How do you evaluate AI startup moats when barriers to entry are near zero?

Jason Kirby (00:01.48) Hey everyone, welcome back to 100 Million Dollar Exits, the show where I interview guests who built and sold $100 million companies. Itamar has actually been on our show before. If you haven't seen that episode, you can check it out up here. But Itamar, I'm so excited to have you back. You know, just for quick reference, you built and sold or sorry, you built and took public Life 360 twice, which I always find is an interesting story, which we talked about in the last podcast. But you're also the GP and solo GP of Recursive Ventures, and you are making early-stage bets, you know, especially in AI. You're in San Francisco, you have a very strong opinion on kind of where the market's going. And I thought it'd be great to have you back on and continue this conversation about AI, what's real, what's not real. And I think what would be the best way to kick things off for people is just tell us, you know, what should be f what founders should really know about what's happening on the ground floor. When it comes to AI, whether it's infrastructure or application layer right now. Itamar - Recursive Ventures (01:04.971) Yeah, well, first of all, Jason, thank you for having me back on the show again. I'm so excited to be here. I really enjoyed the last time, and I can't wait for this time. And also thanks for the audience out there for joining us today. It's been a fun session. Hey Jason, maybe we should I know it's still under wraps- but should the audience know that we just did a deal together? Right? How about, yeah, it's not public yet, right? But but but we can tease, we can tease, we can say, Hey, you know, Idomar and Jason just did deal together and it's an exciting one. Jason Kirby (01:22.92) Wow. It'll b by the time this goes public; it'll probably be done, yes. Jason Kirby (01:33.48) We'll do the virtual high five. Itamar - Recursive Ventures (01:35.886) There you go. Yes. And you know, for us as a venture firm, we we look at you know probably a thousand deals a year seriously, right? And we do eight to ten. So we just hit the jackpot together, baby. It's exciting. yes. Wow, where do I start? I mean, you know, I'm in San Francisco, and this is the epic center of AI, and there's so much stuff happening. And I think let me just start by saying to everybody out there, and especially founders, because you're the you know, the ones doing the heavy lifting and also getting rewarded the most if it works out, is I think This AI revolution is the biggest, most fundamental transformation of humanity and human productivity since the Industrial Revolution. Right? It is a fundamental shift in everything that we do, especially in business and how we're gonna go about it, right? Jason Kirby (02:37.8) So this has been said a lot, and I want to kind of call this out a little bit more and unpack it a little bit more. So it's like we had the dot-com boom, we had, you know, kind of the cloud, we had the mobile, we had these different evolutions. What is it about the AI industry right now that is just so exponential and obvious that this is different than the rest? Itamar - Recursive Ventures (02:59.479) Yeah, so You know, since the invention of the personal computer and then obviously the internet and then SaaS and and all these other things, it's really been about helping humans be more productive with computer systems and and computer system communicating with each other, right? That that's really what we've been doing, and we have significantly enhanced our productivity. With generative AI, it is completely different. It is about, you know, an order-of-magnitude increase in productivity, but also, in many ways, it is replacing human labor, both white-collar collar and blue-collar with machines. That is why it's so fundamental. Right? It's a huge stark difference between, hey, let me give you an app or a mobile web s or a site that's gonna make you better off, versus like, this agent is actually gonna get the work done for you. Right? That is a huge fundamental shift and opens a a a slew of opportunities. One of the opportunities that we're gonna talk about, I think, is is what's called services of software. We look at service organizations, whether it's tax, accounting, legal, so on and so forth, and you really see a future where all of that stuff can be fully Or 95% less automated with AI. That's kind of on the white-collar side. On the blue-collar side, we're obviously seeing things like Waymo. You know, come to San Francisco, you don't need to order an Uber anymore. You just open up the Waymo app and guess what? A car shows up and it drives itself with no driver to wherever you're at. You want to go. So that is a huge fundamental shift between, again, productivity increase and like completely replacing human labor. Jason Kirby (04:14.14) No no. Jason Kirby (04:28.294) And so there's a couple different layers here in the AI space. There's the app, you know, the infrastructure layer, platform layer, application layer. There's these different kind of value creation areas in which value is being created across the AI stack. Kind of what are you betting on? And what are you seeing in the market? Because you know, you have the anthropics, the open AIs, the Googles that are really kind of like taking the lead on the infrastructure side. But what else is happening in the market that you're seeing and that you're b you know bullish on? Itamar - Recursive Ventures (04:56.832) Yeah, so I've just talked about how amazing and awesome AI can be and how I'm a firm believer that ultimately AI plus tech could represent fifty. Five, zero percent of global GDP on a much bigger pie. By the way, tech today is somewhere between 12 to 15% of GDP. So we're talking about a 3X, you know, multiplier here on you know the role of tech in our lives. So now I've talked about how big that is, and now I want to bring it back to the to the ground where we are today. And I think today we're actually in an AI infrastructure bubble where we're getting ahead of ourselves. This will take time. It does it's not gonna happen overnight. And This is for me, this moment in time is a little bit like 1999. It's different, but it's a little bit like 999. The analogy is when Cisco was worth a trillion dollars, building a network box, right? A network box that would help the internet work, right? And then the whole thing blew up, and Cisco took another 17, 18 years to get back to the same market cap. But what came out of that? Google. Meta, right? Amazon. So The the revolution, the huge push into the infrastructure is what enabled the application layer. And the application layer is where I believe and where we're making bets, that most of the money is going to be made. Right? And it's really just human nature. It is the organizations, the application layer in this case, that really cater to the end customer, they are traditionally the ones that end up, you know, capturing most of the value in value creation scenarios like this. Jason Kirby (06:33.298) So infrastructure, define infrastructure. Is that the data center players? Is that the people actually building data centers? Is it the ones kind of contracting, like OpenAI and whatnot? So like define infrastructure for the audience. Itamar - Recursive Ventures (06:46.444) Yeah. So the way we think about this, and I I maybe I if you want we can add it to the show later, we can pull up a slide, but we we think about six tiers in AI. The first one is the energy infrastructure, right? That's obviously infrastructure. The second one is the the the is the chips, right? The semiconductors and and and kinda you know selling them and and NVIDIA and all that stuff. The the the second tier from you know down up to is is the data center, right? the hyperscalers, the neo clouds, all that stuff, right? Then the third tier, which we consider more like an operating system, is the LLMs. It's is it's the foundation model companies themselves. Yes, it is software, what more like AI, more like a statistical model, but We believe that all of these, right, these four layers that I just mentioned: energy, chips, data centers, and LMs are essentially now a commodity. They are becoming a commodity. I'm not saying they are easy to get by today, but when you look at the forecast and where this thing is going, every one of those layers is essentially a commodity. And some people argue that that is incorrect for the LMs, but just look at what's happening, right? Like open source, cheap open source. Jason Kirby (08:01.746) It's like Itamar - Recursive Ventures (08:02.72) Chinese models are now, you know, on par almost with our best frontier labs in the US. Most experts can tell the difference between, you know, chat like GPT 5.5 and and Cloud 4.7, 4.8. They have very similar performances, right? XAI is not that far away either, right? It's all basically converging into kind of the same, you know, same level. That is how commoditized markets look like. So, where's the value? It's in the fourth and fifth tier, right? It is the software infrastructure required to build generative AI applications and obviously add the application layer, both B2B and B2C. That's really where we think that the majority of value is going to be created and that's how we kind of split AI into these different layers. Jason Kirby (08:50.234) And so and it's also more attainable from like, you know, the fund size that you guys are at, the deals that you guys do to kind of be at the application layer for that reason, because they're, you know, they don't need as much capital. And so let's talk about that perspective. And it's just like these crazy large numbers and you know, whether they're valued appropriately or not, you know, TBD, but At the application layer, like the barrier to entry has plummeted. Like for me, like I'm messing around and like built an app for and you know, or take it to market. It took me like a week to build like a fully functioning app by myself. That was never possible without these tools and you know application, you know, these different applications I could do that on now. but it all comes, in my opinion, comes down to distribution at that point. It's like, well, what have you done in pre-distribution? But I'm curious from your perspective, like what are you seeing on the application? Application layer to kind of stand out when the bar to enter is so low. Itamar - Recursive Ventures (09:50.562) Yeah. Well that's a very, very long conversation about Motes, which is at the heart of our thesis of investing at recursive ventures. we we we have sort of an entire framework around What would and here's here's the fundamental question that you have to ask, right? When a CEO presents to us their generative AI application, again, whether it's B2B or B2C, we do both. the first question that we ask ourselves is, okay, why is this, you know, six to twelve months ahead of the competition today? Right? What what is the defendability? What is the mode that that makes it so? One, and two is how would that company be able to sustain that defensible position five years from now? So why in five years are you still Still gonna be one year ahead of the competition, right? It's and and and it's true that for a lot of companies in this era, that is a very, very hard question to answer because many of them are what's called fin rappers on top of LMs. This is actually mostly 2025 news, or even like 2024 news. Most VCs and most founders now understand that, right? Understand that they have to have some competitive moat in order to win. And there's actually quite a few of them. So A lot of the modes that added up in the past, like network effects, marketplaces, right? lock-in effects, where you know your customer has a really high switching cost, these still make a difference, right? If you have one of these and you know, even if there's a hundred competitors but you're really, really hard to replace, then you're not gonna get replaced, and that's great, right? But AI also presents this whole new wave of potentially new modes that people are talking about but not talking about enough, right? The first one, which is obvious that people understand, is proprietary data. And what's really important to understand about AI, which is this generative AI wave, it's it's this non-deterministic statistical model where the quality Itamar - Recursive Ventures (11:48.618) of your data is really the underpinning for the quality of your AI. Like your AI is only as good as the data, right? So if you have proprietary data that other people don't have that pertains to the problem you're trying to solve, no matter what technology Claude spins up next week, you're gonna win. Right? You're gonna win and nobody's gonna be able to pay catch up with you, right? Jason Kirby (11:54.185) Mm-hmm. Itamar - Recursive Ventures (12:11.654) So that's proprietary data, it's obvious. But there's also other ways to create significant competitive advantage, and one of the ones that we spend the most time on is what's called a customer data flywheel, especially combined with what's called reinforcement learning. So if you have pieces of your system that are built by open source models, right? Like Lama and obviously there's all the Chinese ones out there that are doing really great. And you really get reinforcement learning data from your customers saying, hey, you know, this agent plan is like thumbs up, this is great, or no, you should make changes here and step two is wrong and blah blah blah. You can feed that feedback loop, that customer data flywheel back into the model POS training, right? With reinforced Merling and create version one and version 1.1 and version 2 and version 3. And when you're at version 3, anybody who's trying to play catch up and they're still in version 1 are not gonna have the same performance and they're not gonna be able to sell their their you know their wares to to the next customer, right? And that's the That could, you know, that is a moat. And there's other types where we can keep going, but I think it is important for founders more than ever to be thoughtful about, okay, what is my differentiation? What is my moat in this market? Jason Kirby (13:23.4) And so one thing I wanna talk about, you know, literally just got off a call with a founder who, you know, gives me the big pitch, so proud of what he's building, and then like drops the valuation on what they're trying to raise, you know, capital on. I'm like, Good luck. You know, what are you seeing in your you know, in your world right now when it comes to deal making and looking at deals where you're sitting there and like this valuation feels right. Like what are you kind of judging a business based on and a founder based on when it comes to determining these kind of like pre seed seed stage series A valuations? Itamar - Recursive Ventures (13:58.786) So let me start by dropping a a a bombshell in most people because most people think that AI is in a bubble and valuations are off the hook. And they're absolutely right. But they're absolutely right when it comes to the infrastructure. So LLMs are Jason Kirby (14:10.631) Yes. Itamar - Recursive Ventures (14:13.066) In a bubble. If you ask me, I don't believe in the entropic and open AI valuation. I can explain why. I don't believe in it, right? But actually at the application layer, the valuations, for example, for for Recursive Ventures 3, which is the fund that we run today, our average entry point for po post money valuation for you know application layer pre seed companies is twelve point eight million dollar post money. That is actually pretty similar to what we saw in 2020, 2021, before prices went a little bit crazy in 2022 and then Crashed. Like it it's not, and and the VC model actually works with these type of valuations. At seed, you know, we're seeing anywhere between 20 to 40. It really depends on how far the company is. But another characteristics which founders should be aware of is this kind of ha world of have and have nots, right? Like if you're doing if if you have a moat and you have subject matter expertise and you're showing like explosive growth, you're gonna get funded. You're gonna get funded really, really quickly, and you're gonna be very happy with the valuation that you that that that you're gonna get. And but if you're not going from zero to a million in year one and then million to five in year two, which is kind of the high mark these days, right? Then that's the have and have nots. You're gonna be in the have not camp, right? And I'm not sure you're even gonna get funded, right? Not to mention the valuation might not be one that you're gonna get you're gonna be happy with. So at seed, we see this range that's more like 20 to 40, where the good companies are getting, you know, 30, 40, 50 million dollar plus money valuation. right, which I think is a little bit high, and that leads us to the conclusion that the best place for an app AI application there VC to be playing in today is the pre-seed stage. Jason Kirby (15:57.214) I think it's validated from that perspective because you have to kinda get in you have to take that gamble that they are the ones that are gonna get that explosive growth and you get the reward. and when they do, 'cause you get that pretty high, like substantial high markup, in a relatively short period of time. But like let's talk about those founders in the kind of the have not category, 'cause there's a lot more of those than the have the haves. category. I I wanna and I wanna do this like hopefully founders can hear and understand this because like I very much respect your you know approach, your opinion and the deals that you do. And I think you're also very pragmatic and a good sample of what's actually happening. And so I think your opinion matters here. What like for the have nots, like what are these benchmarks that they think they might have been hitting? Maybe they read some blog a couple years ago or something like that that was saying you need to raise at this rate. and all this kind of stuff. But like what's actually happening right now between what's really separating these haves and haves nots when it comes to like specifics of like you meet the boundary like okay but there's this other deal. I'm gonna do that one instead. Itamar - Recursive Ventures (17:02.38) Yeah. So so we've obviously talked about it, so I'm not gonna harp on it more, but velocity, velocity of you know, customer outreach and engagement, velocity of revenue, right? and so that's one. The second thing which is also not new, but I think has been sort of like sharpened here is like Are you on the path to being a category leader? Right? So there's something about AI. I'm this is a theory. I'm and and many might not agree with me, but I think in the age of AI, we're gonna have much more clarity about who what is the best piece of AI or what is the best piece of software. We're gonna have AI evaluating AI, and if you as a customer, let's say you're a buyer of AI system whatever, right? you'd be able to pretty quickly know, okay, these guys have the best agent or the best, you know, harness for the agents that use my models that they already use, right? And it's gonna be unlike you know vendor selection in the past that used to like, you know, okay, touchy feeling, yeah, I like the UX, I like the guys, whatever, this is the the the parity and features. Now you're gonna be able to say, okay, this AI is 99%. Accurate and this AI is 98% accurate. Which one are you gonna buy? Right? So I think that's gonna make decisions or it would make things more obvious. And like who is really the potential winner in a category? And the winner could be a mega winner. Like they can be reinforced to the point that they're kind of massive. We're gonna have potentially less winners, but the winners in this vertical are gonna be massive, right? So you kind of, you know. wanna be able to say, okay, I'm I'm I'm the winner in this vertical, right? I'm the potential winner in this vertical. That's really, really important. And if you do, then you're gonna get funded, you're gonna get amazing you know sort of valuation in in terms for your your funding. The the the the third thing which I think is important and this is again sort of like my opinion and some people don't agree with me. And I got kicked out of Y Combinator because of that, right? Yeah. let's get to the Jason Kirby (19:03.793) Do do share what got you kicked out of Y Combinator? Itamar - Recursive Ventures (19:07.18) So I've written a series of posts about how Y Combinator now is not great for founders because they only care about the one c one winner in their batch, and the 199 other companies are basically screwed because Y Combinator pushed them to a $30 million valuation when they have nothing. and you know, it's just like it's really easy, it's really hard to land off that valuation, right? If you don't get the $5 million of AR in year one. but putting that aside, I think the whole white combinator model is actually under significant threat because this new generation of entrepreneurs is different than the previous generation of entrepreneurs. And what am I what or like the ones that I think are going to be successful. And what I'm what what am I trying to say here is basically Y Combinator has always gone after these kind of 21 year olds and call it dropouts who don't sleep. Don't have kids at home and just code, code, code, right? And they they build software the fastest because they're just around the clock, they're hacking, right? And they're using newer tools and they build better software. So really the competitive advantage was just you know building product as quickly as possible. Now with vibe coding, that barrier is gone. You can build product 10x faster. With half or less the people and anybody can build. So being a kid that quote unquote kid that builds really fast and doesn't sleep at night is no longer a huge competitive advantage, right? What is a competitive advantage in this generation is deep understanding of your customer needs and how their lives could be changed with generative AI. So it's really about domain expertise, understanding the customer. understanding how to rewrite these workflows that they've had before with that they have today with AI in a way that really makes a tangible impact and real ROI. That's an issue that we have with with with AI today. It's like we're lacking ROI yet, right? And I think the people are gonna crack that are not going to be the 20-year-olds that don't sleep at night. It's gonna be the folks that again have been in the industry, have the network, are able to sell, are able to really Itamar - Recursive Ventures (21:18.144) understand the customer at profound levels and create new AI paradigms for them. Jason Kirby (21:23.049) Yeah, I I I appreciate this perspective on YC. I feel like I can counter you on on one piece. I think the time allocation of young people with no kids is still beneficial from the fact that they can chase more customers and like, you know, apply more time to the business than say me who builds an app while at my daughter's friend's birthday party. Well, cool that I could build an app on my phone while like, you know, texting it. It's awesome. But I'm not like hustling like at you know, tw I'm like gonna wrap up this podcast, go spend time with my kids, like, you know, put them to bed. Like there are certain weaknesses I have in terms of time allocation, 'cause I have kids in a family and whatnot. So the twenty two year old in YC spending four months out there will definitely outpace me if I were going head to head from the same starting point. but I have experience, I've been through you know, I might be able to leap through them on other things. Not that I'm Itamar - Recursive Ventures (22:18.184) Absolutely. It's just that my point is that Advantage is getting eroded. It's still there, but it's not what it used to be. Jason Kirby (22:24.477) But I do think you bring up really good point on the thirty million dollar valuation stuff. And I think this is something that founders really have to understand. I'm I'm glad you kind of pinned the kind of twelve million dollar valuation point that you're seeing, which again, for pre seed, but you're chasing like traditional VC structured deals, like they have to have, you know, fund returning potential and you're comfortable coming in at that 12 million dollar price point. What and you know, and you're chasing velocity, you're chasing things that like checks these check these traditional VC boxes. I still see a lot of founders that are like, you know, not hitting those velocity marks and demanding twenty, thirty million dollar valuations because that's what they see online or, you know, especially with YC where they get these massive valuations and they have an edge because they have that network. They basically YC sells to YC. And so you get like an instant early, you know, early adoption of customers, which you know is a competitive advantage to some degree, but what's that churn look like? You know, it's always a question. And you know you sign a one-year contract and do they renew? And that's where these you don't renew, you know, or you don't get the renewals, then those companies start to, you know, fall apart and the valuation is no longer it could be a good business. Could still be very reasonable, but you know, they built the business to go raise a bunch of venture money and they can't now. So I think there's definitely gonna be a lot of you know aftermath in that regard. Itamar - Recursive Ventures (23:37.72) Look, startups are extremely risky to begin with. Like it's a very, very high risk game. When you have nothing seriously going for you yet in the business, and that's okay. That's where we all get started. And you go to YC and you raise on a $30 million post-money safe, right? You're basically significantly increasing your risk on something that's already inherently crazy risky. Because 99% of companies are just not gonna hit out of park day one. The the journey looks looks like this, and sometimes it takes three years to figure it out. So by setting yourself up with the rights housing, your valuation essentially for your business is, you're actually decreasing risk. And what I would say is for a lot of founders, the name of the game is systematically de-risking things in an already very risky environment to the point I I'm not saying don't take risks. You should absolutely take risks. Take very well-measured pointed risks and go in on them while decreasing risk everywhere else, right? So you maximize your ability to take that one risk that you really need to take and prove that one thing that your company is really out there trying to prove, right? So by again, by g getting these really high you know beginning valuations, look, let's let's just put it out there. For most of these YC companies, when they come to demo day with their $30 million save, they have Nothing, right? They have like a few design partners, they're willing to pay them a little bit, they build something, it's a demo, like it's that level, right? So i you're really like if you want to build this company and you wanna maybe give it a a a real shot, you're really shooting yourselves in the foot by taking that 30 million dollar money. And it it sounds so good and yeah, you can look at your now I'm worth five million bucks as a founder because of this violation, but it's not real money. It's not there. It's purely speculative and at your at your expense. Jason Kirby (25:29.598) So could it? Could it be gamified though? And this is something I've talked to founders about and you know, calling it the one and done round, depending on like how what the business is really doing. Could they get enough capital and leverage AI to be lean and kind of be like these, you know, one to five person teams that can get to meaningful revenue? And since it's so safe, they don't really like the investors don't really have the protections of say a price round or like board seats and things of that sort to where they take a little bit of money in and knowing that they're not gonna be worth thirty million, or maybe they they find out later, but essentially just not converting them and running a business and cash flowing. Like what's your take on some of that? Itamar - Recursive Ventures (26:11.212) Yeah, yeah, yeah. We we call that seat strapping and it's an increasingly interesting concept like bootstrapping, but seat strapping is this concept of wow, you only take this one first round, let a million or two or three, whatever you need, and then you're off to the races and you get cash flow positive, you're a profitable business and you never have to pay the piper or the VC again, right? that that is that is it's it's called seat strapping. I I I think yeah you can absolutely do that off YC or anywhere else. What I would say is that first financing round and it it's the terms of that financing round are really not consequential for that. The real question is, can you take a little bit of money and with a very lean and mean team get, you know, the break-even, right? And cash flow positive from there, right? And then keep hyper growing with you know with profits, right? Like investing The business from profits. That is the key question there. Now, whether you raise one on 20 or two on 30, or it actually doesn't take that much capital with today's tools and all the vibe coding and speed on that you can have to figure out whether you can build that sort of seat strapping business. I think the only capital you need is the first one million dollar check from somebody like me to quit your job and hire two people, and then you can really quickly know if this is on the right track to do a to build a seat strap business. And we're the first. To jump all in with blazing guns and say, amazing. Like we would love to back businesses where we're the last check-in. the founder goes all the way right to the IPO, and we don't get diluted by subsequent funding rounds with bigger VCs. That's music to our ears. We're anyway, as a precinct investor, the most aligned with the founder, and we would love to keep that all the way throughout the journey. So again, for us it's ideal. and there's a lot of talk about the about that out you know out there, but it's not fully happening yet. I think it's just starting the seed strapping movement. Jason Kirby (28:09.479) Well, it's too early to kinda tell what materializes, but you know, definitely run into some of these Y C companies out of like the twenty twenty one era that Like now what? What do I do? Like, you know, but I think the seedstrap concept is kinda hot and fresh. Like, I'm curious about I don't know if you heard about pulse eye AI slot backwards, the thirty million dollar raise. like what what I imagine you heard about it. Like what what's your take on a deal like that where it's like solo founder kind of building this, you know, AI agent business. I checked it out. It was like cool at first, like but then I like, I don't know if I'd really chat with the model. But what's your take on that? Itamar - Recursive Ventures (28:44.268) Look I and by the way, I this is polse has great my my friends, they've and Britt Morin, who's Britt Morin is on our board at Life Three Sixty at Offline Ventures, they're great. They're I think the biggest found founders of that thing. I I think the trend and the vibe is absolutely right. Like it's amazing, I think it's gonna happen. I also think it's a little bit too early for that. Like the technology hasn't matured to that extent yet. it does take more people to figure out and what they're trying to do is like a almost like an entire operating system for businesses. I think that takes a lot, in in in in in today's tools. and I think like maybe in in AGI that would be possible, but we're not there today. So as a bet today to what the world could look ten years from now, I think it's a really decent bet. It's a great bet, but But I don't think we're there today. Jason Kirby (29:39.101) You think it was right sized. Like where do you think they came up with like the thirty million dollar number and why was that justified? Itamar - Recursive Ventures (29:46.122) I I don't I don't want to comment. Like I I I know some of the details. But but but look, just maybe maybe one kind of indirect comment. Venture has been successful in the past in sort of crowning winners, right? By injecting a lot of capital and building, you know, a movement and hype around an opportunity. It is one successful recipe to win in venture capital. And I think maybe that's what's happening there. Jason Kirby (29:48.369) Okay, fair enough. You got people involved. Fair enough. Jason Kirby (30:16.168) Got it. No, I I I believe that. That's where it's just like, let's crown the winner now, get it early, give them the right amount of runway to make mistakes and just like throw money at all the problems, which I always find interesting. But then I see, you know, fifty to a hundred other companies again with a similar promise, you know, out there all yeah. which has existed you know, SAS has been doing it for, you know, two decades and like, you know, now they're trying to re you know, bring it into the AI world, but not to this level. Itamar - Recursive Ventures (30:34.358) Yeah. Business in a box we call it. Business in the box. Itamar - Recursive Ventures (30:44.438) I think it's a little bit different because SAS is more like a point solution, either a vertical or horizontal. And now folks are saying, forget about all that stuff. We have one AI that does everything for you, for your business. So I it is it is different than SAS. but at the same time it's kind of, you know, very early to tell if that is really a possibility. Jason Kirby (31:06.954) So when you look at the market now, like you know, my my opinion is that we have a very much barbell distribution. Like there's gonna be these like micro businesses that either give individuals or small teams a great lifestyle. Like whether that's, you know, millionaire, half a million, two million, five million, wherever that range is, and then there's just gonna be this desert Of you know, people maybe trying to cross the chasm, but ultimately then you're gonna have these mega companies where it's all concentrated, where all the talent's concentrated, all the capital's concentrated, and kind of the medium enterprise, middle market, you know, starts to consolidate into either the conglomerates or fade away. Like what's your take on something like that? Do you completely disagree? What what's your take on Itamar - Recursive Ventures (31:51.596) I think it is l look let's start with the basics. Again, I think the pie is gonna be so much bigger, right? And there are completely new areas, like even the deal that we just did together. Again, I can't share the details yet, but it's it's you know, it's in a space that has been almost completely untouched by technology previously, right? There's been almost no technology in that space. Maybe we can just say the space, it's you know, it's Jason Kirby (32:05.746) The port. Jason Kirby (32:19.732) Cattle industry. Itamar - Recursive Ventures (32:20.812) Cattle, cattle, right? Like have you seen any technology in cattle? I mean, there's been a little bit, but yeah, cattle, right? so AI is really expanding the pie, not in terms not just in terms of value capture, but also in terms of the areas that you can go. So I think that that's kind of one piece. And then and then the second piece is, yeah, as I mentioned, I think there's gonna be one or two winners in which one of those verticals, there's just many more verticals, right? So it there will be a concentration, but it's a concentration in its vertical. And and and I think it's true, right? And then there's gonna be a lot of lifestyle businesses. So for venture capital, that means what we have to do and what we're actively doing is actually making more deals, not less deals per fund, right? With the understanding that the VC power law is actually gonna be on steroids in this cycle, right? That if we traditionally have been modeling that a VC winner is a hundred X, maybe now we can model a two hundred X. Jason Kirby (33:09.662) Yeah. Itamar - Recursive Ventures (33:20.33) Right for a winner, right? And that changes the fund dynamics completely when you actually look at that spreadsheet, right? So I think I think that's one thing for VCs. What it means for founders, which I'm guessing most of the people listening to this pod are are are founders, is I think you have to have a much more precise and nuanced understanding of whether what you're building is is really VC scale, right? Is it really on that path to be the category winner or not? Right? Does it have the TAM, the moat, the team? The characteristics to really try to be that one winner takes it all, right? Because that's how it's gonna look at the age of AI. And if that's not the case, then don't take VC money to begin with. Like, don't even talk to me because VCs are gonna put you on the wrong path, right? Inherently, that's what we do. If you are actually not going after a VC scale, VC backable business, and you take VC money, you're screwing up your business, right? Because VCs want that multi-billion dollar exit. And actually, what you have in hand, amazing for you as a founder. Is a cash cow that puts two million dollars of pure profit in your pocket every year. That's awesome. It's just not a VC business. Jason Kirby (34:27.316) So yeah, this is I'm so glad you said this because this is a conversation I have all the time where it's like you have a good business on your hands. If you don't take venture like, well, I want money to do this, this, this is like struggle a little bit more and just get it done without the VC money and you will reap the rewards. Whereas you you will take the risk because so many people change their narrative of what would have been a good business to what they think a VC would want to hear, and that's just like dead in the water. I just feel like, you know, maybe you might like maybe you might get that one or two checks, but then you're gonna get it from like the VC that's gonna put all these ratchets or, you know, controls on you and it's just not gonna be in the best interest for for you. And I just feel like that happens with Itamar - Recursive Ventures (35:08.262) Misalign with the VCs. Misalign with the VCs. This whole VC game is about giving away a piece of your pie to get a much, much bigger pie, right? And if that much bigger pie is not gonna show up, your VCs are gonna be disappointed, to say the least, right? And that that is you shouldn't Get in bed with people that are not inherently aligned, their needs are different than your needs as a partner, right? So that's why, you know, I I push really hard on founders. It's like, why are you even raising VC money for this business? I'm not sure that's the right thing for you. And it's the worst advice I could give because I get compensated by deploying capital and finding winners, right? Is I want to do deals, right? But I don't want to do the wrong deals that for and I don't want founders to be misled by other VCs on whether they should or shouldn't take VC. Jason Kirby (35:59.403) So I'm gonna kinda switch gears. Let's flip the script a little bit. Let's put the VC hat on. You made the bet. The bet didn't it it's clearly not gonna be a fun returner, but it's not a tot dog, it's not like a total write-off. What do you do? Itamar - Recursive Ventures (36:14.168) Well, look, the honest brutal reality is I'm a venture capitalist at the precede stage and we we're really focused on the big winners, right? So What do I really do? Look, I'm always there for my founders, but obviously if it's not in the VC path, then it's very hard to consider additional funding because that's not really our model. That's again, don't take VC money if you don't want to live in the VC model. That's that's what we do, right? so it's gonna be hard to inject more capital. And then I really care about the founders and their own personal success. So I would encourage them to, you know, again, get the business to cash flow break even, make it profitable, build a sustainable business. for the long term and yeah, it's not a fundra returning return for me, but You know, if it's a great business for the founder, then at least that's some comfort, right? For us as as VCs. and you know, maybe you never know, maybe five years from now things change and suddenly it's it's it's a very valuable business and maybe it gets sold for a few hundreds of millions of dollars without raising significant amount of capital and we still make some money off it. So that that is that is a good outcome. in terms of what that means in practicality, look, I'm I'm a little bit cynical about the role of VCs in building companies. I think it's for founders to build companies, it's for founding teams. To build companies. VCs are really there to finance companies and and kind of help with strategic guidance. So I I just focus on that, right? I'm not I'm not gonna sit here and pretend that we recruit all your engineers for you and then we find all your customers. You don't even have to sell them, we just like bring them over and like sign the contract. I think that's that's BS, maybe with some exceptions of mega funds that have hundreds of people, hundreds of people running around helping their companies. Most VCs don't really do that. And as a founder, I would be skeptical when a VC. Jason Kirby (37:44.531) Yeah. Itamar - Recursive Ventures (38:02.077) Tells me that they'll do all that for you because in reality a lot of those promises end up being you know extreme. Jason Kirby (38:08.328) Yeah. Yeah, I could totally resonate with that 'cause their version of like platform is like they have one person kind of just drinking through a fire hose, getting like all the company updates and trying to be like, maybe we can do an update for you and like, you know, maybe make a connection there and then like the founders are like, Well, what about this? You were supposed to help me with this. It's like if you relied on the platform team at a VC firm, that was your first mistake. Itamar - Recursive Ventures (38:35.534) I think I think founders are like experienced founders are kinda disillusioned from that, but maybe this is important an important piece for less experienced founders where the VCs come and say, our platform is like yeah, come on. Jason Kirby (38:48.778) Yeah. Platform's not a differentiator in so many ways. Like, sure, you know, if you're connecting with the other founders and being able to have access to the other people that they've funded, and just like as a sounding board and building relationships and a network, that could be valuable. But like the fact that you're gonna run sales or you know, do those types of things is often not the case. Like you come to VCs to get the capital and blitz the market as fast as possible, and hopefully help them have them help you see around a corner that's not your core focus, like how the capital markets are changing, or you know, that those kind of things could be super valuable. But Itamar - Recursive Ventures (39:20.618) the most important thing, and actually I served founders about this, what is the number of value add? And that is helping you raise your next round. And that's what good VCs are are focused on. So not trying to, you know, Yeah. I mean we've helped over fifty founders raise over five million million dollars. So you need you need VCs, especially the pre seed that are a very strong signal where if they're like I mean, if we're backing with recursive ventures Jason Kirby (39:30.398) Go ahead, say it. Itamar - Recursive Ventures (39:44.672) A lot of Series A VCs on Sandy will take notice, right? They will be like, okay, this is this has been vetted, there's a signal, we know it amar, the recursive team, they know what they're doing. And that that is actually where the real value lies today with VCs at the early stage and this platform stuff is is there's a lot of smoke and mirrors there. Jason Kirby (40:02.696) Yeah. No, I I I I concur on that front. So When you see these VCs, you know, or you look at your own portfolio, you see other VC portfolios, this kind of mid portfolio, you know, concept. You know, the focus from what I'm hearing here is like, you know, a founder comes to you, the you know, it might be offered support, but it's not necessarily the focus at that point. Most VCs, when they give that advice, like, well, go get cash flow positive, and you know that's usually like we're writing you off. Good luck, but we know you're not gonna return the fund as far as signal. Like, would you typically agree with that's the interpretation a founder should have, or do you think it's Yeah, they should look at it differently. Itamar - Recursive Ventures (40:39.595) So Actually it's interesting. Again, it really depends on pre seat VCs versus series A V C. So like series A V Cs they they would go in, they would underwrite the thing, they would write millions of dollars and then go sit on the board, right? And that's actually where complications and potential conflicts arise. So we don't take board seats at recursive ventures. So we're not really pretending to say that we have an active role in steering the company, right? and I think if your business as a founder ends up being more of that, you know, sort of normal business. not a VC backable business. It's actually better that you don't have a board with VCs on, right? Because it's you should really go and run your business. You know what you do best. No VCs know how to run your business better than you, right? We we shouldn't. That's not our job. Right? And then you have a board member that's misaligned because they want this big exit or they want this thing, right? And they don't want to spend time on it anymore because they want to walk away and focus on their winner, right? so that that's that's kind of You know why I think it's very different with pre seed VCs? It's safer to take our money if you end up not going down that kind of VC backable route than it is when you take the big money from the big VCs, you're kind of boom or bust, right? And I think that's very dangerous. Jason Kirby (41:57.323) So let's change gears with our you know a couple minutes here. One thing I wanna also hear your perspective. You know, you're on the boot boots in the ground in San Francisco. Like how fast are things moving there when it comes to talent building? Like what's kind of the latest technology you're seeing be kind of shared and talked about? Like what what's kind of going on in the backdoor conversations of the private dinners, the private events happening out in San Francisco right now? Itamar - Recursive Ventures (42:23.35) It's interesting because I think there's this mismatch. On one hand in San Francisco, things are going explosively fast, like fastest they've ever been, but it's kind of a a cyclical movement. And then when you get out of SF and you go to industries, you go the mainstream, you're kind of Wondering why the rest of the world is not really there, right? And you see this complete mismatch between, gosh, when was it like Q3 or Q4 of last year when another report got published saying that over 90% of enterprise POCs with generative AI are failing. Over 90%. This is this was, you know, typically it's 50-60% of POC fail, POCs fail. Now with generative AI, we're seeing 90% of POC fails, and we're seeing many public company CEOs saying, you know what? But this AI thing hasn't really hit my bottom line yet. I haven't seen you know significant ROI from those investments. And on the flip side, San Francisco is spending a trillion dollars a year on infrastructure buildup. So so I think it's almost cyclical now in SF. Again, I think that is potential indications of a bubble. And the bubble is not that AI is not. Valuable. No, it it is the most valuable thing that we have created in the last few decades, I believe. It's just it's more like a timing mishap, right? It's timing and it's understanding of where the value is actually gonna be accrued, which is again more at the application level, and that's in the first inning, right? so going back to what you're saying, there's a lot happening in San Francisco. I think maybe the biggest single thread that a lot of people are picking up now, and and I think it's real, we're doing it ourselves in in in recursive ventures where we have identified our entire fund. And I can show you what we've built, but we're like really like I have an AI associate now and like a lot of really cool things. is kind of recursive well we run recursive ventures so it's very coincidental and positive that it's now a big thing. Recursive learning, right? Where AI is basically teaching itself in you know in this kind of Itamar - Recursive Ventures (44:35.778) Flywheels, right? And it is increasingly getting better. And there's there's many different ways of doing that. One of them we've talked we touched on briefly, which is reinforcement learning, right? Learning from your customers, learning from their data, and then continuously improving what your agent does, what your LM does. That's kind of one thing. A lot of people are doing it even without training, just with MD files, or with memory files, right? where it's like, okay, you know, I'll I'll just give you an example, right? So, as you know, I I do a lot of I basically build the media company on top of my VC, right? Content on LinkedIn and other places. And even though you know I I edit and and every single thing and I post it myself, I still use a lot of AI to create these posts to begin with, and then I start editing and writing them, right? So for example, in my world, I have this system that every day picks up the five most trendy news items in AI and VC. That's out there, right? And for each one of them, the AI creates a suggested post of what Ilmar would say based on everything I ever wrote, but not just everything I ever wrote, but also the actual data and engagement, right? Like what is resonating, likes, shares, po comments, so on and so forth. Wait, this is just the beginning. Then I actually you know edit these posts and work through them and then eventually editing you know eventually posting them. And the AI learns from each post which posts are performing well from an engagement standpoint and which are not, and then creating suggestions on how to better style and better write the next posts to get even more engagement. That is a classic example of a recursive learning loop where the AI is basically producing stuff and then there's data on the other side and it's continuously, you know, enhancing, improving what it does in its production to to reach your goals. So that's probably the single biggest threat in Silicon Valley right now right now. Jason Kirby (46:37.427) Is that recursive loop where people are coming 'cause like I'll be honest, that's one of the things I've struggled with is like that extra step. It's like one thing to kinda get that initial push of like, here's like the article or here's the content or here's that that one bespoke result that you're asking for, but then it's like how it's like you're not getting smarter. You know, it's like, I'm having to keep going back and asking for the recursion. Itamar - Recursive Ventures (46:56.824) need data to get smarter. So if if you're only pushing AI software, let's call it out there, and you don't really have the data and efficacy, right? Then you can't improve. You have to have that that kind of flywheel, right? and I think that's one area that some startups are missing. Well newly minted startups here in Silicon Valley, they're all on it. Like they all get it. Like that's our way of becoming the best in what we do, basically. Jason Kirby (47:26.251) So that's that's kind of the edge you're seeing today in the newest upstart of companies is you know really building that feedback loop, building that recursive loop where they're constantly like the human doesn't have to sit there and say you missed this. It's it knows it missed it, and then it fills the gap. and I'm starting to see that with the models too. They're they're inherently trying to get better at it. The problem is if you can't interpret the output effectively. Yeah, and you're just hoping that it's doing it right. it turns into a bit of a black box or something. Itamar - Recursive Ventures (47:52.214) Yeah. The the models have been doing this for a while. The top two reasons the LMs have gotten so good in the last two years is one Expert training, expert post training, which is Mercur invisible like all these guys. So that's one way of cracking. And then the other one is reinforcement learning, is just seeing, getting feedback from users and is this thing working or not, right? That those are the top two main reasons why LMs have been progressing so nicely. With that said, I do want to make a bold claim, put it out there that I said, you know, and it's not people have been talking about it, that I think the progress is plateauing with LMs. They're not then it's incremental now. Jason Kirby (48:33.897) Yeah. Yeah, where's AGI guys? Come on. I was promised AGI and it's not here yet. so fun little question I want to ask you, if you can answer in as short as possible. What is your AI stack at Recursive Ventures? Itamar - Recursive Ventures (48:51.094) wow, it's all homegrown. So I mean I I I would share it with you, but like basically, One thing that I figured out is that VC is sort of, at least for me, is more like a lone wolf kinda thing. And instead of, you know, getting four I'm gonna p four other partners and five other associates, I can just start clothing myself with AI. And that's exactly what what what what I've been doing and what we've been doing as a team is basically, you know, build a AI chief of staff that knows everything about a business. And then we built Jack is my AI associate. He writes amazing investment memos, he prepares due diligence for every single call and every single meeting. then on the content. Jason Kirby (49:10.611) Yeah. Itamar - Recursive Ventures (49:27.344) creation side obviously you know I give examples of of what we're doing and and and they are they is kind of overseeing and and and kind of you know Playing around, really doing very all the repetitive jobs, you know, copy this deck, put this in Notion, put a this this kind of note in R CRM, track this company, like all that stuff is now being completely automated with AI. And it's all all homegrown, and we build it ourselves, and it is easier than ever to build, right? And that's a huge advantage also for VCs, which run small businesses that need this level of automation. Jason Kirby (50:00.832) What are you building it on? Itamar - Recursive Ventures (50:02.922) mostly cloud code. and then we again we have our own proprietary system that we coded from the ground up with our skills and our capabilities and it's all running in the cloud. But the engineering itself is mostly nine cloud code. Jason Kirby (50:19.507) Yes. So completely bespoke system designed specifically for you. I've been talking actually you're not the only VC I've talked to. A couple other VCs like you're not soldo GPs but a couple you know, a little bit bigger teams, like Billy's full and I've been I've been interacting with them, which is kinda funny, where it's like you know, Enterplay Adventures is one of They have Ivy, which is complete AI, that sends me an email every day with those like certain updates that are related to me, intros that I could potent you know, could potentially make for me, these different like kind of cool features and functions. And it's just one of the partners who just gave up on excuse me, gave up on like the day to day and just went all in on AI, like doing like twelve hour days. Itamar - Recursive Ventures (50:54.37) I see. Jason Kirby (50:59.667) You know, building this stuff out, which I think is absolutely fascinating. The first couple iterations suck, and now it's starting to like, wow, this is this is very relevant and feels very natural. so it's pretty wild how we're seeing this all come to life. some parting advice for for the AI founder who is, you know, looking to be that venture breakout. You know, what's a what's the final parting advice you can give out to them right now that are in that stage and they're on the grind trying to build the next big thing? Itamar - Recursive Ventures (51:29.166) There's never been a better time to build. There's never been a better time to go big. Now is the time, especially at application layer. We're still in the first inning. So this is an amazing time to start a company. While you're doing that, it's very important to pay notice to a couple of important things. First off, is this venture scale? We talked about it. If it's not, just don't go the wrong path. It's just gonna destroy your company. One. Second thing is it's more important than ever to be differentiated and have a mode, like we talked about. If you don't have that, then It's gonna be impossible to be venture scale, but also you should be aware of your business getting commoditized, even if it's not a venture scale business. So that's that's that's a very important thing. And the the the third one is this kind of barrier that we talked about that no longer existed, right? So I no longer exists. So I've not written a single line of quote for 20 years. I've done a lot of product management roles, right? And that that gives me a leg up, but like anybody can build. So what really matters is that subject matter expertise, what we talked about. And if you have that and you have a clear vision of how your industry can be transformed by AI, I think you're a very well positioned, right? to to win, right? the last maybe thing I would say is like Almost like word of warning, I think because we're in a bubble, you have to take into account that kind of one way about is your tokens are being subsidized by the bubble, right? Like at least in the short term, I think in the long term inference costs are gonna go down, you know, significantly. But in the short term, if the bubble bursts, you could be paying three X for the same tokens because they're not being subsidized by OpenAI and Entropic, right? so when you're thinking about the economics in your business, you're thinking about the operating market. Have a plan B of what happens if your token cost you know goes up 3x, right? have a backup plan if that happens, because we've seen bubbles come and go and we've seen bubbles burst, and this one is probably not necessarily different. Jason Kirby (53:34.399) Very, very good advice. And you know, this brought up one question. Would you back a company today without a CTO? Itamar - Recursive Ventures (53:45.395) potentially, but it really depends on where and and how. So For the enterprise, probably not, unless the CEO is very strong technically, because you know, with between hallucinations and drifts and proprietary data and building systems right and flexibility of architecture and routing and all these issues that are not top of mind for kind of deeper, more complex enterprise products, I think you have to have somebody who's cutting edge in AI and development, right? for B2C companies, maybe. I think it's possible. you know, obviously as you scale and you want to serve tens or hundreds of millions of consumers, you're gonna have to figure out how to architect it right. Those are not easy things to solve. So having a CTO from the get-go is better, but it might not be a necessity. Jason Kirby (54:33.067) Hey tomorrow. Thank you so much for coming back on the show, sharing your insights. Absolutely loved you having back on the show and would love to do this again. So we'll leave the links down below for people of how to find you and learn more about you, and potentially reach out to you if they think they're a good fit for you. So thanks for coming out. Itamar - Recursive Ventures (54:48.994) Follow me on LinkedIn. I write I write daily posts to help founders in the AI era. And I hope you'll find them interesting and helpful for you. So just look me up on Ilamar Novic on I'm the only Ilamar Novics. Just look me up on LinkedIn and follow me and engage as much as you can. Jason Kirby (55:06.068) Awesome. Well, thanks for coming on. I appreciate you. And we'll be in touch soon. Cheers. All right, let me stop the recording. Itamar - Recursive Ventures (55:10.712) Thanks, Jason.