Early Stage AI Investing, Defensibility, and Founder Anti Patterns

00:00 | You know, on this episode of the show, I have with me Itamar Novick. 00:03 | He is the founder and general partner at Recursive Ventures. 00:06 | And we're gonna be talking about being an early AI investor, and how does he kinda tell what traction looks like when there isn't traction. 00:12 | We'll talk a little bit about what those AI moats look like. 00:15 | We'll talk about anti-patterns that affect speed, judgment, and fundraising. 00:18 | That's a good one for founders out there. 00:20 | And also, when we talk about AI startups, what about platform risk and defensibility? 00:24 | Yeah. 00:24 | I'm sure a couple other things in there as well, but, Itamar, thanks for taking the time. 00:28 | I'm here. 00:28 | Excited to be here. 00:29 | Thanks for having me, and thanks for all the listeners that joined us for today. 00:32 | Absolutely. 00:33 | Okay, Recursive Ventures. 00:35 | tell us what kind of VC you guys are. 00:37 | Yeah. 00:38 | That's a good one. 00:38 | So at Recursive Ventures, we lead, 00:41 | We're an institutional VC firm that leads mostly pre-seed rounds in AI companies. 00:47 | now a lot of people, they think about AI as like, oh, ChatGPT, and all these LLMs, and models, and now agents. 00:53 | We kind of see that as sort of the almost like the infrastructure, like the operating system of AI, and we're very much focused on the applications built on top of it, whether they're consumer AI applications or whether they're enterprise, you know, for business type of, agentic applications. 01:08 | That's really what we specialize in. 01:11 | a little bit more of a background maybe about me and the team. 01:13 | So I'm a repeat entrepreneur. 01:14 | My last company I public- It's a company called Life360, a billion dollar market cap, list- NASDAQ traded. 01:21 | And I've been investing for a little bit over 15 years now, while I was building companies at the same time. 01:27 | So been doing startups for 25 years, and always kind of investing in and building. 01:31 | And I've been fortunate to invest in 2 decacorns and 6 unicorns at the inceptions. 01:37 | So been doing this for a while, and has so far been a lot of fun, and, got that lucky, I guess, a few times, so that's us. 01:44 | There you go. 01:45 | All right. 01:46 | I guess, yeah, I'm gonna start off maybe a question how you obviously have been investing for a while. 01:50 | you've been a founder, a serial founder, a- and you invest early. 01:54 | And I guess when you are an early stage investor, pre-seed, seed, talk to us a little bit about what you're looking for, 'cause I think this is probably the highest risk, obviously highest reward, but also the highest risk area- 02:07 | Absolutely 02:07 | to invest in. 02:08 | I mean, absolutely. 02:09 | I mean, we write inception checks, you know, with, with 2, uh, women in a garage with a dog starting a company based on an idea, and we do that all the time. 02:18 | So yeah, very high risk. 02:20 | And, you know, our framework, which I don't think is necessarily special, is really, you can kinda summarize it in, like, team TAM, right? 02:29 | Team TAM mode is the easiest, cleanest way to think about it, and now I try double-clicking that a little bit. 02:35 | So team, that's obvious. 02:37 | That's something that everybody's looking for in venture lands. 02:41 | Like, you wanna have amazing people who are really smart, who are really adaptable, really flexible, are subject matter experts in their space. 02:48 | And specifically when it comes to AI, and we'll talk more about that, it's obviously we've got this completely new cycle starting now with AI. 02:54 | Everything is in flux, and you gotta have really good technologists who know how to harness the power of AI, use it to build better, you know, solutions and services and right? 03:04 | so that's for the team. 03:05 | TAM, I think, is the one thing that founders kinda are not... 03:10 | Sometimes they skip. 03:11 | They, they're not as aligned as they should be with the VCs on that, and we can talk about where and how. 03:14 | But, like, really VCs, I mean, we're here based on the VC power law to try to invest early in what would be a multi-billion dollar business, right? 03:23 | So it's gotta be the total addressable market, the TAM, really has to add up to something that is very, very significant. 03:30 | And it doesn't mean that the other companies that people are building are not good. 03:34 | They could be great. 03:35 | It's just the VC model really pertains to that kind of huge, kinda out, the decacorns, right? 03:41 | That, that's really what we mostly subscribe to. 03:44 | And then I would say where we're a little bit more unique, even though there's been a lot of conversations on that in the last especially 18, 24 months, but we've been kinda looking through that lens long, is having a moat. 03:56 | Is this concept of like, okay, if you can't explain to me what is the mechanism that will keep you ahead of the pack, and there's gonna be a billion competitors, and they're all gonna copy you, and all this stuff is gonna happen if you're successful, then we kind of find it really hard for us to invest. 04:13 | Because we wanna have a, almost like a systematic unfair advantage of like, oh, wow, this company has something that's very, very hard to replicate, and that's why even 15 years from now when they're a public company, they would be able to explain why they can, you know, solidify their position and stay ahead. 04:30 | Question there for you. 04:31 | When we're looking at these early stage companies, and you mentioned, the power law, and it, it's well understood. 04:38 | It's how VCs make money. 04:41 | there's similarities in other industries, as well, so I, it's, it... 04:44 | I think it's a concept that applies. 04:46 | Do you see with AI and this notion that you just don't need as many people, and also with subject matter expertise you can create some very narrow solutions, is that gonna possibly change the way VCs operate? 05:05 | Meaning before you're looking at TAM, you know you gotta hit big. 05:09 | You gotta keep taking big swings. 05:12 | But does the model possibly change if maybe we're not looking at every company being, you know, a multi-billion dollar company, but a very successful company that's just a half a billion dollars, but... 05:24 | and they're doing great, right? 05:25 | They're profitable. 05:26 | I mean, it's a notion that we haven't really come across. 05:28 | But is there any thought that potentially the VC model with the type of companies AI will produce might evolve, might change, might shift? 05:39 | Absolutely. 05:40 | Let me give you a few different angles on it. 05:42 | So the first one is, are we gonna see companies being created where they require significantly less capital to get to scale? 05:48 | And I think the answer is yes, I think that's gonna happen. 05:50 | And I think what that impacts is primarily the big funds, right? 05:55 | So what my fund does is it's... 05:57 | We are the ones that put in the money for you to quit your job at Google DeepMind and go out and start a company or, like, hire your first employee. 06:04 | That funding is not gonna go away. 06:06 | But then if instead of hiring 100 salespeople, you only need to hire 5 because they're all bike coding their way through sales and whatnot, then would you need that $100 million check from Andreessen? 06:17 | I don't know. 06:18 | It begs the question. 06:19 | So that's one kind of angle on that, and it mostly impacts the bigger VCs, the more classic VCs. 06:24 | Another angle of that, which we are totally subscribing to, is this notion of seed strap, 06:30 | Right. 06:31 | So we've had bootstrapping for a while, and it's been associated both with companies that eventually suddenly become venture-backed as they are at scale, and, you know, suddenly they, like... 06:42 | Flavio is a good example. 06:44 | I don't think anybody invested in that until they got to 50 or 100 million revenue, and then they went VCs and go public and what. 06:50 | but seed strapping is this concept of like, okay, I'll just raise a little bit of money, $1 to $3 million from early stage VCs like me, but then I'm done. 06:59 | I know how to scale this up all the way to an IPO and that's it, and then an evolution of bootstrapping. 07:05 | I think w- we're very supportive of that. 07:07 | We've done a few deals like that where the founder is saying, "You guys are last money in," right? 07:11 | And, but I think it's too early to call, and this ties into this kind of narrative of, like, a one-person billion-dollar company, right? 07:22 | I mean, we're talking about it, but have we seen it yet? 07:25 | Nah, not exactly. 07:27 | And is it gonna happen? 07:29 | Well, maybe we can make some bets to try to see if it's a real thing, but I don't think it's proven yet. 07:35 | I think it's a new model that we'll just have to see how it plays out. 07:40 | And then for your last thing around can you build an AI company in a niche that becomes significant enough from a return profile to an investor even though it gets sold for $340 million, et cetera, I would say theoretically yes, but only if the whole stack recalibrates itself. 07:58 | So for example, if I invest at a $20 million post-money safe, and that's cheap actually compared to what's happening in Y Combinator now, where there's two 23-year-old kids with nothing are raising money that's $50 million valuation. 08:11 | That's where the, the, the stack or whatever, like, the, the process break, right? 08:15 | If I could do those deals at a $5 million post-money valuation, and then it ends up getting sold for 300 million without taking significantly more capital, yay. 08:25 | Absolutely, that works. 08:27 | But that's not what happens in reality. 08:29 | What happens in reality is they go raising the $50 million post, which is ridiculous, right? 08:34 | And then it doesn't work, and they end up selling for 100 million or something. 08:37 | And then, like, for the investors, the math doesn't work. 08:40 | And more often than not, it also doesn't work for the founder, which is a whole different conversation on why not to raise money according to YC guidelines on a $50 million post when you have nothing. 08:52 | Yeah, that's... 08:52 | I love 08:52 | That's, that's, that's, that's great insights. 08:55 | I guess I wa- I wanna tie my next question into subject matter expertise, 'cause I am curious. 09:03 | I don't know when we'll get to the billion-dollar one-person company or if it's even feasible. 09:07 | I don't know. 09:07 | Like, that's... 09:08 | It's a great idea in concept. 09:10 | I did have, Ben Seron, 09:12 | who's a solo entrepreneur right now. 09:14 | He's building what will be probably a billion-dollar company all by himself and his team of agents, and he's doing something phenomenal. 09:20 | but also I don't know if everyone can do that. 09:22 | That's, that's so far, you know, a very unique situation. 09:25 | And, and whether that's sustainable. 09:26 | I mean, maybe he goes up to a billion and then goes... 09:28 | I don't know. 09:29 | There's all this, right? 09:30 | That, that- 09:30 | Nothing. 09:31 | But- 09:31 | Yeah, that, that is actually an interesting point, 'cause even if you got there, at some point you might, you might need a second person. 09:38 | Absolutely. 09:39 | A- and I guess when you talk about subject matter experts, so if we're talking about AI, early stage AI, and you're look- looking at the moats and they're coming to you, what are you looking for... 09:50 | Obviously moat is a moat, but you said that moat needs to go beyond the model. 09:54 | I guess can you tell us what that means when you're evaluating these startups? 09:58 | So let's actually separate the conversation to subject matter expertise and then moat, because moat is like its whole big- 10:03 | Absolutely 10:03 | right? 10:04 | So- I think subject matter expertise and being close to the customer and understanding the customer, and then obviously at the back of that having a very significant market motion because you are part of that industry and you know people, you can... 10:18 | then they believe in you. 10:19 | I think that is more important in this cycle than previous. 10:23 | Because in previous cycles, it was hard to build, right? 10:27 | You had to hire the right engineers, and it took time, and the better they are, the faster you get there, and the better experience, better product you have. 10:35 | In this cycle, we are con- continuously moving toward a target goal where building is cheap, really cheap. 10:46 | Anybody can do it. 10:47 | You don't need the best engineers in the world to build an application. 10:50 | Like, you know, if you're building the new LLM, that's a whole different thing, right? 10:53 | But to build an application is becoming increasingly easier. 10:56 | Anybody can do it, and they can do it really fast, and with a very low... 11:00 | So suddenly, 2... 11:03 | I'm going back to the YC kids for some reason. 11:05 | But 2 YC kids really were back because they didn't have a family, and they would work 15 hours a day to code, and they would code the heck out of something in 6 months that people would build in 5 years. 11:18 | That advantage has been eroded completely. 11:20 | It doesn't matter. 11:21 | So what does matter? 11:23 | Subject matter expertise. 11:24 | The same kids who are now gonna go after pick a topic. 11:29 | Treasury management. 11:31 | That used to be a chief. 11:31 | Whatever, I picked a random one. 11:33 | It's like, what the hell do those kids know about that? 11:35 | What kind of credibility do they have about that? 11:36 | Do they understand the workflows? 11:37 | Do they understand the needs? 11:38 | Do they understand the customer? 11:39 | They don't. 11:41 | Previously, we're banking on them building stuff, throwing it in front of people and learning in the process. 11:45 | Now, that's no longer the barrier, so it's really you have to know it. 11:48 | And I think that's why subject matter expertise matters more in this cycle than previous. 11:53 | So that's on subject matter expertise, and that's the kind of founders that we like to back. 11:56 | But on moats, it's a very complicated conversation, but you get the gist of it. 12:02 | Like, okay, if you don't have a moat in this new era, you are gonna be... 12:07 | It's death by 1,000 cuts from competitors who replicate you in no time with no cost. 12:15 | And, a-and it's commodit- it's commoditization, right? 12:18 | Because, hey, I mean, I used to build a piece of software, and I'd sell it $100,000 a year, and people would love it, and then somebody else would come, and they'd say, "Oh, we'll give you this software for 50, but we don't have all these features." And they're like, "Ah, no, we want to stay with 100K." 12:34 | Feature-rich app. 12:37 | Now that's gone. 12:38 | So it's basically a race to the bottom, right, with no moats. 12:41 | Moat is the only thing that really does not have you in this race to the bottom, right? 12:47 | So the classic moat that everybody's thinking about is a data, and that data moat still exists completely. 12:53 | So if you have first-party proprietary data that other people don't have, right, nobody can build a model as good as you. 13:00 | that's inherent because AI, agentic, LM doesn't matter. 13:05 | It's really only as good as your underlying data and how that underlying data pre-preempts the problem you're trying to solve, breadth and width of data, right? 13:13 | If you have other data, who cares? 13:15 | It doesn't solve the problem. 13:16 | If you don't have enough data, who cares? 13:18 | It doesn't solve the problem either. 13:19 | So really, when you look at LLMs today, they're all converging, and it's because they're all using the same dataset and the same techniques. 13:26 | And what they don't have is all this first-party data that applications companies can access and potentially use. 13:32 | So that's the moat that everybody thinks of. 13:35 | But we also think about several other types of moats. 13:38 | probably the most important one is reinforcement learning or recursive learning, and that's, you know, also we print the br- the brand, the fund is Recursive Ventures, and it's been designed that way with the thinking about, okay, we're gonna get to recursion here. 13:50 | Is kind of this concept of, you know, you put a solution, you put an app in front of users, and they're like, "Oh, okay, so my agent wants to do this, and this. 13:59 | Here's the plan," blah, blah, blah. 14:00 | Thumbs up, thumbs down. 14:02 | That seems like a marginal thing. 14:03 | It is not. 14:05 | It is exactly what powers reinforcement learning, which is a huge competition, I mean, they overlap, and they're both being used, to, what's called expert learning, which is what got us here. 14:15 | So what a lot of people don't know is that the... so the hidden gem with a lot of this LLM generative AI, progress in the last few years has really been through what's called expert labeling. 14:27 | Of, like, experts coming in, getting all these prompts with ChatGPT, and saying, "No, no, no, you're wrong. 14:32 | This is the right thing. 14:33 | This is the wrong thing." And that's also why after you embed those experts in your model, their response says, the Chinese come in and copy all that, and it works, right? 14:41 | Because it's already been labeled by the experts. 14:44 | So reinforcement learning sort of gets you that through actually operating in the field, and that is a moat. 14:50 | So if you have a reinforcement learning cycle with your customers, you're gonna get to a better app, to a better agent, right? 14:57 | And that's a big moat. 14:58 | the last thing that... 15:00 | Well, we look at a few others, but we look very heavily at hardware. 15:03 | So hardware has been something that most VCs haven't been emphasizing for a very, very long time, especially in consumer electronic settings. 15:09 | We actually love that, right? 15:11 | Because now you have all this data that nobody else has because you have a wearable, and you've got, you know, vision pr- whatever, right? 15:17 | so that's another big one. 15:19 | And then there's also a lot of our companies that are doing things either with open source data sources and community stuff, or they're signing exclusive deals to access data with other folks that is not available, and they lock those, those data sources in. 15:33 | And it could be from hardware or sensors, it could be from applications, it could be... 15:37 | Like, that's something that's not public, right? 15:39 | And then they're leveraging that to get a, a, a first mover advantage and keep that advantage going through, through business development, right, through, through deals with others. 15:47 | So those are some of the moats that we're, we're, we're looking at. 15:51 | We-- There's more. 15:52 | There's a long list, and we've written a bunch about them. 15:54 | But, 15:55 | what I wanted to emphasize, it goes beyond like, "Oh, I have data. 15:57 | Yeah." 15:58 | Okay? 15:59 | Absolutely. 16:00 | I g- I guess just a question, 'cause you mentioned obviously subject matter expertise. 16:04 | use example of, uh, you know, a couple young kids trying to do something within, the finance, world and, not having the understanding of the workflows, the process, all those things. 16:17 | I- in the age of AI, does subject matter expertise become its own moat if, if your product is built for one of those, like, very particular processes? 16:30 | Yes, because what AI does, and everybody can tap into that, is build the, the, the consensus, right? 16:37 | AI builds the consensus. 16:38 | If you let it build, that's what it's gonna build. 16:41 | And what subject matter experts bring to the table is the insights that are potentially outside of the consensus that matter, right? 16:48 | And that's the difference. 16:48 | The difference is that better user experience, better understanding of the customer, and as it pertains to subject matter experts, especially the more senior folks, it's the go-to-market motion that goes together with that, where they have their Rolodex, they know the people, they're trustworthy, and they're like, "Yeah, we're gonna sell you Y and it's gonna work for you because we understand you as a customer." 17:07 | And I think that's, that is, that's a huge differentiator. 17:10 | Is it a moat? 17:12 | I think it's part of the moat. 17:14 | It's not the whole enchilada. 17:16 | Absolutely. 17:16 | Makes sense. 17:18 | You mentioned pivoting topics, uh, 'cause I, 'cause I do wanna get to this, 'cause I, I, I think, um, you have a, you have a deep, uh, knowledge base in founders and, uh, when we first w- talked, you'd mentioned anti-patterns, and it took me a second to go read your, uh, postings to understand what you're talking about. 17:36 | But I, I, I like this. 17:37 | I'm gonna let you, uh, explain what, uh, the founder anti-patterns mean, and we'll kind of dive into some of those sub areas then. 17:45 | Absolutely. 17:46 | So, so, okay. 17:47 | So I've invested in close to 200 companies over the last 16 years, and then my 2 partners have invested in, like collectively investing over 600 companies, out of which, you know, 2 dozens are unicorns or beyond at this. 18:00 | So, and here's the thing that we saw. 18:01 | It started anecdotally, and then we started seeing it over time with like, you know, actual numbers. 18:06 | There is no single path for success. 18:08 | Every company that's successful kind of charted its own way, figured something out. 18:12 | And if you try to learn from success, and most of the media, most of the world today is really focused on success. 18:18 | It's like, "Oh, let's talk to Aires Ky, 'cause he built Airbnb, and let's learn from him, and let's copy what he did," whatever it is, right? 18:25 | That actually doesn't... 18:27 | It doesn't work because every company's path to success is unique, right? 18:31 | And there's not necessarily that much transferable from that to whatever you're doing. 18:37 | What is repeatable and is potentially also available, a-avoidable is, is, is, is, is risk, is failure, right? 18:46 | So what we started figuring out is like, wow, look, so many of our companies are failing because of A, B, and C, and they're all repeating the same mistakes. 18:53 | And actually, if we could teach our founders to avoid all these mistakes, their odds of success go up a lot. 19:00 | Because there's a significant portion of super early stage company building that actually pertains to reducing risk. 19:08 | And people don't talk about it though. 19:09 | They talk about taking risk. 19:10 | And it's right, but you're taking a risk already by building pre-seed stage startup. 19:16 | That is already a huge risk. 19:17 | You're 95% chance of losing, right, at that point. 19:21 | So actually, one entrepreneurial mindset that works is, okay, how do I systematically reduce risk? 19:26 | And some of that is fundraising. 19:27 | You raise money, you have more cushion, right, for him, right? 19:30 | And so on and so forth. 19:31 | But what we did is we wrote down 80, 85, actually, startup anti-patterns. 19:36 | Those are things that feel at first glance is like brilliant ideas. 19:40 | It's like, oh my God, I'll chase Google as a customer, and if I land Google, it's gonna be amazing, right? 19:46 | And then actually what happens to most startups that do that is they wind up, you know, they end up with a three-year-long process with somebody at Google that they spent half their time and money on, and at the end, somebody at Google just comes in and say, "Oh, we don't care about that," because the executive team or something, right? 20:00 | And this whole while, they haven't been focused on getting the real customers that really buy them and use them and learn from them, right? 20:07 | There's one example of a startup anti-pattern that we call elephant hunting, right? 20:12 | so just to give you an example of like how there's one different lens of looking at startups, like how do I systematically reduce risk as CEO, as founder, team, right? 20:22 | And that's what we write about in a lot of our sort of narratives and thinking are like, okay, you, you're doing great. 20:29 | You're succeeding. 20:30 | We don't need to recap that. 20:31 | It's more like how do we make sure you keep going on that right kind of path, right? 20:34 | But I guess when you're looking at some of these anti-patterns, you mentioned a few of them. 20:38 | When you look at the founders that you're investing in, how do you align those fou- the- 20:43 | these little anti-patterns with people you're investing? 20:45 | Do you look for those- that naturally have this mindset? 20:47 | Do you have to help them understand some of this? 20:49 | 'Cause, I d- I didn't pay attention to this until I had much more gray hair. 20:54 | Yeah. 20:55 | I think it's really embedded in our diligence and how we talk and anything about engaging with Tom, right? 21:00 | So, so, you know, w- we obviously ask the obvious questions like, "Hey, how are you doing? 21:06 | Our sales, revenue, AR," all that stuff. 21:09 | But then we do push on the areas like, "Okay, where do you think you've made a mistake and why, and what could've you done?" 21:14 | Right? 21:14 | Or, "How are you thinking about your strategy moving forward?" And, and because we've got this anti-pattern background, we keep thinking about, "Okay, are you really doing things in ways that we know great," right? 21:26 | Or, "Are you doing things in ways we know are scalable through this anti-pattern lens?" And I would say that naturally has us gravitating more toward backing experienced entrepreneur. 21:35 | I think that's the case. 21:36 | And just give you a random stat, Recursive Venture III, which is the fund I'm running now, is, it is almost 2/3 successful entrepreneurs is whom we back now. 21:46 | Yeah. 21:46 | Obviously, that's a different profile of folks who've been there, done that. 21:51 | They can talk to you about their wins, but at the same time, they can also talk to you about all the mistakes that they survived through. 21:57 | Right? 21:57 | and I think that's really important for us because, you know, there's so much, again, risk in early stage venture and early stage company formation. 22:05 | You can do everything right and the market turns on you or the trend is wrong or Andreessen Horowitz shows up and gives your competitor $250 million and suddenly you're no longer the cool boy, like, cool company. 22:17 | Like, all this stuff can happen and it's completely outside of your control. 22:21 | But it is within your control as a founder to de-risk your business by understanding this anti-pattern, and that's what we're adamant on them as part of our diligence. 22:30 | Fantastic. 22:31 | I guess maybe a final question for you. 22:33 | obviously right now, everyone is focused on AI, and everyone says we're early. 22:39 | also there's trends that come and go, and there's life cycles to all technology booms that we've seen. 22:47 | With AI, I mean, are we... 22:49 | I mean, obviously we're early on. 22:50 | I don't need to, we don't need to talk about that. 22:53 | But as early stage investor, how long is AI the priority? 22:56 | I mean, probably will be for a while. 22:58 | But the reason I'm asking is, some point, the next thing does start creeping up, and obviously as an early stage person, you wanna be at the forefront before everyone's, you know, deep into it. 23:08 | is that something you think about, or at this stage it's, doesn't even matter right 23:12 | I would think... 23:13 | Look, at the end of the day, we do what we do because we love, supporting entrepreneurs, but we are obviously a fiduciary of our LP's money, and we're in this game to make as much money as possible. 23:22 | That's why we get paid. 23:23 | That's what we do. 23:23 | That's our job. 23:24 | So I think it is a must. 23:26 | And, and, you know, I'm, I'm more of a founder, so I think about this dynamically every single day of like, "Oh, should we go into space? 23:33 | Should we go into defense? 23:34 | Should we go all these things?" And there's always the AI is now everything. 23:38 | It's like saying you're investing in AI is like saying, "Oh, I'm investing in software." It's like everywhere. 23:42 | It's in healthcare, it's in spa- what- whatever, right? 23:45 | So I think it's ubiquitous from that standpoint. 23:47 | and I would say, we, yes, we are kind of widening our lens. 23:51 | We are a generalist firm. 23:54 | But I think, let me just throw a couple of things that we believe in because that will help Oren, not just you, but also the people listening. 24:02 | So we think AI is the biggest one we've seen since, and even more than the Industrial Revolution. 24:09 | We think that if tech is now 12, 15% global GDP, AI plus tech can be 50 and a bigger pie, which is just unbelievable, right? 24:17 | That's one. 24:18 | The second thing is we think we're in a blip in time where AI infrastructure is in a bubble. 24:24 | We believe we're in a bubble. 24:26 | And it's a little bit like 1999 when Cisco was worth a trillion dollars selling you a networking box to power the internet and then the whole thing collapsed. 24:33 | But what came out of that? 24:35 | Google, Amazon, eventually Meta. 24:38 | Those are application company. 24:40 | So the folks that end up providing the end solution to the end user, that stand in front of the customer saying, "Here is the thing that's gonna make you amazing," right? 24:50 | "Give you ROI." 24:51 | it's not these LLMs company. 24:53 | And people are getting it wrong. 24:54 | They're like, "Oh, Anthropic is gonna eat all this." No, Anthropic is not gonna build a company to go after 500 vertical and horizontal SOC. 25:02 | It just doesn't exist, right? 25:04 | unless it's AGI, and that's a whole different ballgame, and then we can have a different conversation about humankind, not just about AI, right? 25:12 | so, so that's the second thing that we believe in, that there's a misallocation, both timing, like we're too early for AI to be ready to fully like show time at the enterprise especially. 25:22 | And, a- and we're not focused on investing in the right things. 25:25 | Like more money... 25:26 | Well, you need the data centers to power everything, so I'm not against that. 25:28 | But like putting another trillion dollars into Anthropic to build their LLM is completely useless when the LLM is becoming a commodity and the Chinese are 3 months away. 25:37 | And it's just gonna go s- s- So like what's the point? 25:40 | What I'm saying is, I think the LLM is worth 0 in the long run. 25:45 | It is worth 0, right? 25:47 | And I don't think people are getting the memo on that yet, and that's what we believe, right? 25:51 | So that's, another important part of our belief structure. 25:54 | and because of that, everything, right? 25:55 | So, so VC is gonna change, you know, startups are gonna change, like you alluded to before. 26:01 | And I think that means that if you're not flexible as a venture fund and you think, you keep thinking of like, "Oh, I can just go raise a trillion dollar fund 9 and it's gonna be all the same," no. 26:10 | Like everything is changing. 26:11 | And if you don't change alongside with that You're not gonna make it. 26:15 | So this is the era of entrepreneurial VCs 26:17 | it's not a if question, it's a when question, so that's definitely a interesting, way to end the conversation. 26:22 | Itamar, thank you for joining. 26:23 | Thank you for, sharing with us. 26:25 | If somebody wants to reach out, learn more- 26:27 | follow up question, et cetera, what, what's a good way of getting in touch? 26:30 | Yeah. 26:31 | So the best way is LinkedIn. 26:33 | Find me on LinkedIn, Itamar Novik. 26:34 | I'm the only one with that name on the globe currently. 26:37 | and, I write daily, for founders about founding companies, about VC horror stories, which is all the bad stuff the VCs can do to hurt your business and you should know about it. 26:47 | Follow me on LinkedIn, engage with me, and that's a great way to get to us. 26:52 | Another probably faster way is get a warm introduction through somebody that knows you and knows me or one of my partners. 26:59 | it's not that we don't wanna help and talk to everybody, it's just there's only so many hours in the day, so we prioritize, warm introductions like many other VCs. 27:07 | And I have tens of thousands of connections and more than that in followers. 27:12 | So if you are gonna be able to find a way to get to customers, I believe there's probably a way for you to get to me through somebody who knows me. 27:21 | Good point. 27:22 | Awesome, man. 27:22 | I appreciate your time. 27:23 | Thanks for being on the show. 27:25 | Thanks for having me, Amir. 27:26 | So much fun. 27:27 | See you next time. 27:27 | Thank you. 27:28 | Bye. 27:28 | All right. 27:29 | That's it for the episode. 27:30 | Be back again, different guest, different topic. 27:32 | Until then, 2 things. 27:33 | One, I think Itamar did a great job of talking to us about being- 27:37 | an investor, early investor in AI, and how he views that market, what he looks for. 27:42 | We talked about the founder anti-patterns, which was, I think, a great part of the conversation if you're a founder out there to take away those points as well, and follow Itamar- 27:50 | 'cause I think, he posts a lot about this stuff, and it'd be a great resource. 27:53 | Also share the episode if you can with, another engineer, somebody who's looking to be a founder or a current founder. 27:59 | I'd appreciate that. 28:00 | Also, like, subscribe, comment. 28:02 | Let me know how the show's going for you. 28:03 | Until next time, thank you and goodbye.