The Exchanges

Q What was like the light bulb moment for you? When did you flick into, and this needs to be my priority?

A It's hard to say that there was, there was one moment. Um, we've always had this vision at amplitude of a self-improving product where you have a product that dynamically responds to your user feedback. So it knows what features you like. It knows when you're getting frustrated and stuck. Um, it knows how to change things based on your input as a user. Um, and I had always thought this vision was like, you know, 10 years out if, if, if, you know, if, if even that close. And one of the things that's becoming clear, it was actually a lot closer. Uh, that, that moment was actually probably a lot closer than we had realized, uh, because of what we saw happening on, uh, the coding side with AI. I said, okay, look, we, you know, whether or not a self-improving product is gonna be possible, it's clear at least a step towards it, we're gonna have to go do it. So, and the way to do this is we're gonna have to train the organization on this. I started working with, uh, James, uh, the founder, CEO of Command, as well as Wade, Um, our engineering leader to figure out how do we train the organization on AI, and then we, we came up with an AI week, and unfortunately for a bunch of reasons, we weren't able to actually do that until June, but that was a, that was a key, uh, pivot point. What we did was we got a bunch of the existing leaders in the organization. So, you know, our VPs of product…

AI assessment note: “because of what we saw happening on, uh, the coding side with AI.”

Q I don't think everyone realizes how Amplitude started out. Can you maybe talk about your journey to actually finding the Amplitude idea and maybe reflections on it a decade later?

A So before Amplitude, um, we started this company called Sonalite, which was a voice recognition I was like, early version of Siri, and it had this really amazing demo, actually, where you could, it listened in the background for your voice, and this was before any of the Hey Siri, Hey Alexis stuff. Um, and so it, like, listened in the background on Android phone, and as far as we could tell, we were the first to come out with the technology. We didn't know anything about what made for a successful product, or business, or company, and we were just kind of taking our shot on that. Um, we were just two kids at a, at a college, um, that We're like, okay, what's a problem that seems barely at the edge of possible? Let's, let's, you know, that's probably an interesting place for us to, to go to. Um, you know, and we ended up choosing, choosing voice recognition, not because we were passionate about the space. So we, we did YC with that. We did, uh, went through the whole batch, did demo day, did this amazing demo on stage. We got tons of press written about us. Uh, but the product and the tech, it was just not good enough. It wasn't a good enough, uh, product. And so we ended up right after demo day, deciding to shut that down. Um, and then We had always, uh, built our own analytics in-house, uh, you know, it's like, it's what you do as an engineer. You're always like, I want to bui…

AI assessment note: “So before Amplitude, um, we started this company called Sonalite”

Q I would love to go deeper into that. You mentioned that some of the companies you work with are getting 40% plus response rates, and it's not just brand, it's that they're sending these really well-crafted outbound emails. Um, what exactly are they doing to make these emails and messages so great? Like, any examples that might spring to mind?

A Yeah, I think the, There's two things that, that they do well. The first is creativity on who they actually send the outreach to. And so they'll have pretty creative sourcing strategies or ways of finding talent. Maybe that's people who went to the same high school as you. Maybe that's people who, um, happen to have worked at a similar company previously, and you have some connection through that. So really going deep and thinking about every individual person, why am I actually reaching out to them? And then that also gets reflected in the outreach text. And so, um, for 40% response rates, those are definitely going to be Personalized. And that may mean you spend five minutes on personalizing each outreach message. It takes a good amount of time. Um, but it's also worth it because you're going to be getting in touch with people that are definitely not going to be responding otherwise.

AI assessment note: “There's two things that, that they do well. The first is creativity on who”

Q One question I have, I think that every founder has on their mind is that when it comes to sourcing, isn't effectively everybody sourcing for like the same people and it's highly competitive and how, have you seen any strategies for how can you be creative in your sourcing to find people who might not be so competitive but are equally talented?

A I think being able to find people who are non-obvious on paper, um, but then become obvious throughout your interview process Is a real advantage. It's also really hard. And so I think there's like no kind of single path that, that makes it work. Some things we've seen people do that have worked for them is in some cases just looking through GitHub and basically clicking through contributors to open source projects. In other cases, the Twitter strategy sometimes goes in that direction because, uh, even if it's a Twitter profile you recognize, it might not be like a LinkedIn profile that you would click into. I think there's often a lot of indexing, especially amongst founders for colleges and A lot of great talent may not have gone to a great college, um, especially from, like, your competitive advantage perspective. If they went to a good university and worked at a good company, they're going to get a lot of outreach anyways. If they've maybe only done one of those two things, they might still be really talented, and they might not be getting as much outreach.

AI assessment note: “looking through GitHub and basically clicking through contributors to open source projects.”

Q As it gets towards more of the, you've identified someone you want, you've interviewed them, you know, you want to hire them. You mentioned a lot about sort of convincing them to join you over a bigger company. How hard should you try actually to convince someone? Um, and when do you know if maybe you're trying too hard?

A It's again, a little bit like enterprise sales. Like, Knowing what does the candidate or in the sales comparison, the prospect actually want, what are they really looking for? And you might get signals early on that they really want to be in big tech, and that can be really hard to convince them otherwise, especially if it's like, say, comp package related or stability of job related. It's probably not worth fighting that battle because they are likely going to end up going that direction. I think in the later stages, if they have indicated they want to be at a startup and you've kind of, you know, checked in on that a few times and you're pretty convinced they want to be at a startup, then you should fight really hard. Because then it's you against a different company and, um, you should try to win that.

AI assessment note: “It's probably not worth fighting that battle because they are likely going to end up”

Q than they were on engineers. And you know, if you're coming into YC and you're interviewing and you're like, Hey, we're going to be spending more money on lawyers than engineers. Like my gut is that the YC partners are going to say, okay, like maybe you're not building in the right business here. Like really, are you as a three person team going to be able to be successful?

A Yeah, this was actually really what happened for us. A lot of those application cycles in 20, 21, 20, 22, like, We will look at it and feel like, okay, like what's the domain expertise you need to build something really good here? And it was actually like, if you're building a hardware company, you're building robots, you kind of need to have like some experience building hardware and robots. And we felt that with crypto, it was actually like, you needed to really be paying attention to the regulatory aspects of it and understand securities law. And I know if you're putting like real estate on the blockchain, you need to understand what all of the legal implications of that were. And a lot of the time we'd be in interviews and we just feel like the teams had no understanding of it all. And it was unclear how you would

AI assessment note: “Yeah, this was actually really what happened for us. A lot of those application cycles”

Q Okay, Jesse, so AI had its sort of magic moment with the chat GBT launch. Do you feel like crypto is maybe going through its own ChatGPT moment right now?

A Candidly, no. And I think the reason I say no is because if you ask a friend or family member, I don't think they'd say, oh yeah, I have had a magic moment with crypto. I think instead they might be fearful. They might not actually know what crypto is doing on a day-to-day basis. And so the thing that's powerful about that is that that's the best time for us to be building and the best time for YC entrepreneurs to be trying to break through And the moment is now to break through because we've done the infrastructure work, the tools are in place, the stable coins, the chain, and now it's about putting it together into that magical experience that grows like wildfire across the world.

AI assessment note: “Candidly, no. And I think the reason I say no is because”

Q attention right now is AI, and we're just seeing all these magical AI companies building great products and growing really quickly. It seems like AI and crypto should be a very natural fit with each other. Um, so maybe can you talk a bit about that? Like, where do you see the opportunities for those two technologies intersecting, and maybe how does Coinbase think about the opportunities for founders there?

A I think it is a special thing that crypto and AI are growing up together right now. Because when you look at the problems that AI has, I think you can solve a bunch of them with With crypto. And, and two of them stand out to me. One is in a world where AI proliferates, um, it's gonna be really hard to know like what's real. And crypto provides a level of hardness and verification that I think can connect to AI in a way that solves that problem. Like if you have millions and millions and millions of things being created, being able to use crypto rails to authenticate them and verify them and say, hey, these things are connected. These things are real. I think that that's gonna be a huge unlock. The other one, which I, I think I'm maybe more excited about is that AI is about programmability. It's about enabling agents that at their core are computers and writing software and reading software and consuming software. And what better substrate for those agents to be operating on than money as software, right? If I'm an agent and I want to be sending money to other agents or transacting, I want to be using something that's natively built for me. And I think that's exactly what we're seeing happen with crypto right now is that crypto is plugging in as a platform. That enables agents to transact natively. So instead of going and trying to operate in a browser over the legacy rails, the…

AI assessment note: “crypto is plugging in as a platform. That enables agents to transact natively.”

Q So if we zoom out, like you could, stable coins are essentially just like one type of tokenization. Maybe could you explain what does that actually mean to sort of Tokenize something, and then are there other types of assets you're excited about seeing be tokenized and brought on chain?

A If you think about the existing financial system, um, you, you can kind of think about all of these large classes of assets, right? Stablecoins map to, uh, you know, fiat currencies, right? You have the dollar, you have the euro, you have the yen, you have, you know, hundreds of fiat currencies from all the countries around the world. That's obviously not the only class of assets. You also have stock. You have bonds. You have real estate. Uh, you have all of these complex debt structures. And all of that today, um, sits inside of the existing financial system, right? Maybe, uh, it's the, the, the stock certificates that sit with the DCCC. Uh, you know, it's, it's this whole world of records that map to financial assets that sit inside legacy systems. And so when we're talking about tokenization, I think one of the big swaths of tokenization, I think the one that most people think about when they're thinking about tokenization is basically how do we take all of those asset classes and move them out of the legacy books and records and into this new programmable environment? How do we move them from, ah, the, the, you know, records that maybe started a hundred years ago when we had the first stock certificates into smart contracts that live on base? And that is a massive opportunity. Right? Like, again, there's trillions and trillions of dollars of assets in the world. And I think…

AI assessment note: “take all of those asset classes and move them out of the legacy books”

Q And how does this play out in practice?

A You'll have a product and you go to a new customer site. You, you start working with a new customer and the, the problem that they want you to solve is not a problem that you've ever solved before, but you believe that it's one that with a little bit of work, maybe a lot of work you can solve for this particular customer and you'd be making a huge impact for them. You'd be delivering an outcome to them that would be extremely valuable for them. So you take the product that you have and the FDE with help from the product team figures out how to deliver that outcome, how to build that use case, how to, you know, deliver the piece of software that you've built in a way that actually works for the customer.

AI assessment note: “You'll have a product and you go to a new customer site.”

Q Some, like, Don Draper, like, who wears a suit and has worked in the DOD for 20 years and, like, takes generals out to Steak dinners and things like that. And that's actually not what you guys did, right?

A Well, I mean, there's two angles as one is, uh, we talked to a lot of those people early on and they said, why the hell would I work with a Silicon Valley company when I could work with, you know, a big five defense prime? Uh, and then even when we talked to people who, you know, seemed like they might be successful in this role, it was just very clear to us that they wouldn't mesh with our culture and they wouldn't actually be successful. And when we tried doing something like this, it almost never worked. And so, what we found was very different, and, and I think the difference between sales-led product discovery and FDE-led product discovery is that sales-led product discovery, you're talking to people from the outside. And again, this is important very early on, but it's not as effective as the FDE-led product discovery, where you're solving these problems from the inside. So, you know, the scope of a, of a traditional implementation might be You start with something that's pretty close to what the product does, but you want to be solving one of the key problems that leadership has identified. If you're not solving one of the top five priorities for the CEO, it's probably not going to work. They probably won't have the energy to persist through the much more challenging route of getting effectively a new piece of the product built in a way that worked for them. Then once yo…

AI assessment note: “we talked to a lot of those people early on and they said, why the hell would I work”

Q So in the context of how all these FDE companies price very differently based on outcome, how does that fit in with now the culture doing demos? Because there's this thing in, at least in SAS, or I used to get this pushback from my engineers, demo-driven product development, it would be sort of looked down upon. But in this case, it's different for FDEs, right?

A One of the interesting things that happens there is because you have to go repeatedly show this to new customers, you're forced to give these new demos. But, but actually I think demo-driven development works really well if you have the right kind of product. So, you know, in the early days of Palantir, we actually had one demo. It was a flow where you're, you know, stopping a terrorist plot. And we started this with, you know, just one of our features. And every time we integrated a new feature, we had to think to ourselves, how do I show that this new feature is actually helpful for the analyst who's going through this demo, who's stopping this plot? You know, when we integrated a histogram, we had to say, well, how do we actually use this? How does that work with the existing features that we already had? And we went this, you know, we integrated a map and we had the same question. And if you think about the world from what am I building? Then, you know, you're thinking about your capabilities. You might think of each of these features individually and how to build the best, best version of these features. But when you're building a demo, you're thinking about it from the customers. Perspective and a really good demo is something where you show it to the customer and you are creating desire. In that customer for what you're doing. They have to see what you're doing and just …

AI assessment note: “actually I think demo-driven development works really well if you have the right kind”

Q What did you learn in that first version? Because you, you built a code editor from scratch. You guys haven't done the whole forking yet.

A Yeah. We had the fear of God in us. I mean, we had, people hadn't, hadn't really liked some of the things that we had built for a while. So I think that, you know, we were kind of all in on it and very focused, but what did we learn from that? Um, I think that we learned kind of the first initial set of AI features where, you know, when we started, I think that there was just one key command, and it pulled up this, like, universal remote in the editor, and then you asked it to do something, and then entirely the AI would just figure out, oh, do you, what, what, what exactly do you want it to do? Um, you know, do you want something back that's like a chat response, or do you want, um, like a code suggestion that you can then take, or do you want it to go search around your code base and answer a question, or do you want it to go spin for a really long time or a short time? And there wasn't a lot of control, and I think that we learned, you know, given the tech at the time, Um, at the end of twenty-twenty-two that you actually, it has to, the form factor has to look a bit different, and so we learned kind of the first early AI features that then became part of the core of Cursor from iterating both for ourselves and also giving it to people. I think another thing we learned was, you know, we were very rapidly building a feature complete version of what we want in a normal code ed…

AI assessment note: “we learned kind of the first early AI features that then became part of”

Q That was, uh, awesome talk about all the scaling laws, and recently, Anthropic just launched Cloud Four, which is just available. Curious, uh, how does it change what is possible as all these model releases keep compounding for the next 12 months?

A I think that, ah, we'll be in trouble if it's 12 months before, before an even better model comes out, but, ah, I guess, ah, a few things with, with Claude IV. I think that, with Claude III. VII's Sonnet, Uh, it was already really exciting to use 3.7 for coding, but I think something that everyone noticed was that 3.7 was a little bit too eager. Um, sometimes it just really wanted to make your tests pass, um, and it would do things that, that you don't really want. Uh, there are a lot of like try accepts, things like that. Um, so with cloud four, I think that we've been able to improve the model's ability to act as an agent Specifically for coding, but, but in a lot of other ways, for search, for all kinds of other applications, um, but also improve its supervision, the sort of oversight that I, I mentioned in my talk, so that it, ah, it follows your directions and hopefully improves in, in code quality. I think the other thing that we've worked on is improving its ability to, ah, save and store memories, and we hope to see people leveraging that, because Claude IV can blow through its context window with a very complex task, but can also, ah, store memories as files or records, retrieve them in order to sort of keep doing work across many, many, many context windows. But I guess finally, I think the picture that scaling was paint is one of incremental progress, and so I think …

AI assessment note: “improve the model's ability to act as an agent Specifically for coding”

Q human approval before they would send the reply for a customer. But one thing that has changed just in the spring batch, I think a lot of the AI models are very capable to do tasks end-to-end, to your point of that, which is, ah, remarkable. Founders are selling now directly replacements of full workflows. How have you seen this translate to what you hope the audience here will build?

A I think there are a lot of possibilities. Basically, it's a question of What level of success or performance is, is acceptable. There are some tasks where getting it sort of 70% right is, is good enough, and others where you need 99.9% to, to deploy. I think that, honestly, I think it's probably a lot more fun to build for use cases where, ah, 70 80% is good enough, because then you can really get to the frontier of what AI is capable of, but I think that we're sort of pushing up the, The reliability as well. So I think that, ah, we will see more and more of these tasks. I think that, ah, right now, human AI collaboration is, is going to be the, sort of, most interesting place, because I think that for the most advanced tasks, you're really gonna need humans in the loop. But I do think in the longer term, there'll be more and more tasks that can be fully automated.

AI assessment note: “a lot more fun to build for use cases where, ah, 70 80% is good enough”

Q Now, other question is, you have an extensive training as a physicist, and you're one of the first to really observe this trend with scaling laws, and it probably comes from being a physicist and seeing all these exponentials that happen naturally in nature. How has that training come about with, ah, being able to perform like the best research in the world with, with, with AI?

A I think the thing that was useful from a physics point of view is looking for the biggest picture, most macro trends, and then trying to make them as precise as possible. So I remember meeting, like, kind of brilliant AI researchers who would say things like, learning is converging exponentially, and I would just ask really dumb questions like, are you sure it's an exponential? Could it just be a power law? Is it quadratic? Like, like exactly how is this thing converging? And it's a really dumb kind of simple question to ask, but basically I think there was a lot of fruit to be picked and, and probably still is in trying to make the big trends that you see as precise as possible because that, I don't know, it gives you a lot of tools. It allows you to ask like, What does it really mean to move the needle? I think with scaling laws, the, the holy grail is finding a better slope to the scaling law, because that means that as you put in more compute, you're going to get a bigger and bigger advantage over other AI developers. Um, but until you've sort of made precise what the trend is that you see, you sort of don't know exactly what it means to beat it and, and how much you can beat it by and how to know Systematically whether you're, you're, you're achieving that end. So I think those were kind of the tools that, that I think I used. It wasn't necessarily like literally applying …

AI assessment note: “useful from a physics point of view is looking for the biggest picture”

Q Which is very much, uh, the Jevons paradox. As intelligence becomes better and better, people are gonna want it more. Not that it's driving the cost down, which is this irony, right?

A Yeah, absolutely. I mean, I think that, uh, yeah, that's, that's certainly something that we've seen, that there are certain, uh, certain points where AI becomes accessible enough. That said, um, I think as AI systems become more and more capable, um, and can do more and more of the work that, that we do, it's going to be worth it to pay for, uh, frontier capability. So I think it's a question that I've, Always had and continue to have is kind of like, is all of the value at the frontier, or is there a lot of value with kind of cheaper systems that aren't quite as capable? And I think this sort of time horizon picture is maybe one way of thinking about this. I think that you can do a lot of very simple bite-sized tasks, but I think it's just much more convenient to be able to use an AI model that can do a very complex task end-to-end, Rather than requiring us as humans to sort of orchestrate a much dumber model to break the task down into very, very small slices and put them together. So I do kind of expect that a lot of the value is going to come from the most capable models, but I might be wrong. It might depend, and it might really depend on the capabilities of AI integrators to sort of leverage AI really efficiently.

AI assessment note: “it's going to be worth it to pay for, uh, frontier capability.”

Q And then you realized, Actually, we need to do something else. Like, what was it to actually pivot it to, you know, what it is today?

A It wasn't a straight B line. Um, we, I mean, being programmers ourselves and being inspired by products like Copilot and, uh, also papers like the early codex papers. I remember at the time, one of the things we did to justify to investors that they should kind of like invest in our crazy cat idea Is we did the back of the envelope math for what Codex, the first coding model, costed to train. From my memory, it only cost about 90 K or a hundred K by our calculations. That really surprised, surprised investors at the time and was kind of helpful in us getting enough money to, to pursue the, uh, CAD idea where you had to start training immediately. So we always knew about coding. We were always excited about it. We were always excited about, you know, how AI was going to change coding. We had a little bit of trepidation about going and working on that space. Because there were so many people already doing it, um, and, uh, we thought Copilot was awesome, and, you know, there were dozens of other companies working on it too at the time. When we decided to put aside CAD, which was a little bit of an independent idea, that was sort of the science not really working out, us not really being excited about that domain, the thing that drew us back into coding was our, our personal interest, and the thing that gave us the confidence then to continue with it was, one, seeing the progress t…

AI assessment note: “the thing that drew us back into coding was our, our personal interest”

Q So driver supply and consumer demand. What was that like, you know, sort of over the 10, 12 weeks?

A Yeah. So the number one thing we were, we were very scared about was whether or not consumers would want this product. And it's because delivery is not a new idea. It's been around since horses. It's been around forever. And, and so, Um, and obviously the US is a very highly capitalistic market. So if something doesn't exist, maybe there's a good reason why it doesn't. And so we wanted to make sure, um, from the get go, whether or not consumers would pay us for the service. So that was one big question. The other was whether or not we knew that restaurants had a need for it. We didn't know if they would pay us. Um, that was a second, you know, pillar. And then the third is whether or not, uh, there would exist drivers who would actually Want access to this? And so early on, uh, we did all the deliveries. One of the best parts of doing all the deliveries besides teaching us what are all of the steps within a delivery is what customers wanted. And our earliest customers tended to be, uh, families with young children. It tended to be the mom who, um, made a majority of the decisions when it came to, to meal prep and food. And so They told us what they wanted. They told us what was important. They told us what restaurants we, that they preferred. Um, and, but the most important thing was that they kept coming back in again without our throwing advertising at them, coupons or discou…

AI assessment note: “our earliest customers tended to be, uh, families with young children”

Q At what point did you decide to apply to YC?

A While we were in school. So we, we, we actually, the first two weeks of the summer batch of Uh, were the last couple weeks of our time at Stanford. So we had a few weeks of overlap or a couple weeks of overlap. I remember very specifically my classmates or some of our classmates, you know, planning their exotic vacations to Europe or other, um, interesting areas for, for the summer. And when they asked me what I would be doing, I said, I'd be delivering hummus for my Honda. It was a very different type of, uh, answer, but Yeah, you know, we had a lot of fun. I mean, the, the, the, the earliest days, all four of us did all the deliveries.

AI assessment note: “While we were in school. So we, we, we actually, the first two weeks”

Q You just continued to grow. You kept, you raised, you know, multiple rounds of funding. Your Series C was actually a down round. What was that like? You know, you're growing the business, raising more money. You know, you're fighting off competitors, you know. How did you manage that, and what was that experience like?

A It was very tough. I mean, because on the one hand, you see all of the internal metrics going in the right direction. You're growing organically. You're growing quite quickly organically. You see that the market is perhaps larger than you expect. These are all the positive signs on one side of the equation. On the other hand, Um, to your point, we're also investing in scale because this is a business where, um, you need enough order volume to make the math work. To get there, you have to invest before you get the demand. And so we needed to raise capital. And there were a couple, there were a few years actually in a row, three years in a row, 2016, 1718, where I continued to struggle to raise capital. And I think this was the part that was quite difficult for us, where you have a company whose product seems to be Moving in the right direction across any metric any way you want to cut the data. On the flip side, you know, I'm receiving hundreds of rejections, um, for investment. I think this was, this was certainly one of the most difficult periods, um, so far that we've had to overcome.

AI assessment note: “It was very tough. I mean, because on the one hand, you see all”

Q You spent an hour thinking about the idea. You spent five years working on it. And as I recall, you got some real scale. You guys raised a fair amount of money. Like what's the, what was like the summary or the post-mortem on what happened with Flycar?

A At our peak, we had 17 different airport locations. It was a very, um, we didn't take, like we took a very, uh, sort of asset heavy Sort of approach, right? We had, like, leases at 17 different airport facilities. We had shuttle services to and from the airport. I mean, we were washing and gassing up, like, hundreds of cars every day. Like, it was a very, very operationally intensive business, and, ah, and, and I, I sort of equate it to, like, if you think about, like, making money in a business, um, is like, kind of like, you know, if you have a lemon, it's like squeezing the juice out of the lemon, and like, what's left is, like, the money you make. Flight car was the type of business where, You know, it's the last drop out of the lemon. That's the money you keep. And so if you screw up squeezing the lemon earlier, you don't need to squeeze the rest of it to understand that you are not making any money. Right. And it was a very, very, very sort of poor gross margin business because you have all this fixed expenditure. And so one very key thing is start a business that is a higher margin business. Uh, right. And that's why the idea matters.

AI assessment note: “it was a very, very, very sort of poor gross margin business”

Q What if someone's thinking, wait, but I can never charge that much, my product is bad, or I haven't built much. Like, how do you overcome that?

A What I would say is you'd be surprised how much enterprise software is bad. And it's not, they're not charging twenty-k, they're charging, like, you know, millions. First, uh, and second of all, like, in the big picture, like, 20 K a year, like, just doesn't matter to a company, uh, that's, you know, that has a 102 hundred employees in it. Like, it really doesn't matter, and if they're not willing to pay that, that is telling you something that you should be realizing, which is, you know, maybe you should work on something else, or you need to tweak what you're building. Um, that's the honest answer, and over time, you can, you know, as you get more confident, you can, you have more referenceable customers, you have happy customers, you can charge More over time. Um, but initially just charge enough to know that like it's something rational, there's pain, they're willing to pay, and then just prove that you have something.

AI assessment note: “What I would say is you'd be surprised how much enterprise software is bad.”

Q What was that experience like? You ended up selling that company Uh, it was probably the first time you were managing people and, you know, doing enterprise sales. All of these things were useful lessons from that first experience.

A I mean, it obviously was not a successful company. Um, it was. And so it's a very painful thing to go through, but the rate of experience and education was incredible. Another thing that PG said or quoted somebody else saying, but always stuck with me is your twenties are always an apprenticeship, but you don't know for what, and then you do your real work later. And I did learn quite a lot and I'm very grateful for it. It was like a difficult experience and we never found product market fit really. And we also never like really found a way to get to escape velocity, which is just always hard to do. There is nothing that I, that I have ever heard of that has a higher rate of generalized learning than doing a startup, so it was great in that sense.

AI assessment note: “it's a very painful thing to go through, but the rate of experience and education was incredible”

Q Do you want to talk about level three, four, and five briefly?

A Yeah, so we realized that AGI had become this, like, badly overloaded word, and people meant all kinds of different things, and we tried to just say, okay, here's our best guess, roughly, of the order of things. You have these level one systems, which are these chatbots. There'd be level two that would come, which would be these, this, these reasoners. We think we got there earlier this year, um, with the O-one release. Three is agents, um, ability to go off and do these longer-term tasks, uh, You know, maybe like multiple interactions with an environment, asking people for help when they need it, working together, all of that. And I, I think we're gonna get there faster than people expect. For as innovators, like that's like a scientist, and you know, that's ability to go explore like a not well understood phenomena, Over, like, a long period of time and understand what's just, kind of go just figure it out. And then, and then level five, this is the sort of slightly amorphous, like, do that but at the scale of the whole company or, you know, a whole organization or whatever. That's gonna be a pretty powerful thing.

AI assessment note: “Three is agents... For as innovators, like that's like a scientist... level five”

Q Yeah. And then the other one is this idea of honesty. Like Brad, do you think as a second time founder, it's really possible to get honest feedback from your investors or friends or things like that? Or do people have a hard time being honest with you at that point?

A Yeah, I think it's harder. Um, when you're a first time founder, nobody cares about you because they don't have any sort of relationship with you yet. You know, you talk to some investor and You're just the product and they tell you if the, if they want to invest or not. Um, and, but then after you've been through a process with people and you've built some sort of rapport and relationship, it's now costly to tell them what you really think about what they're working on. It, you know, it, it basically, it, it all just comes from, it's just a continuation of the same, the first startup being carried through into the second one and not having like a clean break and resetting all those relationships. Um, but, but yeah, I definitely, um, Um, just, it's really hard to get, you know, real true feedback from people.

AI assessment note: “Yeah, I think it's harder. Um, when you're a first time founder”

Q Right off the, I mean, like, I can't visualize the pain, though I'm sure there is pain. Um, we are currently starting with a few clients and are basically tailoring our service to them. Cool. How do we avoid the trap of making our service too specific, and how do we balance between the requirements of a few clients with the need to build a platform that is more universal?

A So, I kind of think that, like, when I see a startup like this, I really have this basic question, which is, are you going into this space with knowledge of the space? Like, were you a previous property manager? If so, you should have a lot of insights on the types of problems that are general, that can be applied across the industry, and so you should have a lot of insights on what your initial product should be. Step A. Step B. If you're coming into this with little personal experience, you think there's an opportunity here, but you kind of don't know what it is. I don't know any, I, this is your problem. And like, I don't know a clever way of avoiding it. Um, Just learning the business means you're gonna be building a lot of features that probably are not generally applicable, and that's the cost of learning about this business. So in my mind, like, if you already know about the business, use your gut. If someone asks you to build something that, like, your experience tells you is not general, great. Don't do it. Um, if not, consider the first five customers as you just getting an education on the industry, and it might be a very expensive one. That's fine.

AI assessment note: “if you already know about the business, use your gut... If not, consider the first five customers”

Q Um, how do you launch an MVP of user generated, a user generated content platform? Um, if I make an MVP with minimal content, users will probably leave since there's not too much to consume.

A So, The classic answer to this is, one, does your platform have a lot of replay value? Is there any reason why I should come back daily or weekly? I think that, um, that's a really important question to ask because even if you get a lot of people, if there's no reason for them to come back, it doesn't really matter. I think, two, is there a smaller community that if that community was using the product, they would find it useful, even if others weren't? You know, the classic, I guess now old example of this was Facebook at Harvard, like as long as the Harvard kids were using it, or even a class, even just one of the grades in Harvard was using the product, it was useful. And so therefore they didn't need to have millions of users day one. Um, literally users on the order of like a couple thousand was very useful. So, um, the way that I'd be thinking about this question is one, how frequent do I get value out of it? If it's not that frequent, is there any value I'm creating? Two, is there a smaller population of people who can find this useful even if all your friends aren't using it?

AI assessment note: “is there a smaller community that if that community was using the product, they would find it useful”

Q company is a website for shopping from home-based and small businesses around us. After evaluating what I do, I've realized that I have a small, I have, I have users in traction, but we do not have product market fit. Now the question is, should I continue working on this, especially since margins are thin, or should I quit this and begin again in search for market, product market fit?

A You know, I think this is always the hardest question to answer because I see problems on both sides. There's a set of founders who, if they don't succeed immediately, their mental model is that their company is a failure. But then there's also a set of founders who even after learning two years worth of horrible facts, they still like will not acknowledge that it's not working. Um, I think the best strategy for this, um, the one that's easiest to execute is to time bound it. So what I would say is like, Hey, we're going to try everything as hard as possible to make this work for this period of time, probably no less than six months. And if we haven't figured it out then, then we're gonna be opening ourselves up to, um, pivoting. I find it's easier to time-bound it, because, like, you can, it's easier to be intellectually honest with yourself about, like, have I worked on this really hard for six months versus three? Versus, have I tried all of my ideas? Like, that's a really...

AI assessment note: “I think the best strategy for this, um, the one that's easiest to execute is to time bound it.”

Q Um, should we launch our MVP, uh, for free, or is it better to charge people?

A Let's be clear, you have a service where you're actually, like, There is real cost, meaning, like, your service is helping someone exercise. There is cost to sending someone to that place. There's cost to recruiting them. There's cost to their time. In any situation where you actually have real cost, you should be charging from day one. Um, in a pure software play where you don't have any costs, maybe consider something that's freemium, but, like, you don't want to anchor, like, the number one thing with these, like, marketplaces that are actually moving humans around is that If you're into a situation where you're providing a dollar worth of value and you're only getting 75% cents back, you're gonna get a lot of demand. But you're gonna lose a lot of money, and when you actually equalize the, the demand and the price, you'll lose a lot of customers.

AI assessment note: “In any situation where you actually have real cost, you should be charging from day one.”