How Founders Find Ideas That Nine in Ten People Reject

1. 0:00 – 2:00

   Intro

   1. GTGarry Tan0:00

      If you only want to work on things that are hot, you're going to find yourself working on derivative ideas that end up being obvious, that end up having five, ten, 100 competitors. It's great for that number one, number two, but guess what? Like, number three through number 98 of all the people in that market, their startups are going to die. Nine out of ten people might tell you you're stupid or crazy, but then one out of ten people might be exactly the person who believes what you believe. Run out and try to find things that humans really desperately want and need, and then you'll figure out the rest. (instrumental music) Welcome back to another episode of The Light Cone. As Peter Thiel says, "Competition is for losers." And as we look at all the AI startups we're working with, increasingly, AI competition is back. So, how do you actually deal with it? You know, well, I think if we go back to Zero1 again, uh, what we would say is, we're looking for how do you think from first principles, how do you actually deal with that competition by being contrarian, and right? Harj, how do you think about it?

   2. HTHarj Taggar1:15

      Yeah, something I've been thinking about recently is, um, probably just over a year ago, we talked about how we were finding it, um, easier than ever to fund companies that were looking for a startup idea and that they could pivot and find an idea. And it felt like the two causes of that were, one, there was just so much greenfield. Like, AI was new, um, there were so many verticals to go after that hadn't been picked over yet, and the models themselves were changing. Like, there was such a, there was a step function increase in new models every few months that just caused the idea space to expand. So you could both... There was greenfield and you could always count on a cha- big change coming that would shake things up, um, and create more ideas and I think clearly we've seen the benefit of that. Like, there are so many vertical AI agent companies that are doing tremendously well.

2. 2:00 – 6:22

   AI verticals are becoming more crowded

   1. HTHarj Taggar2:00

      But I kind of feel like the vibe is shifting a little bit now, where when I'm talking to founders, doing office hours, trying to help them find ideas, it's not as easy as, "Hey," like, "go figure out, like, uh, a vertical where there's, like, a workflow to automate, like insurance or banking," because there's multiple startups in each of these verticals now. Um, and there actually hasn't been, like, a model that's shaken things up, um, for a while now. And so, I think it's becoming more important to think about what's your actual, like, unique insight that is going to enable you to find a good idea and what's, like, the contrarian bet you're going to make, um, so that you can actually stand out from all of the competition?

   2. GTGarry Tan2:38

      So, how do you actually discover a secret, if you will? You know, something that you know and believe that nobody else believes yet.

   3. HTHarj Taggar2:46

      M- maybe we could do some case studies. Maybe, like, why don't we talk about some of the companies we've seen, um, that made a contrarian bet when they were coming up with their idea initially? And, like, maybe that will shake out some things people are gonna learn from.

   4. JFJared Friedman2:57

      I think a helpful comparison to the previous technology shifts that we've had in, in the world, like, if you look at the history of when we invented the internet, when we invented, uh, v- the smartphone, each time there was a, you know, roughly two-year window (laughs) where it was really easy, there was, there was essentially like a, like a modern day gold rush, right? Where, like, the, a whole new class of startup ideas, like, just opened up, everybody rushed in, launched all the obvious ideas. And then, you know, roughly two years after that, the obvious ideas begin to have been picked over and you have to, like, look deeper for a, a secret, yeah.

   5. GTGarry Tan3:35

      I mean, I think it's deeper than merely, like, is it obvious or non-obvious. Non-obvious sounds, like, in your body, might feel like, you know, neutral. But actually, non-obvious feels dangerous and scary.

   6. JFJared Friedman3:48

      Mm-hmm.

   7. GTGarry Tan3:48

      Like, I could devote my life to this, waste 10 years on this-

   8. JFJared Friedman3:51

      Yeah.

   9. GTGarry Tan3:51

      ... and have no outcome. And, you know, I guess the way that manifests, I think, is that, um, there's just, like, ways of thinking that are out there that we pick up in media or in, like, sort of the conversations we have with the people around us that, um, we don't examine. So, you know, I was doing office hours recently with, uh, someone in the marketing space and they came to me and said, "You know what? Like, nobody has ever made a company large enough doing, like, just what we're doing." The weird thing was, you know, we have AI now. Like, this is sort of the perfect moment where no one's doing it. That means there's no competition. (laughs) In fact, everyone else who has been in this space, uh, you know, prior to the AI revolution has failed. So, there are just so many dead bodies stacked up around this idea, you know, how do you know that you're not the right person to go and actually blow this wide open? Especially because there is still this giant new capability. You know, we have Age of Intelligence, the rocks can talk, they can think-

   10. HTHarj Taggar4:57

       (laughs)

   11. GTGarry Tan4:57

       ... and they can actually do the work of real human beings. You know, you might be one of the only people who has the guts to go into that space and actually apply this. And his numbers are going up, the customers are sort of pulling it out of him already. Um, sometimes I'll talk to people about his startup and people say, "Oh, I need that tomorrow." So, you know, this is, like, the perfect example of, you know, sometimes the market will be giving you indications of product market fit, but then you'll be using these mental models from, you know, our people on X talking about it, or is TechCrunch talking about it, or, you know, what will my friends say at a party? They'll say, like, "Oh, you're working on a Tarpit idea." So, it sort of interacts with this idea that, like, competition is back and competition is bad. It's that, like, the obvious things, uh, you know, you're just going to pile up 10, 20, like, you know, some of these spaces seem like there are, like, five or six or maybe a dozen or, like, two dozen different startups, all of which have a real shot at it, and then now's sort of the moment to, again, focus on the secret again.

   12. HTHarj Taggar6:02

       You know, Jared, to your point, the, like, new tech platforms create this, like, two-year window. And so we were talking about this earlier, like, with mobile, like, iPhone launches, Android comes soon after it, um, in the immediate, like, year or two before that, like, there are a lot of, like, sort of...

   13. GTGarry Tan6:17

       ... obvious ideas to go after, like photos. Instagram sort of grew out of that era. But,

3. 6:22 – 9:50

   Non-obvious successes

   1. GTGarry Tan6:22

      like, the actual big winners turned out to be companies like Uber, DoorDash and Instacart. Um- Those were so non-obvious. Like folks who weren't around then don't remember how non-obvious that was. Like when the iPhone came out, there was like a million articles, a million like social media posts about like what kind of companies you could build with the iPhone. And like, I don't think a single person thought that like Uber would be the consequence. I, I think the one in this context that's really interesting to talk about actually might be DoorDash because, um, sort of essentially what we're saying is right now, competition is back. There is more competition to find a good idea, AI idea, than ever. DoorDash entered a very crowded space. Like food delivery, food delivery in general had been around for a while. Mobile was probably like, um, a catalyst for even more food delivery apps. So by the time DoorDash launched, you already had Postmates, um, iterations of it actually. I was just talking to- Grubhub and Seamless were huge companies, yeah. There's actually a YC company, Order Ahead, that Gary- Oh, yeah. ... and I worked with that was actually doing quite well at the time. Um, it wasn't actually delivery, it was, um, let you pick food up from the restaurant, which at the time- Yeah, it was just pickup. ... actually, it seemed like a bigger market maybe than food delivery at the time. Right. They were farther ahead. Like they had way more locations, for instance. Yeah. I mean, interestingly like it, I think it happened even before that, with uh, it was, uh, Lyft before and they were called Zimride before. Yeah.

   2. SPSpeaker7:39

      Mm-hmm.

   3. GTGarry Tan7:40

      And then funny enough, YC had a company called RideJoy- Yeah. ... that was exactly a competitor against Zimride. And they were like neck and neck. Neck and neck, yeah. Yeah. And, um, Zimride decided to try to do, uh, peer-to-peer local, uh, rides. You know, Zimride and, uh, RideJoy were sort of both picking off people from Craigslist, and the idea there was like, "Oh, I'm driving to LA this weekend." You know, this platform would match you with other people and help you get l- you know, a ride to Tahoe from SF or LA. Uh, and then suddenly, Zimride realized, "Well wait a second, like what if we actually did this at a much smaller scale because we have smart, everyone..." You know, 70, 80% of people out there started having smartphones. So instead of having this like long drawn-out email back and forth to like meet at this time so we could get a ride and then, you know, I'll pay up for gas money, it became something that could be used every single day, uh, for short-haul rides. And that was like sort of the first moment that people thought, "Well, maybe you could have a lot of people, more or less like a, a mobile workforce, um, entirely driven by the phone." And then I think that's, that happened like sort of contemporaneously to Uber with their black cars. Yeah. But the funny thing was, and maybe this rhymes with, uh, the lot- often the office hours that we have to do all the time. It's like, I remember meeting with the RideJoy guys and I was like, "Hey, this Zimride thing seems like it's working pretty well. We're seeing, um, a lot of other companies that..." You know, I think we had, uh, just met Apoorva Mehta at Instacart and he was doing it for grocery delivery. Basically the market was sort of pulling it out of all of these startups, and that was exactly the moment where, um, interestingly, interestingly I think those founders said, "We're worried about, uh, the laws. It seems illegal and we don't want to do anything illegal." And they weren't wrong. I remember talking to the Lyft founders, um, the week before they launched Lyft, and they were extremely worried that they would go to jail. And they decided to like roll the dice and launch this thing anyway. (laughs) I think a big reason why other people didn't launch Lyft and Uber before was literally that. It's like it was basically illegal to do that and they were worried that they would go to jail. What's funny now is, uh,

4. 9:50 – 18:05

   End users can get regulations changed

   1. GTGarry Tan9:50

      it does seem like the world will sort of change the laws, as it were, (laughs) if it turns out that, um, the end user or the end consumer wins by that much. Totally. I mean, a lot of great startup ideas are sort of in this gray area of like the law is not totally clear, it's a little bit murky whether it's legal or illegal. Mm-hmm. Even OpenAI is like that, right? Yeah. I mean, they, they crawled the entire web without permission. You could argue that that's fair use or you could argue it's like massive copyright theft. I guess all of these sort of buttress this idea that like non-obvious is not merely like, you know, purely intellectually, you know, not clear that this is going to work. It actually is a little bit more subtle than that. It's like, "Oh, I feel like that's, might be a little dangerous." Like there might be something that, you know, I don't feel comfortable doing. Um, and then really, really great founders sort of sense that as actually a signal. Coinbase was like this as well. Were, were the Coinbase folks like ... I mean, that's all- always operated in a gray area of legality. Coinbase to me feels a little bit less... I don't feel like crypto was well understood enough for it to be clearly illegal. Actually they, they actually couldn't take that approach because they actually needed to get a banking partner, um, in order to even be a, like have the service launch, right? So- Yeah. I would say that, uh, Brian started off recognizing that a lot of the use cases maybe way at the beginning part, you had, um, cypherpunks, like people who were like radically into like libertarianism and this idea that you should not have centralized banking. Uh, but they took it to an extreme where they wanted totally anonymous identities. I mean, it was more s- You know, Bitcoin really very early on was a lot more Silk Road- Mm-hmm. Yeah. ... than what it is today. So I actually think that Brian Armstrong early was like the exact opposite of ... You know, I mean, he was contrarian in another way. Like if you hung out with a lot of people who were really, really into Bitcoin, uh, in, you know, 2010, 2011, 2012, the majority of people you ran across were cypherpunks who said, "F the state." Uh, "You know, F the laws." Right. "And we're going to, you know, have sort of this radical freedom through Bitcoin." And there's Brian Armstrong like doing deals with banking companies- Right, yeah. ... and working with regulators. (laughs) Yeah. Yeah. That, that's what his contrarian bet was. It was that it's worth doing all of this like extra work at a, for a time where it wasn't clear the market even wanted it. Like it was very clear to cypherpunks and like Silk Road all wanted like crypto and anonymous payments. It wasn't clear that regular people would, and so why, it didn't seem like a valuable thing at the time to go and talk to a bank and do a partnership and go through all their like KYC and AML laws because like you'd only do that if you thought that regular people would want to trade crypto at some point. Yeah. Those things actually make your product worse, right? Yeah. Like, like, like forcing people to go through KYC, like...... greatly increases the friction, yeah. A- and beyond that, they're like, they're the exact opposite of what the current market, what, wants. (laughs) Like the current market could not have been more enraged about, like, Brian and Coinbase's approach to crypto at the time. Yeah. I mean, you're just gonna run across people who say, like, "It'll never work." But when markets are brand new and nascent, like, you know, often when you're that early, the thing that passes as, you know, obvious is, uh, clearly wrong actually. (laughs) So that might, might have been like a very profound version of that. Before anyone cancels us for te- saying that, uh, you know, founders might get something out of doing illegal things- (laughs) ... like, I actually don't think that people should, by default, go out and try to do illegal things. Like, I think the through line is not that you should do illegal things or, you know, it- it's that you should think about from first principles, what are the things that markets and people need? Totally. And then I think basically in ca- in the case of, you know, UberX and Lyft, uh, it became very clear within like sort of the first few months of that, uh, coming out, especially in San Francisco where literally we have some of the worst taxi cab like sort of infrastructure and really difficult transit. I mean, basically like that came out of necessity. Totally. And then the quality of life inside San Francisco went from here to here the second that you could actually just freely move around very easily without the uncertainty of like call a taxi cab and they would just literally not show up. Not, not show up like half the time, I noticed

   2. DHDiana Hu14:09

      (laughs)

   3. GTGarry Tan14:09

      And then it's like, how do you actually like live in a city and get around? (laughs) Like, you know, that I, I think that, you know, these services literally made San Francisco 10X more livable in a lot of ways. So, you know, I- Yeah. ... I think it's like, don't just run out and do illegal things. No. Like that's actually clearly net bad. Like- (laughs) No, of course. ... it's more what are things that, uh, people users want? And then thinking from first principles also does actually involve thinking about, well, what are the downsides, right? Yeah. Um, like, how do we mitigate those things? I think, I think the through line is that the laws were dumb and didn't make sense. And if you thought about it from first principles, you'd realize that these laws were actually holding, that these particular laws were actually holding back society rather than helping it. 'Cause they were written in an era before the smartphone when like illegal taxi services were actually kind of a scourge of society. 'Cause you'd have like random people driving around, like claiming to be a taxi, but then they'd like kidnap people. (laughs) And like there was no accountability and like tracking. But now that everyone is in like a smartphone-based system, there actually is accountability and tracking. And Uber is a- was like is actually able to operate it safely. And the taxi medallion system that like ruled the country just didn't make any sense anymore in a smartphone era. Yeah, I think that's the key point. The key point is, yeah, don't go out and do things that the law explicitly says you cannot do. But finding laws that were written in a time before some big tech shift that changes everything and just don't reflect reality can be really valuable. I mean, again, going back to crypto, crypto is a prime example of this, where like- Exactly that, yeah. ... you have lots of securities laws that were really well-intentioned and designed to like protect consumers from scams and, um, and buying things they didn't understand. And so like a big part of the SEC rule is like you have to break things up into like brokers and intermediaries and clearinghouses, but a lot of the rules in the crypto world just don't make sense. They, um ... And so they, they're almost have to be like rewritten while there's a lot of legislation going through now. But I think that's, if you can find things where like the laws don't really cover the world as it exists today and you're sort of brave enough to see what happens if you try, that can be fertile ground. Yeah. The reality is like, uh, actually- (clears throat) ... like that's the role of government. Like this is why we have legislature. You know, one of the things we're trying to fight for right now is uh, for instance, open banking. So you know, I think all of us sitting here and most of the people watching, I think would take things like Plaid for granted. Like, of course I want Plaid. Like the, you know, that's my data, that's my money. And so, you know, what do you mean like a giant bank has the right to charge exorbitant fees, uh, to get access to that data? But you know, that's being adjudicated right now. Like the Trump administration actually has to make some decisions and is an open comment period right now around, uh, whether or not, you know, a giant bank can prevent tiny startups like, you know, YC startups or the startups of people are going to start here. Can they charge a crazy amount or use their terms of service to actually block access to this stuff? And so my hope is that, you know, a functioning government will sort of think about all these things. You know, what the banks might be saying in order to get regulatory capture right now is that, "Oh, it's not safe. We're thinking about the safety of the consumer." But when you really dig into it, it's like they just don't want people to be able to switch off of their banks. Yeah. You know, they, they don't want the money to go to a lower fee place. Like this is their moat, right? Yeah. Regulatory capture absolutely is one. And then the thing that is cool about um, a modern democracy, I think people would debate whether we're, we have that on the one hand. (laughs) On the other hand, like, you know, I think on, in individual cases that might be, you know, the wrong thing happens here and there. But in aggregate, like over a long enough time period, if you can get your product and your service into the hands of enough people, you know, we do still live in an demo- democratic society where, you know, representatives will vote and will like change the laws and that's very much good news. So first principles plus democracy equals open markets and freedom, which like what, that's what we're fighting for. But it does happen.

   4. DHDiana Hu18:04

      I guess

5. 18:05 – 25:10

   Finding contrarian ideas, what founders should look for

   1. DHDiana Hu18:05

      one different question is, what are some of the current contrarian things that founders should be looking at right now? Since what we're saying is there's more competition now as opposed to a year ago, there hasn't been, to your point, Harsh, a big model improvement or one preview has been all about a year or so, that was the big stepping function for model capabilities. And there hasn't been anything thus far. I mean, there's been incremental things with 03 and all that, but nothing like that that makes it harder for startups to find new, new ideas.What are the new things that are also in this gray area-

   2. GTGarry Tan18:44

      Well-

   3. DHDiana Hu18:44

      ... that founders can look into?

   4. HTHarj Taggar18:46

      Potentially one framework for finding them at least, it's not like a prescription for exactly what, is to look at what are, like, emerging playbooks for ways to build startups that might be wrong, or it might be time to flip them. But to give you a concrete example, like, DoorDash, I think was a prime example of this. Because when DoorDash started, there was actually another YC company, SpoonRocket, um, doing food delivery where they would actually, like, cook the food, um, in these kitchens that were-

   5. GTGarry Tan19:12

      Yeah.

   6. HTHarj Taggar19:12

      ... spread around the city.

   7. GTGarry Tan19:13

      Yeah, they operated ghost kitchens all over SF.

   8. HTHarj Taggar19:14

      Yeah. Uh, Sprig was another one, right? And actually in that moment, those companies, I think, came out of this meme of, like, a full stack startup. There was this period where it was seen as just building software is not ambitious enough and, like, the big opportunities are going to be in going full stack. Like, don't just build a food delivery app, actually have the kitchen and make the food, and that's where all the opportunities were, and I think that was becoming more established as, like, the playbook around 2014-ish era. And so in a way that, like, DoorDash's contrarian bet was actually saying, "We're just gonna do deliver-... We're just actually going to have, like, an app and a marketplace and we're not going to try to be a full stack startup." Which was obviously the right bet in hindsight. And so what might be some of the- the, sort of, consensus playbooks for building AI companies that have emerged over the last year that, like, might be wr- might be worth taking the other side of?

   9. DHDiana Hu20:02

      I guess when I can think of this has come up, there's the compound startup notion that Parker Conrad from Rippling really made it popular. I think it hasn't been as adopted, I think people try, but it's actually in practice really hard. I think it's one of those that for certain AI startups it is actually possible to execute on it and not have to wait two years to ship the product, which is what I think Ri- Rippling roughly took for the first version. Maybe an example of a startup is, uh, Campfire. This, uh, YC company that's building basically a AI native version for CFOs to compete versus NetSuite. And it turns out that N- NetSuite is a pretty big piece of software and it's very hard to compete. You can't just do, like, um, kind of, like, a point solution in order to be adopted. So, one thing to reevaluate, instead of doing the standard SaaS approach of let's just build the best point solution, let's actually build the whole thing. Which is not something typically done as an advice that we give for early stage startup because it- you just end up not shipping for a long time and not getting customer feedback. Maybe in cases like this, is the right answer. And Campfire, they've been closing a lot of now big accounts and it's taken a bit of time to get there, but it seems to be working.

   10. GTGarry Tan21:20

       Yeah, they're killing NetSuite, which is wild. Like, how can a startup that, you know, has a dozen people kill NetSuite just, like, hard to believe on the one hand, on the other hand, like, this is the timeline we're on, so... The other thing that's cool about Campfire that we're seeing in a lot of enterprise startups is, uh, you know, with CodeGen increasingly you can actually bring the switching cost closer to zero. And that was actually the- you know, that's- how you can- how- you know, how can Campfire switch someone off of NetSuite on a timeframe that matters? Like, even with the idea of a forward deployed engineer, like, before CodeGen, like, it'd be, like, six weeks of, like, writing custom, like, scripts to, like, convert one data schema to another and, you know, often very specific. Like, if- if it's dynamic schema on one end, that's a lot of custom work that's painstaking and if you get something wrong, like, the end result doesn't actually work and then you have a churned customer. So, that's, uh, a very- like, that's some good news for people, especially doing very complex enterprise projects with, like, conversion. Like, we're seeing more and more examples of enterprise sales that just wouldn't work or it would be, you know, six months to actually get someone to say yes and sign, and then another six months to get a data- uh, a data conversion or data integration done. If the demos are very good, like, that six months could be, like, two weeks.

   11. HTHarj Taggar22:40

       Mm-hmm.

   12. GTGarry Tan22:40

       And then if you can CodeGen and get very, very good at a suite of tools to convert, like, data from the schema to yours, you could have, like, time to value in, like, less than a month when it used to take a year. So, that's, like, very good news for startups out there that are doing enterprise.

   13. HTHarj Taggar22:57

       I think forward deployed engineers we've spoken about a lot but, like, that might be one that's worth wondering if that's going to shift back as well. Like, if you think-

   14. GTGarry Tan23:04

       Mm-hmm.

   15. HTHarj Taggar23:04

       ... like Palantir obviously was, like, invented it, it was an incredibly contrarian thing at the time that they did it, this was to sort of like blur the line between consulting and- and software. It's now certainly become the adopted playbook. We talked about it with Bob McGrew a few episodes ago, um-

   16. GTGarry Tan23:17

       And he was actually fairly skeptical of it himself even though he's one of the people who invented it.

   17. HTHarj Taggar23:21

       (laughs)

   18. DHDiana Hu23:21

       (laughs)

   19. GTGarry Tan23:21

       Which is interesting.

   20. HTHarj Taggar23:22

       Uh, uh, remind us how so. What was his-

   21. GTGarry Tan23:24

       Well, he was saying essentially that, like, he thinks it's, like, being greatly overused and he would, like- thinks it should be used very sparingly in very unusual s- situations as opposed to being, like, a default playbook, which is interesting.

   22. HTHarj Taggar23:35

       It has certainly become the default playbook.

   23. GTGarry Tan23:37

       Yeah.

   24. HTHarj Taggar23:37

       It is actually working incredibly well. I think we've seen, like, that companies are growing at, like, aggressive growth rates because they're employing the forward deployed engineer model. But yeah, if- if I had to- if I had to guess one thing, like, if I were trying to just purely pick what's, like, the contr- like, what's the one thing to be contrarian about, like, judged by just what's become the most entrenched default playbook, it might be the forward deployed engineer approach.

   25. DHDiana Hu23:58

       Actually, yeah, you have a good company that is flipping it on its head, Harj, there's, uh, GigaML.

   26. HTHarj Taggar24:05

       Yeah, I think Giga is taking the concept of the forward deployed engineer, which is e- everything you said, that you- what you want to basically transform sort of, like, your customer schema and business logic into sort of your schema and- and logic.

   27. GTGarry Tan24:19

       Yeah.

   28. HTHarj Taggar24:19

       But it's-

   29. GTGarry Tan24:19

       It's like consulting work basically.

   30. HTHarj Taggar24:20

       Right. But instead of a human forward deployed engineer, they- they have- they're just using CodeGen to do it actually, th- they've built their own AI forward deployed engineer. And a big part of the reason they can win deals faster than their competitors is that the human deployed engineer still takes weeks to get the thing going, which is, uh, like, really fast compared to, like, historic en- enterprise consoling arrangements, but, like, the AI FDE can do it in minutes. And so I think that's, like, the kind of the flipping that plate, like, it's not really an FDE at all really, it's actually just, like, product, um, just like your customer inputting in specs and you, like, deliver them a product instantly. But I think th- that could be an example of a contrarian bet that will pay off really well.
6. 25:10 – 33:40

   Flock Safety and selling to local governments

   1. JFJared Friedman25:10

      Flock Safety could be a really interesting one to, to talk about.

   2. GTGarry Tan25:13

      So, the story of Flock Safety, I remember I was still at Initialized, we were looking at companies at demo day, and then funny enough, that morning, um, all the cars on my street in Noe Valley were broken into. And s- you know, it was a, a professional crew, they came in and, uh, broke into every car on our street. They took all the bags out the back of the car, and then, you know, my house at the time had, like, a little alcove that was dark, and so they brought all the bags into our driveway, and then rummaged through them all. And I, you know, had to cancel my meetings that morning to talk to the police and, you know, I had a Nest camera that was like-

   3. SPSpeaker25:49

      Hm.

   4. GTGarry Tan25:49

      ... captured the whole thing. It was like, "You can't do anything," like, it was a pro crew, like, they were three people. It was like with mechanical, like, military-style precision, and the police basically said, "Sorry," you know, "unless you have a license plate, uh, we can't do anything." And so, uh, that made the first principles investment very easy for me because I, you know, Garrett Langley came in, he was, you know, founder from Atlanta, he had a successful exit previously, uh, but this was hardware. So, they were selling a, uh, camera about this ... it basically had a Raspberry Pi with a camera in it, and then a solar array. And, uh, you know, computer vision had, you know, with ImageNet had been around long enough that you could run it at the edge in the device, and solar had gotten progressed just to a point where it was just good enough that you could run these like, you know, sort of in perpetuity. And so that was his pitch, like, "We're going to go to neighborhood groups, like, neighborhood-

   5. SPSpeaker26:44

      Hm.

   6. GTGarry Tan26:44

      ... associations." And, you know, they already had started, I think, in Piedmont and in Atlant- in the Atlanta, larger, greater Atlanta area. There was like, a lot of things that, um, from a VC- VC standpoint, like, VCs didn't like hardware, uh, they don't like things that, you know, sell to s- potentially small markets. I mean, I think-

   7. JFJared Friedman27:04

      Yeah.

   8. GTGarry Tan27:04

      ... in the investment memo I put together, like, I just called that out directly. It was like, "Well, if you multiply out the number of neighborhood groups and the ACV for those na- neighborhood groups, uh, there's no way that..." I mean, maybe the max was like $50 or $60 million a year or something.

   9. JFJared Friedman27:19

      Mm-hmm. Like that, the actual tan of this thing-

   10. GTGarry Tan27:21

       Yeah.

   11. JFJared Friedman27:21

       ... was like, only $50 million a year.

   12. GTGarry Tan27:23

       Right. I mean, speaking as a VC, like, maybe the best exhortation I can give to the people listening is like, "Do not use that." Like-

   13. JFJared Friedman27:31

       Yes.

   14. GTGarry Tan27:31

       ... you know, it's merely, uh, an indicator. It's, like, useful to know, but, like, neither investors nor founders should use that as like, cross it off the list, right? Uh-

   15. JFJared Friedman27:42

       And news being run out of Atlanta, Georgia.

   16. GTGarry Tan27:44

       Right.

   17. JFJared Friedman27:44

       Right? So, it's like three things that make it, made it, like, basically unfundable. (laughs)

   18. GTGarry Tan27:47

       Yeah. But, you know, I mean, that's why, like, I think I was, uh, hanging out with Brian Singerman at Founders Fund for a little bit and, you know, one of the things we always talk about is, the more rules you have about investing, uh, the more ways you can basically talk yourself out of making a lot of money in venture. (laughs)

   19. JFJared Friedman28:04

       (laughs)

   20. GTGarry Tan28:04

       So, I don't know. I mean, to the extent that that is useful for founders, like, I think that that's true. Like, you know, we did a whole episode last episode about how, you know, you should be aware of the seven powers and moats, but it would be the stupidest thing in the world-

   21. JFJared Friedman28:17

       Yeah.

   22. GTGarry Tan28:17

       ... to use that as your only criteria for whether or not you should work on a company. And likewise, you know, I think increasingly even in our office hours, we find ourselves giving that advice with, uh, founders. It's, you know, it is much better to go from first principles from, uh, what does society need? What do users need? You know, what ideas li- could only happen now? And then the rest, like, you can kind of figure out, right? Like, the same calculus you could do for Coinbase very easily, like, the, you know, total market for Bitcoin, you know, was not trillions of dollars. It was l- like probably on the order of tens to hundreds of millions of dollars at the moment that Coinbase really started thinking and coalescing around that. Some of this is the good news. It's that like, the people who actually change everything are the founders, actually. And so if you only want to work on things that are hot, you're going to find yourself working on derivative ideas that end up being obvious, that end up having five, 10, 100 competitors. And then, you know, it's great for that number one, number two, but guess what? Like, number three through number 98 of all the people in that market are going to die. Like, their startups are going to die. So yeah, this is just a, a little bit of a profound version of that story. I mean, Flock Safety today solves 10% of all reported crime in the United States.

   23. JFJared Friedman29:39

       It's crazy. I feel like that's a generalizable lesson of Flock Safety, which is, um, at any time, if you have some startup idea and you go to talk to a bunch of VCs about it, a lot of them will give you feedback, and I could imagine that if Garrett had gone and talked to a bunch of VCs, they would have told him like, "Oh, this isn't VC fundable, selling to local governments, hardware, like, you should do B2B SaaS." It's cool that he didn't, like, he ended up in a market where, like, he had very little competition perhaps because everybody else, like, thought it was too weird.

   24. GTGarry Tan30:07

       I think if you're just really razor focused on your customer and the actual need, like, it just became so obvious. I mean, I literally got a Flock Safety in front of my house and even though it did not end up catching, um, that crack band of thieves, I felt way safe-

   25. JFJared Friedman30:22

       Mm-hmm.

   26. GTGarry Tan30:22

       ... when I had it in front of my house. And, um, I think that that's what happened with their communities. Uh, you know, especially like, I remember doing office hours with Garrett and he's like, "This is insane. We actually caught, um, you know, someone who had kidnapped a kid."

   27. JFJared Friedman30:38

       Whoa.

   28. SPSpeaker30:38

       Wow.

   29. GTGarry Tan30:38

       Right? And so when you look at the actual human impact, I mean, going back to what we were saying earlier, it's like, don't run out and try to do illegal things. Run out and try to find things that humans really desperately want and need, and then you'll figure out the rest, honestly. Like, if you're just-

   30. JFJared Friedman30:54

       Yeah.
7. 33:40 – 36:42

   The sci-fi founder and "impossible" big ideas

   1. GTGarry Tan33:40

      used for it.

   2. JFJared Friedman33:41

      Diana, do you have another example of places to look for contrarian bets?

   3. DHDiana Hu33:44

      I think one category is sort of the sci-fi founder, is really going after ideas that most people are scared to build because they're just so freaking hard and sometimes science and physics, or, uh, laws may need to be rediscovered to be possible. One example that really comes to mind is, uh, OpenAI. It wasn't clear whether AI was going to be a thing when it got s- when Sam started it out at YC, and it took many years. It seemed to be mostly a tinkering project for researchers that were publishing papers, that were doing kind of side quests. I mean, they had the, uh, Rubik's Cube solver, they were solving DOTA, and it was unclear how all these would come together to what OpenAI became today.

   4. JFJared Friedman34:34

      And also something that people forget is that when OpenAI launched, it got mostly negative press. There were some people, like a small, like, group of, like, techno optimist people who thought it was really cool, but by and large, most people, especially the AI researcher establishment in academia and in other companies, mostly just, like, were s- extremely negative on the idea that a bunch of, like-

   5. GTGarry Tan35:01

      20 and 30 somethings.

   6. JFJared Friedman35:02

      ... th- th- thirties, the somethings-

   7. GTGarry Tan35:02

      Really twentysomethings.

   8. JFJared Friedman35:03

      Yeah. Could create AGI. They're like, "You know, we're the experts in this. We've been doing this for, like, 50 years. If there was a way to do it, we would have already-"

   9. DHDiana Hu35:11

      Hm.

   10. JFJared Friedman35:11

       "... done it. These kids don't know anything. They haven't published any papers yet. There's no peer review here."

   11. GTGarry Tan35:16

       Yeah, not publishing papers was a really big critique, and you know, especially around the scaling laws. Like, I think there was this aspect of they're spending how many millions of dollars on GPUs, on projects that, uh, will not cause more papers in the world to arise? Papers were, like, the thing that, uh, they were trying to, you know, sort of, you know, paperclip optimize for, which was totally the wrong thing. Like, the thing that, uh, really great builders optimize for is, like, outcomes for customers and users.

   12. JFJared Friedman35:45

       It was the same thing when Elon started SpaceX. He wasn't the first billionaire to start a space flight company. I think he was, like, the fifth billionaire to, like, try to start a space flight company, and so all the press was like, "Oh, look!"

   13. DHDiana Hu35:54

       (laughs)

   14. JFJared Friedman35:54

       "Another billionaire-"

   15. DHDiana Hu35:55

       Hm.

   16. JFJared Friedman35:55

       "... like, there to squander his fortune on, like, rockets."

   17. GTGarry Tan35:58

       (laughs)

   18. DHDiana Hu35:58

       And the whole idea of building reusable rockets was blasphemous, right?

   19. JFJared Friedman36:02

       Yeah.

   20. DHDiana Hu36:02

       The... I think he went and talked to, um, rocket scientists and they were like, "This is not possible."

   21. JFJared Friedman36:08

       Yeah.

   22. DHDiana Hu36:09

       And after many years and many launches, it didn't work out.

   23. JFJared Friedman36:12

       Yeah, and th- and then every time a rocket blew up, like, there would be another huge wave of negative press. So, like, for both of those companies, like, it required the founders to, like, stick to their guns in the face of most people telling them that they were, like, stupid or crazy for, like, a long time.

   24. GTGarry Tan36:25

       I mean, nine out of ten people might tell you you're stupid or crazy, but then one out of ten people might be exactly the person who believes what you believe, and then you're contrarian and you become right, because it's necessary to actually attract and be a magnet in the world for all the people who agree with you. Uh,

8. 36:42 – 37:42

   Outro

   1. GTGarry Tan36:42

      I hope you all take a moment to really think about, how do you know what is real and correct in the world? And re-examine all of the sources of these things. And if it's coming from users, coming from your own personal experience or the experience that, of people who you directly talk to, uh, great. Like, that's probably pretty good, verifiable stuff that you should use as a substrate of your reality. Uh, but if you're doomscrolling on X, you're, you know, sort of listening to famous people, like... You know, to be frank, even including us.

   2. JFJared Friedman37:14

      (laughs)

   3. GTGarry Tan37:15

      You know, we're all N equals one. The only people that matter are the people who you care about, who have certain problems, and your ability to solve it, and your ability to attract all the other people who want to solve those problems too. So with that, we'll see you next time. (instrumental music)

Episode duration: 37:42