Trevor Blackwell, Co-founder Y Combinator; Founder, Anybots

I'm Jessica Livingston and Carolyn Levy and I are the social radars. In this podcast, we talked to some of the most successful founders in Silicon Valley about how they did it. Carolyn and I have been working together to help thousands of startups at Y Combinator for almost 20 years. Come be a fly on the wall as we talk to founders and learn their true stories. Carolyn, I am so excited because today we have on the show my friend and Y Combinator co-founder Trevor Blackwell. Welcome Trevor. Welcome Trevor. Thanks. Glad you could be here. Yeah, no, I've been wanting to do your podcast until you launched it. We've already heard a lot about you from Paul Graham. Your name crops up a bit here and there. And so we'll definitely have some specific questions that we want to clarify from the Paul Graham interview. But I want to go back a little bit to how did you make it from where you grew up in rural Canada to Harvard's PhD program? Well, so I grew up in Saskatoon, Saskatchewan, north of North Dakota for people with more American geography references. It was a small city and I was like the only, there only like a handful of people that were into computers when I was there in high school. And I'd read about Silicon Valley. I had an Apple, I somehow managed to get me an Apple II computer when I was a kid. Wow. And I read everything about it and about how jobs in Wazniak had created this thing in their garage. And I sort of imagined them like with their fingers freezing in the garage. Because that's what's, can you help us to schedule what your glasses are like? They were in California. You were in Saskatchewan. Well, I'd never been to California. Sunny, Sunnyvale. So I always sort of wanted to go there to see it. But you know, I ended up in Ottawa for college at Carlton University, studied engineering and computer science there. And then I started working for Del Northern Research, which was kind of like Del Labs for Canada. That's the research arm of the Canadian phone company. After college? Yeah. Well, actually during college in the summers and even in the evenings after college. And they were doing a joint project with the guy who ended up being my advisor at Harvard HTCUNG. And so I got to know him through that project. And they put in a good word for me and I got accepted there. So you know, put everything in the car and moved to Boston and started at Harvard. So what did it feel like to show up in bustling Cambridge, Massachusetts from Saskatoon? Yeah, it was an eye opener. All kinds of things I'd never really seen before. But you know, it was like I got there and like, you know, it was this transition to suddenly being in a place where like almost everyone is smarter and more knowledgeable about the subject than I am. So that was, you know, that lit a fire under me at the, at the train catch up. But you must have caught up pretty soon because you met Robert Morris because you were in like getting the PhDs at the same time, right? Yeah, we had the same advisor and offices weren't that far from each other. So Paul had started via web and he was good friends with Robert and he asked Robert, who's the best programmer in your program? And he said, you, you must have caught up very quickly. I guess so. Very modest, you're very modest. I mean, I was good at the practical stuff and it was the theory that I felt insecure about because I'd sort of studied computer engineering rather than computer science. So I could program, but you know, they would keep talking about these complicated theory things that took me well to catch up on. Well Paul described you as basically a programming monster doing prodigious feats of programming, which now that I know you, I believe. But do you remember when you were with Robert in the program and he said, hey, I'm working on the startup, do you remember getting introduced to Paul and the whole via web thing? Yeah. So he introduced me to Paul and Paul, like explain this idea that lots of people would want to sell things on the internet, which I believed because you know, this was late 94 and Amazon sort of existed. There were a couple of other bookstores on the web. I just loved it because I could order any book I wanted if I just knew the title instead of like whatever might be available at the campus bookstore. So I liked the idea of there being more things for sale on the internet and I believed the idea that every catalog company would eventually need to be on the internet, which was a bold idea in 1994. Yeah. And you probably dealt with your fair share of catalogs being like an electrical engineer, right? Wasn't every thing that didn't you have huge catalogs? Well, yeah, I mean, in Saskatoon, there was not a retail store for computers or electronics. You know, order stuff from like Digi Key existed then in the 80s. That's still the standard place to buy electronics components. But you know, at the time it was like in the back of a magazine, there'd be this page with a form to fill out for quantity and price and whatever. You'd send it in with a credit card, with my dad's credit card number. And so yeah, being able to order stuff on the internet seemed great. I'm dying to know what your first meeting with Paul was like. What did you think of him? What was he like? What stage was VIA Web at? Well, what actually got me interested in it was that it was a software business that didn't require writing Windows software and putting it in a box and shipping it with version numbers. So like kids these days don't know how horrible that was. But in the pre-internet days of software, like to sell a piece of software, you would write it and then you would burn it on CDs, put it in a box with a nice glossy cover on it and sell it. And if there was something wrong with it, if there was a bug in it, too bad. The users had to suffer with it until like next year when the next version came out. And because there was no updates, no downloads. So that's a terrible way to write software because knowing that any mistake is going to take a year to fix, you probably want to spend like six months testing every aspect of it. You got to test it on people's weirdly configured computers and with other software. And it was such a grind to just make a simple piece of software that did something. So not many people wanted to be in that business. So and certainly not me. So like when Paul pitched this idea, like we could write all the software on our Unix systems that we liked from using them in grad school and run it all on our own hardware and update it whenever we liked. To me, that's sort of what got me excited about it. Okay, the novelty of the sort of web-based software, basically. You never heard of anyone doing this before. Like this was not a thing you'd ever heard of. No, not really. Like it existed for open source software. But I can't really think of someone who made a business doing software in that way. So how did they really you in? How did they entice you to get involved? Because you were getting your PhD. Yeah, no, I had, I had, yeah, I had official things to work on. Well, Paul showed me this demo of this thing that I'd never seen before, which was using a web browser. But instead of like when you clicked on a link, instead of just like going to a page that that make pointed to, it would do something. It would change something on the system. So he had this, you know, little demo. It was no colors or anything. It was just, you know, like blue links and gray background. But you could sort of create data on the server just by clicking links. And I thought, wow, this is great. Like some day all the software is going to work like this. And so I wanted to just try writing software that way to see if it really was possible. So I wrote, I, you know, built this completely separate thing that just did that. I started it on my birthday, which is why I remember the date. I thought like it's my birthday. I can do whatever I want. I was in at the, in at school, I didn't want to work on my PhD. So I thought, I'll just write one of these things where you can click on links and it does stuff. Did you write it in small talk, Trevor? Yeah, I wrote it in small talk, which is, it was a good language to do it in. And this was also a good language to do it in. Yeah, I wrote it in small talk. And then a couple days later, I showed that to Paul and said, hey, check this out. And so you could sort of see that I'd gotten the idea. And then, you know, we kind of started working on it. It was quite a while before we got official about, you know, stock or paychecks or anything like that. But we sort of, yeah, decided like I would be, I'd be part of this thing. Did you like finishing your PhD or like, what was the timeline in terms of? if you're getting really involved in V.O. Web or graduating, like what happened there? They overlapped in the most difficult, stressful way. I did finish my PhD. I graduated on the same day we signed the acquisition deal with Yahoo. Don't you remember that? I do now from Paul's interview. Yeah. Yeah, I had to duck out of the ceremony in my crimson robes running through hard times. I was in Harvard Square to like, "Sign this thing." No way. I mean, you literally ducked out of your graduation. Yeah. But, you know, like months before that, I realized like, actually, I really just like, I like building this thing. I want to keep working on this company. I think it's going to be great. But, you know, I said I'd like do a PhD. So I got to write up a thesis. So yeah, I took a sort of a bunch of the things I'd been working on and put them all into one big file and started editing them until it kind of made sense. For your dissertation, you mean? Yeah. Yeah. Looking back on your dissertation, does it feel like it was a fascinating one? Or did you just say, "I have to write a dissertation. I'm working on this kind of thing. Let's make this work." I think there's some value in it. Like it definitely is like there's sort of one core idea and then there's a couple of weekly related examples of how to use it. And because those are just the things I'd been working on. And like I had this idea for a while and thought, you know, this would be a good way of, it's a way of measuring the performance of systems. So when you're trying to make a computer go faster or some software go faster, you got to measure it. Lots of ways of getting fooled because computers just go faster and slower sometimes for mysterious reasons. And so I had this way of making more robust performance measurements sort of in the back of my head for a few years and then just tinkering around with things I'd used it on a couple of projects and so I thought, "Okay, that's going to be my thesis. I've got two chapters already." Does this have anything to do with the bogometer or nothing to do with the bogometer? No, not really. Maybe the name bogometer sort of comes from that. Well, because, okay, so the bogometer, this was another like new idea that no software company had probably ever had this opportunity before. But we had software that was running on our servers and the disks were clicking away. And you know, you could sort of like notice in the middle of the day the disks would be clicking away more like disks used to click. I remember. And we thought, "Let's make a big mechanical dial that has like a clock hand that goes back and forth that shows how many hits per second and how many orders we're getting." I mean, at first we were getting like zero orders per hour, maybe one. So that dial didn't have much range. So I made this thing with one dial that was like the number of hits per second. So I had a motor and a big pointer on it and a laser print of scale so I could change it when we started growing. So I put all these things together and it was and connected it to a serial port on one of the computers and it showed us how many hits per second we were going. And so it ended up being always one of the little highlights of the tour. Like we'd have investors or potential employees or something or customers coming and we'd say, "Look, we're getting 4.5 hits per second on our servers." Why is it called the bogometer? Like what's the bow? Well, because I thought, like Paul sort of suggested this idea and I thought, "Oh, that's a bogus. It doesn't matter how many hits we're getting like this second as long as we know day to day." So I called it the bogometer and then I could actually ended up being a really good idea. Because sometimes something would go wrong. The internet would go down or something and all the day I also would go, "Zzzzzzzz. And you could hear them because they made a little whizzing sound." So you'd say, "Whoa, something's going wrong." Now, was the bogometer in your office which actually was also the server room? I put it, well, maybe one version was inside, but I put it on the wall outside so people would sort of look in and there'd be the rack of slightly wobbly servers. But then this, because it grew to have I think eight dials on it showing all kinds of things like revenue and you know, and then-- Oh, let me pause you on that theme. So because I know we're going to talk about robots soon and I'm just wondering like the building knobs and dials. Is this very early robot stuff or had you already been building robot stuff as like a hobby on the side? Oh, I started building robots, you know, in middle school. Didn't do a lot for my social popularity, but it was fun. And like there were pretty clunky early on. Like the first one was like a big old oil drum, you know, one of those sort of green and white striped drums and I put some holes in the bottom and had these army surplus motors and like a box with switches on it to control it and I eventually put an arm on it. And so I've been building robots, you know, sort of what I always wanted to do, I think. Okay. So the Bogometer was not like your debut robotic thing. This is like hobby that goes way back. Okay. No, no. No. One of Trevor's many amazing like side projects that probably could be a standalone product in a company. Right. But he just like whips it up on the side. It would be a good startup. I keep hoping one of the, you know, metrics as a service companies will have that as an option like a wall display for your office with real-time stats. Do they not have that? Yeah, I would think someone would. I haven't seen it. Yeah, it goes into the office anymore. Right. You heard it from your first, great new idea. Trevor, do you remember when the power went out in Cambridge and you guys had to keep the servers again, I'm going to reiterate they were in your office. Yeah. They're like tower, tower servers sitting on one of those like wire racks so they kind of wobbled a lot whenever someone locked by. Wasn't it hot and it's a hot, right? Like wouldn't it be hot in there? They did. Yeah. No, they were, they were quite hot. We had, you know, and like with one, it wasn't so bad. At some point we had maybe six of them and then so then I got like a window air conditioner and that wasn't enough. So I stacked like, ended up with stacked four window air conditioners to try to keep this room, which was like, you know, we were in, we were in, we were right in Harvard Square, which is a great location because it was, you know, just a few minutes walk from my grad school office and their restaurants and shops. But you know, this thing had been a house, I guess, in a long time ago and you know, someone converted the third, four to be an office and it had been like a professional house, right? Yeah, old, yeah, old, wouldn't probably 1900 or so structure and it had a bar in the bottom. It's Irish bar. That's fun. I got a little loud at night. Um, so you were, your, your room wasn't your office also called the hot tub because it got so hot. It was, it was hot. Okay. And all those servers looking away in the fans, and the bogometers going back and forth. So then there's a massive power outage in Cambridge and tell me what happened relating to the generator. Yeah, it was, it was a, it was a big power outage because some transformer blew up underground and it was, they said it would be like maybe a day or more to fix it. So we had a battery backup system. I was surprisingly legit of us to have that, but it's, you know, it was good for half an hour. And so in that half an hour, I thought, okay, I'm going to drive out to Costco by a generator and get back here in half an hour. Costco is actually pretty far. But so yeah, so I zoomed over to Costco got, you know, this horrible gas generator, you know, a generator or something, maybe 10 horsepower and lugged it up the stairs and put it out on the fire escape. I was going to say wood building gas power generators just sounds like a disaster in the making. Yeah. Yeah, I'd like filled it up at the gas station just like, you know, pour gas in it. It never can or anything. Somehow we managed to get it, we managed to cut over to the generator and get it, get it going. But it made such a racket that the people downstairs, like this lady downstairs who are, I don't know if she was exactly our landlord, but she was the landlord's agent anyway, came and told us like, no, no, no, you can't have that thing up here, you know, they are extremely loud. Those gas power generators are extremely, yeah, especially like, you know, it's like a power failure in the 90s and the city is suddenly very quiet. You can see the stars, I was a bomb. What did you do? Well, also, if you don't have a gas can, how were you proposed to refill it? You know, one hour at a time, I guess. You moved the generator. Did you not? Yeah, we moved the generator downstairs, I think. And at some point, I think that's what I was going to say. We hadn't had it on long enough to recharge the batteries and we lost power. But then eventually, a couple hours later, power came back on and it was okay. But at that time, we could see it on the Bogometer. How much revenue per hour? And it was real numbers. It was tens of thousands of dollars an hour. And our merchants were losing that if our servers were down. And they would call us and yell at us, like suitably angry for how much money they were losing. The upside of this kind of software is you didn't have to have these boxes with yearly releases of the new version. But the downside is like something goes wrong and you gotta get up in the middle of the night and fix it. Like I wore the pager for basically the whole history of Vueweb. You know, there were lots of things that went wrong. So a lot of late nights. Late nights in the server room. Yeah. But the generator, I mean, what I remember from this story and that's why I wanted you to tell it, wasn't it like on the sidewalk with like 20 different extension cords going all the way up to the third floor into the window through the office to the other side of the back of the office where the servers were like tight as a pin. Like so tight and like chest level because there was no slack. Yeah. Like it's tension. Sorry. I'm sure the story won't seem that good on the podcast, but it's like one of my favorites. I just love it. Were there any electrical engineering adventures when you were at Vueweb? Because Paul said you used to do a lot of crazy things in the electrical engineering department. A little bit, a lot of our users wanted their orders faxed because the people in the warehouse probably didn't have email. So someone would place an order online and our system would fax it to a fax machine in the shipping warehouse. And you know, at first we just had one fax line, but that soon became not nearly enough. And so we had this big rack of fax modems, you know, 15 or more. All sort of wired up to the phone, the phone closet. We were pushing the limits of how many phone lines you could get into one of these old three story walk ups. So there's that, you know, various adventures in like trying to keep the servers up all the time. Like we started off. We had one server and if it was down like nobody even noticed, you know, before we had users. And then just gradually, gradually it started getting more critical that we couldn't have the servers be down. So at some point we thought we should get a battery backup, you know, because if the power fails, it's not just that it's down, but at least at that time, like shutting those kind of computers down without, you know, that running a shutdown command could corrupt the disk and then you could be in big trouble. So we got this battery backup thing. But, you know, the computers are already plugged into the wall and we wanted to plug them into the battery backup system. So there's no proper way of doing it. But I made this cord that had like a mail plug on both ends and inject, you know, so I could sort of bridge between the power strips that were running the servers and the output of this UPS. You know, battery backup system and surprisingly it did not blow up. Yeah, I was going to say that sounds really scary. That's scary. Oh, I got to you say, okay, it's worked out. That's for the listeners. That's like the famous Trevor line. Oh, okay, it should work now in his Canadian accent. And how many would you say 50% of the time it would work? Yeah. And the thing is you guys couldn't move because you're all still at Harvard. So like you can't go move to more convenient location to put your booming business into. You know, we went around and looked. We looked at, you know, legit looking offices like, you know, glass towers out, you know, on Route 128 and they were so soulless and you're him. Yeah. You know, I probably could have put up with it, but Paul's pretty sensitive to the environment. Yeah, the atmosphere and he just wasn't going to do it. He had his routine and he probably, I mean, I didn't know Paul during V.O. But I know for sure he had a routine where he like probably got his breakfast, went into program and wanted to like walk to and from work. Did not want to commute out to Route 95. What was Paul like to work with back then? Like I think the defining, the defining dynamic was that he would say, you would insist that there was a right way of doing something and I'd be like, oh, come on. No, I can't be the right way. And then like I go around and think about it for a while and then actually realize he was right. And you know, if that only took an hour, it was no problem. But sometimes I'd like build the other version of it, you know, and show it to him. And you know, it was still wrong. And eventually I had to realize like, yeah, actually he was right. So. So that's familiar. Actually, I have to share with you Paul described you this morning as an extremely powerful engine with a rather small record. I love that. I know I just love that. Oh my gosh. Okay, so Yahoo, as we heard, you signed the deal with Yahoo to get acquired on the day you got your PhD. You move out then to Mountain View. Yeah, so it was clear this was going to be the next chapter of our lives in California. So, you know, we moved out there, bought a place like we had to move a dozen employees to had to relocate, you know, had to start using Yahoo's server infrastructure instead of our racks in the walk up. And so yeah, it was like a very busy summer. Were you excited to go to California? I was. Yeah. As, yeah, like as soon as I came out and we were driving around and there's like Apple and there's Hewlett Packard and you know like Silicon graphics and all these great companies who's stuff I've been using for years and years. And just like there they are and the employees are walking in and out. So. And let's not forget though that Yahoo back then in 1998 was like ground zero of the internet bubble, right? Yeah, no, they were hot. Well, they got a lot hotter after they bought us. Not not just because of you know, maybe we helped a little. But yeah, they were they were the the darling of the internet. So what was it like working there? Did it feel like big company yet or did it still feel start-upy? Well, at first it felt really quite start-upy. It was you know, small number of people just doing the work. They had sort of crappy offices, you know, two story industrial park and Santa Clara. And they basically left us alone to just like you know, we had a we had a business that was working the technology work. They just left us alone to do that. But we could you know, we had more resources. We could get more we could get hardware and you know much much more internet bandwidth and. And then you know, they started asking us to integrate with some other systems, which which made sense like using the Yahoo login and stuff like that. And that was all good. But it did start. I mean, it definitely turned into a big company pretty quickly. I think they felt that that was like they were supposed to be a big company. And so they had to start doing all these big company things, even though no one there really had much experience with doing big company things in a way that worked. So you know, like one of the things that was frustrating was like at Vue Web, we just talked to our users. We knew dozens of them by name and we could just email them or call them up and ask them questions about like what should the software do you know if this happens. And so but Yahoo had this group of like user interaction research. And so they you know, they wanted to sort of do all the user interviews. And you know, they would produce these nice reports, you know, with statistically valid observations from users. And you know, we'd get them like two or three months after we had asked the question. Which by then I'd probably already completely changed with the software did already. So yeah, it was like that was that worked poorly. I mean, maybe it worked okay for you know, Yahoo's core business of you know, search and Internet directory, but it didn't really worked very well for us. What happens and you give them that feedback are they like, oh yeah, thanks. And they just keep doing their same thing. Well, you know, I could have been a better, you know, a corporate person. I wasn't very good at like I would just get frustrated and do it my own way rather than try to change the system. I did once try to change something big there, which was like at some point I thought, you know, basically the core product for. [BLANK_AUDIO] that Vioeb became, which became Yahoo Store. I thought, this is, you know, like, it feels like it's 90% done. I'm mostly doing maintenance. I wanna try something new. And so my idea was that, you know, we had all these merchants who we knew would want to buy ads online for their stuff, you know, to pointing to their store. And which, like, wasn't really a thing then. You know, at the time the banner ads were like, you know, for, you know, TV channels or gasoline stations or like the same thing you would see on TV. - Mm-hmm. - It was like TV ads or magazine ads. But I figured like, most of our merchants would spend a thousand dollars a month on, you know, advertising their store and driving traffic to it. So I wanted to make, you know, that advertising product. Google later succeeded with AdWords, where you can just buy like a few thousand dollars of ads with a credit card. But at the time in order to buy ads, like you had to go talk to an ad salesman and they weren't interested in talking with you unless you were making, you know, at least a hundred thousand dollar ad buy. - Yeah. - So there's no way you could just buy a few ads with a credit card. - So did you build something? - I did. I built it and it worked great. And then, so like I built this prototype and went and showed it to the ad sales people. And it was like, look, you know, we can now sell ads to this whole new category of users. And they didn't get it. It was like a, it was sort of a case study and crossing the chasm kind of thing where they had this existing business of selling ads, you know, like multi million dollar ad buys to big companies, to proctor and gamble and advertising toothpaste and whatnot. And so this idea of like, you're gonna sell a thousand dollars of ads and I would be like, no, but like they're, we'll do that a thousand times and it'll be real money. But they didn't quite get it. And then, but the big problem was, like, I had to show a price for the ads, right, to buy them. And I mean, this is a holdover from legacy media that there's like a list price, which is like 10 times higher than anyone ever pays. It's like the rate on the inside of the hotel room door, or it, you know, like it cannot exceed that rate. And so they put this ridiculous rate up there. And then everyone gets charged less. So, you know, like the official rate for the ads would be like $80 per thousand ads per CPM. And but the actual price was like $8. That's what any of these big companies would get a deal for. And so they wouldn't let me put $8. I had to put $80. And, you know, because so that kind of killed it. You know, at the time, it was, like, it would have been a tiny fraction of their revenue. So it was hard to get anyone, you know, excited enough about this to like change this pretty big policy and reveal what the real price might be. Because, you know, a few suckers did pay the high price. Oh, interesting. I have on my list of questions. Why didn't Yahoo become Google question mark? [LAUGHTER] Oh, man. They could have-- Oh, I think that sort of explains part-- They should have bought Google. You know, they would have been able to buy Google for under a billion dollars at one point. And that would have been a great, great deal. But, yeah, didn't happen. You stayed at Yahoo sort of longer than a lot of people from VioWeb. Right. I stayed about three and a half years. At first, partly out of-- well, what turned me off was this frustration with not being able to launch a new product. Because at VioWeb, like, we-- Paul and I would have talked about it over lunch, and we would have launched it in the afternoon. That's the way we did things. Yeah. And then, like, doing it there, I spent, like, months trying to talk to the right people and get people on board. And eventually, it didn't even work. So what a frustrating thing that was. Did you try other projects or just that one? That was the-- no, that was the main one that turned me off, working for a big company. So then you decide to go and start any bots. Because I just want the listeners to know, like, Trevor was responsible for a really big scientific achievement, which was making the first dynamically balancing biped robot. Tell us what the dynamically balancing biped robot is and how it all got started. Well, after I decided I didn't want to work for a big company, I thought, OK, well, what do I want to do? I want to build something. And so I kind of made a list of all the things I thought that science fiction had been promising us for a long time and still didn't exist, and that I could contribute to. Like, it didn't need new physics or anything. And robots seemed like the obvious answer. So I left Yahoo on a Friday and started a robot company the next Monday. In retrospect, I encourage people to take a break between things. Because it's fine. Like, because I was all energized by this exciting new idea. But it had been a long time since I didn't break from working on VIA Web and PhD simultaneously. And then trying to transition to Yahoo. And then-- Yeah, it's a long grind. Yeah, so I started-- I wanted to build a walking robot. Like, two-legged human-sized walking robot. Because they have them in science fiction, and we don't have them in real life. So something needs fixing there. And pick up where you left off in middle school. Oil can. Yeah, basically. Did you immediately rent the space at 320 pioneer at that point or did it take a little while before you got to that? I started in the basement. So I bought a bunch of machine tools and aluminum and made prototype legs all sort of in my basement. And at some point, I needed a little more space. I subled a place from tell me, if you remember them. They did voice recognition. Yeah. And so I got this-- I don't know, it got a huge amount of space, cheap, because it was the dark days of late 2001. And yeah, so I started trying to get this robot to walk. And so the first version-- I had the hardware. It was two legs with a hip and a knee and an ankle and each. And so the very first thing I got working was, I built myself a little miniature version, sort of a hand finger-sized version of this robot with sensors on it. And I had the big robot sitting over there. And I could move the legs and the little robot. And the big robot would move the same way. So I thought, well, I wonder if I can just learn to make it walk like this, just by moving my fingers. So I spent hours and hours practicing and could never get it. It would just fell over faster than I could react to it. I could kind of make it stand and balance with deep concentration. As hard as balancing a ruler on your finger or something, you've got to be really on top of it. And yeah, so the manual control wasn't going to work. So I went back to working on the full software control for it. Can you explain, though, just for the listeners about, what does a dynamically balancing biped robot mean? So well, there are a lot of ways of building a walking robot. But the way that had been demonstrated before by Honda, in particular, the Honda Asimodo worked like this, was it had pretty big square feet. And the feet were quite stiff, meaning like the joint at the ankle was quite stiff, so that you could take this robot and pose it in a pose such that if you drew a perpendicular up from the center of the feet, it would go through the center of gravity. And then you could just like with the robot fixed in this pose, you could set it on the ground and take your hands off and it would balance. OK. It's a little-- it wouldn't be hard to push it over, but it would be stable. And so that's sort of straightforward to do with computational geometry. You can calculate where the perpendicular of things goes in three-dimensional space and where the center of gravity of the robot is. So that was the way that it worked, is that sort of geometrical thing where at every point in time, the perpendicular from the center of the foot that's on the floor always goes through the center of gravity. And that is a particular kind of look to the walk. It's very smooth and stable. But you have to have the legs bent because you can't sort of bounce up and down. And it has a particular kind of look to it. But it really can't deal with soft terrain, with walking on grass or anything. So I wanted to build the thing that walks like people do. And then the Honda one, they'd always put like a flat surface down when they did demos of it. Yeah. They brought their own stage with carefully constructed layers of-- you know, cross-members under the stage, so it had the same stiffness everywhere. Yeah, I wanted to build something that could just walk anywhere people can. So it had to balance dynamically. And also I wanted it to be able to run and jump and do things like that. So the other problem with the kind of stiff jointed robot is you can't jump, because you know, each of the joints, like if you think about the knee joint, there's the knee joint and then a gearbox and then a motor. So if you try to land a jump on that thing, there's just going to be gear teeth flying everywhere. So I built mine with compressed air power, used pneumatic cylinders for all the joints. Which, I mean, what's good about them is, well, they're springy and kind of the same amount of force per unit area that muscle produces, but 50 pounds per square inch is what you get from the cross section of human muscle. And indestructible, you could jump and bounce and, you know, kick the ground, you know, over and over again, and it wouldn't do any damage. Because I'd also seen, you know, robotics programs, like I'd seen some work at Carnegie Mellon, where they were building a walking thing closer to the the Asimov style. And everyone was terrified to actually do anything with this robot. Like they had this nice piece of hardware. But they knew that like if the software did the wrong thing, it would just like break it. And then they'd be spending the rest of the month like replacing gear teeth. I just have to ask this question because as you're talking, what's going through my head is all of those YouTube videos, who's all seeing a Boston Dynamics. Right. And all of their dogs and their bipedal and all that, you know, and I'm just, as you're talking, I'm thinking like, yeah, no exactly, we was talking about because we've all seen these videos. When you watch those videos, are you like, yeah, I figured that out a long time ago. Like what's, what is your emotional reaction to watching? Well, Boston Dynamics is, I mean, the stuff in the last two years is way ahead of where I got to. Sure. It's pretty impressive. Do you ever feel like, yeah, I figured that out before anyone else was really doing that? Or like, were they simultaneous, Lee figuring out, I think they were simultaneously working on something pretty similar. And I think there's a time like when I was ahead and then they were ahead and I was ahead and they were ahead. But were you guys keeping tabs on each other? Like did you knew what they were doing? They knew what you were okay. Okay. So it was just a standard sort of mildly competitive situation. Yeah. Well, I mean, at the time they weren't commercial. They were just DARPA funded and, you know, we're publishing and I was releasing videos on YouTube. And, you know, so it was all sort of, I don't know, not super open. Like neither of us open source in our software, but we were happy to demonstrate what we have. Well, then my other question for you would be, did you have a commercial use case in mind while you were doing all this tinkering or you're like, hey, I'm just going to figure it out, because it's such an interesting thing for me to do. And I'll figure out the business case later. I didn't have, no, I didn't have a very clear idea of what it would be good for. I mean, I thought, you know, someday, there are going to be robots and at least some of them ought to be able to walk because, you know, they need to go outside and so on. But no, there was, there wasn't really a commercial commercial case for it. And at some point I realized like, I need more money. I got to find someone to sponsor this. And so I started talking to, to, well, started trying to see who would buy it. So, I mean, the military wants walking robots. I kind of didn't want to work with the military combination of like vague pacifism and allergy to bureaucracy. For the military, the use case for walking robots is, you know, if you're operating in a theater where you can't tell who the, who the soldiers are versus civilians, the only way to know for sure is to send a soldier in and see who shoots at them. And so if you can do that with a walking robot that looks enough like a soldier that people will shoot at it, then you can tell who the bad guys are. Well, when I met you Trevor, and by the way, you play a very important role in my life, because when I came to that party where I met Paul for the first time, you were the person that I talked to the longest at the party. And you had just been featured for your robot for Dexter Pratt fall, I think, on like CNN. You were interviewed on CNN and had a little clip. And at the party, you would come in from California just to go to the party. And you're like, you're whipped out your computer. And you're like, let me show you the robot. And I remember it was so fascinating. It was carrying a cup of coffee and not spilling it. Do you remember this Trevor? Okay, that, yeah. I remember that, that era. So what point was that in the in Dexter's? Well, that was quite a bit later. That was, um, okay. Yeah. I mean, so I eventually got it kind of walking, not, not real well, like not well enough that you'd be comfortable having it, you know, walking on a sidewalk in a crowd of people or something. But, you know, it was balancing and walking and could deal with, you know, somewhat non-flat floors and obstacles and so on. And you could shove it a little bit and it would recover. And then I started working on the top half of the robot on the arms. And I, you know, I thought I was really foresighted by building the top half with the same bolt pattern. So eventually, I'd just be able to bolt it onto the legs. But to test out the top half of the robot, like, you know, the legs were still kind of unreliable and, you know, it was sort of like a bit of a miracle every time it worked properly. And so I built a wheeled base for the, for the, you know, to mount the arms on. And so like that worked really well. So we had a machine that, you know, it was still tele-operated. There was an operator, you know, wearing a special suit and a special glove that would, you know, follow his movements and make the robot hand do the same thing. The robot hand was Monty, right? Yeah, that's right. There is the dexter, the walking robot and there was Monty, the upper torso on wheels robot. Which, Carolyn, I don't think you were rounded dinners at this time, but actually, Dexter and Monty played like a very important part of Y-combinator. And we'll get to that in a second about how we started Y-combinator. But the robots, when we used to have these dinners, when no one had heard of Y-combinator, and we'd like convince someone important to come to a dinner, we'd be like, there's nothing to see, right? It's a bunch of startups. We'd say, come back and see the robots. And it would take up like an hour. It was so fancy because they were really cool. It distinguished us from the typical VC firm. Yep. They're kind of terrifying. Yeah, it was very, it was, they were a little scary, unique experience. They were scary. They were so big and metal and they could definitely turn. Yeah, very, terminator vibes for sure. Yeah, we put a hole in one of the walls with runaway robot. Oh, that office. Well, I want to ask you about this. So you're working on the robots. And do you remember when Paul in 2005, he must have emailed you to say, we want to start something up where we invest, you know, in startups. Do you remember the email? He was kind of trying to convince you to do Y-C with him because he really wanted to work with you and Robert again. Yeah, he phoned me, which was unusual. Usually you just send a one sentence email. He didn't even remember talking to him like, yeah, I remember talking to him out in the parking lot for some reason. And you know, he pitched me this idea that we should be the kind of investors that we wished we'd had. You know, like when we started VIO app, we didn't really know anything about business and certainly didn't know anything about how to get investors. And we talked to some sort of official investors and they wanted things like monthly cash flow projections and we had no idea like how to even do that. And you know, we eventually found some investors like Paul's art teachers' husband. I got to do a land investor. Yeah, Julian. It was your first investor at VIO app. It was great. He was helpful. He incorporated us and he was a lawyer so he knew how to deal with things. And then we had some other investors who are kind of a nuisance. But mostly like we just didn't, it was such an opaque business. We thought, let's just make, you know, I mean, this is Paul's idea. Let's make easy investment. That's the standard way of getting a company going. So when he called you, what did he say? You won't have to do much except what? Yeah, what? He knew you were working on any bots. Well, at the time it was, it was, you know, let's just try this for a summer. We'll put out a call for startups on his website. We'll get a bunch of applications. We'll read them. So I think you wanted me to read the applications and see if their technology was going to work. The applications were PDF documents. Do you remember? Yeah, well, I turned them into PDFs. I think the first version was, we gave them a template and they would email it in to us. special mailbox and then I wrote some software to reformat it. So we weren't just looking at a wall of text and at least had the questions bolded and so on. Carol, and that's the kind of thing that Trevor used to do. We'd be like, this doesn't seem to be working efficiently. And Trevor would be like, okay, let me just whip something up. And so yeah, that's just makes sense that you wrote something that turned the individual emails into PDFs. Okay. Yeah. So and we, you know, we printed them all out. And as we, like, we had 300 or more. And I remember sitting there with these huge stacks of them where we, you know, have, we were sort of treating them like, like assignments or something. And we, you know, write our numerical scores on them. So Paul said, we're going to do this project over the summer, which we called the summer founders program. And you believed in the idea and you thought, okay, this will be a fun thing to do and get to hang out with Paul and Robert. And of course I was involved too. But do you remember like coming for the interviews and stuff in Cambridge? Yeah. That was so exciting. You know, like there were all these people that had this, had these good ideas of things to start. And so, you know, like, I get the, yeah. So I remember coming out to Cambridge and we'd, we'd interview, we'd interview at the time we did like half hour interviews, right? Oh, yeah, even longer maybe. We might have had 45 minute slots or something. I think there were 45 minute slots, but how would like, we never would have known that before we did it. These are the kind of things you only learn by doing it. So like we, we learn, like we eventually learn the painful ones. We rarely changed our opinion after the first five minutes. And so that's just keeping the one where you knew like this is a now, the next like 40 minutes was really painful. Yeah. We were recently interviewing the Stripe brothers, John and Patrick. And they basically said, how weird is it that in your very first batch, you got, you know, the founders of Twitch, the found ultimately Twitch, the founders of Reddit, these are Sam Altman. Like, was that shocking to you that like the people we got were so good in that first batch? Well, I guess that's what we were hoping for as we'd get good founders. I didn't have a preconceived idea really for how, how many good ones we should expect. I think a lot of investors think you have to get, you know, nearly 100% and everyone that goes wrong is like, you got to do a post mortem and understand how you screwed up. Yeah. And, you know, we knew from the beginning that it wasn't going to be like that, that, that, you know, maybe only half of them would, would work out. Oh, I think we had lower expectations than that Trevor. I think we were psyched just to get like some decent programmers and see what happens. We weren't expecting half of them to succeed. Yeah. Did you come out that first summer? Did you come out and spend some time at all in a tent, any of the dinners and like meet some of the first founders? Yeah, I think I went to more than half the dinners that first time it was a bit of a strain traveling back and forth, but it was worth it. Like the dinner is where the most energizing thing in my life at the time. You know, because it was really interesting guys working on cool stuff. And, you know, I liked Reddit the product. I liked many of the, many of the products. I liked Kiko, the, the calendar that Emmett Sheer and Justin started that eventually turned into Twitch. A lot. I know. A lot of past later. For a series of pivots. Yes. So do you remember then we felt like this is pretty interesting this summer founders program? Do you remember one Paul and I came out to California and said, well, Trevor, we don't want anyone else to be the Y Combinator of Silicon Valley. So we're going to do a three month program in Silicon Valley. How about we borrow some of your office? Yeah. How hard could it be to turn this, you know, low ceiling, droid about lab into a startup dinner theater? You're very accommodating. You're, I mean, you did have a big warehouse kind of thing. You had a lot of space, right? Yeah. I had a lot of space. It had been this warehouse for like religious book shipping company. And, you know, so there's this big warehouse. But then, you know, the office part of it was pretty drab, you know, with like the very acoustic ceilings and stuff. Green carpet. Green. That still lives in one room. I mean, that's still it. That was all over the place here. And the whole place was the green carpet and the low huntsy acoustic ceilings and the fluorescent lights. Yeah. So you said, okay. And we said, we just need a little space. And then, yeah. Well, then Kate got to work on designing it because she'd designed a lot of the Cambridge office. So Kate Quartot, who was an architect by trade, oversaw the illegal renovation of part of Trevor's office space. As far as I know, it was completely legal. At some point, the city disagreed and shut it down to the city, find out. Yeah. Well, this, I think you were still in Cambridge or something, but the city, like found out that we were doing this, you know, pretty major renovation on the inside. As I understand it, they usually get tips from the people taking away the dumpsters. Like they see all this construction debris and they don't see a construction permit. My God. Okay, keep going. So they shut us down and put the like this big red stop work order on the front door. And so we couldn't get, you know, we couldn't get any of the sort of official contractors in, like, like a union, the electrician or something would come. Kate got some of her friends. Less, less formal friends contractors to keep working. But we couldn't get the electrical work done. So I actually wired up a lot of the lights in order to like for that first dinner, you know, we had like people coming, you know, on the first Tuesday night. And, you know, like they shut shut down the work just like a week or so before. And we have a light. So I was like up on one of those skyjack things wiring up the lights. I'm in denial that this trauma had happened. I'm now remembering this. So Kate had gotten pretty far though Trevor, right? And like we'd ripped out the hung ceilings, ripped out a lot of the carpet. But like a few weeks before the city got wind busted the construction. And then that's when we started locking the door. And like some people still came in and painted you did the electrical work. Is that what happened? Yeah. Yeah. And that's why we were so nervous because we'd already been busted, right? Yeah. Right. And you know, we were probably pushing the limits on our fire occupancy permit to for some of those things. I want to unhear a lot of those. Just kidding. Maybe like change something. And so the paint was drying on the walls. Yeah, we had to tell people not to like mean on the walls that first night. Yeah, because the paint was still drying. Oh my goodness. And that's when you know, part of the YC dinner experience was come to the back of the room and see these robots of Trevor's. And it was really part of the whole charm of early YC. Yeah. Wow. Do you remember any other crazy stories from the first batch in Mountain View? Yeah. What are your favorite anecdotes from those early days? You know, I remember the Twitch founders, you know, like that started off as something called Justin TV, which was basically Justin Khan had a camera strapped to his head. And a giant backpack full of computer stuff. Because at the time, you know, that you couldn't just stream video over the internet from a phone or anything. You had to have this like complicated system. So they had this backpack with like four different cell phone antennas in it. And that that Kyle voted built and then Kyle went on to found crews and it's a self driving car company. But so I remember them like bringing this, they had this, this backpack full of stuff. And something went wrong with it. And so they were like using my lab bench and soldering iron and so on to try to get it running again. And then this was in the very early days of streaming. When people thought that calling the police on a streamer would be extremely funny. And well, the first time they did it, they called, they just ordered a bunch of pizzas for Justin. I remember. Wasn't it during a YC dinner? It was during a dinner. So, so, yeah. So Justin was there at the dinner streaming somehow the dinner and they thought, "How? Order these pizzas and we'll get to see them." You know, see him be all confused by the pizza delivery. But actually like everything was hungry and the pizza was going to be so pretty. Someone else bought you all pizzas. No downside was very annoyed because the pizza delivery guys showed up like during the dinner. You know, we have it. We had dinner already and we had a speaker coming and this pizza delivery guys like, "Okay, I need payment." Of course, the pranksters didn't know I thought the pranksters had like 30 pizzas and expected me to pay, which I did because I was like, oh, it's Justin TV again. There was always stuff like that happening. There were definitely like some crazy stories I feel like that happened within those walls. The toilets always broke down. We only had two toilets for everyone. And that was fine when we only had 10 startups, but yeah, because we do. Right, right. This thing was basically like an office for 20 people plus a warehouse. So it was at its limits when we were having like 100 founders. And then when we did the demo days, we'd have like 300 investors show up. And I'll try to park their BMWs on this little side street. It's called a sec. And there'd be one bathroom and we eventually got portapoddies in the parking lot for them. Is that true? Which we're not really bad at class. Well, just people like know where to wash your hands. What about the secret third bathroom in the very back? Is that the sneaky bathroom? Is that just for any bots people at the time? Like you didn't let founders back there. Yeah. Right. It's also trying to run a robot company in that space. Right. I was very protective of the sneaky bathroom. But by the way, I say it was the sneaky bathroom. At first, we finally redid it, of course, illegally. But it was, it was really gross. It was like, yeah, it was like the warehouse bathroom. Yeah, it was the warehouse bathroom was really dirty with like old pants and stuff. It's where we go, you know, pour out paints or something. Yeah. Like paint cans and stuff. And I remember, ashton Kutcher, who was an excellent investor, by the way. And he was coming to our demo days very early on. He was married to Demi Moore at the time. And he brought her to one of the demo days. And I remember meeting her, she, the line at demo day of the two bathrooms in front, which were the, quote, nice bathrooms, was out the door. And I remember saying, well, I do have a special bathroom in the back. I was like, but it's really gross. And God bless her. She went to this nasty bathroom in the back with garbage and stuff. And I was so embarrassed. I'm like Demi Moore just used the sneaky bathroom. And I'm so nervous. It's funny. Oh, there's motivation to clean it up a little. Yeah. But there were definitely some funny stories in that place. So Paul told the story about you and note cards. And I feel like we have to get your side of that story. So do you want to talk about the note we have? I don't, I don't think it's nearly as eccentric as Paul portrays it. At the time I had, I had this idea. And it actually came from, from John McFee, the author. He has a book about his process. Maybe it was just a New Yorker article, How He Rights Books. And so he said he'd go do all the research and put it on note cards and then spread all these note cards out on the floor, like this whole floor of his office until he felt like he had everything in the order that he wanted to write about it in. And then he just go through it and write it. So I thought, I should do this for my thesis. I'll just make notes on note cards of like every idea. I'll take my class notes on them even. And at some point, like I'll just find these two cards from disparate classes and there will be some great idea that comes from putting them next to each other. Did that ever happen? A great idea. No. No, it was actually a terrible system. It was bad for, it was bad for ticking notes because actually a note card, like a five by seven note cards, a little too small for a whole idea of information, especially for a math class or something with a lot of formulas, derivations. As you've described it, I agree with you, not eccentric, like unusual, but not eccentric. But I thought the Paul was making it sound like these were no cards you would bring two parties and you would use them as discussion topics when you were just meeting people and chatting with them like it was like they were like, you know, so did you ever do that? Let me pull out my family. Yeah. Let me pull out my card on this side. Like, did you ever do that? We need to know. No, I don't think I ever did that. I wonder what Paul can do. Okay. What about not an party, but in the office, wherever, like hold on while I pull out this note card. Yeah, that I probably never consult the note cards, basically. Yeah. Yeah. Okay. Well, if all they had on them were interesting academic ideas versus like cocktail conversation ideas and I agree it's less eccentric than I was led to believe in the Paul. Graham story or what about administrative administrative day? Like, didn't you have a notebook on your cards on them? Oh, that's just the checklist manifesto right there. Yeah. No, checklist is good. Checklists are good. That didn't last long the whole card thing. No, no. Trevor, speaking of being eccentric, do you remember when you were on the Segway, um, Polo team and was was on your team can tell us a little bit. Yeah. Yeah. When the Segway came out, I thought, I thought it was an interesting idea. It kind of looked really dorky, but it's, it's, like, it's a new kind of transportation. Like no one had done that design before. It's different from a bicycle or something. And I had this fascination with balance from walking robots. I thought, I know how to make one of those. But I made one. And you know, like in a couple of weeks, I had something that worked. And it was, it was pretty clunky. Like it had a little potentiometer for steering. So like you had to, you know, it had these big handles for driving, but, you know, this little knob for steering. I wrote about that on the internet and that got a bunch of hits on my website. And I got invited to this Segway Polo group. There's this group of Silicon Valley nerds, including Steve Wozniak, who would play polo on Segways. So they had like polo mallets and a ball and would drive around on the on the Segways. And like it sounds, it's about as nerdy as it sounds, I guess, but it was surprisingly athletic, actually. Like it was, it actually believed that. Yeah, like like polo on horseback is athletic too. Yeah, for sure. You got to be pretty strong. Well, that I believe. Yeah, it's, but on the Segway. Yeah, because you're, you know, leaning and crashing. Were you playing on your Seg well, which you reverse engine? Yeah. Did you have your right? Right. So I went to one of these, I went to one of these things and brought my home built one, which had some advantages. Like it was, it could, it could spin around really fast. The Segway, like you turn the thing and it goes, it turns around like six, like five seconds to do a rotation. And we just go, and fling you off in the process. Right. Well, I didn't, I didn't put any lemon on it in the software. Like I didn't know what the limit should be. So I just didn't put one. So it would go as fast as the motors could rotate. And so yeah, I took it to a couple of these games and eventually they said, like, like, first of all, it's like, you know, lots of exposed metal. Whereas a proper Segway is all sort of pad is like the, you know, soft plastic around it. And like it's way faster. So wait, they kicked you off the team because you had a technological advantage. So instead of just like wanting you on their team, they were like, yeah, this guy can't play it all. Like forget it, drop him. That is not cool. Yeah, they should have respected the fact that you built it. Why didn't they put in orders for you to build them their machines? Well, actually, we should ask Trevor what he's doing now. I mean, like you, so you, so you're not doing, so any bots eventually, what happened any bots actually? You know, I just finished packing it up and put it in a shipping container and shipped it to England. And so the robots are coming. They're coming. They're going to live in your barn. Oh, I can't wait to see them. They're going to come along last friends. Yeah. That's actually said. But you're still working on some robots. So just software for robots. Yeah. Okay. Trying to write software that will sort of learn to control the body better. Like there have been all these advances in machine learning in the last two years. And some of them have been applied to robots, but not very many. So that's kind of what I'm working on is trying to figure out how to apply this machine learning stuff, which now works to controlling robots. Well, if there's anyone who can do that, I bet it's you. I'm glad you're still working on it. I didn't know that that was still happening. So I think that's awesome. Yeah. Well, maybe I'll figure it out. Awesome, Trevor. I had so much fun catching up with you and getting your perspective. Because often, you know, I'll tell the story of YC and I kind of know what Paul thinks about it too, but I have never really asked you what you thought, you know, about some of the stuff that happened in the early days. And of course, hearing about Monty and Dexter, the robots. And I'm glad you're still working on robot related stuff because it's just amazing. And I think this is going to be a really fun episode when it launches. It's like a full circle. Yeah. Yeah. So thank you for coming on the show. And it's great catching up. with you. Yeah, you two. Thanks for having me. Thanks Trevor. See you. All right. Bye. Bye. Oh, Carolyn, that was so much fun talking to Trevor. Yeah, it's great to talk to him. And I think he cleared up the note card story, although Paul's version is funny. I know. But I actually thought that Segway Polo is the price is pretty funny. It's pretty funny story. Segway Polo. I'm so glad he played on that nerdy Segway Polo team because the whole reason I was able to get Steve Wozniak to be interviewed for founders at work was through Trevor's connection. Oh, no kidding. That was my connection. I didn't know Steve Wozniak. Yeah. Yeah. So that was very critical. Yeah. And also, I mean, not to make a joke out of it. But when I met Paul at his party, like I said, Trevor was the person I talked to the longest and all about his robot. That's their endipity. Like what if he hadn't shut up at that party? I mean, I know I might have like left out the back door thinking nobody here. I want to talk to right, right, right, right. That was very lucky. One thing that didn't come up was just how much fun he is to work with. And like in the interviews, I'll never forget. I mean, we have some long days with the interviews as you know, and you get kind of punch drunk. And Trevor had this knack of every time, like saying something that would make me go into like fits of giggles and just completely lose it makes me laugh so hard. Yeah. And he's also like saved my bacon several times because he's he's the person you call if your computer crashes and you can't even log on or you can't get on. He's like resurrected my your laptop before. Oh, that's so nice. Yeah. He just can get stuff done like that. That's that's awesome. And I'm glad to learn more a little bit about his robots because they were really important at the time. And he never got any publicity really for him. And I actually didn't try, you know, I was just thinking I'm happy that he gets his, you know, robot family back to him in England. But I'm sad they're not going to be in our space anymore because they're definitely a part of the YC folklore from my perspective since I saw them also the first time I came to see you guys. I got a robot tour also. So we've all been through it. And the fact that they're not going to be there anymore is a little bit sad. But that's okay. Well, sorry, well, you'll just have to come visit us in England. No problem. I will do that. All right. This was a great episode and I can't wait for it to come out. That sounds great. See you soon. See you soon. Bye. Bye.