Lessons from the Early Days at Uber and Advice for Founders with Kevin Novak
46:37 Share

Hey, everybody. This is Ben Kesnoka, co-founder and partner at Village Global, a network-driven venture firm. And this is our podcast, where we go deep on all things business and technology with world-leading experts.

Hello, everyone. Ben Kaznoka here from Village Global. I'm here with my colleague, Maya Fry, and Kevin Novak, partner at Rackhouse Ventures. Prior to becoming a VC, Kevin has had a long career as an operator and data science leader, most notably head of data science at Uber. Kevin, welcome to the podcast.

Thanks for having me. So, Kevin, let's start by hearing about your background. Tell us about your career to date as a leader in the Valley and what's the mission behind Rackhouse? Yeah, absolutely. I mean, you mentioned I've had a career operating as a data scientist, sort of an early stage data guy for about a decade. I was in academia well before then.

But it really kind of developed a lot of expertise and frankly, a lot of hard-won experience building startups in these early phases. And the motivation I had to kind of

contribute that expertise to the next generation of companies as well as a lot of perspectives I've developed on sort of where I thought opportunities were going to be created for the next generation of AI companies got me really interested in investing. I was in academia because I like to teach. What were you teaching? What kind of topics? I'm a nuclear physicist by training way back in the day. And it was really interesting because I

In a way, one, nuclear physics is essentially solving physics programs in programming languages. So I was doing effectively the same job as a data scientist. But the interesting thing about physics, and I think that kind of informs a lot of why I love data science so much, was in academia, the way graduate students kind of earn their keep is they end up teaching physics labs, right? And I was low man on the totem pole.

and so got asked to teach the physics labs for the medical students, so for pre-med kids. And pre-med was the worst group because they all had to get A's, they all have to have like 4.0s to get into med school, and physics is completely irrelevant to a career in medicine, right? It's like one of these weird bureaucratic like box checking exercises. So everyone there was like in the definition of just like,

you know, maliciously compliant. Everybody was apathetic. They did not care, but they just had to care enough to get a good grade. And man, like apathy is like soul destroying. Like to anybody who's mission driven, like apathy will kill you. And what I loved about data and data science was it was such a new field or kind of a new conceptualization of this approach that like

Nobody was in data science, still is, because they're like, I am a data scientist because my father was a data scientist and his father was like, we come from a family of data scientists. Like, everybody started out doing something else. Right. And they all kind of like consciously step back, like critically evaluated the path they were on and then like opted in and said, like, look, this is the path I want to go chase.

By the way, I love the phrase maliciously compliant. So that's an outstanding phrase. Now, my question on data science are, I assume you know DJ Patil, my old friend of mine, friend of Village. Now he's a VC actually himself. Now, did he invent the title data scientist?

Was he the first to have the title? Like, what's the history? Because you're right, this isn't an occupation that existed for a long time. Yeah, like, when I was joining the industry, it was, like, him and then, like, Jeff Hammerbacher. And I think that they...

Those were kind of the people who were like really focused on coining it. In fact, like when I joined Uber, I didn't have the title of data scientist. I think I gave like we got to like pick our title. So I was like a computational like algorithms engineer or something. Right. And within like three months, like that article came out about data scientists sexiest job of the 21st century or something. And you realize like that was kind of what.

It was. But yeah, I mean, DJ was was one of the still is, I mean, like one of the most people in the field, but was very much kind of leading the church. Yeah. So so tell us what was your job at Uber and just talk us through some of the insights you gained from your role there. And what were some of the takeaways that you've carried with you?

So I joined Uber as like their 21st, 22nd employee, depending on how you count. I was their second data hire. And what data science at Uber meant was essentially we worked with the engineering and product teams. And there was a bunch of sort of math problems and pretty complex machine learning problems relating to just this idea of delivering like

reliable, like consistently reliable, scalable transportation. And so I was basically given a pretty broad mandate. Travis had like a wishlist of like, here are the 15 things we need to build. They all require more math than anybody else on the team. Like see what you can do to kind of figure it out. And it was really my first startup experience. I'd done some entrepreneurial stuff back in the Midwest, but this was like my true first Bay Area experience.

experience. And I remember, especially coming out of academia where like a fast moving project gets done in like low single digit number of years in academia, right? It's all three to four year type time horizons.

My first big lesson at Uber was that like just this reality that you can always get more done more quickly than you ever think is possible. Right. Like one of my first big projects at Uber was inventing and launching dynamic pricing, right? Surge pricing and, and,

I remember in academia, I was working on like my thesis was around simulators. And basically this idea of like building models of complex systems. And so I was super jazzed because I could tell Travis, look, like I'm really good at this. Give me like four guys in six months and we'll build a system of like how traffic flows in a city and like we'll throw in some economic terms. And then once we've got the simulator,

Then I can like you at the workbench to start testing dynamic pricing theory. So like, this is easy. Like give me four guys, six months plus another two or three to implement. Like we'll have this done in eight, nine months. And Travis is like, it's just you, you have three weeks, figure it out kind of thing. Right. And, and,

And I realized, like, coming back to him, I'm like, well, then I have to kind of, like, demo this live. Like, we're going to, like, ship a beta. And he's like, yeah, uh-huh. Like, that's kind of the way we do things around here. I mean, and it's such a great anecdote. It reminds me of the Reid Hoffman line, you know, if you're not embarrassed by the first version of...

of the software that you ship, you haven't shipped, you ship too late. Were you embarrassed by the first version of dynamic pricing at Uber? I wouldn't say I was embarrassed. I was certainly stressed in terms of, it was one of those things where one of the first features I built and tested was,

the ability to like rapidly turn it off. And the thing where one of the, and the idea was like, well, we're going to do this live. I have some really strong hypotheses about how this should work. You know, economics is not exactly a new field, but there's going to be a ton of implementation details and a bunch of guesses. So,

I ran it personally, the Nant and Grayson project for several years and basically worked every Friday and Saturday night for those three years just to make sure the system was behaving as expected.

And those early phases were definitely up till 4 a.m. kind of thing, just making sure that that last Uber at the end of the night... So you're monitoring Uber ride traffic. So there really wasn't God-motored Uber after all, huh? You were checking out where everyone's going. I mean, it wasn't actively directly controlled. But for example, we have guardrails around prices couldn't get too high and stuff. And so you just wanted to...

you didn't want prices sort of pegged against the maximum price threshold in purpose who would eat. And you'd end up with sometimes just because of the quirks of how Uber sort of dispatch system worked, the system obviously needs a pretty current view of the transportation network. So if we would have upstream data logs or new features rolled out kind of irresponsibly, I was always just like, look,

Dynamic pricing involves like money. As soon as money gets involved, everybody's emotionally involved. So let's try and take extra precision, extra caution on any of those types of features. And just as a bit of a side but related topic, dynamic pricing, I think it's awesome. And that's a free market at work.

It is a controversial idea in some political factions, and we don't actually see dynamic pricing at play in a lot of parts of the economy. I think Amazon tried something for a hot minute where supply and demand was driving prices on items, and there was a huge backlash there.

How do you think about the politics of dynamic pricing? Yeah. I mean, I think it's funny. Wendy's recently kind of announced something about surging Baconators, and I think they walked it back within like 48 hours. They were going to try to charge more in the morning for bacon. That's right.

Well, they were trying to dynamically price food like during a lunch rush kind of thing. And immediately my inbox blew up. And I think I'm not even sure it made it 24 hours kind of thing. They kind of touched that particular third rail.

And that's pretty funny. We'll charge more during breakfast, lunch and dinner as a restaurant. We think through some complex math, we've determined that people tend to eat the most during 1130 a.m. and 1 p.m. It took a lot of science to get there. Exactly. And so like.

But what I thought, I still don't think people fully kind of appreciate this about like dynamic pricing and Uber was ride share is,

As a sort of as a, as a group behavior, sort of a fleet behavior is like, what does that mean? Call a stochastic system. It's very path dependent. The state of the kind of state of where cars are, then given point in time. Is a function of a bunch of decisions of drivers deciding to relocate or not relocate you requesting or not requesting and.

Because it has that stochastic nature, the actual decisions you make, some of which can be economically influenced, have a strong sort of collective group behavior, right? And so...

What we realized pretty early on was when demand starts kind of outstripping supply, people are pulling cars out from like your particular neighborhood faster than cars can come in. Right. So you're in your ETA start going up. So your ETAs are three minutes and then they're kind of five minutes and then they're 15 minutes. And the longer an ETA is, the longer your trip takes as a whole.

the fewer trips that that driver can then do sort of collectively as a group. And so like dynamic pricing was actually about basically economically shaping both drivers decisions of like, should I relocate to this neighborhood and

and riders decisions of whether they should request or not to actually keep like the ETAs low enough that we maximize the number of trips per hour, right? There's sort of this second order like maximum trips per hour effect that like you kind of don't appreciate until you think about the fleet as a whole. And that second order effect led to like if we weren't surging, you would do 30, 40, 50% fewer trips per hour just because of like that dependency.

And that effect of avoiding that penalty was like way more valuable than the surcharge. And so like these like deeply stochastic systems are actually really good candidates for pricing because, you know, more trips led to more riders getting where they wanted to go. You know, but there's like no, there's no path dependency in making Whoppers or something, right? Like it's all, it's just sort of that effect doesn't exist and like pricing dynamically makes, has less of a big outcome on the bottom line.

So let's go back then to the moment you go to Travis and say, give me nine months, a team of five, and I'll build something awesome. And he says, how about just you and how about three weeks? So as you then went through that experience, what have you learned about building and assembling great engineering teams and leading them with a sense of urgency? And how applicable are some of the lessons from Uber to other startups when you're working with early stage founders? Yeah.

For sure. I mean, what I always kind of solved for were employees who were collaborative enough that they could work in a group, but were also individually skilled enough that they could kind of dig themselves out of any problem they got stuck in. They could kind of unblock themselves.

So I really like these kind of heterogeneous skill sets of what you can like in the data world. If you can like do analytics and SQL, you can do advanced machine learning. You could write production code well enough that the engineering team kind of didn't hate it.

And you could present, you know, verbally or sort of in writing to a point where a smart non-data person can kind of understand what you want to say. Like, you were my unicorn candidate, right? You were everyone's unicorn candidate kind of component. But every role that we hired for, and you were on the engineering team and data teams, was always thinking through, like, where are the places? And they're usually unique to an organization of, like,

where individuals get stuck if they don't have this skill. Like, what are the skills you need to kind of unstick yourself? And those are what we always interviewed for. That's what we selected for. And then over time, as the organizations got a little bit more complex and developed and mature, when we were developing platforms or sort of tooling to support data scientists, engineers,

It was like, how do we build mechanisms or systems such that the number of sticking points that any given person has is minimized, right? Like, why do you need to have like five cross-functional skills? Well, some of that can probably be ameliorated with technology, right? If you have a good data catalog or a good development system, you don't need to be like the world's best backend engineer or data engineer to do this job. So it was kind of a...

Making it easier, like creating systems that are more forgiving and make it easier to unstick yourself. But if you always start with the idea of let's build a team of people who could, if they had to ship their entire project start to finish, we ended up with much more agility, just sort of institutionally.

Maya, do you want to jump in? Yeah, Kevin, based on your experience at Uber and just in terms of how many decisions was based on data, how are you now thinking about advising founders when they might not have a lot of data to begin with, especially pre-seeded seed stage? How do you advise them on how to structure their teams? Do they need someone on their team that's the founding member that knows about data, how to think about data, how to...

build the infrastructure to make decisions when you have the right data. How are you thinking about advising founders on that? Yeah. I mean, one of the things that is kind of, kind of sounds kind of intuitive coming from like a data investor, but I actually think one of the common failure cases is almost being like too data driven, right. As you're getting up and going, like, again, we should be one of surge pricing on three, four weeks of development. Like it was all kind of like gut feel and like first principles thinking, right. So, yeah,

What I usually emphasize with these teams are like, are if you, one, you can ship products where you don't necessarily need like a terabyte of information to like train your model, like build off the shelf, but you should be building a virtuous data cycle in whatever you do, right? As far as, so like in the example of search pricing, we,

We built like a pretty rigorous, even right out of the gate, sort of analytics pipeline to collect user behavior. Like, did you see the screen? How long did you spend looking at the screen? Did you...

Did you press the button or not, like to request or not, right? All of which was like literally like building from first principles like price elasticity curves once you have 1500 data points or 2000 data points. So it was sort of this idea of if you have data to get going, great, but don't necessarily let that stop you as long as you are collecting the information such that you can kind of progressively get smarter over time.

And then the other component is recognize that shipping a data product without a deep training set is riskier. So going back to our single surge, make sure you think about what does the switch it off use case look like. In surge, it was easy. You just kind of turn off the algorithm and everybody gets base prices. We maybe have a less good night in terms of revenue than we might have otherwise, but

Ultimately, the product still functions. Like, you know, build an off switch is kind of the fundamentals here. But if you're not collecting that information and you haven't kind of thought about, like, how you kind of wire it back into the next generation, then you can't, like, kind of ever basically ascend the, like, complexity curve. So getting that cycle going is super critical.

What are a couple of examples of some startups you've invested in over the last few years, Kevin, that you think nailed the data strategy from the beginning?

Yeah, I would say, I mean, one that really comes to mind, we worked with them in Fun One, is this company, Fathom Video. They're a kind of note-taking app, automated note-taking transcription. They kind of compete in that space. And one of their really sort of novel, their first novel sort of differentiating factor was Fathom.

It's a plugin for Zoom and they have a couple of standalone components. And then you get kind of a sidebar where you can annotate a video with like, you know, it's kind of designed for like user research. So a positive feedback, negative feedback, follow up, you know, you can kind of customize it, but users can annotate video, right? And one of the things that I thought was so brilliant about this product was users, and then you have like a workflow where these annotations can sort of be useful, but users

Along the way, users are basically annotating this video feed and giving

the Fathom team, thousands, millions of examples of what a positive feedback looks like as far as in a human interaction. You're creating this label data set such that now you can, they've rolled out, there's later versions of this where a lot of this annotation is now automated. The idea of saying, "Great, we can actually figure out without you annotating whether this was a piece of positive feedback or not." I love that complexity building because

As we go into this world of generative AI, this idea of extreme automation and the tech intuiting what we wanted and sometimes there's these complex situations.

That's an incredibly valuable proposition that they can build off of. So I think it's a great transition to the broader AI landscape. There's so much to talk about with respect to AI, but I think there is consensus, right, that data, having access to unique data or being able to bring proprietary data to bear in any Gen AI context is very valuable for startups today. When you look out at all of these

people who've reinvented, refashioned themselves as AI investors in 2024. And you think about your history in the data world, what's sort of most frustrating to you or what do people most misunderstand or get wrong as they think about the intersection of AI and data or just AI investing in general? Like are there particular areas or theses that you think are overhyped or underhyped or just plain more complicated than people make them out to be?

Yeah, I mean, I would say in general, I probably fall. It's more nuanced than underhyped or overhyped, but I tend to skew kind of the AI is a bit overhyped at this moment on average. The, you know, I think my take in all of this is that, well, there's a couple of things that are really driving a lot of how I'm thinking about the current kind of cycle. One is.

I think it's pretty unique in history, certainly in the last maybe four or five kind of hype cycles across industries where the first product, the first real breakthrough product was effectively an API. Like you're kind of seeing that with LLMs right now, OpenAI, all of these sort of big foundational model companies. Yes, there's a conversational interface, but like it's basically an API.

And I think that that's that plus the reality that I think a lot of the sort of current or incumbent market leaders and their various verticals became the market leaders by sort of disrupting a prior generation of companies that were a little bit asleep at the switch during a platform shift, whether it was sort of desktop to mobile or on-prem to cloud.

But I think those two factors are leading to a situation where incumbents are a lot more aggressive at responding to sort of LLM or AI native upstarts. And it's easier than ever before for the Intuits and Microsofts of the world to just like immediately integrate really like

features into their product offerings, they're at parity with anything a startup could share. So when I'm investing and I encourage people to think about when they're starting companies, like if you are founding a generative AI company and you're looking for investment,

you are implicitly making the statement that the incumbents in this space for the problem i'm solving are not going to be able to respond right and you better really think thoughtfully about that like is that necessarily true that you know the zen desks or intuits of the world are not going to be able to implement loms and if so um you better have a damn good reason you know why they can't come after your particular customer so i think i think the you know going back to the like

sustaining versus disrupting technology, I think a lot of generative AI is going to be sustaining. Your value will accrete to the incumbent. Not to say that there aren't huge opportunities for sort of disruptors to come in, but you got to think a lot harder than the typical founder is right now. That definitely comes to mind. I think the other component of all of this is, especially when you talk about like round pricing and unit economics,

I think founders tend to forget how long it takes for a company to go public and like how many hype cycles you are going through, right? As far as like, if you are raising your pre-seed now, median time to exit is 10 to 12 years. I think I saw it was about 11 years. Like generative AI will be uncool at some point in that 11 years and maybe they'll come back around to being cool again. And so like building your cap table and like not kind of over-optimizing for either rounds too big or, or,

or optimizing for the long people on the cap table are not going to serve you well on that 11-year journey. Let me ask one question, then, Myles, turn it over to you. Kevin, just to restate what I heard you say a second ago, which is there are these incumbents in the tech industry today. Of course, there are always incumbents. Every startup ever always has incumbents they're confronting. But perhaps some of the current AI opportunities that startups are going after

will be particularly fraught given the cost to compute or other infrastructure advantages that the big tech companies have. So we ought to be, have some caution around that. And then you're also making this point, which I think we agree with very much around, um, uh, some of the pricing of these venture rounds is out of whack. And there might also be an unrealistic timeline expectation from founders. Um, and, and the, the, the, how they're connected is that, uh, when you raise, uh,

money in an early round at too high a price, you're setting yourself up for difficult subsequent fundraisers. And some of those fundraisers may happen in a macro environment or VC culture environment where there's not as much of a hype cycle. And so you might be setting yourself up for kind of a painful down round.

That's right. Or you put your company at sort of avoidable existential peril. But yeah, I mean, going back to the kind of incumbent note, some of it is like cost of compute. But I think it's more fundamental than that, where maybe in other situations, you have this sort of disruptor who shows up and says, like, look...

you know, I want to ship something that feels like, let's use Uber, for example, like I want to ship something which is like a connected experience for the kind of feels like taxis or like the classic limo experience, but it's way more dynamic. It's way more digitally native. And we can like basically kind of unlock this idea of like abandoning personal car ownership or something. Right. And when I think about the incumbents,

I'm just going to bet that like they haven't thought about the problem in this way or like with sort of sufficient rigor. And I'm going to get so big so quickly that like by the time they figure this out, like I'm already off to the races. Right. Like, I don't think that that kind of idea of saying like,

Like if I'm doing like generative AI for tax preparation or something, right? And like, I'm going to bet that Intuit hasn't thought about putting LLMs in their products, right? Like that is not a bet that like, if the strategy is that complex, like that's not a bet I want to take as an investor. Like, you know, that kind of awareness gap doesn't exist in generative AI right now. So it's got to be more rigorous than that.

Yeah, Kevin, it's interesting because we were having a conversation internally about what our definition of AI native is, like what it really means to be an AI native company.

Is there an open gap in the market that incumbents are just not going to restructure their companies and products to do? So the point, right, into it can go out, go ahead and integrate an LLM and you can have an NLP based system to do your taxes or whatever it might be versus a company coming in and being AI native from the start. So having an open advantage to rebuilding their entire company from scratch.

leveraging LLMs from the very beginning of their product. And I'm curious if you have any thoughts on that, especially considering Uber had a really open window. They encountered a lot of regulation, but they had a lot of data. A lot of their remote was because of how much data they were collecting and using that.

So I'm curious over your time, like maybe two questions. One, what's your definition of what an AI native company looks like, especially considering the data advantage that a company might have versus incumbents? And maybe second to that is like, how does that develop over time? Like how do you see maybe net new startups making sure that they have some competitive advantage in two to five years? Sure. Yeah. I mean, I think that there's sort of a kind of known, like coming out of the Clayton Christensen sort of innovators dilemma world of,

of where I think generative AI, especially like at the vertical application layer can really develop. AI native to me is we're building a product that pursues a customer segment or a customer profile

which is fundamentally unserviceable by the incumbent offerings, right? Like in terms of, you know, one of our investments, for example, in fund one is this company Fishtail that's doing automation of trade finance. So they've developed a pipeline where they can do automatic risk scoring. They've got kind of a document automation process.

And what it's led to is sort of a cost of servicing, a cost of originating a loan, which is so small that they can pursue trade sizes, like below the capabilities of these sort of incumbent sort of big banks that are a little less automated.

And so it's like literally this is a customer that like we just kind of own because we have this sort of fundamentally different AI native tech stack. I think that there are every sort of new emergent technology will create some set of opportunities that kind of feel like that. I think generally there will be no exception. I think the other thing which is maybe a little more unique to especially LLMs is the ability to kind of introduce customers

called the modern technology stack, the modern data stack.

where the fundamental mode of interaction is something other than I'm sitting at a computer interacting with a piece of software on a monitor. You know, like you've seen, I think we're early in this, but voice seems to be like a pretty, like an effective, viable channel to interact with somebody who's chatbots. I think visual is going to be another component of this. So I think that there are lots of professions out there where they have been like pretty, you

hesitant to adopt maybe V3 SaaS or even mobile apps as a form factor just because it doesn't pass the real life test where voice or video, or maybe you take a mobile experience and you supercharge it with through LLMs. I think those new modalities might create some real greenfield opportunities for a whole bunch of other professions. Kevin, let's talk about founders who are at the very beginning of their journey.

Because one of the things that Village and Rackhouse is doing together this year is opening up an EIR offering for founders who are thinking about the kind of company they want to build at the very beginning of their ideation process. And we're going to put in place some structure for them to help them flesh out their idea, find great co-founders, and go from negative one to zero.

And we can maybe each share some of our lessons and insights for how we like to support founders at this very fragile stage. And of course, ultimately, this is an open call for those of you who are ideating on an idea or trying to think about how to get started. Do reach out to us. We'd love to talk to you about whether you're a fit for this Rackhouse Village EIR program. But Kevin,

sort of more generally when you reflect on founders who you back at day zero, that very formative part of the journey, what are some of the things that you're looking for in the founders that you back? And what are perhaps some of the frequent mistakes that folks make as they're coming up with an idea, as they're doing early customer development, as they're looking for a co-founder, any of the above? - Yeah, I'm super excited to talk about this. This is like one of my favorite projects. I know it's been a few months in the making,

So, yeah, I mean, when I was working with founders and sort of with our direct investment strategy, I always told everybody, like, send me two people and a complete pitch deck, you know, we're ready to kind of talk. And what I meant by that is, one, I obviously want the team and the sort of basic idea of

But when I think about a complete pitch deck, it's basically like I want every slide of that sort of metaphorical deck filled out. So give me the problem you're chasing, give me the solution you want to pursue. You know, think about your business model. You know, obviously, caveat on sort of maturity, but like you think about your go to market, have a fundraising plan put together. Like you got to have kind of all of it strung together. I think.

this EIR program gives us the ability, you know, the Village Network, the Rackhouse Network, the Rackhouse team to work with folks who are maybe have half of the pitch deck completed, right? Or they've got the problem but they don't quite know the solution or they've got a product or they're trying to figure out like the right way to get to that customer they have in mind. So when I'm working with founders, you think about the sort of teams that we want to work with on the Rackhouse side of things,

Like, one, the things that we can't manufacture for you are sort of passion for the problem space or sort of the value system you want to pursue. So I love working with these people where they're like, I don't know what it is, but it's got involved like AI and be good for climate or something. Great. Like that eliminates 80% of the known universe. Like we can help you get through the last 20%.

And the other thing I think that I'm really looking for in this space is like, is I guess you call it like kind of psychological toughness. So I'm saying like, look, early startups are not,

necessarily hard by and large because you're trying to like solve a research project, which has never before been solved. Like that. I mean, there's always exceptions here, but the typical early stage company is hard because the sheer amount of like uncertainty and ambiguity you have to navigate of just like, look, every one of my slides, there's a bunch of white space on it. Like how do I like manufacture some confidence and like figure out a way to just triage, like breaking through all of this uncertainty requires a

a certain toughness, a certain optimism, we can help with that. And I can definitely help you work through all that white space, you know, a hell of a lot more efficiently than you might on your own, but you got to kind of have the guts to run, run through that whole process. And then the last thing I say, just because I'm, we're an AI investor, data investors, sort of a passion for, for leveraging data in any flavor. I think one of the sort of common misconceptions I see with founders is

and we've talked about this with investors is that like the only good ideas right now are generated or relating to generative ai i think we're seeing huge opportunities in every other flavor of ai like we're using logistic progression or

Another thing that I'm really excited about right now are data-enabled marketplaces. So how do you help pools of supply and demand find each other more effectively by just standardizing a data model? That's still a data company. It still fundamentally leverages a lot of our experience. You never have to say the word LLM in your whole founding career if you want to. Yeah.

So that's great. And let's, you know, one thing that's always of interest to me when I reflect on my own journeys as an entrepreneur and also backing lots of founders at inception is this sort of delicate process of looking for co-founders,

and also baking an idea. And one of the things that seems to be true about really talented, ambitious people, whether or not they're the CEO of a company, is that in the early moments of a startup, they tend to like to have input

on the idea, right? Like people tend to be more bought in to projects that they help co-author. And so I think a common question or dilemma that I've been in my own career and I see lots of founders be in is they're kind of a solo, it's one guy or one gal who's just in their head thinking about ideas and they're out there networking and talking, coming up with different paths forward.

And they're thinking about both the business idea and their kind of founder dating, right? They're talking to potential co-founders and thinking about the idea. And if you go too far down the path

with an idea before having a co-founder, the challenge is you could fully bake the idea. And then by the time you find your technical co-founder or business co-founder or whatever, whoever your complimentary partner is, there's not really an opportunity for that person to co-own it with you. And that can be okay. It could just be a very imbalanced power dynamic among co-founders that certainly can work, but that's a trade-off that you're accepting.

But then there are also founders who they'll really prioritize the co-founder path. They'll say, first things first, I'm going to lock arms with someone who I think is high integrity, who I could see myself working with for years and years and years. And we're going to have a general space that we're interested in. Maybe it's AI and climate. Kevin, to use the example you just threw out there. And we're both generally interested in that. But then we're going to go down the idea maze together.

as co-founders. - Right. - And there's pros and cons to that approach as well. What one challenge is that if there's not sort of somebody who's clearly leading on the idea, it can kind of take a long time and it can kind of be diluted.

you know, it can be the worst of all worlds instead of the best single idea as you kind of perpetually compromise and as you try to find something that you can both agree to work on. I'm curious what this prompts for you and what sort of best practices you see if you meet a solo founder at day zero. How do you talk to them about finding co-founders but also fully baking their idea? Yeah, I would say first,

First, the first thing that definitely comes to mind is on the founding journey. It's true whether you're working with us in the eye or not. Like, do not fall too in love with any one idea or methodology or even co-founding for that matter. Like, you should have...

passion and interest, but like there is a lot of risk in like in basically falling so in love with like your own nonsense that you kind of forget to figure out if it's a good business, right? Like to take the idea or like I've always wanted to start a company with this person without really actually being like, do we actually make sense as a co-founding team, right? So one, like I think a little bit of flexibility, a little bit of kind of looseness in this is just pragmatically good advice. I think

What to me is really interesting about the, about the process you outlined, I think we can really support both in the ZIR program. You are a solo founder, you're a solo founder looking for a co-founder, you're a co-founding team looking for an idea. Like there's a lot of ways you can find, be helpful here.

is whenever you're kind of looking for that co-founder, I do think that there are some things you have to just talk about compatibility on. Like, do we have the same values? Are we excited about the same thing? Do I want to work on the things you want to work on for, again, that like 10, 11 year kind of litmus test? Do we complement each other well? All of those are kind of just more observable fit components.

And I do think that with these ideas, the best opportunities I've seen for co-founding teams is like when you come into work one day and you're saying like, man, I've really thought about like,

how we could use AI to model power grids better, to use this AI for climate thing. And I've just been really nerding out on the math of that. And your co-founder is like, "Well, actually, that's exactly right because renewable energy farms are really struggling to figure out where to tighten the grid." And you start building off of each other's ideas.

Like the best thing is like you're both going in the same direction and both like bringing something to the business plan that like you didn't think of. That's the best sort of cooperation. Yeah, I think it's a great point. And, you know, one thing I've been reflecting on, like when I think about more contrarian beliefs that I hold, one of them might be that I think actually mission is a little overrated in the early days of a startup. Like what I find is that in the early days of a startup, founders, you know,

get most animated by the dynamism of their co-founders and like, are we all jamming together? And I actually think there's a pretty wide band of things that people can get excited to work on. Like it doesn't have to be this particular problem. I'm really excited about this particular itch to scratch. Like, no, actually, I think people can often be more flexible on that than they realize if they're super stoked to be working, you know, in the foxhole with a couple other guys or gals who they really respect and feel like they can build with.

And then at scale, if a company is working, people are passionate about it. I mean, there are people who are more passionate about working at what some would say are super boring enterprise software companies than there are people who work at the Red Cross, right? One organization has a really inspiring mission, the other maybe not so much, but if the culture is revving and in this case, the company's being successful and you like your colleagues, you can get fired up to go to work every day.

And so I think actually whether it's really early stage or really late stage, I actually think mission might matter a little less than some people think. But for purposes of this conversation and the program that we're working on together with these really formation stage founders, I would encourage folks to think really hard about their co-founders, unless they want to be a solo founder, that can work. But really think about who you want to lock arms with.

And agree on a swath of ideas that you might want to pursue together, but don't get overly attached too early to any particular idea. Not just because you're not going to allow your co-founders to co-bake it with you, but also because almost certainly once you come into contact with the reality, once you leave the building, as Steve Blank used to say, and go talk to real prospective customers, the idea is going to change and alliterate. And one of the things that

great precedence seed investors like Rackhouse or Village can do with you is help you do that iteration. And

And there will almost certainly be iteration. So don't overly stress, I think, on day zero about the particularity of the idea, but do hold a really high bar and do get really excited about who you're going to go into business with. 100%. And I think that what you were saying there at the end about this sort of efficiency or speed of iteration is one of the things I think that this program can be really great at. One of the sort of mental models I've thought about, even supporting our existing company's

is the like 30 day to 30 minute arbitrage of like working with experts. So funding companies is all about problem solving, right? And there are lots of problems that you run into where you don't know the answer and like you can work it, like you're smart people and it's sort of tractable, but it's gonna take you 30 days, right?

And if you have access to the right people and the right experts who can just kind of help you skip ahead to the answer to the solution, you can solve this problem in 30 minutes and buy yourself 29 days, 23 hours and 30 minutes worth of savings, right? Like Rackhousing Village to me, I think are really good at like maximizing the number of 30 day to 30 minute arbitrages. And you will just iterate through so many more cycles through that kind of time savings and time efficiency. That's why I'm so jazzed about doing this. Yeah.

Kevin, you know, I'm curious just on that like 30, 60, 90 day, you know, upon EIR coming in, one great way to, you know, eliminate ideas is looking at data, trying to get signal, trying to see if you have momentum, whether or not to pivot, you know, data doesn't lie. Ideally, you know, you could be objective about what to pursue based on how many people are adopting something. And so,

I'm curious for early stage founders, how much of conviction is just looking at data and seeing whether or not something is actually working versus like instinctual conviction, right? Like I think I know I have something, it just might take a bit of time and out of the scope of a six month program, see adoption later on.

So I really do think data can be a huge signpost. I'm sort of a big believer of like, if it can be measured, you should at least attempt to measure it. Or I think one of the other common misconceptions is people who are like fundamentally unwilling to like pay for data. And I'm not saying that you got to go like buy, you know, a crunch-based download or a post a pitch book download for 30 grand, but like,

Oh, if I paid a freelancer like $2,000, I bet you that they could probably pull some really relevant data or something like that kind of like pay to acquire. In the grand scheme of building your company, the two grand you spend on that freelancer is going to be so insanely high in terms of ROI. So in general, I'm like, look, if there is an ability to measure it, let's measure it.

I think you need to condition that with enough wisdom, and maybe this is where we come in, to recognize that not everything you think is measurable or not everything is going to be data-driven about launching a company, right? Like, I remember...

one of the really sort of early scaling challenges in Uber where we would, we had a bunch of dashboards that were kind of tracking like organic app downloads or like if an app opened, even if we weren't in your market, right? And so we were like, we built this whole pipeline, like being like app opens by G up, right?

And so we were, we had this huge leaderboard of like, okay, here are the markets we're in and here are the leading contenders. And we were like getting really wrapped around the axle of like, do we launch Seattle 8th and then Chicago 9th or Chicago 8th and Seattle 9th?

And I remember Travis came in and was like, "Gang, look, if we do this right, we are going to be in every city in the United States. So in the grand scheme of things, Chicago versus Seattle or Seattle versus Chicago, this is not what we need to be working. Everybody back to work, just pick a decision and go." Kind of thing, right?

So I think this idea of like be data driven, but recognize sometimes it's just like the time he's been doing the analysis is going to be more costly than like than the times that's at risk figuring out the results of this analysis. And just get back to work. Speed is also an incredibly powerful weapon for entrepreneurs.

And I love your use of the word wisdom there. It reminds me of a meditation master who once told me that, you know, I could read the Buddha's words in a book, but it wouldn't come as wisdom. It comes as knowledge.

Wisdom has to be learned firsthand by studying the nature of your own experience in moment to moment. And I think there is something to this idea that data on a sheet is knowledge, but wisdom is a thing that can contextualize it or that can determine how consequential it should be. Or it's wisdom that can unlock data paralysis and analysis of like, we're going back and forth, the Uber example you just mentioned. So I think that is...

It is the role of great advisors. It is the role of great investors, founder friends who hopefully have done it more than you have, or at least as many times as you have, you the founder. And to have those people on your team in the room in the formative days of a startup can really be invaluable.

I agree 100%. Kevin, thanks so much for the time. Great to have you on the Village Global Podcast. And we're excited to kick off the Village Rackhouse EIR program. We'll post links and follow up in the show notes and look forward, founders, to meeting many of you in the weeks and months ahead. Yeah, thanks for having me. I'm so excited to launch this program with you guys. Thanks so much, Kevin.

Thanks so much for listening to the Village Global podcast. You can check us out online at villageglobal.vc. We'd love to hear from you, your feedback, your ideas, your inspirations. You can email us at hello at villageglobal.vc.