Building the Berkshire Hathaway of AI Services | Brennan Pothetes, CEO of Infinity Constellation

Brennan Pothetes is the CEO of Infinity Constellation, an AI-native holding company. They've raised $17M from Freestyle Capital, Charlie Songhurst, and others. 

Brennan's favorite books: Zero to One (Author: Peter Thiel)

(00:01) Lightning Bolt Moment – Origin of Infinity Constellation
(03:48) Transitioning from Founder to Holdco CEO
(07:00) How Infinity Incubates vs Acquires Companies
(09:05) Infinity vs Traditional Venture Studios
(10:58) Aligning Incentives: Founders, Holdco, and Investors
(16:32) Target Markets and How They’re Chosen
(20:27) When a Company Graduates from the Holdco
(23:04) Resource Allocation and “Too Many Toddlers” Problem
(26:29) Shared Infrastructure and the Code Commons
(29:48) Model Choices, Full Stack AI, and the Infinity Playbook
(32:05) First 3 Hires at a Portfolio Company
(33:53) Who is the Real Competition?
(37:03) Infinity’s 5-Year Vision and AI Trends
(38:37) Rapid Fire Round

Transcript

Prateek Joshi (00:01.557)
Brennan, thank you so much for joining me today.

Brennan (00:05.048)
Thanks for having me.

Prateek Joshi (00:06.941)
Let's start with the lightning bolt moment of Infinity Constellation. So what did you see in the world? What problem did you see in the world that convinced you to launch the company?

Brennan (00:22.318)
Well, I think it's sort of like it came from a couple of different angles, which I think is pretty interesting. So Francis Pedraza is the founder and chairman of Invisible, which is one of the world's largest AI data training companies. I was actually their first enterprise customer many years ago. And then he sort of was starting to encounter the innovators dilemma at Invisible because it had gotten so big, but he saw all these like blue ocean AI service companies that could be started like Invisible that

they weren't able to all get after, right? Because like a big company has to stay really focused on what they do and what they do is, you know, some of the best AI training in the game. And, and basically at the same time, I had just built an insurance company called Butter that I had sold. And, you know, I was looking at building AI to automate a lot of like humans out of financial services, because all the biggest problems that I found in financial services through my whole career.

was because we hired loads and loads of people. So I generally believe that you can increase GDP and increase better products by having more efficient humans in the process. And so I had sort of looked at this being like, I really want to build some AI service companies. And Francis was like, I want to find a way to do this. And so we really got together and created Infinity Constellation as a hold-co for applied AI companies delivering full stack services. And what we found out from both our journeys was that

Um, if you think of like a lot of SaaS companies, they're trying to make you be a better baker. So they're like, here's a mixer to make you make a better cake. Go make that cake. Um, we're like, don't worry about being a baker. Here's a cake. And I think that that works really well in AI because a lot of people want to use AI, but these models are improving really rapidly. So it's like a new innovation every month. And so even people that are in the business, it's really hard to stay on top of like, what's the best tool for the best thing? How do you build the best workflow?

And so, you know, I usually have people, anecdotally throw out to me, well, like AI hallucinates or like Sam Altman says AI always hallucinates. I'm like, yeah, but like your employees always get things wrong too. You don't just stop like hiring people because of that. And like this tech is very impactful and like literally month over month, it's getting better and better and better. And people are investing trillions and making it better. So like, why would you not use it in services? And so all that sort of like culminated with Francis and I creating infinity constellation.

Brennan (02:47.054)
to really just build like a trillion dollar hold co that supports AI service companies. And so we have companies like radiance, which is an AI brand design agency. And yeah, we use a lot of AI and the founder Jacqueline's incredible. But they, have really great designers and end of the day as a customer, you just want really good brand design or we have like zero hiring, which is our our AI RPO. So it's a high velocity recruiting as a service and

They basically in the industry, they normally take 75 days to find and place a candidate at high volume. They do it in 22 because they're using AI workflows, but end of the day, we're very like outcome focused with all of our customers across all of our products. So like we have eight other companies like that. and we've gone zero to, above, above 7 million and under 12 months of revenue. And so we're very highly efficient. growing really fast. And so I think that's like, really, we're now at a point where we've proven our whole co model is really working for AI services. So that's like.

That's why that's a bit of the story of what we've built and what we're up to.

Prateek Joshi (03:48.353)
Wonderful. That's a great introduction. going outside the company for a second and just the mechanics of running a hold code. So if you go from a fintech or a SaaS company to a holding company, it's just a very different mechanic. So maybe can you compare and contrast running a company, like a software company, running a hold code? Like what mental models do you have to update to run this company?

Brennan (04:05.858)
Definitely.

Brennan (04:18.092)
I mean, I think the first is like, some of your instincts are wrong as a founder. So as like a software founder, I'm in like founder mode. I'm in the weeds. There's a problem. I want to fix it. I want to get the data. as a holdco CEO, you have to be a lot more like, you know, the goat Charlie Munger. And you actually like, you know, a lot of people are running holdco these days and they try to like get in the weeds with their businesses. You actually need to like abstract yourself from that and really like, you can make investment.

You can de-invest, you can back CEOs, you can change the teams, but end of the day, have to really be more of a, I try to have Infinity be more almost like a Citadel for AI services or like a data quant hedge fund where we should have really good data on underwriting companies, really good data on quantitatively how we think these companies are performing. And then that helps us allocate and helps us manage our portfolio. And that, you have to be somewhat abstracted.

So in a lot of cases, like I'm playing more of a board member these days with our eight companies. I'm not in like, you know, knocking on doors, closing sales, building out products, doing stuff that I like really am great at as a founder. And that's like, that's quite hard because if I, if I pop too much into founder mode, I kind of become shadow CEO of a company. And then that disempowers the founder who's usually better at their job. And so the thing that like keeps me up at night.

is that most investors advice is wrong for founders. And now that I'm like a Holdco CEO and not, you know, a SaaS founder or a FinTech founder, I have to recognize that I fall more into like the investor Holdco camp and they know more than me. And I'm actually here to more identify risks and be a great support and make allocation decisions. And so that's like, that's really like the different mindset and like, you know, Francis and I are still like super involved in all the companies.

But we're definitely involved in different ways. We're pushing on the risks of the business and is it positioned correctly? And most of our businesses that are really ripping, they're positioned perfectly. So how are they counter positioned against legacy services industries in the market? And so we spend a lot of time really making sure we've nailed the vision and strategy and that we have the right players in seats. Which is, I think as a founder, you do some of that, but you also are a lot more tactical around just shoveling the dirt.

Brennan (06:37.004)
which is like a very, very important job and like getting that granular data and how to pivot and iterate your product. And my product now is infinity constellation. So I'm iterating a lot of like, well, who are the right founders to attract? What are the right businesses to back? How fast should we go? When do we do a double down? Those are all like my iterations, whereas like, you know, running a company, running a fintech is a little different. So

Prateek Joshi (07:00.957)
Now going into this analogy of you're running an AI native work or halfway that builds like YC. going into the weeds of operations a little bit, you raised 17 million, launched eight AI service startups on day one. Can you walk us through how, take an example and do you,

incubate these companies or do you find founders are already working and then you acquire them like Berkshire does? How does a company get formed here?

Brennan (07:34.86)
It's very much like inception. So we have a blueprint, which like invisible is Palantir for operations. And that's why it does AI training and applied AI at the fortune 500. And we basically looked at that blueprint that mixes humans plus the technology and how that wins. then we say, wow, like we'd really love to do something in the EA business because, you know, we think there's a lot of alpha to be made in EAs. So then we find like an awesome founder like Jen Sermarco.

to run Everest and she really comes up with the positioning, runs the company as the CEO. We're more like junior co-founders to her in that. Or we have like Stuart Lacey, who's a repeat founder. And he really came up with the idea of Labyrinth, which is reg tech acceleration. So we do that for prop tech and actually like nuclear power and energy where we help get approvals faster using AI. And so he really drove that, but like we showed him our blueprint and then we co-crafted a memo together.

around how we see the positioning, what power he's using from like the seven powers is one of our business bibles. And then we just, we spend a lot of time thinking about the business frameworks that we do, and then we share learnings of other companies, but it's very much like the founders driving it. So we're building these businesses from scratch. We are like qualifying the people we're working with, but they're deciding if they want to be in the trenches building with us. And we're deciding if we want to build with them. Right. So it's very much like, you know, inception.

Prateek Joshi (08:59.233)
This looks similar to Venture Studio, but I'm assuming there are a couple of differences. So can you just talk about compare and contrast to a Venture Studio model?

Brennan (09:05.795)
Definitely.

Brennan (09:09.964)
Yeah, I think it comes down to incentives. like a venture studio and I are both creating businesses, but the end customer of a venture studio, I believe is your, is the venture fund that's investing in the idea. So like if I was an MD of a venture studio, I'm trying to create a bunch of ideas that are really fundable that then can go return a good MOIC for my fund. And then I can raise fund too and keep doing it. Right. So like their ultimate goal is to like create companies that are attractive, lucrative investments.

And liquidate funds to their shareholder base, which they're LPs. For me, as a Holdco, my shareholders have backed us to create companies, but a Holdco exists forever. So like we're a Delaware C. I don't have a seven year time horizon. I don't have to liquidate all my assets. So if I create 10 businesses and two are breakout businesses, my incentive is not to market them to a seed stage fund and make really investable companies. My incentive is to make really, really, really investable businesses.

in the long term. like that means the businesses have super good predictable cashflow and they're real businesses driving real EV. And that little nuance of like, I don't necessarily need to jump a hype machine to build a business. I need to build a really good AI full stack business within my thesis. And so I, then when that business like goes big, I'm on the same side of like the founders and the late stage investors, because I'm going to own part of that business indefinitely.

and that's, and that's, think a key incentive change and why we actually looked at doing a venture studio model. And ultimately, like we decided because we don't need large amount of capital to start a lot of these AI service businesses. like, we really want to have long-term ownership in this and have like a really long-term growth horizon. And so we don't want to just be like capped at a seven year fund cycle. Does that make sense?

Prateek Joshi (10:58.655)
Yeah, and the companies that you start, obviously you get a CEO, you got a team, they get equity, obviously you have equity in the company. Then the people who have backed you, Infinity Constellation, can you just talk about how the portfolio CEO, you and your backers, like how are the incentives aligned? Because I'm sure your backers have time horizons even though you don't. So how do you think about that?

Brennan (11:25.996)
Yeah. Yeah. So how we get liquidity, I heard how we get liquidity for our investors and then how our founders are aligned to that. the step one is like, we are backed by funds, family offices, big and small. And, you know, the first, our goal is like any other venture backed company where it's like, we want to go IPO or we want to do secondaries for our investors when they like need to get liquidity. like, that's pretty basic. It's like drive EV.

And I think driving EV is like how predictable and sticky my cashflow is. So if I build really good service businesses that have low churn, high margin, that are good growth, that creates very predictable cash flows. And then based off of that, I'm going to be able to like return money for shareholders. So that's how we stay like aligned at a macro level, very tactically how incentives work. And I want to, I don't want to talk like numbers, but I'll be like kind of vague here. All of our founders have shares in infinity as a hold co, but they also have shares in their existing business.

And that creates like a good dual incentive where they're willing to help all the other companies, but they're also willing to like, they have a nice little hedge to build with us and to like build in the right way. Um, and, ultimately, like, let's say you're a business that is a double, is a singular double. Like you said, I was a founder to be a home run, but let's say you're a singular double. Um, if you're a profitable business, which all our businesses are unit economic profitable, and you're making money, we'll happily keep you in the hold co indefinitely. And so as a founder, that could be a great asset.

for you and your family and for the business versus like, so you're aligned with us aligned with your shareholders. on the other end, if you hit a home run, we actually get to control the time that you get to come out and do your series B. So like, we'll happily launch you out of the hold co we'll co-lead your future rounds. And that's aligned with my shareholders because that's, that's creating value for them. That's aligned with the founder and that actually de-risk the series B, which is like one of the hardest rounds in venture right now to go and get.

because you can really time it right. You can have the right milestones. You can build your business in the right way. And that's like, that's the big message I think around like what we're doing at Infinity is like incentives matter and how we've lined up our incentives with our team and with our shareholder base is like really strong. And like, you know, I definitely have like done some of this, but Francis is like the brainchild here. He like lives and breathes governance all day and structure. And he like just sits here and thinks on the stuff.

Brennan (13:47.094)
And he's been talking about it with me for years since I was like a first customer of Invisible. And so that, like we've really built in the right flywheels to keep everyone aligned. And that's like a monger thing, right? Like if you focus on incentive alignment, then you do a lot less management.

Prateek Joshi (13:57.601)
Yeah

Prateek Joshi (14:02.911)
Yeah, yeah. I think it's giving the portfolio founders a stake in the parent company, the holding company. That's a very interesting angle and rare and also it's aligns them long term and that's a very interesting move. in addition to obviously they're running the portfolio company though they have a good stake in that, but it's a very interesting angle. And also I think when you talk about incentives reminds me of obviously, know,

Charlie Munger has talked a lot about this, Warren Buffett, but also Renaissance technology, Rentech, the hedge for the most famous, listening to a podcast episode where they put a lot of thought into figuring out what incentive mechanics will help us retain the best people. And many people are tenure, like they stay there for a couple of decades, if longer, the average tenure. because of...

The reason is that the incentive source structure is well aligned. The longer you stay, the more you make. also you have a stake in the company. It's very net-net. think incentives matter a lot. And I think you made a very good point.

Brennan (15:07.512)
Yeah. mean, my favorite of this is Citadel. Like I think they've nailed it. like, like really when France said I were creating infinity, we were like, and there's like a slide in our deck somewhere. but like, it's like the Citadel for AI services. And it's really in our head. like, who are the best founders by category? Just like who's the best portfolio manager in equity or like in debt. Right. And let's like, let's bring them in, let's have them create with us, but then let's give them like so much equity.

Prateek Joshi (15:10.293)
Yeah, there we go.

Brennan (15:34.882)
that, and then our infrastructure is so good. It's like, why would I ever go and like do this outside of here? Because this is making me so much money and, and, so great. And so that's like really like what we're trying. We took a lot of inspiration from that. And I think like the more, the more time I've spent digging into like Citadel, digging into like these like quant hedge funds, like I think there's an opportunity in venture and in company creation to be a lot more data-driven and a lot more like quant hedge fundy about it.

in like very positive ways that I don't think we've been as data driven as we could have in the last like 20 or so years. I know I haven't. And so now it's like kind of a thing that I'm very focused on, because I think if you can get the incentives right and then you can measure it with a lot of high fidelity data, which like AI and AI agents allow you to do now, because you don't need to like spend as much time manually getting data together, then you can as an operator and a capital allocator make a lot better decisions. so anyway.

Prateek Joshi (16:32.64)
I'm a huge admirer of the founder of Citadel, Ken Griffin. And the more I've read about him, I've listened to his style, it's incredible how much heart he has put into company building, which is incredible. right, going into the target market that you're pointing your energy towards. So professional services, massive market, multi-trillion dollars. Now, when you look at what

Brennan (16:37.283)
Yes.

Prateek Joshi (17:01.319)
sections or what areas to choose to build companies in. What characteristics of a given sub-vertical excite you? Like what makes you go, okay, that is a good target market to attack.

Brennan (17:14.828)
You know, this answer changes a lot. When I first started out, I'd be like, okay, accounting, right? Or like it's things that I knew. And I was like, this, this could be a thing. How it's actually shaken out is it's so much like our rate limiting factor is, and the reason why I do these podcasts is because I'm trying to find more founders and I'm trying to find more builders because our rate limiting factor is awesome people to build with us. And so I actually usually let.

Prateek Joshi (17:17.29)
haha

Brennan (17:42.934)
the awesome person that wants to come build with us, pick a category, pick a market, and then we start a process of finding human centric service businesses that haven't been really impacted by tech and really go from there. And so what's so funny is there's someone I'm talking to right now that is pitching us pool cleaning roll-ups and it's an IoT device with pool cleaning route. You could do a roll up around it. And then instead of using

really skilled labor, you can use unskilled labor because the IoT device is handling, distributing all the chemicals. So it's really like route management. So it makes it a very easy roll-up opportunity with lots of EBITDA. I never would have thought pool cleaning would have been something I would have entertained or thought there would be EBITDA in there for us, but there is, right? And that's because that person is an expert in IoT and an expert in pools. And that's been something that's really, really interesting. So it goes back to sort of like inception, but...

If you really look like from a macro standpoint on like how I underwrite it, I think until we have really big breakthroughs in robotics, stuff like plumbing and like propane and HVAC electrical, like that's harder, right? And I think people are rolling up HVAC right now, but like a lot of those businesses are power law businesses where 20 % of the market owns 80 % of the market or some degree. And I think until you have robots,

you're not going to see a lot of EBITDA margin pulled out of that. But businesses that, like I love design because those are a lot of people who work behind computer screens. you look at a design agency, they trade at like one to two X EBITDA because they're so customer centric, they're project based because their revenue is not really durable. So you could go find someone awesome like Jaclyn, who's our CEO of Radiance, and then go and build a repeat revenue model where

You can lock customers in, use AI to expand the margins and then really compete both from an investment profile, but also from a customer quality profile. like really like that. Right. And so that's sort of been like the rinse and repeat of a lot of what we've done is like what's being done behind a computer screen. What has cashflow? What's stuff that like you're using? I'm using like a lot of the times I'm like, okay, I'm creating a business. There's a new business I want to create an IT services. Why? Cause we started having to hire a bunch of IT people because we're

Brennan (20:00.386)
all around the world and we need laptops and we need all sorts of things. And then we're like, wow, this is like a cost center, but actually I bet this could be a really great service. And I bet AI could really help this service. So I start talking to IT administrators. And I was even talking to one that managed Madison Square Garden and like, it's a very manual business, still not using a lot of tech, even though they're implementing technology. And I'm like, okay, that seems really ripe. And so that's like kind of how the discovery process works.

Prateek Joshi (20:27.85)
Now on day one of building a company, the portfolio company, obviously the founder is in-house, everything is in-house. What metrics or what signals do you use to identify, okay, this company is ready to graduate out and be a standalone company. And maybe part B to the question is, you mentioned for breakout companies, you would lead series B.

Where does the capital come from? Do do SPVs or where does the capital come from for those larger rounds?

Brennan (20:59.384)
So we're super new. So we're literally like a year old. So we're, just walking on that road now with a couple of our breakout companies. And so the first is we always default to doing what's right for the founder and right for the business. So they drive that. So like, if you're one of our top companies and you're like, look, I want, I can really deploy 10 million and I want to go and like bring in a specific fund to do that. They can go do that. And then our board approves it. Right. So like, we always default to like what's right for the business. And that's like our golden rule that

Like Charlie Songhurst is on our board and he's like one of our big investors and he's incredible. Shout out Charlie. Go raise an angel check from him if you can. He's great. But he was like, hey, we should always default to like do what's right for the business and then everything else will follow that. And that's something Francis and I've really stuck by. And so that's how we think about it for the business. And then for us, like when we co-lead it out, we basically raise out a very like favorable multiple.

And we have like a really broad investor network. So like we don't necessarily have cashflow issues or access to capital issues, especially for like our top performing companies. And so, we'll either raise that hold co to support that, or we'll like do an SPV to support that. one of the two, but we're, think we're still like in the wind a little bit as like, what's better because like the, you know, to bring you in some form of podcasts. So it can't really be confidentially.

But I think it's a good, healthy debate that I'm fine having publicly, which is you have some of your best companies. They really need to drive when they want to take that external capital. And from one end, I could make a really great argument that never raising for them and owning as much as I can is really great. But on another argument, if I'm thinking like a quant hedge fund person, I'm like, wow, I wouldn't mind. This is becoming a power law thing for me. I wouldn't mind risk sharing some of this and then using

basically starting more companies to even out the portfolio. And so I think that that's a very ongoing debate that we have as a board and as a team. And we have these ongoing conversations with the founders as well.

Prateek Joshi (23:04.993)
That's great. when you're early and you're building this out, and how do you address the too many toddlers in the house problem? Meaning, how do you allocate resources? Because I'm sure everyone wants the max resources. So how do you allocate resources? And what's the roughly what's the optimal number of simultaneous portfolio companies you can realistically help? Because in an ideal world, of course, you would want to have 1,000, but you can't.

help a thousand all at once. So how do you what's the optimal number and then you're like okay we're gonna start graduating some of these out because now it's too many.

Brennan (23:41.272)
So every investor I bring on and every board member we invite to be part of the board, I say, I'm a dad that has eight kids and I need lots of uncles and aunties to help me raise the eight kids. like me and the platform team of Infinity are a rate limiting factor. Cause like you're right, our model is great. The data is showing that it's working. So like what's the rate limiting factor for us just like going to a hundred companies, right? And I think the first is like great people that want to build with this.

And then the second is like, we manage a lot of back office for the company. like finance, HR, legal, IT, that's benefits to the founders. So like, we also need to like dog food and use AI and scale ourselves up to support more companies. And then we never want to drop underwriting standards. So like, we want to make sure that that's, that we're holding that quality bar really high, but that we're also like iterating and positioning our current businesses appropriately to their max potential. And then how we handle internal resource allocation.

Most of our, again, this goes back to incentives. like if you and I were to go raise a venture round and let's say we raised a hundred million or five million, it doesn't matter. You burn that in 12 to 18 months every time. For our hold co, it's a little different. So the founders come to us and when they start the idea, they write an investment memo in that investment memo. They set out what capital they need based on what milestones my board pushes back on that or agrees. They approve it. And, and basically as long as you have positive unit economics and

the board continues to approve of the vision and the direction you're going, you get access to more capital. And that's how the business works today. And what's crazy is all our businesses have gotten to like really, really good venture milestones with very de minimis amounts of capital, because like they're not incentivized to go and take that 2 million or that a hundred million and burn it in 12 to 18 months to get to the next cycle. They're incentivized to build product and sell and disrupt the service industry. So it's like a little bit of a different vibe. there's not...

It's not like a knife fight for capital. It's more of like who's earned the right to get to the next stage of capital. And like one of our key cultural pillars that we set as a group with the CEOs was that headcount kills. And so like, we're all actively trying to hire like as few people as possible. Like Francis and I would like love like a one person CEO. Like I was talking to someone out of MIT yesterday about like an autonomous company and how to make a one person CEO company. And I think it'll happen in our lifetime.

Brennan (26:01.548)
So like most of our founders are competitive and like they don't want to go and burn enough. They don't want to go burn 5 million in 12 months and hire a ton of people like the businesses of five years ago. They want to go and build these like really efficient companies. And so it's just a bit of it and like incentive differences that that around the allocation. And then, on the business side, it's like a large, like largely an underwriting and then rate limiting factors of like how many businesses we get when, but it's a lot.

Prateek Joshi (26:29.161)
Let's talk infrastructure. You mentioned earlier that you help with back office and accounting, HR, tax, which is good. Now on the technology infrastructure, what resources are shared? Because AI is a common thread, I'm sure the bunch of needs are overlapping. So can you talk about what technology infrastructure is shared across all the companies?

Brennan (26:51.522)
So initially when we thought out, we actually thought there'd be a centralized tech team. And it's funny, I've gotten this question like multiple times today. So it must just be the energy of today is asking about this. we initially were like, well, we'll build like AI agent infrastructure centrally and then it'll fork and like da, da, da, da, da. And then what happened is like, just like a venture portfolio, our portfolio started getting power law where like one of our companies really started to break out. And then it became like, well, all the tech resources should just go to that company.

because that'll return this. And then we were like in our own innovators dilemma out of Holdco, but my value as a Holdco is to create more businesses. So then I was like, okay, this needs to be solved. And so really what we've done is now we ring fence tech teams. Every tech team that gets started has technical founder. That's like an applied AI person. Every team has ring fenced engineering resources, just like a venture company. And so we want that because we want the IP siloed and investible.

The second reason is that all the businesses, even though they're AI services, are really different. like Everest requires really different tech for its EAs than Radiance as a designer, then like Unlimited for like AI, VC analysts, right? And so like, even though they're all like AI agents or AI services, they're like actually quite different and they're using different foundational models or using different workflows. So they're bespoke. Now, then if you pop back on your CEO Holdco hat, you're like, okay, I've just lost all my network effects.

between the tech, between all the companies, because now they're all siloed. So what our Adam Haney, shout out to him on my team, he was one of the early engineers at Invisible and is X Meta. He came up with the idea of creating what we call Code Commons, which is any of the company's developers with permission from their board can post code to the other, to basically our own private open source library to infinity.

And then as other teams adopt it and fork it and use it, they get shares of Infinity Holdco. So let's say you make a really good authentication app or an authentication workflow. You put that because that's not necessarily strategic. Everybody needs it, but it's really key. You can put that in code comments and then everyone can fork it and prove it. And then you get shares. So our board approved basically a share bounty for all the engineers to like come to basically like commit code and share code to each other. And so that way, like, you know, company A company B they're

Brennan (29:13.282)
their investors are okay. And so like they have approved what goes open source in infinity. But then we also are still getting like the network effect of sharing the technology. And then every two weeks, all our developers get together and basically like do demo with like the coolest stuff. And it's actually gotten like kind of competitive. Like today, they actually happened like, I think today. And one of the CEOs told me is like, yeah, my developers came to me after this meeting was like, how do I get more like...

How do I get more tickets in this week? Like I want to ship these features by like the next two week sprint, cause I want to show them this. So it's like, I don't know. There's like healthy competition among the teams.

Prateek Joshi (29:48.435)
Now, because of the way Holtcore is structured, how does it influence your choices around model hosting or going full-stack, vertical AI or building proprietary models, fine-tuning, a bunch of these little considerations, call it Infinity Playbook. How does this influence your thinking and do you kind of...

coach or train your founders to do things a certain way versus old school SaaS playbook where you just hired a whole bunch of people and put more money towards it.

Brennan (30:19.982)
I mean, our playbook very much is like an orchestration playbook. like our thesis is we control distribution, we deliver full stack products, and then we don't, we haven't built a foundational model yet. I think we would and under the right circumstances with the right talent, it's just like really expensive. And I think they're going to get commoditized, right? And I think the ones that are really big are really big. And, and I'd rather be on like their side of the table as an orchestration partner. That's leveraging a bunch of different models because let's say you're building

great example, you're building a full stack AI company for real estate. You might need a voice AI agent. You definitely need a PDF reader. You definitely need someone that's like a couple of NA end flows, like connecting into like Yardi and a couple other systems. Like there's a bunch of stuff and those might all be really different pieces of technology from different providers. And so like, Hey Jen might be the best video AI. And then there might be...

The model that's the best, like, Mistral the other month was like an amazing benchmark for PDF extraction. like Google Gemini has come in and been like awesome on video rendering. And so I think it's up to the founder to pick the right model for their service stack. So for one, we play with everybody and we're an orchestration and distribution stack. So that's, that's the second. Now a founder could come and pitch me a foundational model and I could be really interested in it, but I'm probably like not the one out there. Like we're not the ones out there like funding and building that per se.

because I think that takes a lot of compute. I think that takes a lot of expertise. And we would look at it, but we just haven't yet. And so I really empower the founder, but I steer them towards what's been working for us, which is own the distribution, play with everybody, and iterate like crazy. And I think that that's the right sauce.

Prateek Joshi (32:05.739)
What are the first three hires a portfolio company would make? No, I know it depends on the company and everything, but on average, what are the first three hires and in what order?

Brennan (32:17.91)
I like putting them together as like a founding team or like a founding unit. But I think the first founding unit is like, need the business thinker, you need like the business system thinker slash visionary slash salesperson. That's usually the CEO. And then you need the person that's like the cracked AI workflow kind of person. and that's usually like the VP of engineering or the head of engineering of the team. And then I think depending on like the product or company, so like our company lightning is a direct consumer.

Philosophy app so we like digitized a bunch of philosophy books and then you can go on and get your own like philosophical DNA test You should do it One of the founding partners of that is a Rhodes scholar Zohar who's like an amazing philosopher and he's a rabbi and just shout out shout out to Zohar He's great But he's like put his DNA in that in that business around like philosophy and it's really important for that business to have someone with that credibility as well as this the tech person as well as like the CEO visionary salesperson and so

I think each one is a bit different, but like, just depends. And I could see in like biotech, you'd have that same flying formation, but then you'd have someone that's like, you know, a virologist or someone who's like, you know, the deep subject matter expert there, but like typically those are sort of the two flying formations we go with. And then I won't ever start a company like we'll help people find that awesome AI person. Cause we've got a really strong network or we'll help source a CEO to like a bit, like an AI person that wants to build something.

but we'll never start a company without a CEO and we'll never start a company without a tech person. Because that was some, we've done both before and it was like not happy times.

Prateek Joshi (33:53.941)
As you're building this, in many different directions from which competition might emerge, but really in your heart, your deepening heart, who's the real competition here and why?

Brennan (34:07.938)
That's a great question.

I'll give you two answers. like the first is like a philosophical answer. So one of my favorite stories to tell is about Otis elevators. Do you know why Otis elevators is everywhere in New York?

Prateek Joshi (34:23.531)
Go for it.

Brennan (34:24.716)
Yeah. So he invented the emergency brake on the elevator. And when he did that, he was like, great. I built something amazing. People will now adopt elevators because people are scared to be in elevators. Let me build it and they'll come. What happened is no one bought an elevator still. So he went to the world fair, stood on a wooden platform and cut a rope in front of a bunch of press to show that his brake could stop and basically work and be safe. And that's called a trust factor. And I was like debating this with someone earlier the other day. It's like,

We can all debate at how fast the tech is growing right now, which I think it's like the fastest it's ever been in my lifetime. And like, will that, you know, if you go to AI 2027, there's a bunch of different scenarios, but like, when will we hit, when will we hit AGI and can we hit AGI? That's a great debate, but there's another factor in this, which is like the human trust component. So like no matter how fast technology moves, I think our trust moves on a different vector and a different scale. And so like we've had the technology to have self-driving cars for a while now.

We don't have self-driving cars, right? Like we have the technology for flying cars. We don't have flying cars, right? Well, it's because we like kind of don't trust each other to like zip around and flying cars or like, you know, I was in a Waymo in Austin, Texas the other week. was a delightful experience, but then my wife took it and she got like stuck in a circular loop and then someone had to like call in and drive the car away. and I think it's a work in progress, but there's like a trust factor to the tech. And so I think our biggest competition philosophically is that trust factor.

And that's what AI services does is it allows us to sort of hack that trust factor curve and end up just like delivering like you believe in AI, you want a great product like design or an EA, you can get it, right? Boom. That hacks that equation. Very tactically, I think more and more people are starting to figure out that the full stack AI is the play. So like Y Combinator is looking for full stack AI companies right now.

There's multiple like tier one VCs that have raised hundreds of millions to do AI kind of like roll ups in the services space where the thesis is like, let's go buy a hundred million in XYZ business and then convert AI into them. And I think that like, they're all kind of like chasing the same thing we are, but like we're counter position where like, we really believe that like we need to create our business and then do it within our hold cow. And I think that that, makes us like a little bit different, not to say like, I think at the right time I would consider roll ups and other things, but that's my.

Brennan (36:48.354)
That's my two answers. So philosophically, we compete against trust or compete for trust. And then very practically, I think there's a lot of people doing like PE type roll-ups as well as like YCE is coming at this now.

Prateek Joshi (37:03.091)
All right, final question. Five years from now, what does Infinity look like practically in terms of the number of companies, maybe ARR, what's the whole co-head count? And also to achieve that, what AI advancements that are happening now are the most exciting to you?

Brennan (37:21.974)
I don't know about the tactical stuff. I would hope we have, with a cultural value like headcount kills, I hope we have low headcount to employee. I want a really high revenue to employee ratio, so I'll give that. And then for us, my goal is to first get us to 20-something companies, which I think gets us to about 100 million in ARR or about a billion dollars in EV. And then it's really going to be like, what do we need for the public markets? And that's really what I'm focused on. But I want to show that we're not just a whole co-full of assets.

I want to show that we're actually like an engine for creating new AI startups and in the services space. And that I think will put a higher multiple on the business and be really valuable. So like, we'll also have our own tech to create companies, which we're working on now. We're going to have a lot of different businesses in different verticals. I think we're going be able to go after more and more attractive founders and be like, Hey, instead of like going raise venture, come build with us. Like this is faster. This is better outcome for you. and

That's really how I see it. And ultimately, think Infinity could be worth trillions in EV because this is the first time ever in our lifetime that we're really approaching a 200-year-old service industry. Everything from business valuations to pool cleaners can get disrupted right now with technology. And I want us to be the ones to do it.

Prateek Joshi (38:37.505)
Amazing. All right. With that, we're at the rapid fire round. I'll ask a series of questions and would love to hear your answers in 15 seconds or less. You ready? All right. Question number one. Question number one. What's your favorite book?

Brennan (38:45.944)
Great. Let's do it.

Brennan (38:52.076)
My favorite book. Zero to one, Teal.

Prateek Joshi (38:59.809)
Alright next question, which historical figure do you admire the most and why?

Brennan (39:06.382)
probably either Alexander the Great, because I'm Greek, or Augustus Caesar, because he just like, he built Rome like really prolifically and his biography is really, really crazy.

Prateek Joshi (39:22.143)
Yeah, I really appreciate those references because I'm a big, big history nerd and I know exactly where that might come from. So that's amazing. All right, next question. What has been an important but overlooked AI trend in the last 12 months?

Brennan (39:45.164)
I think the human trust factor, like we're so fixated on how fast this technology can develop and compound, but so many people I talk to are still like, they're still basing like their experience with AI off using chat GPT from two years ago being like, it hallucinates. I'm like, how do you like, you know, like you can connect to Anthropic to your email and just be like, you know, who have I not emailed back in the last four days? And it like returns it like it's pretty incredible.

Prateek Joshi (40:09.633)
Right, Amazing. What's the one thing about service-based companies that most people don't get?

Brennan (40:24.778)
I how essential they are to our economy and our GDP. I think we really focus a lot on like Fortune 50. And as founders, you're like trying to sell into like Fortune 500 or Fang companies. And really like a lot of these businesses that are service businesses that are our neighbors businesses make up a huge portion of GDP around the world. And I think we're in this like moment where AI can create like massive GDP expansion and enable a whole new generation of entrepreneurs.

And I don't think that's like baked into any economic forecasts or anything. Like people are focused on like job losses at like IBM or like JP Morgan or whatever. They're not like thinking like, wow, like my neighbor, like we could all create businesses that are going to be better. And like our neighborhood GDP is going to go up. And I think that that's pretty powerful.

Prateek Joshi (41:11.615)
What separates great AI products from the merely good ones?

Brennan (41:17.74)
I don't know if I know that one yet, but I really liked the product experience of Claude. And I think you can tell that they're just like, really, like a really thoughtful product team. And they've done a lot to care about like the design and the user experience of it. And so for me, like, just because I was part of the team was simple back in the day, they, they built a lot of design for banking and I worked with some amazing designers there and I still just, I'm a sucker for really good design.

like user design.

Prateek Joshi (41:48.737)
Yeah, I think design is becoming extremely important because the baseline has grown so much that everyone expects a product to look nice and cool and fun and that's the baseline. So, fantastic. What have you changed your mind on recently?

Brennan (42:08.162)
That's a great question. I used to be very like, can de-risk investments in the US and start companies only in the US. And I think there's actually like such massive blue ocean opportunities, like in the middle East, in the Gulf region, South America, Europe. I think that there's a lot of AI service opportunities all around the world. And I've sort of like changed my tune on that now where I've like flipped around with my board saying like, actually I want to do more stuff outside the US because

In some cases, I think there could even be bigger markets unlocked. So like I think Dubai and the Middle East and Abu Dhabi, like there's actually very big sentiment from those governments to like adopt AI in a real way. And they're very serious about it. And like they have, you know, massive plans and massive investment that they want to put behind it. And I think that there's like big, big things to be done.

Prateek Joshi (42:56.705)
Right. That's actually a good one. Because in half my life I was in India and here. think you have more of an appreciation of what outside US can. It's a huge market. You're right. I think it's a huge market. I think unless you actually, yeah.

Brennan (43:11.982)
Well, it's always so hard to like, for me as an entrepreneur in the U S like I, it's in San Francisco or New York or wherever, I can like background the founder. know the laws, I know the country, I know the investors. When you go into a new region, like, I don't know all those things. So it's like, I think you have to find new ways to de-risk it, but like, there's also, you know, in these investment equations, there's always the IRR question. So it's like the beta goes up, but the IRR can also go up quite a bit, especially if you're first mover.

And like I spent time in India, like I think India is going to be an amazing AI services market and AI market in general. And I think it'll, what's, what'll be very curious is like where they start building a lot of data centers for compute and then where there'll be like local models and local regions doing different things. Because I see that whole industry getting more and more fragmented, which then will incentivize, you know, further startup growth and all those markets.

Prateek Joshi (44:07.494)
Next question, what's your wildest AI prediction for the next 12 months?

Brennan (44:12.428)
my wildest AI prediction.

Brennan (44:23.82)
I don't know. Honestly, they've kind of all been hit. And I keep getting really surprised. And I'm like, for the first time in my startup career, I keep underestimating it because I'm so skeptical about tech. I'm like, well, it probably can't be a VC fund analyst. And then my team just busts out and they have an AI that's doing LP reporting. And I'm like, whoa, OK, that's legit. And so it's probably, I think, the wildest take of some tech I've seen.

Prateek Joshi (44:46.251)
Yeah.

Brennan (44:52.704)
is I think that there's like now an opportunity to literally have an AI avatar on zoom or like I could be an AI avatar right now talking to you in a podcast. And I think like probably the next six to 10 months, like we'll get to a point where it'll be that and you won't even realize it. And I think that that could be very interesting for like social situations or interviewing or all sorts of things, but it also creates like some virtual identity issues. So that's some, that's some stuff I'm thinking about right now.

Prateek Joshi (45:21.377)
All right, final question. What's your number one advice to founders who are starting out today?

Brennan (45:29.614)
Founding is your biggest personal growth journey. It's not about making money, it's about growing. And if you approach it that way, you're gonna end up finding what you love and just do that. And for me, it's never been about, I mean, legally, I have a fiduciary obligation to make my investors as much money as possible. But personally, it's all about my own personal growth. And that's why I'm addicted to doing this, because I'm growing and make myself better. And I honestly like...

The, I think we have a once in a lifetime opportunity right now. So that if you're a builder and you can come and change and create, like you can help change the world's GDP with implementing AI and service businesses and like transform your country, your town, your region. And I think like we, those of us who can create businesses, like have the obligation to do that right now.

Prateek Joshi (46:19.073)
Amazing. Brennan, this has been a brilliant discussion and also very unique because I think in the 175 plus episodes I've done, I think you're the first like hold co person running an actual AI hold co. this is a very, the mechanics are amazing. It's one, it's just brilliant. So thank you so much for coming onto the show and sharing your insights.

Brennan (46:29.454)
The first.

Brennan (46:37.912)
Thank you for having me.