Building Zipline: From launch disaster to drone-delivery giant | Keller Cliffton (Co-founder, CEO)

Keller: So what are the chances this is going to work? I said it's about 1%.

Brett: For today's episode of In Depth, I'm joined by Keller Cliffton, co-founder and CEO of Zipline, the autonomous drone delivery and logistics company.

Keller: So many of the problems that humanity has to solve, they require going and getting your hands dirty in the real world. Like, so much of the stuff that's going to be critical for the future of the United States is figuring out how to build physical stuff.

Brett: In an era where so much company building is about bits, Keller is obsessed with atoms, the real world unfancy work it takes to move things from point A to point B reliably and at scale. But Zipline isn't just a drone company. It's an end-to-end logistics system. And the hard part is actually everything around the aircraft.

Keller: But it turns out the aircraft is only about 15% of the complexity of designing an autonomous logistics system.

Brett: In our conversation, Keller shares the principles that made Zipline work when it shouldn't have.

Keller: I talk to so many hardware entrepreneurs who are starting in seed stage or series A, and they always say a number. I'm always like, "It's going to cost 10 times that much money to build that product."

Brett: Things like how to build teams of heat seeking missiles for pain when the odds of success are 1%.

Keller: We want to divide the company into small special forces teams that have very clear goals and are operating with high levels of agency, but also accountability.

Brett: Why the only way to learn in the physical world is through fast iteration and feedback and how Zipline survived the 10X rule that breaks most hardware economics.

Keller: He was like, "Keller, do you know when the right time to fire someone is?" I said no, but I was kind of expecting him to say something very cliche and he ended up saying, "The first time the thought crosses your mind."

Brett: Let's dive in. When you reflect on yourself as a founder and the entrepreneurial people that you've recruited and developed, what's your thinking on nature versus nurture for input drivers to your overall success?

Keller: I think that one thing that we've learned the hard way again and again and again at Zipline is just that it's very easy in any hiring process to focus on specific experience. And the reality is whenever we do that wind up, we almost always end up regretting it because especially for a startup like in the early phases, you're always... By the time you actually recruit someone and onboard them and have them up and running, the job has changed. There's a different priority inside the company and we're like, "Oh, actually nevermind. This thing we hired you to do is no longer relevant. We actually really need you to go own this thing over here." And we just ended up realizing that it's much more powerful to hire for innate characteristics, almost always. And there's like exceptions to every rule, but in general, if we could focus on these innate characteristics that would be indicative of someone growing super fast at the company, we'd almost always win. And so the four innate characteristics that Zipline focuses on are practical problem solvers, fast learners, low ego, mission-driven. If we can find people who have those four things, these are the four things that we think are pretty hard to teach people. So I guess to the nature-nurture point, these things are probably a lot more nature than nurture. Particularly practical problem solver might sound like an obvious thing. It's totally not. I would say this is present in less than 10% of humans, but people who are like predator natural practical problem solvers, people who can simplify a problem down, realize, "Well, these are the only two things that are going to drive the solution," and then as quickly as possible build initial versions of that thing, get it working, learn by doing in the real world, and then iterate their way to something like extraordinarily good. We had no idea what it looked like at its end state. There was no good role models to use from a regulatory or business model perspective. So we were figuring everything out from scratch, and that meant that we had to be really humble and curious and just like get something into the real world extremely fast. It would be super scrappy. And we would just learn by doing.

Brett: Do you think that's globally correct across startups? When you think about your friends who are running companies, do you think that insight is correct for everyone or it tends your guess is it's more correct only in the context of Zipline?

Keller: The thing that I could speak to with confidence is that like hardware companies super, super struggle with this. I mean, Zipline, we've been building the company for the last 12 years and we've seen many different kind of like super cycles where these companies raise hundreds of millions of dollars or billions of dollars and then they quickly burn through that money and then they go bankrupt. Especially in robotics, for example, there's a real predilection to just go and sit in a bunker and build something super cool and kind of convince yourself that like the metrics you're looking at are what matters, or you're kind of building really cool hardware and you're like, "The use cases will come." And Zipline was always just way more paranoid. We were like, "We have to build the simplest possible thing in less than a year and get people to pay us for it." And then once we're in the real world iterating and operating, we will learn by doing. And if we can get customers to pay us more money for certain things, that'll be a good sign. If there are maintenance problems or reliability problems, we'll have to solve them day in and day out. Just getting as quickly as possible into that mode of like rubber meeting the road is really hard for hardware companies to do. It's much easier for these companies to continually do these like update videos on X where you show the hardware doing something really cool once. And that can be very impressive looking. And in fact, I think a lot of investors can be like, "Oh wow, this is really working." But of course, I would say the dark truth about robotics is like anybody can make a video of the robot doing a thing once. The challenge is always like, can you solve the reliability and generalizability problems so you can put the thing in the real world and have it do what it's supposed to do day in and day out in a way that's reliable and useful for customers.

Brett: You hinted this a second ago, but what's kind of your working theory on each of them and in the sense of how much the role of nature or nurture plays. It seems like your gut response is its nature sort of more nature and then sort of genetics drive it, but like so you're not a believer that like kids that are in Montessori educations or things that get people to focus on real world problem solving, not regurgitating facts matters for this in your experience. And we're talking about the first one.

Keller: I guess my opinion is probably that like I have three kids at home, three kids under four years old. So I'm certainly struck watching them that like all kids are born with this like incredible curiosity and like joy of learning and ability to just go and play and do stuff and see what happens and kind of like probe the physical world in that way and learn about stuff. I would say in general, we feel like we're working against how much of the educational system is trying to kill that in kids or teach them not to do that, teach them not to take risk, not to play, to kill the curiosity. I guess I'm just speaking for myself. My educational background, especially through middle school and high school, I felt like it was almost teaching me none of the things that would have been useful to me in my career, what I ended up doing. I learned way more working, making seven bucks an hour working in a restaurant for 40 hours a week all through high school. And I learned way more in that job that's relevant to my job today than I did in high school. When Zipline is interviewing for people, we really don't care about their grades or where they go to college or what extracurricular clubs they're a part of. We're super, super focused on what are the extracurricular things that they're doing. There's a mechanical engineer on the team who's very near and dear to my my heart. He's been at Zipline for almost a decade. And I remember interviewing him being like, "Well, what did you do?" He's coming from the University of Washington. He was like, "Oh, I built a submarine." I was like, "Oh, okay. What do you mean you built a submarine?" He's like, "Well, we figured out all the different kind of major design considerations and we had to design the seals and figure out all the different subsystems." And I was like, "So what'd you do with it?" He's like, "I got into it and I took it down to 50 feet in Washington." I was like, "You're hired." It's like literally submarines are so dangerous. I don't know that much, but I know that submarines are incredibly dangerous. And the engineering behind the pressures even at 50 feet are pretty extreme in terms of how you design stuff like that. The fact that this 20 year old totally risked his own life based on the strength of his own engineering to go down to 50 feet on Washington, you know? So we're always looking for those kinds of extracurriculars. It's like, "Did someone participate in Design/Build/Fly?" These are kind of specific, very hands-on, first robotics, solar car. There's some kind of F1 college programs. In high school, a lot of folks are applying to Zipline and they have insane garage projects that they're doing as 15 or 16 year old kids. I guess whatever that is is the thing that Zipline is typically interested in and is most predictive of success. We hire a lot of folks out of high school these days. We've hired some kids who are 15 or 16 years old.

Brett: In as full-time employees?

Keller: Well, they start as interns. I mean, different examples. I mean, I remember there was someone who joined Zipline when he was 15. He applied to his clearly prodigy level robotics talent. His parents were helping, I think, him manage some of the implications with school. And so he started as an intern for three months. And then COVID hit and so schools were shut down and so we kind of decided, "Hey, just work with us full time." And so he actually worked full time. And then he got into Stanford early admission and calling his mom and saying, "Hey, we're going to offer him a job for $180,000 to lead a team of mechanical engineers at Zipline." And she was like, "Sounds good to me." So then he actually just ended up joining us full time, delayed matriculating to college for a year. He ended up going for a year and then dropping out to start his own company, which is now funded by Sequoia called Reflect Orbital. They build space mirrors. So in general, I would say Zipline is having more and more luck hiring earlier and earlier. We can teach people most of what they need to know [inaudible 00:09:47].

Brett: Explain more about your working theory of why it makes a lot of sense for you. Because most people would say hiring a 17-year-old like, "They need to bake more. They're not ready to be an employee of my company."

Keller: Yeah, we find it to be wrong. I think people, again, not all 17-year-olds, but the 17-year-olds who are really self-motivated and really excited and wanting to be part of this future, I think a lot of them also have this feeling that the future is going to play out over the next four years. And if they go to college, four years from now, too much of this will have happened. There are lots of people at Zipline, by the way, who I'm thinking of who kind of fall into this category. And then yeah, the other thing is just that I think it's a knowledge probably on their behalfs that a lot of the universities these days are not teaching them the things that they would actually need that would make them super valuable at a small scale startup. The best way to get those skills is to go do. And then the flip side is that a lot of times these are the hardest working, most maniacal, insane people at the company who just want to be part of really hardcore teams operating in hardcore ways, moving lightning fast. They're usually blown away that Zipline will give them almost like crazy levels of responsibility with regard to a new product, and that's how we've designed the entire company.

Brett: How do you find those kids?

Keller: They find us. Yeah. I mean, we just are typically getting emails. And the first thing we're always asking is just like, "What have you built? What have you built? What have you built in the garage? What cool projects have you worked on with your friends? What most inspires you?" Some of the answers are insane. The submarine or someone who joined Zipline recently when he was 16 had designed a full GPS visual inertial odometry system for a quad copier that he had 3D printed using Nvidia GPUs running on board. He did this while he was at a boarding school. It's like you're hired. You know what I mean? So anyway, I don't know. I guess the high level thing is that I think success in startups, extremely important nature and nurture, but I think the nurturing is only happening by doing. I think this attitude that kids are learning the right things in college or in grad school is largely incorrect. I think the only way to develop those skills is to go and build things. You can do that in your garage. You can do that with a small team at a startup.

Brett: To the point that you're making when you're making these hiring decisions is a vast majority of the signal, the body of work that they've produced said differently. If you took any of these 17 year olds and you could not ask them what they've done, you brought them to Zipline for an onsite, could you set up in a set of practicals or something that could give you as much signal as looking at their body of work or no it's very hard to do that?

Keller: No, I think you can. I mean, for sure, we want as much of the interview process as possible to just be applied. So with the software-

Brett: Yeah. What does that look like?

Keller: Could be anything. When we're hiring technicians, we're going to build a test jig with them for a specific piece of hardware and we're going to show them the machine shop and say like, "Go for it. Let's build this test gig together over the next day." For a software engineer, it's like pair programming with them. For a mechanical engineer, we might be going through a design review for a specific part or subsystem and asking them to design something on a whiteboard with us or think through how to do a lot of different calculations to figure out if something's going to fail in a critical way. But yeah, for sure, I think we just want as much of it to be applied as possible. Supplying's very non-traditional in terms of the way that we... We're willing to break every rule in recruiting. I think a lot of people have this sense that there's one default way you should do it. We do all kinds of things that probably maybe look kind of crazy to certain candidates, but we're happy, for example, to be like, "Hey, do you just want to come and work with us for a week or just come work with us for two weeks?" We're like, "We can get someone to start with us for a month." And it's like there's no degree to which we can't just like, "Hey, let's just work together quickly and learn by doing." People will figure out pretty quickly if this is their life's work and this is where they want to be versus not.

Brett: So what's the role of experience?

Keller: I think experience in leadership and experience in terms of how to lead small special forces teams in very hardcore ways is extremely valuable, but a lot of other kinds of experience are probably becoming a lot less valuable just because the pace of technological change is so fast. I mean, hiring someone who's like, "Oh, well, they have a PhD in simultaneous localization and mapping from CMU five years ago," that was a hot area of robotics five years ago. It's not as relevant today. There's just like new approaches. I think that there are certain things that are really hard to train from a leadership and management perspective, but when it comes to technology, I think there's no replacing just people who are at the forefront and doing the work themselves. It's the reason that when we look at leaders throughout Zipline, there are a number of different things that we hold leaders accountable for and how we judge the performance of leaders, but one of the things is still technical depth and we do not want any middle managers at the company. I think middle managers would be way too quickly. Every aspect of the company is changing so fast that the only way you can keep your priors up to date is by doing the work yourself. And so the leaders of our recruiting team are running large recruiting searches themselves. Any leader at the company is spending a significant percentage of the time.

Brett: Doing the IC work.

Keller: ... doing the actual IC work.

Brett: What's your thinking on why leadership and management and leading a team, like that experience is valuable there? The counterintuitive thing would be like, that should be easier to figure out than some incredibly hard science technical breakthrough related to power management that you could have a 17-year-old working.

Keller: Yeah, I don't know. We find it to be the opposite. It's interesting.

Brett: Work through it. What's your best guess?

Keller: I think that a lot of the really hard technical problems are more tractable and can be solved in more practical ways and take less time to learn.

Brett: Than the human problems?

Keller: Yeah, than some of the human problems, for sure. And maybe it's just Zipline's perspective, but we're in a mode where we have a lot of extraordinarily brilliant people who are early in their careers growing exponentially fast. So it's usually not like... Our growth is usually not constrained by people who are like really brilliant and able to tackle hard technical problems. A lot more of our growth is constrained by like, "Do we have people who know how to be world-class hiring managers or do we know people who can be strong enough leaders that they kind of know how to play every note on the... They can be extremely positive, but also like radically candid and tell someone when it isn't good enough and someone who knows how to fire people fast enough?" That skillset is far more rare.

Brett: Is there a thread that ties the people that you have to ask to leave together or they're "not a fit" for a myriad of reasons?

Keller: Well, definitely one of them is like really bad behaviors that they learn at big companies, which is kind of like related to what we've been talking about. I definitely think a lot of big companies are actively training people on the exact wrong behaviors for them to succeed.

Brett: And that's like generic bureaucracy, political nonsense, things that have nothing to do with accomplishing work or it's another...

Keller: I think that's a big one. I think a lack of knowledge about how it feels to operate in desperate circumstances. I think a lot of people are trained on how to not lose rather than on how to win. If you're like in a big company like Apple, and I mean, Apple has already won, so the main goal for people is to not lose. And then when it comes to like, how do you do HR for a team? The main goal is like, "Let's just not get sued." And so that leads to a number of behaviors from an HR perspective that work for a huge company at that scale. Or maybe not, but depending on your argument. For a startup, like a small startup where the goal is like, "We are default dead. The only way this works is if we can shoot this incredibly narrow gap and get to this like extraordinarily revolutionary product." And the chances of you achieving that are not that good in the best of scenarios, but it's like we have to basically... It's hard to describe, but it's basically like playing to win versus playing not to lose. I think that if you're playing to win, you're like, "All right. Well, we're going to have to take a lot of different risks across the board in order to get to a scenario where we can actually do this impossible thing." But it means that we're not making decisions about how do we reduce risk for the company in every possible way. How do we just make sure that something doesn't go terribly wrong? It's like, in fact, we will fail nine times before we can figure out how to do it the 10th time. A friend of mine sent out this email the other day talking about talent that really succeeds at startups that I think was profound. He called them heat-seeking missiles for pain. That is the perfect definition of people who really succeed at startups. These are people who are customer-obsessed that they kind of always have their antenna up and they have really good spidey senses like, "Something feels really screwed up over here." They can then instantly zero in on it and like ask enough stupid questions to figure out, "Okay. Yeah, this process is broken inside our company or this product is not working correctly." And then they can basically like pull together whatever resources are necessary and be like, "Let's work through the weekend and fix this in a maniacal way and then instantly get into the real world and start learning by doing and iterating and figuring out what parts of that hypothesis were correct and not." And then they can just do that in every part of the company again and again incrementally. That's a very, very hard thing. It's risky to do. A lot of times it's not your job. So you need someone whose attitude is, "It's not not my job," which is something we say at Zipline all the time. Someone who's willing to crack some eggs. The person who's doing that might piss off some coworkers because they're like, "Why are you all up in my business? I'm supposed to be owning this." But yeah, those heat-seeking missiles for pain are worth their weight in gold for small startups. And I think generally that's just not what we're teaching at bigger companies.

Brett: So if you're interviewing somebody or sitting down with them and you're trying to understand or they're playing to win or playing not to lose, can you do that in high accuracy or you really need to either reference them in a particular way or see them do a work trial?

Keller: I think all of the above. Yeah. I mean, speaking of references, one thing that Zipline does is just always amazes me. We still do blind references for every single person who joins the company. I think this now maybe seems like out of date or I've heard from certain people that they think, "Oh, that's not something you're supposed to do from a recruiting perspective."

Brett: Oh, I'm obsessed with it.

Keller: Sounds insane to me, but blind references are hard. It actually takes a lot of time. You got to go network your way to people who've worked with this person in the past. And they obviously can't be provided references. A provided reference is a paid reference as far as I'm concerned. They're just going to tell you what you want to hear. Whereas getting to a blind reference and actually having an honest conversation about someone, like nine out of 10 times, they're like, "This person is incredible. I didn't know they were on the market. I want to hire them myself." I'm often asking like, "If you were going to go start your own thing today, would this be one of the first five people you would hire?" If the answer is yes, we've just learned something so, so strong and positive. In terms of conviction that this person is as special as we thought, maybe 10% of the time you realize the person is leaving a trail of destruction behind them in their career. And that's a very valuable thing to know about people that can sometimes be hard. There are some people who interview incredibly well and just leave a trail of destruction behind. And there are people at Zipline... I mean, Zipline has extremely strong tenure. There are so many people who have been at the company eight, nine, 10 years have been building day in and day out with me as we've kind of learned and grown together. There are also people who totally didn't do the right thing and didn't succeed and also maybe didn't leave under very good circumstances either where we have to remove them in painful ways that are damaging to the company. And then I'll see those people get hired by a competitor or by another startup and nobody called me. So it's always like, I just know that there are lots of companies that don't do blind references. It always surprises me.

Brett: What about in terms of when you're actually interviewing them? Do you spend a lot of time basically asking them for examples of the past and that line of questioning, or no, do you do something else?

Keller: I mean, it depends. If we're talking about an IC who's joining us out of college or even out of high school, or if we're talking about someone who are managing to lead in 100 person company, the interview is going to look really different. On leadership hiring, we spend a lot of time actually going through LinkedIn specifically with the person looking at all the people they've hired in the past. So we're like, "Oh, great. You led this team. So let's talk about who are the three best people you hired and the three worst people that you hired." And it's like, "Let's actually pull up their names and look at them on LinkedIn. And let's think through, would we have hired those same people? Would we have been as impressed by them as they would have been? Do they have the ability to admit where they screwed up?" As crazy as this sounds for a lot of startups, there are some people who are so jumpy that they never have to deal with the consequences of their own stupid decisions. This is one of the reasons I love meeting people who have like, "Oh yeah, this person started when this company was 20 people and it grew to a thousand people and then they stayed with it all the way until bankruptcy." There are plenty of startup stories like that. And when you meet someone who went through that entire journey, it's like, "Man, good on you." Talk about sticking with something [inaudible 00:22:16].

Brett: And it's atypical. There's a lot of 24 months here, 24 months there.

Keller: Exactly. So it's the opposite. Exactly. So they're obviously fair weather sailors and then there are people who are like, "Man, they went through multiple really hard up and down cycles at a startup." Those people are so incredibly valuable. You know that they will stick with it when things get tough, which always happens with any ambitious idea. And especially when it comes to leaders, there are certain people who have kind of gone on these leadership tracks, but they've been bouncing between roles at different companies. Two years here, two years here, two years here. And you ask them, they're like, "Oh yeah, I grew this team from 20 to 80 people, and then I took this other role at this other company." And you realize probably half the people they hired weren't good, but they didn't have to sever the consequences of that. Anybody who's had to grow a team really fast, like 4X the size of a team over a year, they spend the next year realizing all the stupidity-

Brett: Cleaning up.

Keller: Yeah, cleaning up. And so it's like very valuable to have a sense for like, not only do they hire world-class talent, which is one of the main things we're trying to evaluate, but also have they had to realize their own mistakes and then clean up their own mistakes. So having someone who's like, "Oh yeah, I tripled the size of the team this year, but then I realized all these cultural problems and a quarter of the people I hired were toxic and didn't work out and I had to figure out how to remove them from the team and fix it, turn around the culture," that's one of the things that we interview very carefully and it's very, very valuable. On the world class talent side, a lot of the things you often want to know is like, "Great. So what did those people go on to do?" Zipline was just hiring a senior operations leader a year ago and I remember talking to him, he was coming from this big team at Tesla and he had done a lot of hiring on that team. And I think it was about 100 person team when he was kind of leading a big chunk of it and he was like, "Oh yeah, I had to hire like the first mechanical engineer onto the team." I was like, "Great. So what does that mechanical engineer do now?" He's like, "Oh, the mechanical engineer got promoted five times. Now he leads the entire division and he has 5,000 people reporting to him." Done. That's the information we need to just get a sense for like, does this person know how to identify and then recruit and then retain world-class talent, people who are going to grow super fast in their own careers?

Brett: So given the way you describe the type of people you want at Zipline and the way you want them working, it feels like it would just turn into Lord at the Flies. And all these people doing all these different things, stepping on each other and it would just be one big mess. It's obviously not the case. How do you organize all these people, many of which are probably disagreeable or struggled at work well because when they see a problem, they just have to go solve a problem. Or like, given the types of people you described, you must have had to figure out how do you organize them, particularly as you have more and more people. The normal thing is we're going to put in process, we want to put in structure, we want to standardize, how have you worked through that?

Keller: You always obviously have a balance, but I mean, a couple of things we've already talked about. One is you make sure to hire people with low ego. I think that's kind of a necessary precondition for this sort of a thing to work. You want people who are ambitious, but who put the needs of the team above their own needs. And we believe very strongly, I mean, one of Zipline's cultural values, especially for leadership is this idea of servant leadership. So it's just like, there's no fancy versions of anything for anyone in the company. People always find it weird. They come in and spend time in our offices and I sit amongst the electrical engineering team. I get confused for... I mean, I had an intern sit down at the desk next to me a couple months ago and she was like, "Oh, what's your job here?" And I said, "Oh, I clean the toilets and do whatever else is required of me." And she must have thought it was weird because she must have gone and asked someone else like, "Why does he have this weird answer?" And then she came over an hour later and was kind of embarrassed, but she thought I was an intern. The way we all sit together and work together, I just think there's a very powerful feeling of servant leadership. We want leaders who constantly are willing to do shit work themselves. Nothing is beneath a leader at Zipline. We're willing to plunge the toilets, be down in our hands and needs cleaning shit up and either at the test sites or in the office, the most unfancy work we are all willing to do. And we're constantly willing to put... The needs of the team come above everything else, including the ego of the leader. It's a rare thing. It's a hard thing to find. And I think especially in the Bay Area, you have a lot more people who have learned opposite behaviors of that in order to get ahead of other kinds of companies.

Brett: What else is a part of the way that you've organized all of these people that works?

Keller: I think that you have to basically be really clear that we want to divide the company into small special forces teams that have very clear goals and are operating with high levels of agency, but also accountability. So we were definitely inspired. Netflix talks about this idea of freedom and responsibility. I think that's a really powerful concept and Zipline definitely has kind of like stolen it and adopted it.

Brett: How do you manage that where it feels like there's so many interdependencies?

Keller: Yeah.

Brett: And so it's not like here's this little startup, it's single threaded. It feels so highly integrated; hardware, software, regulatory, on and on and on.

Keller: It's not perfect. I mean, a year and a half ago, Zipline was incredibly scary position. I mean, the company was possibly running out of money. We were six months behind launching this next generation technology that was really, really important for the company's future. And we have this joke. I mean, I've heard SpaceX talks about itself in this way, which is that we specialize in turning the impossible into the merely late, but Zipline was definitely fully in that mode. We're desperately trying to get this product launched. We didn't hit the cost targets and we didn't hit the timeline. At the end of November, just a couple months ago, Zipline has now launched this next generation product and it came in exactly on target, on time, just getting better at this as we go. Both accelerating the timelines, but lining up the goals and accountability across a program so that each of these teams can operate in a highly entrepreneurial way, but you make sure that at the end of the day, everything adds up to the full product that you need.

Brett: What's the full list of leadership attributes?

Keller: Yeah. I guess we may as well talk about this publicly. I've never talked about that publicly. Actually, I ended up... It's new. We have only started thinking about it more carefully. And it's because I think I started thinking more carefully about what did I need in executives at Zipline, and I found myself having the same conversation again and again and again with different leaders trying to describe what were the different failure modes and where were we seeing excellence? And then as soon as we wrote it down for executives, we kind of realized like, "Oh wow, it's really practical. We should totally do this for everybody." The five things we look for are strong magnet for world-class talent. I mean, we've talked about it, but does someone really know what it takes at every layer to hire someone who's like top 1% in their field and then recruit them and retain them? It's not an easier trivial thing. The second thing is they know how to challenge teams to greatness. And by that, the way that we talk about that, we mean both positive and negative. Leaders that are great at challenging to greatness, they have to know how to really rally and inspire a team and people on that team will be like, "Oh man, I would follow that person to the moon." But at the same time, there's a strong degree of radical candor, direct feedback and telling someone when something isn't good enough. You just have to be able to be like, "This doesn't meet the bar. It's not good enough. I'll work with you this weekend to fix it, but we must fix it and fast." So leaders who can play every note on the piano is kind of how we think about it. You got to be able to be on the really positive optimistic side, like, "You did absolutely amazing. This is world-class work, really inspiring and rallying." But also like, "This is not nearly good enough. We have to fix it immediately." And then the third thing is just being great at performance management and removing people from the team fast if necessary. Fourth thing is technical depth, which we talked about a little bit. And then the fifth thing is entrepreneurial drive and ownership mindset.

Brett: What's your least squishy way to articulate entrepreneurial drive?

Keller: I think it's like heat-seeking missile for pain, which we kind of talked about. I think it's a deep sense of you act in every possible sense like it is your company. It obviously is for people at Zipline. I mean, we all have ownership in this thing that we're building together. But does the person behave truly like an owner of the thing that we're building together? Do they have this attitude of like, you spend every dollar as though it's coming out of your own pocket? And then do you also have this sense of maniacal urgency? Like, way better to do the thing tonight than to wait 'till tomorrow or wait until Monday. And even more than that, I think there's this sense of like it's not not your job. If something needs to be fixed, we will fix it. It doesn't matter whose title says what.

Brett: What have you found, just sort of in closing out maybe this section on talent, is malleable in people? Meaning you have somebody who is not meeting your standard of performance in any part of the business where you come up with a fictitious example. What have you found can actually be shaped versus, "No, we got to get rid of them. Not going to work. We can't develop them to greatness."

Keller: I mean, I think there's lots of small tactical stuff where it's like, this role isn't right or the scope isn't right or like the overall work setup isn't right. And one of the biggest things that we found is that there have been periods at Zipline where we've hired too many people early in their careers who were not in the office five days a week. And we've now stopped doing that because especially for people early in their careers, I mean really for everybody, I mean, Zipline is like a very, very in-person culture. We have to be, we're building hardware. There's so much complicated multidisciplinary integration that we have to do on a day-to-day basis. But yeah, I think all of that stuff can be fixed if you're just honest about it and you're addressing it every single day. Zipline has an attitude just that like, don't wait for like some performance cycle to give someone feedback. Give them the feedback that day. If we walk out of a meeting and I as a leader think something wasn't quite good enough, immediately just grab that leader and be like, "Hey, that didn't go the way I was hoping. Here's why. Can we do better tomorrow? Let's do an iteration tonight." Treat people like adults, not like children, which I think is something that tech companies actually struggle with, especially over the last three or four years. And people want direct feedback and I think giving it fast makes it way more likely that the person can actually improve. And all of those things I just described, those are different than the innate characteristics, which I think are way harder to teach. But those things can be fixed very quickly and should be able to be fixed on an hourly or daily basis. You shouldn't have to wait a month or a quarter to go through like a performance eval cycle and get the person that feedback.

Brett: Have you found that there's a specific way to let people to go, let people go when they should, meaning you talk to almost all CEOs, regardless of-

Keller: Everybody's too slow?

Brett: Right. I wish I'd took it. Do you think you've done very well there or still it's too slow and it's just there's some law of physics going on here?

Keller: Zipline has gotten better at this as we have matured as leaders and we've gotten better as a company. Something that Alfred Lynn, board member and our partner from Sequoia told me very early on, was... We're kind of having this conversation about whether I should make some difficult decision or something. And he's like, "Keller, do you know when the right time to fire someone is? " I said, "No." But I was kind of expecting him to say something very cliche and he ended up saying, "The first time the thought crosses your mind." Which when I heard that, I was like, "That is the most sociopathic thing I have ever heard. There is no chance that that is true." Every year that has passed since he told me that 10 years ago, I've realized there is more truth to that statement than I realized.

Brett: Why do you think that is?

Keller: Because you never, ever, ever end up... When you're kind of hemming and hawing, you're, "Oh, it's me, not them. It's the role. We got off to the wrong start. We didn't get the onboarding right," there are a million excuses that you'll make to not have to kind of engage with the hard truth that this is not the right person for the role. And I think the reality is for the people who are going to be at the company who are like clearly A and A+ and who are going to be at the company for the next five, 10 years, who are going to have completely exponential impact on the company's mission, the thought never crosses your mind. I think that's the intuition here. It's that spending a bunch of time debating in your mind, "Oh, is this person a C- or are they actually a C+?" It's just not that good of a use of time.

Brett: Let me shift gears a little bit. A long time ago, Elon put this secret master plan for Tesla out that he wrote up that was basically, "We're going to start with a roadster and blah, blah, blah, blah, blah. And this is how the sort of set of things we're going to do are unfold and we're going to start with something really expensive and ultimately make something that's mass produced. It's going to look like this, this and this and broad strokes." When you think about the Zipline path, how much of it did you understand at effectively time zero? Like you obviously will talk a little bit about this and you've talked a lot about the trajectory of the company in the past, but basically you started doing blood delivery in Rwanda. You can correct me, but broad strokes that. And you're on this mission to move most things around the country and/or the world at some point. How much of that at time zero was the scaffolding mapped out in your head versus the chess game was being played in real time you were figuring things out?

Keller: I would say very little. I have this flag over my desk that says, "We do this not because it is easy, but because we thought that it would be easy." Have you seen that flag? It's so true for Zipline. It's crazy. I mean, when we got started, what we were proposing to do was illegal in the United States. So I remember talking to some early investors and everybody's like, "Well, it's illegal. I'm not going to invest in something that's illegal." I think we had this high level vision for what this could be at scale, it felt like someone should build an automated logistics system for earth. And we had seen all of these robots running around inside warehouses that could pick up a shelf, move it to a human picker who would pick something off the shelf, put it into a box and ship it. But we just remember seeing all those systems thinking like, "Man, someone's going to build that for outside the warehouse." And that would be one of the most important companies on earth if you could build that kind of teleportation system that would kind of operate like the internet, but could actually move things in the physical world. And we had that idea in 2013. It was an extremely naive idea. We had no idea what the hell we were talking about. I remember talking to Alfred, our partner from Sequoia at the time, and Alfred's like, "Okay. Keller, you know nothing about aviation, you know nothing about logistics, you know nothing about healthcare," which is where we were proposing to start, he's like, "What makes you think this is going to go well?" The simple reality is that it's hard to say. I guess we were focusing on this area that we had instincts of like, "Oh, this feels like it's going to be important five to 10 years from now." And so I think we were willing to go and do whatever weird, scrappy things we had to do to just get started quickly in order to learn by doing and sort of like build up momentum, build experience.

Brett: But so the rough shape of what the product is today was not outlined as like the end state. So before you committed to sort of a toy robot iPhone product that the company started with, and then you kind of refounded the company around this, so maybe we could talk just briefly about how you knew that wasn't going to work. hat did you have before you committed to taking a first tangible step or making a big bet? What was figured out?

Keller: I mean, what did we have? We had nothing. We were unemployed. We had nothing better to do with our time. And as you mentioned, we had raised this small seed round and then we had raised a series A on this idea of building robots for the home. But the more I looked at it, it seemed like that was not going to be a good... I mean, in 2013, the world was not ready for that. That was not going to be a good market. We were competing directly against iPhones for people's attention from an entertainment perspective. It just didn't feel like it was going to be possible.

Brett: And was that easy? Because again, your personality type and problem solving makes me feel like the first reaction is, "Okay, it's kind of working a little bit. I can fix this. I can make this work." Or was it, "No, this is going to work. Let's move on." It wasn't something you wrestled with.

Keller: I think it's a combination of things. I mean, one, we started the company when I was 23. This was the first idea we had. We had built a couple prototypes. It was helpful that I could see there were a couple companies that were farther along on that path than we were. These are companies that people haven't heard of, but like Sphero or Anki. These companies had raised hundreds of millions of dollars. And I could look at the products that they were building and be like, "I just don't think that that's going to be that valuable." And I think we could kind of tell that they were selling a lot of the product, but they weren't being used by customers nearly enough. And so we kind of ended up, I just had this sense of like, "Man, if I'm going to spend the next 10 years of my life working on something, we should find something that is insanely valuable and inspiring. I want to work on something that's big enough that it would be worth a decade of my life." And so I think that we knew we were interested in robotics. We had a team of like five people who I thought the world of. And so I think we were still in search of like the problem that would really inspire us and we had nothing else to do with ourselves.

Brett: When you think about refounding the company in that moment and you're raising money to do that because you ran out, what was the thing that you were raising money for? What level of definition had you figured out?

Keller: I mean, not only did we not have very much figured out, but we were also totally desperate because it felt like the company was going to die. We needed a vision. The reality is when we started building this new version of the company focusing on logistics, we knew nothing about any of this. We didn't know anything about aviation. We didn't know anything about logistics. We didn't know anything about healthcare, which was kind of where we thought it would make sense to start. It helps to be unemployed and desperate, which is basically what it felt like at that time. We had seen early indications of like, we had seen these robots running around inside Amazon warehouses. For example, there was this company Kiva that was acquired by Amazon for a billion dollars and they designed these orange robots that would run around inside warehouses and they'd bring a shelf to a human picker. The human could then take something off the shelf and ship it. And we remember seeing that. I remember seeing that and thinking like, "Someone's going to build with that for outside the warehouse," which is obviously extremely simple and also naive idea. But like, what did that really mean? It seemed like that was a huge thing to go figure out over the next decade. Could you figure out how to build an automated logistics system for earth that would function more like the internet, but in terms of sending things around in the physical world? And maybe we almost had to be naive in order to even think that something like that was possible at that time, but I think it felt to me like maybe you're sitting at the bottom of a giant mountain and most of the mountain is covered in fog, but you can see the top of the mountain. I felt like it was so obvious that somebody was going to go do that. That was the holy grail. That would be one of the most important robotics companies to be built over the next decade. It felt inevitable that it was all going to be automated. We could see that people were using 4,000 pound gas combustion vehicles, driving things around in the US to deliver something that weighs like five pounds to your house. I mean, it's just the economics are broken, it's super slow, it's unsafe, it creates a ton of pollution and traffic. It just seemed inevitable that someone was going to build something better, but we had no idea. The question of, "Okay. So how do you get from the bottom of the mountain to that point?" was completely unknown to us. And we had no money and very little experience. So it was pretty important to pick something that was in our opinion and also it was illegal. Small problem when you're talking to early investors trying to convince them to fund this, like it did not look good. And so instead it was like, "Look, we're early in our careers. If we're going to invest the next decade of our lives, let's pick something." I wanted to make sure to pick something where even if it was going to be super hard and there were like a lot of unknown tactical steps on how to get there, but I was like, it was indisputable that if you could build that thing, it would be hugely important for humanity and it would save a lot of lives.

Brett: But the thing that you picked, was it like divine intervention? You woke up one morning and it just, "This seems like it makes sense"?

Keller: I think we were really interested in logistics because we knew we wanted boring and repetitive. As a roboticist in 2013, what kind of use case are you looking for? You're looking for boring and repetitive. So like what is the most boring and repetitive industry on earth? It's probably either manufacturing or logistics. In fact, we were looking at both of them, kind of interested in both of them. And it felt to me like more people were doing robotics and manufacturing and very few people were really thinking intelligently about automating logistics. And logistics is extremely boring and extremely repetitive. We just do the same thing again and again and again and again and again. Day in, day out to your house or to a business or to a hospital. And we felt like that is something that robotics ought to be really good at doing. And then the more we thought about it, it felt like there was probably a really valuable company to be built there if you could figure it out. Like that kind of company would bring logistics to billions of people who don't have good access today. It could save a lot of lives. You could build a very high margin product in a world where today it all depends on like CapEx and expensive vehicles and unionized labor. And if you could transition that to AI and autonomy, you could build a totally different kind of business that would be very successful. So yeah, again, but these are just the naive, totally clueless thoughts of like a couple 24 year olds who have like are about to be unemployed. That's the way to think about it. And so then I think as we started iterating, then it was like, "Okay, yeah, that's like a pretty cool vision, this idea of automating logistics. If you could move things around, if you could kind of like teleport them, but like it's illegal. No one is going to allow us to do this." And so then we started thinking, "Okay. So we got to convince some countries, somewhere on earth to be partners with us and give us"-

Brett: And did that just popped to mind?

Keller: Yeah, basically.

Brett: Really?

Keller: I mean, it seemed impossible to get the FAA to do this at that time.

Brett: And you knew that it had to be via the air?

Keller: Yeah. I believe-

Brett: It didn't go down like there were a bunch of the delivery robots you see today?

Keller: By far to us, the top of Everest was like, it was very obvious, this stuff does not want to be traveling on roads for a number of reasons, which we can talk about. But yeah, I think the more we were thinking about it is like very clear, "Okay. So if you're going to design this, again, yeah, aerial autonomous logistics system that could deliver things 10 times faster, half the cost, zero emission, we got to find one country that's willing to work with us from a regulatory perspective and make exemptions and exceptions to allow us to start very fast."

Brett: And at this time, was your mindset like, "Oh, if the company's about to go out of business, this seems interesting. Let's take a crack at it"? Or you were like, "No, come hell or high water, this is going to happen. We know it for a fact." What was your mindset?

Keller: No. Completely the first.

Brett: It was low. It was-

Keller: Yeah.

Brett: ... "Let's see what happened."

Keller: Oh, I remember going-

Brett: [inaudible 00:45:30].

Keller: We were like 15 people. I mean, the team can confirm. We had no money, so we took the team to Little Shanghai, which is this... I don't know if you're... It's in San Mateo. It's a super hole in the wall, great Chinese restaurant, but very hole in the wall. And the whole team sat at one table because that's how big we were. And I remember this is over Christmas, someone was like, "So what are the chances this is going to work?" I said, "It's about 1%." And I think people thought that was sort of grim, but I was trying to explain it. I'm like, "Yeah, guys, it's like 1% of a gazillion dollars. Yeah, this is totally worth us working on. And by the way, if this doesn't work, we're all going to be unemployed anyways. What's so hard?"

Brett: We'll have some great stories.

Keller: Yeah, we'll have some good stories to tell. So I think we definitely thought this was highly unlikely to work. And then we spent two years building early versions of the system.

Brett: Wait. But explain how you went from, "Okay, FAA's not going to fly."

Keller: Definitely not.

Brett: It's unintended. What was that to your first site?

Keller: From that, we then went and spent two years. So first of all, we moved out of our office in San Francisco. Guess who moved in?

Brett: It's a great office.

Keller: Yeah. "Hope you guys liked it."

Brett: You have these big ass fans.

Keller: "Hope you guys liked it." It was too expensive for us at the time, but-

Brett: You had these stupid big ass fans in the ceiling that we had to take out. Pretty annoying.

Keller: Oh, yeah. We didn't put those in. Yeah, those were in before.

Brett: Terrible.

Keller: So yeah, first round moved in after us, took over the lease. But we decided, got to get out of San Francisco. Like, "If we're going to go do this, we got to go be at ground zero." So we actually moved the entire company to these trailers that we basically installed in the middle of a farm in Half Moon Bay where we could do all of engineering and manufacturing and flight tests together in one place. We moved the entire company there.

Brett: But how'd you even know about any of that?

Keller: Well, we knew that if we were going to build stuff that needed to fly, we weren't going to do that from 10th and full somewhere.

Brett: Right. But I'm saying even like you don't have avionics experience or whatever, right?

Keller: Yeah. No, but I mean, it didn't seem that crazy to us.

Brett: Just figure it out.

Keller: We were pretty sure we could build something that could fly. So we moved out there. And of course it was harder. I mean, we crashed a lot for the first couple years. We had no idea what we were doing.

Brett: And you did this before you knew that a government was going to let you do this?

Keller: Yeah. We did it in parallel. We were starting to build the very first version of the thing. And then in parallel, we were flying all over the world talking to these different governments, trying to figure out who would be crazy enough to trust this team of 15 or 20 nerds to help build a new kind of autonomous logistics system for their country.

Brett: So how did you even get a government to meet with this company that's about to run out of money and knows nothing about any of these topics?

Keller: I think-

Brett: You can't just knock on a door, I guess.

Keller: Well, we just flew there and showed up. And then you walked to the Ministry of Health because we had a feeling that we should start in healthcare. Well, we just felt like, "Okay, if you're trying to get a regulator to give you permission, you better have a really good reason." And for us, the best reason was like, "This is literally going to save the lives of moms and kids. If you don't do this delivery, that person is definitely going to die."

Brett: [inaudible 00:48:23]?

Keller: Is there some risk involved in us flying from here to there? Yes. We can quantify and validate that risk over time together, Zipline and the regulator. But if we don't do this delivery, that person is 100% going to lose their life. That felt to us like the strongest possible argument we could have.

Brett: Did you think of any other ideas or that was it?

Keller: Not really. Yeah. We were pretty focused on this idea of healthcare logistics. We thought that would be the right place to start. You could focus in countries that had a centralized government-run healthcare system. So you could get the whole system through a single partnership with the government.

Brett: And the idea was that this was going to allow us to validate everything?

Keller: Yeah. The idea was this would be a really powerful first use case for the technology because it could save a lot of lives. It could save the government money and it would be the perfect argument to be made why this should be allowed from a regulatory perspective. And I still remember meeting the Minister of Health of Rwanda at the time saying, "Oh, we're going to build this complicated logistics system and it's going to deliver all medical products to every hospital and health facility in the country." And she said, "Keller, shut up. Just blood." And that was the best advice the plan ever got. I mean, every expert we talked to in global public healthcare told us this idea was never going to work. We were completely off. It was a waste of our time. We would never sign contracts. The unit economics would never work. It would never work reliably. We'd never figure out manufacturing, blah, blah, blah. We'd never get regulatory.

Brett: Why did that not discourage you? Did she tell you that in the first meeting that you had with her? She was like, "Okay, let's go do this," and then you had to go do it or there was a whole set of things?

Keller: No, no, it was basically that. Yeah. I mean, she was excited. And then she suggested I meet with the president. And the president of Rwanda is a really special... He's a very special leader. He's very admired, probably the most admired leader across all of Africa. Highly technocratic, highly entrepreneurial, really pushing the government to move lightning fast and innovate.

Brett: Very Singaporean type?

Keller: Very Singaporean. Singapore for sure is their role model. They are the Singapore of Africa. He was willing to make this bet on us, again, just as this crazy team of 20 people.

Brett: And do you think it's because they saw something special in you or it was such a problem that they're like, "Why not let this see if they can... They had nothing to lose and no one else was going to solve it"? Or did you manage to convince them that this was actually going to work and you were credible for this and this reason?

Keller: I think it's a couple things, which I've only now learned much later were the case. One, the morning of the day that I met with the Minister of Health in Rwanda, there was this tragedy that occurred where a mom had severe blood loss during birth and they didn't have the blood products even though they should have at the hospital. And so then the doctor got into the car with a nurse and drove to the nearest blood transfusion facility. The traffic was super, super bad, so it took them like three hours to get there. Once they got there, that facility happened to also be out of stock of the specific blood products they needed. They had to drive another two hours to another place. I think they broke down along the way. They finally got there. They had to wait in a long line to get the products. They finally got the products, drove back to the hospital, and the mom had passed away. And there had been this long email thread, apparently. I only found out about this like a year ago. A long email thread that had been circulated amongst the executives at the Ministry of Health saying, "How could this have happened? So many things had to go wrong for this to happen." And then it just happened to be the case that like on that day I walked in and was talking about this totally stupid, naive idea. And so maybe for her, it just sort of felt like this idea is probably stupid, it's probably not going to work, but maybe there's like some sense of like divine destiny or whatever that's like, "They're here for a specific reason and I could point them at this problem." So I think there's just luck. There's just luck involved. We showed up on the right day meeting her. And I think the other thing that I can't confirm, but I have now understanding a little bit more suspected is that think about it, I was showing up so clueless, like wearing tennis shoes and a hoodie to a meeting with the president of a country. I had no idea how to even address him. I had no idea how to address the Minister of Health all the proper honorifics and things like that. I mean, I'm 24 years old. I looked like such an alien. And when you think about the kinds of companies that typically work with a Minister of Health or with a president, you've got a perfect GE salesperson with like polished shoes, nice suit, probably a specific age range, not technical, selling them MRI machines or a bridge or a power plant or who knows. But those projects are often boondoggles and often don't get built and they are not like founder-led for all these reasons. And so I think that in that case, it may have helped us that we looked like... They probably didn't know quite what to make of us, but we were totally different. We were like 24 year old engineers who were building the technology ourselves and who were telling them we were going to like move to Rwanda and not stop until it was done. And so despite the fact that you probably could have easily been like, "Well, who are these 24 year old idiots and how quickly can I get them out of my office?" I think instead they may have kind of had the opposite reaction, which is like, "If all of these GE salespeople who come in here and we sign these huge contracts and then nothing ever gets delivered for five years, maybe we should try to take a bet on a engineering-driven startup with a really clear vision." There's a good advice I would say for young founders is raise money on the dream, not on the reality. And so I think that we did do, we raised money before we were into the specifics. We had like a cool demo that we could show in Half Moon Bay at our headquarters. We were out there trying to figure out like which country would let us do this for the first time. We raised a small round of financing, and that was by far the hardest round of financing the company has ever raised. I mean, we were like begging and pleading we would literally take money from anyone. In fact, we did take money from a couple sketchy characters. That's the only round we're like, literally it didn't matter what your name was like.

Brett: The wire.

Keller: Yeah. If you're ready to give us money, we will take it because we need to raise $2 million. And it's so incredibly stressful getting that round to come together and very, very, very painful. But we raised this small round of financing. We convinced Rwanda to launch with us. We had signed this contract, and then we launched. And the launch was a complete disaster.

Brett: But how long did it take from when you signed the contract to when you tried to do your first one?

Keller: Probably like three months or six months.

Brett: And you had like a prototype-y version in California.

Keller: We have the product mostly ready.

Brett: And you moved the whole company there?

Keller: We had the product mostly ready. We didn't move the whole company, but a lot of us went and just lived there. I went and lived there. Maggie, who had joined the team at that time, moved herself to Rwanda, built a lot of the initial operating software systems that we needed to do maintenance and inventory management. A lot of the basic stuff that we had no idea how to do. Ryan, my co-founder, went out and spent months on end there. And then we also built this team of extraordinary Rwandans. So we hired full-time to start leading the distribution center and leading operations.

Brett: Why was it a disaster?

Keller: Because we were clueless. We had no idea what we were doing. Everything was designed incorrectly. All of the hardwares were super unreliable. Many of the basic design assumptions we had made were incorrect.

Brett: Was that knowable before you started or was only knowable after having done the wrong thing? What did you actually do and did it feel like-

Keller: Well, I mean, certain things... No, I guess not everything. But so much stuff we hadn't even thought about. We were really focused on designing a really cool aircraft, but it turns out the aircraft is only about 15% of the complexity of designing an autonomous logistics system. We suddenly had to be really good at inventory management. We had to be really good at cold chain. We had to be really good at maintenance. We had to be really good at air traffic control. We had to be really good at vehicle to vehicle communications. And we had to be really good at weather management and weather simulation and prediction. We had to be really good at ground infrastructure, which ground infra was definitely one of the things that was most painful.

Brett: What's ground infrastructure?

Keller: Just like all of the recovery system and the launcher that we were using to get the airplane into the air and then to get the airplane out of the air. If you see early versions of the ground infrastructure that Zipline designed, you would know why it wasn't very reliable. We were literally using deep sea fishing poles that we bought from Walmart and then we paid a bouncy castle company to make these big inflatable pads for us and that's what we launched on. So when we say pretty crappy first product, it's like, this is... I mean, come on.

Brett: It's obvious.

Keller: You can't get any more scrappy than that, and that is what we deployed. And I remember when we were about to be launching in Rwanda, I was out there, we had the president coming the next day. When the president shows up, they not only had built this huge tent and they had all New York Times and BBC and Al Jazeera, all the major global media were going to be there to film the president, launching the first autonomous aircraft. And Abdul and I, Abdul's the first person we hired in Rwanda, or the launcher had completely destroyed itself. It destroyed itself every third launch, basically. The launcher had completely destroyed itself. We're on our backs in the dirt with screwdrivers in our mouths trying to take this thing apart and put it back together and fix it. And meanwhile, there are like five to 10 special forces, like military guys, like Navy SEAL looking guys who wherever the president goes, they show up 24 hours ahead. So it's like the middle of the night, we are on a video call back to the US trying to have engineers in the US explain to us how to take this thing apart and put it back together, make it work. These soldiers are watching us, "You guys are completely fucked.' They were watching us. "What are we even seeing here? This is not going to end well." Real world is infinitely complicated, and there was no way to learn it by just going and doing it. We had tested these systems pretty extensively in one very specific place on the coast of California and Half Moon Bay. We had not tested all of the other stuff like inventory management, cold chain, air traffic control software. None of that was tested. We were building that all from scratch as we went. So suffice it to say, the president showed up, launched an aircraft, the aircraft flew off into the horizon. I thought we're probably never going to see it again. Half hour later, the aircraft came back over the horizon and landed. We were trying not to act as shocked and surprised as everybody else in the audience, but we were just as shocked.

Brett: So it goes off and you're thinking, "Well..."

Keller: That's it. I mean, even in that moment, I was still very much thinking, "Zipline is probably going to fail. This is not going to work. That aircraft is going to crash. It's not going to come back." We were highly uncertain. We didn't know. Miraculously, the aircraft came back and landed and we acted like it was totally normal. We really, I think, did a good job of acting like [inaudible 00:59:03].

Brett: And people are blown away by this, right?

Keller: People are freaking out. I mean, it looked insane. It looked like true science fiction technology. This aircraft had just autonomously flown over the horizon, delivered something to a hospital and come back. And we acted like it's totally normal.

Brett: [inaudible 00:59:16] everything?

Keller: Yeah.

Brett: And what did the Navy SEAL guys... Did you look at their faces?

Keller: I don't remember the Navy SEAL guys, but I do remember the aircraft came in and landed. We built these fences around our distribution centers. And so there's this big fence all the way around. And it was literally as far as the eye could see three different layers of kits on the fence for hundreds of yards. It was thousands of little kids, two, three, four year olds, and then the next layer was like the 10-year olds and the next layer was like the 16-year olds. And I remember the president watched the aircraft land and we were kind of talking about where this could all go, and he pointed all of them on the fence. He said, "Those are all the future engineers of Rwanda." And it kind of made me think like, "Man, this guy really gets it. Like, he has this bigger plan in mind." Why is he working with us? He knows that for Rwanda to win, they have to innovate, they have to be the Singapore here and take risk and be the proof of concept for these new kinds of infrastructure and robotics technology. But anyway, suffice it to say we launched and he was blown away and we spent two hours talking with him and brainstorming about with the future and then he left and then the system promptly broke. Everything broke again for the next nine months. And we spent nine months basically pulling all-nighters, trying to fix all these different systems, trying to make them work reliably. The good thing was we'd only added one hospital.

Brett: [inaudible 01:00:37] problem?

Keller: We'd only added one hospital. We were only serving one hospital at that time.

Brett: So you had a distribution center and one hospital.

Keller: And one hospital.

Brett: That was what you were trying to get to be [inaudible 01:00:43].

Keller: That's it. That's all we had to do, was just make it work for one hospital. And

Brett: It's basically hard for the reasons that we talked about. It's just infinite complexity in the real world environment.

Keller: So much stuff.

Brett: Weather, wind, on and on and on.

Keller: Weather, maintenance, hardware reliability, supply chains, packaging, air traffic control, communications architecture, cold chain, inventory management, supply chains of making sure you get the right products to us so that we could have them to deliver.

Brett: Was there an 80/20 dynamic where there was 20% of the things that drove a disproportionate amount of the pioneering that you had to do?

Keller: For sure. Yeah, I would say so. But it's just you can't really predict what those things are going to be.

Brett: Right. So you have to just...

Keller: You just have to do them and then iterate incredibly fast.

Brett: So then once nine months later you got to...

Keller: That one hospital started working correctly.

Brett: In a reliable way?

Keller: Yeah. Then we immediately expanded over three months to the other 20 hospitals.

Brett: And that was vastly easier?

Keller: It was way easier. It was pretty easy to add the 20 hospitals. So in three months, it took us nine months to just serve that one hospital. And then once that was reliable, we expanded to 20 more hospitals. So in the first year, we actually did what we told the government we were going to do, which was deliver blood to 21 different hospitals. Easy peasy. We ended up adding 20 more hospitals. Then we expanded to 50, then 100, then 400, then 1,000.

Brett: Across multiple distribution centers.

Keller: Soon to be across multiple distribution centers and then across multiple countries. Then we got to 2,000. Today we serve 5,000 hospitals and health facilities globally. It's become the largest commercial autonomous system on earth of any kind, ground or air. We just crossed 135 million commercial autonomous miles. And yeah, I mean, the company, today Zipline saves about 17,000 lives a year. And we're expecting to grow another 5X over the next two or three years. That's all healthcare.

Brett: [inaudible 01:02:28].

Keller: I'm just talking about the healthcare side. Yeah.

Brett: Yeah.

Keller: Yeah. We expanded from blood to delivering vaccines and then transfusions and infusions and cancer products, insulin, all the different programmatic drugs. So really, I talked to so many hardware entrepreneurs who are starting in seed stage or series A scale. I'm always looking at the product and thinking about manufacturing and the hardware development process. I'm always like, "Cool, how much do you think it's going to cost to build that?" And they always say a number. I'm always like, "It's going to cost 10 times that much money to build that product." And we never changed the price. We just lost a ton of money on every delivery for the first two years. And then we just drove that cost down from 10X to 6X to 4X to [inaudible 01:03:10].

Brett: And it's manufacturing of the actual aircraft-

Keller: That's right.

Brett: ... gets a disproportionate amount of cost?

Keller: Everything. Yeah. The bomb cost of the aircraft, the volume of flight, labor, land, ground infrastructure. It's really everything.

Brett: And it's just a game of inches of just...

Keller: Game of inches. Yeah. Every day you're just trying to make 1% basically.

Brett: [inaudible 01:03:29].

Keller: We're now at 1/3 of X in terms of the overall cost. So it's really cool. We went from 10X to 1/3 of X over eight years.

Brett: Walk through the path from what you were doing outside of the United States in mainly healthcare use cases to what you've been doing in the United States. How did you know it was the right time and what was the thinking behind it?

Keller: So the crazy thing is that as we were just scaling and scaling and scaling outside the US, Zipline almost mistakenly became the largest commercial autonomous system on earth. And as we were growing, a lot of big companies in the US started noticing what we were doing and started saying, "We really want that here." We started signing these big contracts with big players in the US. We work with healthcare systems like Cleveland Clinic and Memorial Hermann, Ohio Health, Michigan Health, Michigan Medicine, Texas Health Resources, many other amazing hospital systems. And then we closed a large contract with Walmart where we started delivering for Walmart directly to homes. We also work with a lot of different restaurants. And so all of it was just around enabling instant teleportation like logistics from any hospital or retailer or restaurant directly to customer home. And so we're able to launch with all these [inaudible 01:04:38]-

Brett: Isn't going back to reality is infinitely complicated, like delivering a burrito to somebody in Texas is different than delivering blood to a hospital? Or once you got everything working, it doesn't really matter what you're doing?

Keller: Doesn't really matter. Yeah. I mean, this kind of technology is generally kind of payload agnostic. We're just delivering from a building to a home. Same thing with UPS. I mean, UPS is delivering all kinds of... They have a huge healthcare logistics business. They deliver all kinds of different packages for all kinds of people all over the world. So similar thing here. It did mean that we had to build an entirely new kind of technology platform. Like, we had to go-

Brett: Why is that?

Keller: Well, because the problem of delivering to homes is just way different. The way we were delivering for the healthcare logistics business originally outside the US, we were flying hundreds of miles to make these deliveries to very rural areas. Focusing on the US, delivering to homes meant we needed to deliver in a highly precise, controlled, gentle way, be able to put things like with dinner plate level accuracy directly onto your doorstep. And we need to do it while being very, very quiet and very, very cost-effective. The vision for Platform 2 was to build something that could be 10 times as fast as instant delivery in the US or using a car to deliver something and half the cost and zero emission. And that's what we launched at the beginning of last year.

Brett: Why did you want to do that at that point in time? It seems like with your original platform instead of use cases, you had a lot of market to go after. Was there anything that made you decide that this is the exact right moment to make this huge bet on an entire new platform, different use cases?

Keller: I think it just felt like this is by far when you look at logistics, this is like a market that is 100 times the size. When you look at rural healthcare logistics, but then you look at home delivery. Home delivery is a market that is 100 to a thousand times as large. And so if our goal was to provide a logistics service that served every human on earth equally, we had to do this. And so the moment we thought it was possible technologically, especially when we felt this very strong customer pull from a lot of the biggest companies in the US asking us to do this, seemed really obvious to go do it.

Brett: So I wanted to wrap up. When you think about other entrepreneurs going and building companies that have similar parts to it; hardware, breakthrough technology, regulatory, these sort of very unique styles of businesses, what have you figured out now through the last 13, 14 years that you would sort of share with them that might be useful? There's a lot of things where you just had incredible luck. There's a lot of things that are probably Zipline-specific that you can't just take out and drop in another company. But what about a few of the things that you definitely figured out that might help them in some way?

Keller: I mean, there's the trope hardware is hard. I think that is totally true. If you didn't have mental problems before the hardware startup, you will definitely have mental problems after the hardware startup. Nobody could emerge from this process normal.

Brett: So you should just start with mental problems?

Keller: Yeah, you may as well, you know? Yeah. I think that really getting a lot of these products to scale and learning what you need to learn, getting a couple iterations and figuring out manufacturing and supply chain and logistics and maintenance, these kinds of things, takes a decade. So you can look at SpaceX or Tesla or Zipline or even a company like Rivian, these companies all take a decade to get to meaningful scale. And so I think A, you probably have to raise a lot of money for that company. B, you need an amazing mission because you're going to need investors to be willing and team members to be willing to stick with this problem for like an entire decade before you really see like, "Okay, this is now on like a... It's growing at exit velocity or scape velocity." I think that the flip side of that though is that a lot of the biggest companies that are going to be built over the next... Probably most of the biggest companies that are going to be built over the next 10 years are all going to be hardware companies. And I think even when you look at AI today, I think there's this like interesting just observation, which is that like, so many of the problems that humanity has to solve, it's almost like the problems that are most crying out to be solved are the very unfancy ones. They require like going and getting your hands dirty in the real world, whether it's building like power plants or whether it's building new kinds of natural gas turbines for data centers or whether it's building autonomous logistics or it's fixing manufacturing in the US and onshoring a lot of those jobs and a lot of that innovation again. It just seems like a lot of people are much more focused. They're like trying to find these things to do in the knowledge economy, but like so much of the stuff that's going to be critical for the future of the United States is figuring out how to build physical stuff again. You have this like strong feeling if you drive around, John Collison talked a lot about this idea of like when you walk around a city and you even just look at a bench, you've seen that Tweet. It's like someone made that bench and designed it and funded it and actually built it. It's like a passion project for someone. But think about like the bridges and the tunnels and the airports. These things were all built by our grandparents for us and they're now often like kind of falling apart, but we've stopped building them. I think it's like really exciting and important to think about what is it going to take for the US to get back to building those things. There are hundreds of hardware startups waiting to get built in that vein, they're going to be super important for America's future.

Brett: And do you think if somebody wants to go after those unsexy, important, real world problems and they're 18 or 19, they should just go do it or they should go do X to prepare themself to be the right person to go build the next Zipline in some other market, solve some other problem.

Keller: I think probably both are good options, but I do think it's really hard to... I mean, there's no school that you can go to prepare for this, and that you can only learn by doing. So you either learn by joining a really exciting, fast growing startup that's working on these kinds of problems, or you define your own problem and go work on it from scratch.

Brett: Good place to end.

Keller: Cool. Yeah.

Brett: Thank you so much for doing this.

Keller: Yeah. Thanks, Brett.

Brett: I really appreciate it. And thanks for going over a little bit.

Keller: Yeah, of course.