Every Driving School Said No

Startups get the party, startups get the party started. Startups get the party, startups get the party started. You want tech and tacos, tacos and tech? You know we got tacos and tech is on smash. Fold a green guacamole, keep accountants counting beans. Contador cuentan frijoles, founders trying to raise some queso. La vida San Diego, school of rapping, rapping fire. Tortillas around jalapeños in these modern times. A craft beer and coffee, my coding yoga posse. Fold a Body origami. We be surfing Coronado, eating tacos El Gordo. The tacos standin' puesto, lucha libre and sombrero. Gotta shout out Tech Coast Angels by the ocean, watch the sunset. La Jolla shores, entrepreneurs tryna get funded when San Diego Startup Week is poppin' in the summer. Oceanside to Tijuana, SD be full of hustlers. From couch surfing to penthouses right on the beach. Tech leaders who believe stars are within reach. We interviewin' founders who be shootin' for the moon. Tech, SD Tacos with your host, Neal Bloom. Come on! Startups get the party, startups get the party started. Startups get the party, startups get the party started. Startups get the party, startups get the party started. You want tech and tacos, tacos and tech, you know we got it, hey! Startups get the party, startups get the party started. Hey! Startups get the party, startups get the party started. Hey! Startups get the party, startups get the party started. You want tech and tacos, tacos and tech, You know we got it. Hey.

And welcome back to the Tacos and Tech podcast. I'm your host, Neal Bloom, and I've got on some of my favorite and oldest friends in the, the tech business itself, Matt and Nigel, co-founders and CTO and CEO respective of Coastline. Hey guys, thanks for being on the pod.

Great to see you. Thanks for having us.

Thanks, Neal. Good to be here.

Man, I cannot wait to tell this story, but also what a long time coming to have you guys on the pod. So thanks for being patient for us to tell it at the right time and the right place. Uh, this is, you'll hear everyone listening, quite a moment, uh, to talk about driving and the state of cars and the state of vehicles. And, uh, And also about the state of timing as well too, and just where companies are in their journey. So why don't you guys, uh, Matt, why don't you kick us off and tell us high level, what is Coastline?

So Coastline is on a mission to eradicate car crashes, and the first thing that we are doing towards that goal is we created a driving school. Uh, so Coastline Academy is a real live driving school. We teach Hundreds of in-car lessons every day. We are across 80 states, we have 80 locations. And when I tell people that we are a tech company that has a driving school, they often say, oh, so it's online? And yes, it is online. But we are actually physically in the car with new drivers teaching them how to drive. So Coastline is a driving school, and we are founded by 2 tech people. So excited to talk more about how we use technology to innovate in this driving school space.

Awesome.

Uh, well, we've got a lot to talk about. We'll, we'll hit as much as we can, uh, along the journey, because it's been a journey for sure that I'm fortunate enough to, I think, been there since the beginning, I think since the early days, um, and, uh, of Coast.ai. The original .ai domain name of close to 10 years. But Matt, you and I actually go even further back than that from the running days, the trail running days that I think your family's getting, having a second wind in, which is pretty awesome to see, right?

My wife is, yes. She has gotten the trail running bug back, but I, I am unfortunately missing this belt. But yeah, we, I mean, Neal, you and I first met probably covered in dirt at some trailhead in the valley of Los Angeles, about to go on some ridiculous length run, which hopefully we did without too many injuries.

We're, you know, we're still here. We're not, neither of us are standing, so we can't tell, but we're definitely sitting and not in too much pain today. But yeah, I think almost 15 years.

ago.

Pretty awesome. Yeah. To have met you in that regard. And you and Nigel go back way further than that, I believe, right? Nigel, when did you guys first meet?

I, it was probably right around the same time. So I was working at an ad tech company and Matt was my biggest client, client in the sense of partnering and working together on, on some pretty cool like product stuff. And I just immediately knew like, this is a guy I wanna work with. And I guess just one thing led to another over several years and then we, we became founders together. Amazing.

You probably started collaborating by night on one of Matt's many projects, whether it was fantasy football, AI predictor or something else that Matt has been tinkering with for so long and, and probably deserves a whole other podcast of what is, what's in Matt's, you know, nighttime tinkering bin for sure. But what was it? What was the, do you always remember the first project you guys started actually working on together and where, when that actually kicked off?

Yeah, so we, we built this thing called Traffic Tractor, which was We both had super long commutes. And so we were trying to optimize basically not only our own travel, but like, where could you move to and optimize a super commute based on traffic prediction? And so, yeah, I mean, it was super cool. And, you know, we would be kind of tinkering on this thing at night, and it's like, It's just so much fun to work with someone who's super smart, super hardworking, and like, just like gets fired up about just building stuff.

I love that the first thing was, was car-related, traffic and commute-related. So already see a theme here for sure. Where you guys weren't in the same city yet then, right? You were remote from each other even from then, Matt?

Yeah, I was in Los Angeles. Yeah, working out here. And Nigel was up in the Bay Area. And so we had actually, because I was his client, we, I was actually up in the Bay Area for something. I don't even remember what conference or something. And I was looking for some space to work. And Nigel was like, oh, why don't you come, you know, by our office? And that's when the tinkering began and the tinkering has not stopped. But we did. Nigel actually convinced me and my family to move up to the Bay Area. That's how, that's how deep into the tinker we got.

Yeah, I remember that. I remember visiting you guys at Palo Alto, right?

Yeah.

Yeah. Yeah. Yeah. That's awesome. So where, okay. So did traffic, did that turn into the thing that you actually went on or you started a clean slate and said, let's actually start something from scratch?

Yeah, we, we decided we're gonna start something completely from scratch. But I feel like with that and just a couple of other things that we were like ideas around and just sort of things we were noodling on, I think there was just, there was something about cars and, and, and car travel that was like a real hook for both of us. Just, I think for, I don't think it was even explicit. It was just, there was something That, that, that was this sort of point of connection and of mutual interest that kind of had us both lean in on this whole sort of car accidents as a, as a, as a real problem and, and something maybe we could do something about.

And you both are coming from obviously different industries and different technical abilities backgrounds-wise. Matt, like obviously deep technical, both product and engineering. Take, take me through a few of the things you'd worked on in the past because I know there's some sports related, take us up to speed on kind of the things that you had touched on and were utilizing to think about, you know, Coach that you eventually started.

Yeah, I mean, it's been a long journey, right? And I think that any, anybody who's been in the business of creating with software has similar stories, right? I started in deep tech, actually in fiber optics, working at Nortel Networks, rest in peace, up in Canada. And that's where I got this like obsession with software. And, you know, when I, when I could visualize, because I was actually working on orchestration software and automation for testing of fiber optics. And I honestly was totally not qualified to be doing that. And somebody took a chance on me. They gave me a huge book. It was like an O'Reilly How to Write Perl book. I was still in high school and I just started flipping through and it was like, yeah, I think I could do this. And so I just started building things back then. And so I worked at Nortel for a while. They blew up in the dot-com bubble. I was always on the side working on, you know, my own projects. I was super big into skiing. So I actually created a social network for skiers. Just because selfishly I wanted to meet other people who were skiing and there was no Facebook yet. MySpace was, I don't even know if that was, I don't think there was a MySpace yet. So I really missed the mark on that because I made the social network only for skiers. And I wish I had thought a little bigger at the time. But anyhow, even then—

Not even all skiers.

Yeah, at least all skiers, right? Because this is only for free skiers. But that really did take off. And, you know, that kind of launched my career as, you know, more than just a person with a book on how to write Perl and started thinking about product and marketing and connecting people and that sort of thing. So, you know, one thing led to another from there. I ended up working in sports. So I worked at USA Today Sports. I oversaw all of our product for, you know, our sports app and all of our fantasy stuff, which you mentioned earlier. And around that time is when deep learning was really starting to become a thing, right? AlexNet had come out and that just like kicked off a whole new obsession. And not long after that, Nigel and I met and we were working on big data stuff together with sports. We did some fantasy stuff. And then you tie in the AI part, which we weren't even calling it AI back then. But it was just like this magical assortment of technologies that all came together at the same time at the right time.

Yeah.

Amazing. I remember being in a Pick'Ems league with you where you were, you know, definitely testing some of your coding abilities versus our terrible picking abilities. Did you ever—

I think I got 3rd that year. So not— apparently my AI was not, not good enough.

Is it still in work? Is there, is there still, is the model still learning?

You know, it's been retired because now you can pretty much just tell ChatGPT, hey, go, you know, yeah, go find the answer for me.

Okay. And Nigel, how about you on your backstory a little bit? Where are some of the industries that you'd touched and kind of utilized leaning forward?

Yeah, so I mean, I've always been sort of a founder. And my first startup out of business school, I was one of 3 founders and I was the technical founder of that group because we're all sort of business school classmates. So basically just enough kind of technical understanding and capability to like barely get by. So when Matt and I kind of met, it was like, okay, so like we can have a conversation about sort of the technologies that we have to make choices about, or the sort of the technical choices that we need to make. But I'm wholly like unqualified to actually do any of that work. But I think that's like, that's been really helpful. And in terms of industries, That startup was a fintech startup. I was in a health tech biotech startup right after that. And then I was in ad tech. So I sort of moved around industry-wise, but I'd say my, like, as a generalist, my sort of area of expertise, if you want to call it that, is in sort of operations and business development. And so I've sort of picked up those roles in a startup environment until we can hire somebody to do a better job than I can.

Wow, spanning many industries, biotech, fintech. I love that. I don't think I knew that part about you. So we'll have to talk further about those pieces on what to pull the thread on a little bit of that. But Take us through a little bit of the founding story, like who approached who and who laid the first line? Was it the first line of code or first camera acquired? Give us a little foreshadowing there. How did it begin?

Yeah, thanks for that visual memory. We actually just saw some photos from those first days pop up on my phone the other day and I was like, wow, that was so cool. So, I mean, Nigel and I, you know, he convinced me to move my family up there. So we knew that we wanted to work together, right? And we both had this passion for cars and the computer vision thing, like I was mentioning, that was a bit more foreshadowing, was so wildly fascinating at that time, right? People were really just figuring out how deep learning could be applied to visual problems. And, you know, Nigel and I would have these long conversations about how, how these image networks worked and, you know, how they could be applied to our, our real job and then how we could apply them to things on the side. And then if you remember, there was like hot dog, not hot dog in that Silicon Valley episode. We were actually building Exactly stuff like that just because it was cool, right? It was like, you know, you got a little old Android phone and it could identify hot dog, not hot dog. Like, how cool is that? It's amazingly quaint that that was not that long ago, right? And so we just kind of iterated from there. And Nigel had gotten a Tesla, I think, or maybe you were thinking about getting one. And we were so enamored with this idea that you could use computer vision to control hardware. It wasn't just software. And, you know, Nigel is selling himself a little bit short. He has a lot of hardware chops. His ability to hack on physical, you know, robotics is actually quite impressive. And so with me going into this, you know, computer vision, deep learning stuff, Nigel with his you know, physical chops and then our passion for cars, it kind of just all came together in our first great idea that we started with Coast.ai, which I'll throw over to Nigel to describe what the product was. Yeah.

So the, the idea was that, you know, you can spend, at that time it was like, you know, $90 grand for a Tesla that has sort of the very earliest version of you know, adaptive cruise control and, you know, lane centering and that kind of thing. And so how do we very cheaply bring that technology to older cars? And so our, the idea was using cheap hardware, which the first iteration was like a Raspberry Pi with a camera connected to a laptop, but then quickly became like a modified handset, like a sort of phone in a box, basically, and then use that for the computer vision and deep learning to then control a vehicle. And so we kind of essentially like hacked the cruise control to let it be controlled by this sort of hardware device. And it was like, So like such an amazing feeling. I don't know, Matt, if you remember that first time driving down Carmel Valley Road where we like turned it on and like it worked, you know, we're in, we're in this like, we're in my sort of turbocharged Miata and we're like, this is the one that we're gonna do as our like our test vehicle. And all of a sudden it's like it sees the car ahead pulling ahead and boom, it starts accelerating. We're like, wow, that is so cool.

That's awesome.

I remember— when was it? Okay, so I remember coming through one of your Palo Alto houses, seeing— it must have been obviously Matt's, uh, seeing the rig watching TV for some training model. How long did it take you from that to go from a handset to just, we need image classification times a billion?

Yeah, that— I can't remember. It was like Newt the Mute.

Yeah.

So, being compulsive, obsessive hackers, we thought it would be cool. So, I have this aversion to commercials to the point where if I'm watching sports and the commercial comes on, I've always got to mute it. It just like gives me this weird anxiety to hear the sound of commercials. And conversely, I love the sound of sports. So I would often have sports on, even if I wasn't watching it, but then the commercials would grate at me. And so we figured, hey, if we can do self-driving and we can do hot dog, not hot dog, we can surely do commercial, not commercial. And so we built a commercial auto-muting device that Nigel made that, like, he basically made a remote in the box with a Raspberry Pi.

Wow.

And I trained the model, and as soon as the commercial would come on, it would mute the TV. And then when the hockey or the football game came back on, it would pump the volume. So that's probably what you were remembering.

Yeah, and it, it worked a good 90% of the time, you know, as soon as it would see the green of the field, that was, that was like instantly it would work. But some of the— there were some edge cases that threw it off, and it was just like a like a fun project to try to like, how can we stretch some of the things that we're doing right now? And just to try to figure out, okay, where are the limits? Like, what, what's, um, how can this be applied? It was very much like a, I don't know, we were in like the research phase basically.

And at that time it was, I mean, even, uh, you know, devices, chips and devices were so low powered relative to the huge models that were coming out. So at the time, I think it was like InceptionNet or Inception V3 that was the state of the art. And one of the challenges we had was how do you run this on a Raspberry Pi? And like, you know, you couldn't, but what were the limits, right? And that had a direct application to finding the limits of what we can put in a vehicle.

Well, so take us through some of that. I mean, you're in the car, it's accelerating, kind of, you're like, we have adaptive cruise control, right? It's kind of what it feels like. You built it yourself. So you feel like we're already at an OEM level, uh, like Ford, what's up? Uh, where do you start pushing the bounds and start feeling the limits? Uh, as you, was it a technical thing or were there other pieces of it? I'll kick it to you, Matt, first.

Yeah. So the technical journey was really 2 parts. The first part was pure research, no limits. What can we actually do? Like, can we even tell the car to accelerate and decelerate? And this is an era also of vehicles where they didn't have the same CAN bus as they do now. There weren't video cameras in all cars, right? We're talking a Prius from 2007. You know what? It's a bare-bones mechanical system at that point. So it was very much like, you know, we would just rip up open the floorboard and find some wires. And Nigel would be like, this one, this one, what does this do? Oh, accelerator, found it. So it was a lot of that. And, you know, some of those early photos I was looking at, there's wires everywhere, right? Because we would have multiple cameras running. Because part of it also was, what data do we train on? Right? Of course, we've seen the magic that Tesla has been able to do. Because they have this infinite supply of data. And for us, we were literally just taking road trips with— we built an app that was just a capture app that would just basically, I think it was 10 frames a second, it would record the car's acceleration, deceleration, had an accelerometer in it. And so we just had these huge databases, I mean, relatively huge at the time, of just still images. And all the telemetry that went along with it. So anyways, that was the first phase. And then we pretty quickly moved into phase 2, which was, hey, we think we can actually put all this stuff on a phone, right? And then it was like, what is the highest-powered phone we could buy from China that runs Android that we can shove all this on?

Yeah. And I think what was going on in parallel is, you know, Matt and I were both super aligned that we didn't want to build like a research company or like a forever venture capital backed company. Like we were super focused on like building things that people want and we wanted to make sure that we were solving technical problems. But then we also had this just big question of, well, what is Like, what is actually gonna fit into somebody's life? Like, what is someone gonna actually put in their car? Just because we can doesn't mean they will. And so as we were developing these sort of each iteration of this like product, and it was sort of coming to life, and we were sort of saying, okay, we can actually do this. We can make a, make a really useful product here. You know, it's not perfect yet, but you can just sort of You can see where it's going. And okay, we can do this. We, we started doing the customer development in parallel, and our, our assumption was the primary users of this device are going to be super commuters with old cars and new drivers with old cars. Super commuters, because you want the convenience of adaptive cruise control. new drivers? Because in addition to adaptive cruise control, it's braking warnings, lane departure warnings, you know, things that brand new Mercedes and Teslas at the time have. But, you know, the average American is not gifting their kid a brand new Tesla or Mercedes. And so how can we bring some of that technology to the new drivers? And so that was You know, we started having dozens and then hundreds of conversations with people we thought were going to be the users of this product. It was things like, you know, I'd sit in a cafe with a sign on my laptop, you know, $10, you know, gift card for this coffee shop if you come and like tell me about, you know, let me ask you questions or whatever. And that was the process that led us to the realization that no one was gonna buy this thing that we were making.

I mean, the fact that, you know, you really did the customer development way early in the process is huge. Take, take us, give us a few new nuggets that like, what did you expect to hear versus what did you hear?

Yeah. So As a, as a like car guy myself, I'm super used to modifying my car. You know, it's, it's totally normal, you know, whether it's, you know, my first Volkswagen as a 16-year-old putting in a, you know, an updated, upgraded stereo to like turbocharging my Miata or whatever. It's like, it's very intuitive to me. And so this was one of the lines of inquiry of like, Will someone buy something that has to be added to their vehicle? And it's not quite at the level of simplicity of like a radar detector that you just, or a dash cam that you just stick on the windshield. It has to integrate into the vehicle. So it needs an installer. It needs somebody who's willing to have something installed in their vehicle. And so without even specifically asking that question, just sort of doing this sort of exploratory customer development.

Mm-hmm.

This came up as a theme where people are just, they're just not going to do this. You know, maybe 1 in 20 people would be, oh yeah, sure, I'd take my car into a shop and have this installed, but the 19 in 20 would not.

Hmm.

And that was a disappointment. But, you know, so much better to get that disappointment way early than after we had invested in the development of like a fully baked product and were actually taking it to market.

That's interesting. I feel like that has— I'm not a car guy, but everyone around me has always been modding their cars, doing aftermarket stuff, not only just the stereo face, but actually putting in subwoofers, right? Or actually lift kits. Was that just more prevalent and it's just died off over time and generations, you think? Nigel?

I think it's still common. It's just niche. And for this to work as a, you know, as a rapidly scaling startup, it can't be niche because then you, if it's niche, you've got the intersection of people who are willing to modify their cars and people who are teaching their kids to drive or whatever. And then the, it's just a sliver in the middle. And, you know, we needed like a pretty big market. And I guess the third part of that Venn diagram is cars that we had developed an integration for. And we figured, okay, if we can do 20 models that will cover, you know, 50% of vehicles, but still that's 20 integrations to get to 50. You've got to do like 100 integrations to get to 75. And so You know, every time you sort of slice the, the, uh, the, the addressable market down by one of those factors, all of a sudden now we're selling to this like tiny market.

And Matt, did it feel in parallel? Was there a technical like hurdle also from image classification and the models and that part, or no, it really felt like you could get there technically and it was a more market pull, pull or push thing?

Yeah. You know, I think we didn't get the opportunity to fail on the technical side. Because we made the call early enough in the process that we were still riding high, right? It was like, okay, now we need to make this more efficient. We need to make a good user experience. We need to make it smaller. All of these optimize it. Like we had proven that we could do it. So then it was just optimization after that. But thankfully, we did all this market research, right? It's like, you know, I know we're about to talk about the second biggest pivot in tech history here. But it was, yeah, it was like, dang, we know we can do it. And yet the market's telling us not to. So, you know, let's keep marching forward towards something that is a valuable startup.

Okay. Interesting. Well, I remember definitely the change at the time, but I was not in the room. Like, take us through it. What was the biggest Let's call it, you know, second biggest, maybe first biggest hardware pivot of all time. That's for sure.

So I think the first thing was like Matt and I talking through at length, wow, are we really going to not do this thing that we quit our jobs to do? We've got, you know, I had basically a baby at home. We were like, you know, we had left a pretty comfortable career and okay, we've gotta do, you know, okay, we're not gonna do this. What else can we do? And that sort of kicked off a process that felt like years, but actually it was only weeks where we dug back into all that customer development, sort of reviewed all the notes, which at the time you actually had to do with your eyes. And, you know, continuing to do sort of at least 10 user interviews a week where we would kind of take a seed of an idea from our historical kind of interviews and try to apply it into those next user interviews to try to understand, okay, Focusing on new drivers, like, what is the thing? If you're not gonna buy this hardware product, what could it be? And we very rapidly cycled through a coaching app, a, you know, a sort of technology provided to the driver education industry. You know, some of these like little pieces where we're like, okay, where can we apply things that we're good at to help fulfill the mission that we really started the company with, of eradicating car crashes. And we were doing 2 types of customer development. One is talking to people and the other was testing online. So coaching app. Okay, we have this idea for coaching app. People say it's a great idea. Okay, let's throw up a website. Let's start running some Google ads. And see, will someone pay us $10 a month to mail them a phone that they, they install, you know, stick on their windshield and then provides, you know, automated coaching. Now, automated coaching, by the way, was like us downloading the video and watching it and paying a driving instructor to say, oh yeah, you should have, you know, came to a full stop here or whatever. But it's like a, hey, you know, we know that we could develop this into an automated thing, but we're not gonna invest all that engineering time. Just to find out that nobody is gonna buy it. And sure enough, you know, even people who paid, we shipped them the phone and only half of them even installed it. And so, you know, we kept going through these cycles of realizing this, you know, have an idea, try to very quickly test it and see if this is something people are actually gonna wanna buy.

Huh, interesting. I love all the testing and maybe you know, nerdingness that you guys do. That's, that's the technical background that you guys obviously bring to this from a variety of angles. Um, how did it, how did it become a driving school? Take us through it, Matt. Like, when did it officially, you know, go from coast.ai to the driving school that is Coast?

Yeah, so I think one of the themes that kept emerging, and it's why we did the coaching and why we did these other, other things, is that It was new drivers who we saw the biggest opportunity to have an influence on towards our mission rather than the super commuters and, you know, folks that are more established in their driving lives. And one of the nuggets that kept coming up that we didn't notice at the time, but looking back at all the interviews that we had done, all the feedback we had collected, We found these nuggets of people essentially describing to us how bad their learning to drive experience was. And it was bad in ways that felt so solvable to us. Like, you shouldn't have to bring cash to a driving lesson. That's ridiculous. We have Airbnb and we have Uber and we have Lyft and people are, you know, bringing cash to their lessons. You know, instructors shouldn't be doing errands on their driving lessons. You should be able to see a schedule on the internet so that you can, you know, book your lesson. And so we just— I think it was Nigel who said, hey, I think there's something here, to which I'm pretty sure I said, I don't think I want to start a driving school. Like, that is a little crazy. Um, but we did. And so we ran a smoke test, and this was the magic moment, is that we, we put up a website. To be clear, we did not pretend to be a driving school, right? We— it's a regulated industry, but we decided to be a reseller, a seller of driving lessons, right? In order to do this smoke test. And it was immediately successful. Immediately. So we put a chatbot, I think it was Intercom at the time, and people started messaging us, hey, I need a lesson tomorrow. Hey, how much are lessons? And we're like, people are trying to give us money right now. And it's been like an hour.

Yeah. And then, and we didn't, at that point even, we weren't thinking about actually becoming a driving school. We thought, okay, we can be a service provider to this industry. We can help with putting schedules online. We can help with, you know, some of the safety kind of features for families in the vehicles. And as these leads started coming in, we were like, okay, great, someone wants a lesson in Salinas, someone wants a lesson in Santa Rosa. So I'd pick up the phone and start calling all these driving schools saying, hey, I've got a customer for you. Do you want this customer? And like everyone said no, they were like, we're full, we're good. You know, we've got all of our stuff figured out. I don't need any help. I'm like, what? What's happening here?

That's amazing. And of course, naturally the next day you just bought 100 Priuses.

No, pretty close.

Go ahead, Matt.

Well, I think that's the point where Nigel was like, I think we should do this ourselves.

But did you even know what it took to— had you even researched what driving school world was, or what did you have to go and learn and do to become an actual driving school?

Well, Matt and I had to become driving instructors. Uh, that was part of the process. Um, and we had to get licensed as a school. Which, you know, we had to deal with the DMV and it took way longer than, you know, a startup length of time to, to go through that process. But yeah, Matt, what else?

Yeah, well, and I think like that shows how magic that moment was, how surprised we were on both sides of the market that for one, people were ready to give us money and they had— we weren't even a real— we weren't even a driving school, right? We weren't even pretending to be. And they were like, You know, how much is it? Let me give you my money. And then on the other side, the driving schools were like, we're all full. We don't want— we don't want customers.

Yeah.

And so it was just like, okay, like, we would be crazy not to do this, right? And then of course, we went and we looked at all their driving school websites. And that's when we discovered, you know, more about the lack of software in general in this industry, and the fragmentation of this industry. There were so many driving schools. Each one of them very small. And so we were motivated enough to spend more time than we ever imagined we would becoming a driving school. And, you know, it's one of those things where if we had known it was going to be that hard and take that long, we for sure would have tried something else. But we were convinced. And so, yeah, we, we did eventually get our driving school license. Nigel and I did become driving instructors ourselves. And we hung the sign on the door saying we're open. Of course, we didn't have a door. It was a website. But then I guess the real magic started.

And you decided to not name your company AAA Driving School, the other driving school, to be the first letter in the alphabet on some Like printed phone book, apparently. So.

That's funny. Some of the driving schools really did tell us because we were, you know, doing research at the beginning. We were looking, maybe we could buy a school instead. And there were several people whose marketing strategy was every year they buy an ad in the Yellow Pages.

That's the moat. Yep. Totally. Turns out the moat may be the DMV. I don't know. Or just many other pieces. So in the end, how— okay, so you spin up a driving school, the 2 of you are the first instructors, I'm guessing. How many lessons did you think you gave in the first year? Do you remember?

Um, I ended up giving just over 100 lessons total, and I think probably 3/4 of that was in the first year.

Whoa, amazing. And Matt too, I'm guessing, probably pretty soon.

Yeah. I think I quit driving instruction at number 100.

Wow.

Yeah, I mean, that's a lot of— that's— I mean, it's the customer development, obviously, that you guys put the time into. But then—

Yeah, incredible to be in the car with the customers and seeing it firsthand. And I mean, seeing the joy on people's faces when they learn a new skill and the transformation of the parents from being so nervous the first time to You know, running out to be like, oh, you know, my kid's been practicing and I can't believe she went on the freeway. And it was, it was a really like, I know I speak for both of us. We're so grateful that we were just even able to do it right, to be driving instructors for any amount of time.

As soon as you got behind the wheel, as soon as you created a whole new experience, was it apparent that there is scale built into this and that you guys can, you know, not just monetize and build revenue, but there is something bigger here. Or yeah, take us through that of once you, you know, because obviously you guys are 1+ 1= 3 mentality all the time. Like how quickly was it like we can actually do this and scale it?

Yeah, I mean, we, we, we wouldn't have done it if we weren't confident we could scale it. So sort of the memo that we wrote at the time was basically we're going to build the first national driving school. What would that take? You know, I think the timeline was a little more ambitious at the time than how it's turned out to be. But ultimately, you know, we designed for scale from the very beginning. And as we were going through the process of getting licensed and having our first lessons and such, we were also building a software that would power the industry, um, or power us in the industry. And then, you know, when we're out there on the road teaching a lesson, you know, we would take all that sort of feedback from the student and the parents and our own experience as instructors to build into the software in order to make this something that really could scale to, you know, hundreds and thousands of instructors.

Yeah, it really was like driving instructor by day and like business builders by night. Because it was every single day we'd get home from teaching lessons, jump in Slack. Hey, you know, this person had this feedback. I noticed this, this thing with the schedule is not working. And so we would write the code and update the software accordingly. And we were also answering phones, right? I mean, we were also the entire support team.

Customer service, 100%.

Customer service. And so same thing though. I mean, I, I can't imagine having done it any other way.

Wow.

Imagine what Granola AI running through all those meetings turned into Claude on auto loop, just building back in. But anyway, we digress. That's what you're building. Take us through some of the tech though that was built along the way, right? Like, Matt, like, obviously you were iterating on feedback loops, and what did that do in terms of what you built out that eventually kind of powered so many pieces of Coastline?

Yeah, so, you know, the driving school industry felt untouched by software at the time, and the irony of that to us is not just that software could be used to provide a better experience for customers, But building a driving school of any size requires seriously sophisticated logistics and operations. And, you know, when you look at a driving school from, from a distance, it's pretty hard to see that, right? Because I think most driving schools— well, I don't think, I know most driving schools are an owner with a couple instructors. The owner is answering the phone. They're probably doing their scheduling now, maybe on Google Docs or with some off-the-shelf software. But it's very easy to manage a team in a car when there's 3 people in 3 cars. Once it gets to any sort of scale, those operations become— it's an absolute complexity explosion. And so the software that we were writing The day one software was very much, how do we sell our value proposition to customers? How do we make sure people understand that we're serious about this? We're real because we were brand new. But then in parallel with that, it was, how do we make sure that instructors' schedules are getting filled in an efficient way? How do we make sure that instructors have the information that they need when they need it so that they don't have to be using pen and paper. We had instructors in multiple geographies. How do we make sure that people don't have to be driving, you know, past each other or meeting centrally? And so it was this dual purpose of how do we give a great experience and how do we make sure the operations from day one is all codified in a software platform?

Did you, Nigel, did you start to see that obviously kick in and then Kind of the scale was able to unlock, say, okay, now we can do this in other geos, regions. Like where, where did the, where did the expansion mentality go from there?

Yeah, so I think the, the, the first iteration was very much a mesh network and that's still a big part of, of how we operate today. But it was really like each node was a single instructor and with a single car. And so, you know, starting in the Bay Area, you know, we covered San Jose and Palo Alto with 2 nodes, and then, and Monterey with another node, and then Gilroy with another node. But then also you go to East Bay with another node, and then with each of those, okay, how do you cover new customers? But then how do you make sure the right instructor ends up giving the lesson to the right student. And so the whole sort of predictive booking flow where we would only surface sort of pre-optimized lesson times to somebody to book so that you don't have somebody driving from San Jose to the East Bay and then back to San Jose again, you know, getting that dialed in meant that when we wanted to expand from San Francisco to Oakland, We could have, you know, we could have the San Francisco instructor cover Oakland lessons just right up until the point where there was enough supply of instructor hours in Oakland to cover those hours, and then no more. And then, okay, well, now we can expand to Walnut Creek, and same thing. And so the expansion strategy initially was very much add a node. at the edge of the service area of the previous node. So there's overlap, but then also expansion to new territory. And that was basically our California expansion strategy. So bit by bit, sort of instructor by instructor, we expanded to cover 95% of zip codes and then doubled down on places where demand exceeds supply.

And so from, from these nodes became state coverage, right? When did— and then take me through, gimme a little more color on like how, you know, what that actually looked like in actuality versus when you started to realize we could expand state by state. 'Cause that's a whole different, uh, I would imagine kind of piece on that too. You wanna keep going, Nigel?

Yeah. So, um, The expansion within California was relatively seamless because of this sort of node-by-node mesh network expansion. And then as soon as you jump to another state, that whole thing is broken, right? Because licensing is at a state level. And so you can't just have somebody from like Barstow, California, drive over the border and teach in Las Vegas because you can't cross that line. So I think we approached that phase with sort of the founder naivete of, hey, we can just keep running the same playbook here and we'll just open up in another state, another state, another state. And what we didn't give enough credit to early that we give a lot of credit to now is it's not just a different regulatory agency and a different licensing process, but it's a different, you know, it's different package types, it's different reporting requirements. It's a, you know, different structure of learning in, in many cases. And so we shifted from sort of organic node-by-node expansion to expansion via acquisition, where If we can acquire a school that's already operating in the state, then we've got a head start. We get a shortcut to licensing. They've got a process already for how to serve students best in that state. And so after a couple of sort of really challenging starts in the state expansion, we moved to a sort of acquisition-only sort of geographic expansion strategy.

And I remember the Prius being like the, the company ambassador of choice. Is that still, is that still a thing? The Prius is, is loud and proud screaming Coastline.

It's Corollas now because we had the, the Prius C was our favorite car. And when they discontinued that, they basically, the Corolla Hybrid was the replacement. So that's our, that's our car.

Okay, amazing. That makes sense. I see them around. I still, I like, I love seeing the Coastline stickers out and about in Southern California. And now you're, you know, across the nation, right? From coast to coast. Ha!

Yep.

Which is pretty cool. What have you learned now from like, you know, national, national coverage, if you will?

Yeah, I mean, I'll let Matt speak to how the engineering process has evolved, but from an operations standpoint, You know, it's, it's far more about having locations now than it is about having individual instructors with individual vehicles, and then being able to support those instructors with layers of management and having, you know, instead of a sort of scrappy approach to how we serve customers, it's a more well-thought-out approach to how we support our team of instructors. And so, you know, the logistical complexity still continues to increase as we scale. But also we're sort of filling in the gaps with kind of operational structure That kind of makes Coastline more of a place to work that, you know, you would expect to show up at work instead of like showing up at a startup and, and having your, you know, founding team teaching, you know, lessons beside you and, and, and whatnot.

Matt, what do you think? What, what's changed technically now to help scale it across the country?

Yeah, well, I mean, I can't believe we've gotten this far without talking about AI of the modern variety. But yeah, I mean, I think what Nigel maybe glossed over a little bit is that the complexity of operating in multiple states with different models and different package types and different consumer tastes, it shows up everywhere.

Right.

And, you know, even across managers, across our code, right? We have a lot of code that is just there because in state X, they have to do things one way and then a completely different way in state Y, which also makes it so we can't buy a lot of off-the-shelf software. So we end up having to build a lot of it ourselves. But thankfully, with AI agents, you know, we can talk about it on the engineering side, but even on the operations side, Being able to use AI to highlight opportunities, highlight issues, help our managers make decisions so that they don't have to, you know, get so tied down with busywork so that managers can focus on, you know, talking to their people and talking to each other and, you know, let AI agents do the nitty-gritty of, you know, there's an issue here, there's something to look out for here. That's been a massive help. Like, I can't imagine us doing this now without the assistance of AI at sort of every level.

I'm sure. Yeah. And luckily AI today is at its worst state it'll ever be. So it should theoretically only get better there. Similarly with autonomous vehicles these days are also at the worst they've ever been as well. Uh, but you know what hasn't really changed is human adoption over the ages, right? You guys were at the cusp of adoption of technology into vehicles, and people are now starting to be at the cusp of, am I going to give up touching a steering wheel?

Where—

what's your— I mean, you guys are right there. You're talking, you're teaching people who literally could be the last to ever learn to drive. So you think, at least as a 16-year-old, like, Give us the real take. What are you hearing from inside the driving lesson from people of what they're learning? Why don't you go first, Nigel?

Yeah, I think very little has changed in terms of the student who's showing up and learning to drive. They're doing so because they want to learn that life skill, and it's going to unlock an amount, a level of freedom for them. I think what changes is how somebody who rides in a Waymo thinks about how their, you know, young child is going to sort of— what their life is going to be like when they turn 16. And, you know, I think it's very cool seeing how far technology has come. You know, back in 2016 when we were sitting where we were sitting, it felt like Autonomous driving was right around the corner, and now it's like, yes, it's literally right around the corner because there's a Waymo, you know, just on the way to pick you up. But the path from that to replacing all road trips or all sort of all travel by automobile with autonomous, I think, is It's still decades away, and it's just, it's, it's, it adds this level of convenience that, you know, is going to continue to get cheaper and more readily available. And every bit of convenience makes you think about, you know, what your mode of transportation is going to be on a particular day. But from a sort of lifelong standpoint, I think this is a skill the vast majority of the population are still going to need to have a couple decades from now.

Yeah, it's really interesting, right? It feels like it's right on the corner, yet also 10 years away where you give up your steering wheel and, or your driver's license, I guess, right? That's what we're really talking about. Are you really going to not need a license because you're like, I don't need it from the backseat, Matt, technically. Do you think that obviously the tech, tech has increased, has, has definitely advanced, but not human adoption, I guess. Where, what's your take on that?

Yeah. And, you know, you said it, you know, we were there at the cusp. And so our mission of eradicating car crashes is, I think, furthered by the adoption of self-driving vehicles. So I think Nigel and I both come at this from— we think it's a good thing, right? And I would love to see more advances and more expansion. From the technology standpoint, as far as Coastline is concerned, you know, we're, we're going to continue building technology that helps people learn how to drive, that helps people be safe on the road regardless of, you know, what they're on the road with. And so, I don't know, I think there's a lot of opportunity for us. I also think, you know, I have maybe 2 or 3 friends who've ever been in a Waymo, which is crazy because like, if they were here, I would take it, right? And so, I do think that the adoption curve looks faster in LA, San Diego, San Francisco, Austin than it does to the rest of the country. So in terms of, you know, people wanting a license, sure, at some point it's going to come to an end, right? Whether that's 10, 50, or 100 years, you know, who knows? But I do think that there's a lot of really exciting, you know, technological innovations happening and a lot of opportunity for us to be a part of that.

For sure. Yeah. I mean, I think just One, just your awareness, customer, you know, just your mentality around being at the forefront with customers and building that into every day. You're gonna iterate with that, that's for sure. Um, so it's all very exciting. Obviously, one thing that will never get iterated is the taco, and maybe the way that it gets delivered to you might get changed. Maybe you can, you know, one day teleport one from me to you. Sure. But if you and I want to go get tacos, we still got to jump in a vehicle and like meet each other there somewhere or another, and then we're going to actually eat the taco. So luckily, this podcast will still be around to talk about tacos even when when we're figuring out how we can take a spaceship to meet each other. How are we going to do it? Or what is what's the tacos that we're going to eat, Matt? What's your favorite? Do you have a favorite one? Is it nearby? Is it from nostalgic childhood? A Canadian taco.

I grew up in. I grew up in Canada, where the closest thing that I had to a taco outside of Taco Bell was a beaver tail that we would fold up, right, and kind of eat like a taco.

Okay.

So, I don't know, I feel like I'm gonna get canceled out of San Diego tech, but I'll say the beaver tail folded like a taco, or if I'm down here in Southern California, I just— Costco sells a great taco pack, right? It comes with chicken and good salsa, and you can eat like 10 of them and just feel great about yourself. I'm going with that.

I love it. We've never had the, uh, the Kirkland Special mentioned on here, or, or the Beavertail as well too. So 2 good new ones into the taco database for sure. And now I gotta go. Well, they don't sell the taco on the outside though. It's not in the food mart. It's got, it's like the frozen, the frozen pack. Okay.

It's, it's fresh in the fridge area.

Oh, all right. Yeah.

Yeah. And it's like they give you the lettuce and the slaw and, you know, it's like a deconstructed taco and you put it together yourself at home.

Love it.

All right.

I'll see it. I'll see it at Costco. Yeah. Nigel, how about you?

So this is not going to go in your database because I don't know what taco place this was, but we have a manager of our San Diego market, and we did a, you know, a trip to kind of meet some of the instructors and whatnot. And he chose a taco place in San Diego, and it was just one of the— it was like a grocery store, like a Mexican grocery store with the taco, like A place in the back and it was so good. And, but I don't know.

I don't know where it is. Can you tell us the city? We will, we will figure it out for sure. Because I already know, I'm already thinking in my head, it's Encinitas. Oh, I know which one it is. If it is—

It was Encinitas.

All right.

It's been named for sure. For sure.

I can't believe you, you called that. Yeah, that's awesome. Yeah.

There's a famous one that's legendary.

Uh, yeah.

Right off the 5 there. Hold on, we're doing a little searching. Um, but yes, classic. The classic spots are sometimes not where you expect. In fact, there's a few Sombrero chains inside some gas stations that are legendary, named in a few Blink-182 songs. So, you know, they can name them. Uh, you can go find Sombreros inside Inside a taco station, but inside a market, of course. You know, they're chopping up some fresh meat right there. That's for sure. Okay, it is the— I'm pretty sure it's El Napolitano inside their mark, I think. Anyway, we're putting in— we'll double-check. We will verify our sources. We'll use AI. To obviously verify that we did all of our right work there as well too. But obviously that means we need to go have one of those tacos together, guys. So next time you're in San Diego or next time I'm in your guys' hood, uh, tacos on us or tacos at Costco. That's— Guys, it's been amazing to be along for a ride, uh, on your guys' journey here, and I'm excited for it to continue on. You know, who knows where this is headed? I think in the future maybe there's a you're teaching people to drive other things, maybe beyond cars, maybe flying cars and things there too. So the future is wide open, and you guys are at the forefront of it, and couldn't be more thrilled for what you guys are building.

So congrats, and keep it up.

Awesome.

Thanks, Neal. Cheers. Tacos and techies on smash. Boulder Green, guacamole. Keep accountants counting beans. Contador cuentan frijoles. Founders trying to raise some queso. La vida San Diego. School of rapping, rapping fire. Tortillas, round jalapeños. In these modern times, a craft beer and coffee. My coding yoga posse, fold the body origami. We be surfing Coronado. Tacos El Gordo, the taco stand in Puesto Lucha Libre en Sombrero. Gotta shout out Tech Coast Angels by the ocean, watch the sunset. La Jolla Shores entrepreneurs trying to get funded when San Diego Startup Week is popping in the summer. Oceanside to Tijuana, SD be full of hustlers from couch surfing to penthouses right on the beach. Tech leaders who believe startups Are within reach. We interviewing founders who be shooting for the moon. #SDTechos with your host Neal Bloom. Come on! Startups get the party, startups get the party started. Startups get the party, startups get the party started. Startups get the party, startups get the party started. You want tech and tacos, tacos and tech, you know we got it. Hey! Startups get the party, startups get the party Startups started. get the party started. Hey! Startups get the party, startups get the party started.

Hey!

Startups get the party, startups get the party started. You want tech and tacos, tacos and tech? You know we got it, hey!