people in the near term. It's interesting also to the point on coding, like – [17:21] It's... [17:22] easier because it's like, you know, [17:23] zeros and ones, but [17:26] when I was doing some research on synthetic data, you know, I was listening to every podcast imaginable that people have used or tried to work with synthetic data. And I've talked to people about it because I,

[17:38] There is a cap to the data that we have access to, and there's lots of unstructured data out there that's not good, but – [17:45] in terms of the use of actual synthetic data and code and code training with these models, it's like really like, that's where I found, um, [17:53] was the best use case and [17:55] I believe it was particularly with a podcast with the poolside AI founder. And he was he was talking about this. Look, I don't know if it's like still accurate because it's been a couple of months and things move so fast. Right. Yeah. I've started to see that catch, you know, like the most amount of traction, at least like on the investment and like product side. So I guess like. [18:19] in terms of the state of the tech markets, [18:23] How do you see this evolving in the next, you know, [18:27] Quarter we're in the nice trough of summer and so some people are kind of taking their foot off the gas, but some people are getting ready for the fall. [18:36] I think – I mean, Q2 is pretty active. [18:39] in AI. I mean, there were big deals. I think XAI, Elon's AI company probably drove [18:46] Some of that, but I think we're going to find... [18:48] Q2 to be super... [18:51] Super active. [18:53] Um... [18:56] Yeah, I mean, I think the numbers are always weird because the foundation model investment dollars – [19:02] you know, [19:03] We're so large and if that slows and isn't supported as much by hyperscalers and if

[19:09] You know, they already have the money they need. [19:12] then it might look smaller sort of in the next couple of months, even if I do think there's going to be super active investment. [19:19] in applications and infrastructure and, [19:21] in AI. [19:22] So I think – [19:24] Things are going strong, but there's disillusionment from investors in – [19:29] foundation models. [19:32] A little bit lagged, but... [19:33] I think with [19:36] in this particular summit, it was... [19:38] It's really fun because you had these breakout groups. And so we were able to, you know, break out into groups of like 10, 12 people. And there was a moderator, someone who came in with questions and themes and that kind of thing. And I joined Grace Isford from... [19:54] Lux, she's a partner at Lux, and we talked about AI agents, and she really wanted to ask, like, okay, what is the largest scale AI agent that you've seen put to work? Yeah, I didn't get to see all of them, so I'm actually super interested to hear how the conversation went. You have the videos, come on. Well, we didn't record all the sessions, but yeah, I think hers, we did a little video about it, yeah. [20:19] Yeah, it was good. I could feel the camera on me, like, half the time. Uh, but, uh... [20:26] Not because I'm interesting, just because there is a camera there. Yeah, yeah, yeah. [20:30] uh, [20:31] Okay, so yeah, but that conversation was cool. It was a mix of founders, it was a mix of investors. Turned out a little bit inconclusive. We're still too early. But there were use cases within...

[20:43] service chat bots and that kind of thing, but no one has really seen a mass scale AI agent put to work yet. There were some really interesting companies in the mix and we got to hear from things. [20:56] But... [20:57] like, it was, again, like, we're in, like, one of these, like, hype... [21:02] and then delay because everyone's working kind of modes and the data is being trained. People are creating new models or making it more efficient. But to your point of like, I'd say like the largest splashes, um, [21:18] in the space of really just [21:20] kind of been funding announcements so far. Like, we haven't really seen... [21:24] We've seen stuff in terms of incremental [21:27] chats and some really beautiful creation products, but within the enterprise, I feel like it's still pretty early. I mean, tech is obviously ruled by hype cycles and boom and bust cycles. [21:39] The nightmare to me would be that this is like crypto where you see tons of infrastructure investment. [21:46] And then the applications sort of don't emerge. And then you're like, oh, why do we build out all this stuff for a sector that, you know, you don't need all these sophisticated sort of second order NFT companies if NFTs themselves aren't. [22:01] going to hold on. I mean, I'm, you know, I didn't build a crypto conference. I built an AI conference. I'm much more bullish about large language models. I think [22:08] Part of what got everybody excited in the first place is you can – [22:11] See chat GPT. It is.

[22:14] Pretty amazing. You can see some of the videos that get created. You can, [22:18] Play with Suno. Like, I still think they need to... [22:21] level up some, but... [22:23] you can see sort of the promise much more than it feels like a real – [22:27] technological change. I think, you know, a big challenge for all these companies is just [22:33] it's not mobile, right? There isn't this new distribution mechanism that all of a sudden you're in front of all these new customers. It's like the same people are using the same devices and getting their technology, right? [22:45] in the same places, but the hope is that the tech is so much better that somehow you can create a new company. But it certainly gives Google and Microsoft a lot of power where it's like, oh, you're still going to search? Well, we'll just improve search a little bit with this LLM stuff, and then maybe there won't be sort of a relevant startup. Or even Facebook, who knows sort of clearly investing a lot in the actual technology industry. [23:12] And then, you know, Apple's striking these partnerships where they're clearly at the driver's seat. So I think, you know, we still live in this world where – [23:21] The big tech companies have... [23:23] such huge advantages of distribution that the question will be like, is, are the startups, you know, [23:31] innovative enough and new enough. [23:33] that they can overcome sort of that, [23:35] huge gap in distribution. [23:39] Or will they get acquihired by the big recovery team? Yeah, I know. Yeah, literally. The inflection in the debt deals are definitely on my mind. It's sort of the idea that –

[23:51] You know, yeah, super valuable companies are getting... [23:55] Their exec teams and everybody are getting hired away. [23:58] Do you have any predictions on that? Do you think that's going to occur more or less? Yes. [24:03] Yeah, I mean, I think the antitrust environment is just so harsh. [24:08] that... [24:10] I think it'll keep going. [24:12] I almost wonder, will investors... [24:15] put in terms to try and [24:17] Make it difficult. Like in some – it seems like with inflection and adept, you know, investors are basically – [24:23] getting their money back and maybe, [24:26] given how those companies were doing, getting your money back is a pretty good deal. And so, [24:31] Maybe investors are happy about it, but... [24:33] It's definitely... [24:35] sort of a dangerous situation where, [24:38] for startup world where [24:40] executives can go get like [24:41] sweet jobs at big tech companies and, [24:44] get a good salary and sort of [24:48] forget about this startup that they just promised everybody in the world, you know, all their investors was going to change the world. Um, [24:55] So it's definitely... [24:57] sort of not how the model is supposed to work, but [25:00] It is... [25:02] I think, an outcome of a brutal antitrust review world. [25:08] Are there any specific applications or kind of like areas that you – [25:14] and your team, a newcomer, are squared on that you're thinking, [25:19] you want to place your bets on for the next couple months.

[25:22] quarters, like really just kind of unpack and see what evolves. It could be like hard tech. Well, we did a big dive into robotics. I think there's a lot of energy there. [25:32] in robotics right now, I think. [25:35] I mean, [25:37] I think one way investors are thinking about this is [25:41] We need to build companies that are different enough from open AI. So we need sort of specialization. We'd like... [25:48] to have some sort of model... [25:50] And so I think robotics is an area... [25:53] Whereas physical intelligence, maybe you can [25:56] Escape. [25:57] even building the robots yourself, but you can build models and other technology, [26:02] to operate [26:04] robots. So I think [26:05] Robotics is an area... [26:07] with lots of investment. [26:09] I mean, we're working on a piece that maybe will be out. [26:12] By the time this runs on [26:15] sort of life sciences bio, I think, you know, this sort of drug discovery world, [26:20] is definitely... [26:21] Really interesting. [26:23] I mean, I'm bullish in stuff like – [26:26] Law and Harvey. We saw... [26:31] I'm blanking on the company's name, but Thomson Reuters did a big acquisition already in the legal realm. So I think... [26:41] I think this was something that came up at the event. It's certainly not an original idea, but... [26:45] of mine, but [26:47] Maybe it was Pat [26:49] Yeah, I think Pat Grady said... [26:51] that he thinks AI is going to replace AI,

[26:53] or supplement service professionals. So it's like lawyers and consultants in areas where you would pay a lot of money [27:03] for services, now you're going to have [27:05] language models come in and either [27:07] make those workers much more efficient, [27:10] or [27:11] replace them all together and [27:13] I think that makes a lot of sense. So bullish on areas where AI can... [27:18] really cut down the cost of stuff like law. [27:22] consulting, [27:24] banking. I'm sure a lot of people would love to have their lawyer fees decreased. [27:29] - Right, I mean, even if you can sort of [27:32] you know, you whip up documents and then have them reviewed. I mean, [27:36] So much of the law seems to be... [27:38] reinventing the wheel over and over again. Certainly the law firms have... [27:43] you know, all their old documents. And so... [27:46] It becomes easier for their customers to say, hey, [27:50] we can do most of this ourselves. [27:52] Certainly that's a threat. [27:55] So you mentioned Harvey, you've mentioned Suno. Are there any other names of companies that we should keep an eye out for? [28:03] Oh, this is a tough one. You know, I'm sort of in some ways a lagging indicator following what's buzzy. I mentioned... [28:10] Physical intelligence. I mean, we had, you know, [28:13] Runways, [28:14] Continues to be super interesting. [28:17] Um... [28:19] I'm pretty sure you guys at the conference showed [28:22] their new model before anybody else saw it.

[28:25] I wasn't even sure it was that before. Yeah, did we break it? We should have played that up. You kind of broke it. I know it got released into the public like two days or whatever. Yeah, yeah, yeah. Nice. Yeah, great. Love it. [28:40] Um, [28:42] Yeah, I mean, I think the video creation companies generally – [28:46] are really interesting. Yeah, I don't have a long list. [28:51] At the moment. OK, well, if you think of any, we will add it to the notes because you are the Oracle of technology news. Don't forget that. I talk to smart people and steal their ideas professionally. You know, I'm not I don't see you. I'm not making investments. You know, I'm trying to. [29:11] sort of track where the best investors are looking. So in some ways I'm going to say, oh, well, we're the hot. [29:18] seed rounds. I mean, I am most excited for like [29:23] cool consumer use cases and it, you know, [29:29] Still to come, I think, clearly. [29:31] I mean, clearly you pull out ridiculous lists of individuals to come and speak at every one of your summits, and hopefully I can continue to come to them. [29:41] Because every time I walk away and I'm like, how did I just get access front row to this conversation? But – [29:49] So given that, like every single one was very well curated of having, you know, [29:54] the voice of reason or someone who is like super influential in a certain category. Are there any particular individuals that, you know, you haven't had at the summit that you would want to have on the summit? Oh, you know, Sam Altman made an appearance at our speaker dinner in November, but I have not had him on stage. That would certainly be fun.

[30:13] I wrote a... [30:14] somewhat skeptical, fairly skeptical, Sam Altman essay column during his ouster. So I don't know if I'm necessarily their friend number one. I sort of want Ahmad at Stability to come back. I mean, obviously, you know, Stability has been in crisis lately. [30:37] But Cerebral Valley 1, [30:40] He was certainly the interview that everybody was at their edge of the [30:45] edge of their seats. You know, I was [30:47] It was a big question mark as to him and his background, and we certainly had sort of a – [30:52] feisty, lively conversation and coming out of it, [30:56] I talked to various founders off the record. I was like, what do you think of him? And some were like, he's a genius. He's going to change the world. And others were like, he's a huckster. But it would still be fun to hear from him sort of after the whole thing [31:12] experience. Um, [31:14] Thank you. [31:15] You know, Dario at Anthropic, you know, is certainly – [31:19] top of my list. And then, I mean, we liked sort of the big strategic players. So Jensen, I think getting into the [31:30] You know, my network is super deep in startups and VC, so I feel like we've, you know, we had – [31:36] Ali at Databricks speak at both, literally our first one in March when we were like nobody. And then in November, and that's been...

[31:45] key to the brand, but, you know, I'd written a bunch about Databricks. I'd scooped a bunch of their rounds, so I thought, I think Ali was sort of, like, intrigued by the whole thing. So, anyway, my point is just that, you know, I feel like we've done really well and have [31:57] good relationships with private companies. And so I'm trying to crack, you know, the public companies certainly. And Snowflake, I think was a big move there. We had someone from, [32:08] a senior researcher from Meta. [32:10] you know, Datadog CEO, but I certainly want to hear a little bit more about, [32:16] from the juggernauts, even as, you know, we keep our focus on... [32:21] people building [32:22] What's next? [32:23] And it's hard to get to those people because then their media training really comes through. And they might not share everything. Right. I mean, we certainly want – I thought Sridhar – I mean, you – correct me if you disagree. I think he was pretty open for like a public company CEO. I don't know. I thought it was like a fun conversation. I was a little worried like, oh, last conversation of the day. Is he going to be super guarded? But I thought it was fairly – [32:53] forthcoming. Tell me if you think that would be. I agree. No, I agree. [32:56] Yeah, yeah, I think he was great. He actually... [32:59] like shared a lot. [33:00] I poked him about Databricks a little bit. Yeah, exactly. I asked him about the CEO switch, you know, and then I did ask him about AGI at the end. So we had a pretty lively conversation. [33:13] conversation he I think he dissed Databricks a little bit yeah

[33:18] Okay, so maybe my public CEO generalization is not entirely accurate because you just disproved. No, you're right. You're right. But I'm being defensive. I'm saying, no, there's a chance. But yeah, I think a key thing, honestly... [33:33] in sort of conference organizing is just, [33:36] To sell the conference, obviously, you want these great names. But then on the day of the conference, [33:41] People actually want a great conversation, and I think the big names can be overrated. So there's always this... [33:48] give and take of... [33:50] Oh my God, you have so-and-so versus... [33:53] You know, yeah, we should talk about it. I thought the guy at Databricks, what's his name? Jonathan Frankel. Yeah, yeah. I'd never met him. Like, credit to John Turro at Madrona who put together that panel. [34:08] got Jonathan Frankel there. He was like a firecracker. He was like [34:12] the spiciest guy of the day and not like one of the headline names we are, you know, selling the event around, but I thought he was amazing. So, so certainly you want, you know, [34:22] you know, people who are going to sort of speak their mind. [34:25] Agreed on that. He was my favorite conversation because you get these like three very technical individuals talking about open source AI and he is drilling down on everyone. And it was so unexpected. He was very transparent that he thought he's like, I'm a Databricks shareholder and I'm doing this for Databricks. It's not some like... [34:46] benevolent sort of act. He sort of went against the,

[34:51] the open source sort of, I don't know, religion and said, oh, maybe there are some things we shouldn't release because it's sort of dangerous and open source. He wasn't calling for regulation, but some. [35:02] sort of self-policing. [35:03] he like [35:04] Totally pushed back on the idea that foundation models would become these like $10 billion projects. [35:11] And therefore, you know, the implication of John's question was, if they're $10 billion projects, how can open source stay competitive? Are you going to invest that much? [35:20] But I think, you know, the panelists were skeptical of – [35:24] of the premise and pushback. I also thought Jonathan sort of, you know, put some heat on llama on stage, just about what are their plans. And, you know, certainly, you know, [35:34] We didn't get a complete answer. [35:37] I don't think we know whether the next llama will be open source or what Facebook's – [35:42] plan [35:43] Here is, but I'd love to have Mark Zuckerberg come and talk better. Chris Cox are another one of their leaders. Because, I mean, I do think... [35:53] Lama is increasingly like... [35:55] an elephant in the room. And certainly all the companies on the application layer who are building anything but a foundation model are thrilled that [36:04] uh, [36:05] Meta's doing everything they can to undercut [36:09] the open AI pricing and, you know, the cost of using foundation models. Hopefully we can get you Mark Zuckerberg. Yeah, yeah, take any help I can get. [36:20] Okay.

[36:21] Um, so I guess as you plan out the next one, is the next one going to be in November? Yeah, it's, it's going to be in November. I haven't announced the date yet, but it. [36:31] I mean, basically, we do it. [36:33] The week before Thanksgiving in November in San Francisco. [36:37] That was a really good one. I will definitely try to make it. [36:41] Yeah, we'd love to have you. I like to have... [36:45] I mean, we want new people, but I like the idea that they're like OGs and like people who, you know, were there at the beginning. And like I do want to make more of a deal at the fourth one of people who have been. [36:56] You haven't been to all of them, but there are people who've been to all of them. I will collect a little piece of merch from every single summit. You have the hat. It's like people collect their Taylor Swift bands. I know. I've got the hat. Great. [37:12] I went to somebody's office who had like our last hat, you know, out on their desk. That's honestly when you've made it, when you're like a status symbol. Yeah, I was thrilled. Yeah. [37:23] It's good. Yeah, it's good. It's good content. They're very intimate settings. Like the speakers are actually like walking around and people are talking with them afterwards. Yeah, it's a very high quality summit. I've spoken about it many times in sorcery. So I will continue to emphasize that. I appreciate it. [37:41] So I guess to switch gears now that we're kind of wrapping up, [37:45] you have reported over the last month, you know, a little bit more of like an undercurrent of this like VC downturn. I'm really curious. I think Madeline might've written the piece. But so like, where are we at in terms of like the fund,

[38:02] Cycles. Emerging managers are kind of still trying to raise their funds. Right. It's all really... [38:10] Well, I think we were in a brutal downturn outside of AI. Like in some ways, AI gave us something else to talk about while the rest of... [38:18] startup world was destroyed, you know, and particularly... [38:22] software as a service, which had been the bread and butter of all tech. And we have all these unicorns. [38:29] that raised in 2021 that [38:32] are in big trouble and who knows what's going to happen to them. So I think there's been this bad downturn that, [38:39] AI somewhat helped [38:41] cover over but [38:42] Not totally. [38:44] I think the other... [38:45] big theme is that LPs have flocked to safety or like to these sort of mega funds with, you know, Andreessen raising a ton of money, General Catalyst, Lightspeed, [38:57] Sequoia, you know, and NEA, you know, they're these big, big funds. And I think it's really been hard. [39:04] for emerging managers. And I think especially if you raised a fund, [39:08] You know, if your first fund invests a lot in 2021, I mean – [39:11] Not the first person to say this. That's brutal. You know, it seems like the fair thing would be you get – [39:18] judge relative to the vintage. But at the end of the day, you know, people want, uh, [39:24] returns, you know, they want to see that you've exited stuff and these big funds, even if they're [39:29] spraying and praying, uh, [39:32] have exits that they can point to. - And then they compare it to the public markets, and they're like, well, why would we invest here when we can put it into large caps that are

[39:44] concentrated to 10 companies. Which also is not [39:49] particularly. Right. Yeah. I mean, I think Dan Primack and Axios had something about, you know, isn't like the Yale model dead or so, you know, they're, you know, just this idea that, [39:59] All the endowments would go super. [40:01] Huge into venture. [40:03] I think is more of an open question. Um, [40:07] On the other hand, you know, it feels like there's endless sovereign wealth money. And so, you know, whenever there's like, oh, you know, all the smart money or whatever is a little bit more apprehensive about it. [40:18] Venture. [40:19] There's other money. [40:21] to come in, uh, [40:23] But, you know, long term, I think I'm still – [40:27] Super bullish on venture capital. You know, it obviously has... [40:31] ebbs and flows. [40:34] And I would rather be an LP putting money in a fund – [40:38] today than one putting money in a fund that was going to invest in a [40:43] in 2021. So, you know, it's always hard to be [40:48] counter-cyclical, but the best LPs are [40:52] I think I posted this in like the last sorcery, but probably one of the best conversations I've listened to in a while was. [41:01] on the Liquidity This Week in Startups podcast with David Weisberg. And he interviewed someone at Sequoia and someone at the Hewlett... [41:11] What was it? [41:12] I think it's like the Hewlett Foundation.

[41:16] She just explained it so well. She's an LP. She's like, look, we invest into innovation. We invest into this category because we invest into innovation, and that's the future, and that's why VC exists now. [41:29] and just simplified it down to that. And I thought that was perfect. That was great. You know, she's [41:35] defending the category in general and not calling out all the little details of cycles, right? Like, look, like at the end of the day, we're investing in innovation in whatever shape it, you know, [41:46] Shifts to is always random. We get caught with things left and right and shiny objects, but at the end of the day, like, you're investing into, like, [41:56] the most talented individuals and [41:58] hope that even through pivots, they create some really incredible technology and products. [42:04] Yeah, but just to be annoying, I mean, prices still matter, and, you know, there's innovation. [42:09] in public tech companies. And as you were sort of alluding to before, at some point, there needs to be a reconciliation between the two. And if all your private managers are just marking up their returns and not providing actual exits, maybe you go, you know, to the liquid environment of the public markets. So, you know, competition makes it all work. And so, so yeah, certainly there's a lot of innovation in private startups, but [42:38] needs to be at desirable prices for investors. [42:43] I agree with that. I'm very all for

[42:46] concentrated, disciplined investments. So just so that's clear, I don't believe in any of these 100X multiples. I think that's crazy. You need to return dollars back. But yeah, it's... [43:01] Hopefully, [43:03] some returns come back through these vintages. [43:06] As wild of a markup that they had, hopefully there are some great resources. I mean, we do. I mean, the other – we're in an IPO crisis at the moment. You know, they're just – it's really bad, and if there aren't – [43:20] If, you know, antitrust is a barrier to M&A... [43:24] Uh, and, uh, [43:26] There are no IPOs, then... [43:28] Private companies certainly will continue to look [43:30] Worse and worse. So, yeah, people can say... [43:34] They want to bet on innovation, but... [43:36] innovation has to be able to show that it can exit. [43:40] Okay, so as we close it out, [43:43] What are you most excited for this year? [43:46] Cerebral Valley. No, November. No, I'm most excited to the break between two busy conferences. What am I most excited for in the business or in technology themes or give me a little more? What kind of excitement do you want? [44:02] It could be business technology. It could be newcomer. Maybe you're hiring more people. Maybe you're, you know, creating an AI chat bot. I don't know. Yeah, I... [44:12] We did it. [44:14] We had like a...

[44:16] sort of [44:17] you know, just out of college, engineer try to do like... [44:22] news aggregation for us. I don't know. [44:25] Wasn't wasn't my expertise, but I, you know, I'd like to continue to experiment those things, but that's that's not a priority. [44:33] I mean, honestly, I'm excited. [44:35] to be reporting. It's like, I love hosting events. Um, [44:41] But, you know, certainly organizing them takes a lot of energy and, [44:45] I feel like nothing more satisfying than just... [44:48] writing great [44:49] pieces for the newsletter. So yeah, I think, and you know, I like podcasting. I, I've sort of had a podcast at various points and want to, uh, uh, [44:59] figure that one out. [45:01] get back into it. So I think, [45:03] I'm excited about the bread and butter of [45:06] doing the editorial work and writing and podcasting and learning about everything that's happening in startups and venture capital. [45:14] That's a perfect answer. [45:17] To more journalism. Yeah. I love it. I like it. You know, that's what I want to do. [45:22] And maybe you can make little mini summit series through your podcasts. Yeah, I know. Well, we did a banking summit, you know, in March. You know, we're always thinking of new ideas. I mean, the challenge, you know… [45:34] of any founder is, you know, [45:37] Well, I think in particular, unlike a startup, you know, instead of having technology that you build it and then, you know, maybe you go sell it or whatever. [45:45] I'm very much involved in recreating the podcast and the product. You understand this writing a newsletter. It's not like you make it once and you have it forever. You have to keep making it. That's why media is a questionable business model. So it's always a tradeoff between sort of focusing on –

[46:05] you know, doing great work and then also sort of building the company and thinking of new conferences and all that. You know, I hired... [46:12] Maddie Renbarger [46:14] uh, [46:15] in January. So have one full-time reporter, have one full-time business person. Certainly, I mean, my ambition here is to continue to hire reporters, you know, [46:26] Part of what's satisfying about this whole thing and trying to make a good business out of journalism is that like it gives back and you have more journalists and you have good reporting. You know, I fundamentally just believe [46:39] that there are a lot of secrets in the world and there's a lot of stuff about, [46:42] "how business works that most people [46:44] don't get access to, and so I think there's plenty of information that's sort of substantive about business. [46:51] to publish and would love just, yeah. [46:53] grow the business with my own journalism and with other companies. [46:56] reporters working on him. [46:58] Well, I'm excited for you. And I'm excited to hang along for the ride. Yeah, thanks for having me. This has been great.