Latest / Cheeky Pint / Garrett Langley of Flock Safety on building technology to solve crime

Transcript

Garrett Langley of Flock Safety on building technology to solve crime · Cheeky Pint

• [00:00:12.06] John Collison Cheers. [00:00:12.19] Garrett Langley Yes, cheers. [00:00:14.05] Garrett Langley Likewise, thanks for having me. [00:00:18.04] John Collison Maybe start by describing, how does the Flock product work? [00:00:24.04] Garrett Langley Maybe we rewind all the way back because it's evolved. Eight years ago, living in Atlanta, and there's a… Fun fact: if you're in a place like Atlanta or Memphis or pick a town in the southeast, if you just pull 10 F-150 door handles, some—let's call it three out of 10—will be unlocked and one out of 10 will have a firearm in the glove box. Which is like, regardless of the firearm, your point of view on it, it's like— [00:00:54.19] John Collison It shouldn't be unlocked. [00:00:54.21] Garrett Langley You should keep it in a safe, not in a glove box. That's just really bad. That's what people do. If you're a gang member, and you're trying to obtain a firearm, the easiest way is just to drive into a neighborhood, six kids—because they're kids—jump out and start pulling door handles. Not even breaking into the car, just pulling door handles. This happened in my neighborhood. Someone got a gun, someone posts on Nextdoor, "Oh, my gosh. I forgot my gun in my car, and it's now gone." The Atlanta Police Department comes, and the Major was fairly apathetic. "Hey, sorry. Good luck. We're not going to…" To your Bosch thing, they're not going to fingerprint the vehicle. No one's been hurt. For most major cities— [00:01:36.20] John Collison There's no stakeout happening for this. [00:01:38.07] Garrett Langley There's no stakeout, there's no, "We're going to go do all this research." For most major cities, if a human is not physically hurt, the crime goes to the bottom of the list. I found that really frustrating. This should be really easy. "These people drove into our neighborhood, they stole a firearm, we should go after these people." He was like, "We don't have any information." "What do you need?" "We need a license plate." "Okay, great." [00:02:05.01] Garrett Langley I was an electrical engineer, so I called a buddy of mine who studied computer science with me at Georgia Tech, and I was like, "Look, we're going to go build this camera that's going to track license plates." That's it. That's what the Atlanta Police Department said. [00:02:15.05] Garrett Langley Okay, great. We built this thing. If you ever come to Atlanta office, we still have the original camera on a pedestal that looks like garbage. I'm not a mechanical engineer. We put it up, and all it did was track every car that came into our neighborhood. But as you can imagine, after 30 days, you start to be able to know, that person clearly lives here. They're in the neighborhood twice a day, four times a day. [00:02:37.08] Garrett Langley About two months later, another car gets broken into, another firearm stolen. The same Major comes back, and I was like, "Oh, by the way, here's the only car that doesn't live in our neighborhood that was here last night." That car gets put on what we call a BOLO, "be on the lookout" for. Couple hours later, they find that vehicle. The gun's in the car. The person goes to jail. [00:03:00.18] Garrett Langley What was really interesting is... We were very proud of ourselves. This was not a business at the time, it was a project. The five o'clock news wanted to do the story. I'm on the five o'clock news and the next morning, I had five emails from neighborhoods. I think we might be the only company that for the first, let's call it 20 or 30, 40 million of ARR, strictly driven off of five o'clock news. That was our only growth channel. [00:03:27.09] John Collison Yeah, that actual media appearance did— [00:03:30.23] Garrett Langley Every time we solved a crime, we'd be like, "Hey, do you want to cover the story at five o'clock?" They'd be like, "We'd love to. We love talking about crime." [00:03:37.12] John Collison Because it's the local news. [00:03:38.11] Garrett Langley It's local news. We're not going to go to The New York Times. Nobody cares about a stolen sofa, or a stolen dog. [00:03:45.10] John Collison Okay, but that's interesting. Did you start with… Because I associate Flock today with being plugged into the stolen car databases. But you started without that, just looking up suspicious stuff. [00:03:58.03] Garrett Langley Yeah, well, if you think about it, in the classic, I don't know who gets credit for it, there's 400,000 neighborhoods in America, and there's no consolidation of providing safety for those communities. When I say neighborhood, that is a legally-binding organization that can sign a contract. Not like you and me just put 20 bucks together and say, "Let's buy this thing on the street." That felt really interesting. [00:04:23.03] Garrett Langley Then the other interesting insight that we developed which took time is, I imagine at your house, you have a security system, you might have a gate to your house, you probably have a dog, you have all these things, right? They do two things that actually are not helpful. They tell you a crime has happened, they don't actually help you solve it. [00:04:40.10] Garrett Langley Then what about your neighbor? What we'd realize is that every single security system was focused on the individual. But I think about my life in Atlanta, there was just a crazy random act of violence a couple of years ago. My wife wouldn't run outside for a year. It had nothing to do with us. It was a mile away. The whole premise became that you had to build a safety system for a community, because that's actually what you feel. You live in San Francisco, you get it. Even if you've never been the victim, if crime is up, it's really bad. It was all neighborhoods for the first three or four years of the business. [00:05:13.12] John Collison Give me the stats on Flock today. Just how many cameras out there, has crime stopped, or whatever. And maybe just describe the product as it exists today. [00:05:21.14] Garrett Langley Now it's a much more sophisticated product in that sense. Last year we helped clear just north of a million crimes in America. [00:05:31.01] John Collison What does "help clear" mean? [00:05:32.05] Garrett Langley We want to be clear that similar to probably how you feel in that you help businesses grow, but you didn't do the hard stuff. You made it easy. You made it really easy for me to grow Flock by making payments simple. I don't actually have to do any of the hard work of chasing bad people, putting my life at risk. We write code in design circuits. It's not hard work. We like to be very careful of not overstating our value. We were involved in the arrest, in clearance, or successful arrest of over a million crimes. In a lot of those cases, we are the end, the beginning, and everything outside of the human putting handcuffs on. In some cases, we might be the tip of the spear. [00:06:14.18] Garrett Langley For more sophisticated investigations like the Brown shooter and the MIT shooter, there was a tip on Reddit, but then Flock was the way they found them. But we weren't the tip, so we can't take full credit. [00:06:27.07] Garrett Langley Maybe I'll walk you through a recent example. I won't name the city because the case is still going through prosecution. There was a 911 call. It was a major city in America. Our system hears 911 calls, we can tap into 911 calls. This is wild if you think about 911. Today when you call 911 in San Francisco, let's say you get about three million 911 calls a year. A human picks up every single time. A human manually listens, manually types in information. Imagine if you were running Stripe and every single lead you received was human-routed. You'd say, "This is crazy." That's how we work in cities today. We hear the 911 call in real time. What that allows us to do is then figure out what are they talking about? Is there any interesting information that the system could find beneficial? [00:07:12.04] Garrett Langley In this case, we had heard that it was an attempted homicide. Someone was bleeding on the street. All they could remember was that the suspect was wearing white converse sneakers. Now we have a product called FlockOS that allows us to integrate all the cameras in a city, whether they're Flock-developed or not. That 911 call pops up, the operator is like, "Oh, my gosh. There's a 911 call right there." They can listen to the call. They're like, "This is a really violent situation. There's a privately-owned camera. I can tap in that." Double click that camera. They can use one of our products called FreeForm. They can say, "I'm looking for any individuals in the last 30 minutes that are wearing white converse sneakers." [00:07:49.06] Garrett Langley They then find the individual. They can then push that video to the nearest officer. We run on the dash of the police vehicles. Then that person gets arrested. If you think about the way it used to work, that case never would have been solved. It had been weeks, maybe months, and become a cold case. In this case, this person was from a 911 call to an arrest in about 17 minutes. We do that on people. We do that on vehicles. In that same version of the story... [00:08:16.21] Garrett Langley I'll give you one other story, and then I'll give you a sense of it. In a town in Colorado, there was an armed robbery of a Levi's outlet. Got to get your jeans one way or another. This is a funny one, actually. They call 911. We're plugged into 911. The real-time crime center operator hears it's an armed robbery. They also hear two things. The person has already fled the store, and they drove away in a white van that looks like a black and blue cow. [00:08:43.15] Garrett Langley Mistake number one is criminal, "Don't drive a weird car." Now, this police department also has our drone that we build here in America. They immediately click a button, the drone automatically flies at 400 feet to the Levi's outlet. That drone knows what to look for. It has the visual nomenclature of a white van with black and blue spray paint. Using that same FreeForm, we can start to look for it. [00:09:07.23] Garrett Langley We get another 911 call from another outlet that's having the same time. The drone is already in the air, zooms over, drone has eyes on it. That video feed can now be sent straight to the nearest officer. In a traditional response, you're going to come in hot, blue lights flashing, someone's going to get hurt. This case, the drone's 400 feet up in the air. You have no idea the drone's there. The guy drives home. As soon as he pulls in his driveway, two cops pin him in. Safe, tactical apprehension. It's drones, it's some computer vision, it's cameras that track cars. It's quite a robust portfolio. [00:09:43.05] John Collison What is Flock by the numbers today? How many cameras, how many drones, how many law enforcement agencies, how many individual entities? [00:09:52.11] Garrett Langley Yeah, so it's just over 6,000 cities. It's well north of 50% of America that's covered by Flock. [00:09:59.23] John Collison By population? [00:10:00.14] Garrett Langley Yeah, by population. There's about 17,000 cities in the US, and we have all the big ones outside of Manhattan at this point. That's like the last one to go get. I'm not sure if that's going to happen anytime soon, the current situation there. Yes, it's pretty well-deployed. Then the business has gone from zero to about 500 million in ARR in the last seven years. Building cameras, drones, and selling them to the government, which is like three strikes, you're out. [00:10:29.12] John Collison That's incredible. I feel like something non-American listeners mightn't realize is... I grew up in Ireland, a country of five million people, and there is just, a police force, the Garda Síochána, the national— [00:10:42.07] Garrett Langley For the whole country? [00:10:42.13] John Collison Exactly, police force for the whole country. That is a relatively common model in lots of places. In the United States, it's very localized. Just maybe you want to talk about the average size of it. I don't feel like what you're providing is, in some ways, scale economies and helping police forces coordinate with other agencies. [00:11:04.11] Garrett Langley There's a dynamic that people don't understand which... You're right. Ireland, most of the EU, Australia has six police departments. I mean, let's call it effectively the same. It's a really big country. It's a lot of land. Most of South America... America is really the only country that operates under a model where local municipalities provide law enforcement service. It's very rare. [00:11:26.23] Garrett Langley The benefit, the pro is very straightforward, which is like you actually have a good chance of knowing the police officer that has your territory and has your patch of dirt. The downside is criminals don't really care where cities start and stop. They don't care where states start and stop. Yet historically, law enforcement agencies were incapable of sharing information. To give you a sense of timeline, the cloud was only recently made legal in Florida, as recently as 2022. If you were a law enforcement agency, you could not host your data in the cloud until 2022. Maryland was 2023. [00:12:04.20] John Collison Any coordination between agencies was very manual. [00:12:07.21] Garrett Langley All phone calls, faxing files. One of the big unlocks of Flock was this realization that you could drive better collaboration because if you're a low-level criminal who isn't maybe changing cities... But you look at some of the most successful cases we've helped work, it's like multiple states. We did a huge human trafficking bust. It was 76 people arrested across four states, all organized on Flock. [00:12:40.12] John Collison It's really hard for one agency to wrap their hands around that whole thing. [00:12:44.10] Garrett Langley No, at that point you've got six different local cities, three different state agencies, and the US Marshals all collaborating inside of Flock. That's how it should work. Because in Ireland, that is actually the only option, because you're the same police department. It's a very unique problem that America has created. [00:13:03.16] John Collison Law enforcement coordination. [00:13:05.00] Garrett Langley Yeah. It shouldn't be controversial, but it's actually a very controversial topic. [00:13:10.18] John Collison What's controversial about it? I'm not aware of this controversy. [00:13:15.04] Garrett Langley This is an interesting question of your trust in government probably follows some type of logarithmic expectation or linear expectation of the farther you are removed from those people, the least trust you have. Most people's trust of the federal government is very low. State, a little bit lower. Local, full trust. [00:13:41.13] Garrett Langley There's this interesting dynamic where, let's talk about a most recent case. My guess is that most people don't have a lot of trust for certain federal agencies. But when stuff goes really bad, like the Guthrie kidnapping, it's the federal government that runs the investigation. They are the only ones with the actual resources to work these cases. It's the FBI that's running that investigation. The Brown shooter, the FBI. The Mar-a-Lago assassination attempt, the Secret Service. It goes down, but people have this big distrust, and so we actually see it in places like California, there is state legislation that law enforcement is not allowed to collaborate with the federal authorities. [00:14:17.23] John Collison Because of all the— [00:14:18.13] Garrett Langley On certain topics, because it's just different politics. We have to sit in the middle. When it's clearly legislated, like in California, it's actually fine. That's the law. That's actually very simple. It's in other states where it's opinion. That's like, "Oh, that's really messy," because we should just write laws. You operate in a regulated space. Regulation, when done properly, just defines the rules of engagement. Sometimes it's better or worse, but at least everyone knows these are the rules. [00:14:48.15] John Collison Yes. As I think about where Flock grew up and the helping with flagging stolen vehicles, and a very vehicle-oriented product, how does that work at a technical level? Like you guys... There was a database for stolen cars already, But that's an example of prior cross-law... Like, who maintains that database? How does it work? [00:15:16.13] Garrett Langley Yeah, it's a good question. You think about, there's a couple of fun engineering problems that Flock had to solve. The first is how do you do what we just described, read a license plate, track a car on solar power with a 5G backhaul? [00:15:31.20] John Collison Which is the camera that you have installed on the traffic lights. [00:15:34.13] Garrett Langley The camera that we've designed. Yeah, exactly. But you very rarely have fiber or power where you want to put a camera. [00:15:39.02] John Collison You want a self-sufficient box that you can totally put up in an intersection. [00:15:41.12] Garrett Langley Totally infrastructure free. And be able to track a car going 100 miles an hour and do some level of computer vision on top of it. That was a pretty fun problem because if you put, let's say you wanted to put a GPU in it, you can't run on solar power then. So then it's like, "Well then what do you do on the edge? What do you do on the—" [00:15:59.13] John Collison You can't buy any GPUs these days. [00:16:01.00] Garrett Langley Yeah, that would have been a huge disaster for us. But the FBI maintains a list called the NCIC, which is about a quarter to half a million known vehicles with other warrants. It could be Amber, the Amber Alert system, the Silver Alert system. We saw, sadly, like a ton of Silver Alerts too. I think we did just over a thousand Amber and Silver Alerts last year that we helped clear. It's like missing seniors and missing kids. Pretty sad stuff. We have a direct integration with the FBI for that. Then at a local level, let's say like the Bay Area. There is a Bay Area hot list that we maintain with those agencies. [00:16:45.09] John Collison Oh, wait, so I assumed that there was a single integrated… When a car is called in as stolen, I assumed that went somewhere integrated. [00:16:54.05] Garrett Langley This is scary. Let's say your car is stolen and you're in South San Francisco. They will immediately put it in the Bay Area Flock hot list immediately. It will take 24 hours to make it to the FBI hot list. That's a CSV file that gets sent around on FTP servers across the US. [00:17:15.13] John Collison As much as the economy runs. We're very used to that model. [00:17:18.05] Garrett Langley Used to that stuff. [00:17:18.05] John Collison Yeah, exactly. [00:17:19.15] Garrett Langley The real-time happens on Flock and then there's about a day lag to make it nationwide. [00:17:23.20] John Collison Okay. Local entities have some local lists and those propagate to a national database, which is an FBI database. What you guys are doing is making that real time, which obviously, if a crime is unfolding in real time, is a big deal. [00:17:36.16] Garrett Langley Exactly. [00:17:37.06] John Collison Yeah. [00:17:38.10] Garrett Langley It's not good to wait a day. [00:17:40.01] John Collison Yes. [00:17:41.17] Garrett Langley You can drive really far in 24 hours. [00:17:43.05] John Collison It turns out, yes. Then you had some great stuff before about just a fraction of… Stolen cars are bad news, the fraction of crimes that involve stolen cars. [00:17:52.08] Garrett Langley There's an interesting phenomenon where you don't... No one really steals a car for fun. There's not much you can actually do. [00:17:59.16] John Collison Is that true, like all the sideshows and stuff? Isn't there a little bit of stealing cars for fun? [00:18:03.07] Garrett Langley Well, the sideshows are definitely... They look fun, never participated in it. [00:18:07.08] John Collison Neither have I, just want to make clear here. It's my official statement on the matter. [00:18:10.21] Garrett Langley I haven't ridden a dirt bike in a sideshow, but I have watched plenty of videos online. That's fun. You're typically not stealing a car. You're typically driving your own car in those. [00:18:17.17] John Collison Oh, okay. [00:18:18.03] Garrett Langley You typically... While there are some people, I guess, who would steal a car and just drive it for fun, the main use case is when my car was stolen... My car was stolen, and then they used it to go rob three CVSs, did a bunch of drugs in the car, and then ditched it. That's like a normal... You steal a car to go do something bad. Or you steal a car to go shoot someone, because you don't want to shoot someone in your own car. If you're thinking ahead, you want to do it in a stolen car. [00:18:44.05] John Collison We've all watched Pulp Fiction. [00:18:45.12] Garrett Langley Yeah, this is how you do this, right? We do a bunch of other interesting things. We have this feature called "cold-plating", where if we detect that the vehicle, from an AI perspective, the make and model doesn't match the DMV record, that clearly the person's driving with a stolen license plate. That's another thing you can do. It's easier than stealing a car, you steal someone's tag. [00:19:05.21] Garrett Langley Then you got to go down to, "We've got to steal the same license plate from the same type of car." But then if we detect an anomaly that the same car is in two places at once, we'll flag that. That's kind of weird. That's like a glitch in the Matrix. It's like we have all these, and we would call them more like anomaly detections where it's like, "This isn't illegal yet, but it sure is kind of weird." [00:19:25.01] John Collison Yes. It provides enough reason to pull someone over. [00:19:31.03] Garrett Langley Yes. Depending on the size of the city, it's enough to dispatch an officer. I think if you're a major city, like a San Francisco, you have enough going on. But if you're a foster city, you're a smaller town where thankfully you don't have violent criminals running around every day, you're like, "Oh, that's weird. We have two cars with the same license plate in different parts of the city. We should go see what's going on." [00:19:51.18] John Collison It strikes me that a way you can think about what Flock is doing is not changing any of the norms around privacy but really expanding law enforcement's bandwidth to go deal with stuff. And so I don't think anyone really has an expectation of a right to privacy of the car they are driving just being observed from a distance or their license plate being readable when out on a public road. But if a car is called in and stolen, rather than officers manually looking for that car and that license plate, instead you can just much more quickly get it. And yet this kind of stuff ends up super controversial from, I don't say super controversial, but it ends up controversial from a privacy point of view. And so, I don't know, is that take too generous? Is there a steelman on the other side? Why is it— [00:20:50.10] Garrett Langley There is. I think you're right on the controversy. I would articulate it as like, if you're building a business that impacts millions of people's lives, it's going to be controversy of some degree. Like, I'm sure people… I'm sure there's someone who hates Stripe. I don't know why they would, but I'm sure that person exists. It's just like they're someone who hates Walmart. It's like they're trying to sell cheap groceries. Why do you… It's like people hate every company. Maybe they hate us more, I don't know. I think there's a few things that make it... For the steelman argument. One is you can see it. [00:21:22.11] Garrett Langley I bet you if we pulled up your iPhone, and we looked at the number of apps that you've given full-time local location services, it would shock both of us. And then if we looked at the number of data brokers who then leverage that data to sell you ads, we'd be really shocked. And I think if we rewound 30 years ago and said, imagine these private companies tracked your location in real time and sold that to advertisers, we'd be like, "That is unacceptable." But because we can't see it, we let it go. [00:21:51.23] Garrett Langley We have this perception of anonymity or being anonymous. [00:21:56.01] John Collison Anonymity. [00:21:56.17] Garrett Langley Anonymity. Thank you. Yet, you're an engineer. How many data points do you need to triangulate where someone is? How about where they sleep and where they work? And I know exactly who you are. [00:22:06.05] John Collison Yes. [00:22:07.16] Garrett Langley And so, there is that one piece, which is because we operate in the physical world, we are held to a higher standard, which I think is a shame because personal opinion, what I do online, it's way more interesting and telling about my personality and my life than what I do in the real world. I go from work to home and to kids' birthday parties on the weekends. That's my life. So there's that piece. [00:22:29.15] Garrett Langley The second piece is there is an appropriate debate of what level of privacy erosion are we willing to take for an increase in safety. You and I both choose to live in cities that have governments, so we've already chosen to remove some of our privacy. We drive on public roads where, constitutionally, we have no expectation of privacy because we're using the government's roads with the government's license plate. We have a driver's license in our pocket. Driver's licenses are a modern thing. That's not a… That hasn't been around for 200 years. It's a newer concept of a driver's license. So we've accepted all of those. [00:23:03.10] Garrett Langley The articulation that I would have is like, I want my kids to be safe. And as long as there's accountability of how the information is used, which we do, every single action in our system, the audit is stored in perpetuity and it's publicly available. For us, we really believe in this concept of a certain data retention, which limits the level of abuse. It doesn't eliminate abuse. There's still going to be abuse, but it limits that, hey, if it's seven days, 14 days, 30 days of data, it's just not that interesting. Relative to what you see with data brokers online, where they have your entire internet history stored forever. [00:23:38.20] John Collison You're interested in fresh data. That's what's— [00:23:40.19] Garrett Langley Yeah, but criminals, that's what's— [00:23:42.11] Garrett Langley Yeah, there's a really fast drop off of where have you been in the last couple of days, versus like where were you a year ago. It's very irrelevant in a criminal investigation. [00:23:51.12] John Collison Yes. [00:23:52.06] Garrett Langley Or it's protected behind a warrant, which is fine. [00:23:54.01] John Collison All right. You talked about cell phone locations and your cell phone having your GPS location. I'm reminded of the phenomenon where it feels like the balance of power and the offense-defense mix within crime fighting tends to change over the years as new technologies are invented and criminals get wise to fingerprinting exists and things like that. It feels like cell phone locations have had a huge impact on crime fighting. Not even GPS location, but just the coarse cell tower data. And in particular, in murder cases, it's had a huge impact. I'm curious just what are the other major trends going on, from a technology point of view, in how crime gets fought. [00:24:38.21] Garrett Langley Yeah, that's CDRs would be the nomenclature in the law enforcement. [00:24:43.07] John Collison What's CDR? [00:24:44.01] Garrett Langley Cell phone data dump or CDR— [00:24:46.07] John Collison Cell data records? [00:24:47.06] Garrett Langley Something like that, yeah. But to your point, it's the broader, like you hit these three towers, therefore you're in this general vicinity. That is one of the best-kept secrets for law enforcement of how to solve crime, which people dumbly commit crime with their cell phone in their pocket every time. It's very rare. I think I've heard of one case in the last 12 months where someone was smart enough to commit a homicide and left their cell phone at home by design. It was a very intentional, they were very organized. A couple other ones. [00:25:16.23] John Collison How did they get caught? [00:25:18.16] Garrett Langley It's actually Flock, obviously. So they were smart enough to leave their cell phone. They were dumb enough... So they drove from LA up to San Francisco general area. They hit a bunch of Flock cameras along the way. They didn't pick up their cell phone until they went back to LAX, and we have cameras all around LAX, and they finally made that one mistake. It was a completely cold case until they got a hit on the car pulling into LAX, then got the cell phone, could do the whole thing. This is a good case. [00:25:49.13] John Collison And I guess the thing you can do is you can know that a cell phone's in a car because you can— [00:25:52.17] Garrett Langley We don't touch that, but the law enforcement— [00:25:53.23] John Collison Sorry, law enforcement can cross-reference cell phone location with Flock. [00:25:58.13] Garrett Langley So they knew that the vehicle was in a certain parking lot. They can then use… They have to have a warrant at this point. Warrant to pull what cell phones were in that area and then start to build the case. But to your question, I'd say the current phenomena that is causing a ton of problem domestically is drones. This is an example where it's asymmetric warfare that law enforcement is being put up against where criminals have no rules. [00:26:21.20] Garrett Langley So here's a good example. In one of the counties we work with, it's an affluent-ish county in the Virginia area, South American cartels fly illegal drones through these neighborhoods. They'll flip on night vision, look through houses to see if anyone's home, then go break in. They don't want a confrontation. They just want to see what's going on. [00:26:44.07] John Collison They're using drones to case out— [00:26:45.18] Garrett Langley To scope out. Yes. [00:26:46.18] John Collison Wow. [00:26:47.05] Garrett Langley And this is where the downside is. Law enforcement has just recently in the last year been allowed to fly beyond visual line of sight. These guys have been operating for years. [00:26:55.06] John Collison Sure. [00:26:56.10] Garrett Langley They jury-rig an LTE modem to a cheap drone. They can fly it anywhere, however they want. And then legally, even to today, law enforcement is not allowed to engage that drone. They can't take it down. It's FAA airspace. It's not law enforcement airspace. [00:27:12.14] John Collison Sure, yeah. [00:27:13.17] Garrett Langley And so they just sit there. That's a problem in neighborhoods, that's a problem in prisons. It's probably the number one problem we have in our prison system today... Well, we have a lot of problems in prisons. One of the leading problems for enforcement officers is you've got these pretty incredible, actually, from an engineering perspective, drones that are carrying 10, 20, 30 pound payloads flying them over the prison walls, and they'll literally dangle it down to the prison cell and the person will reach out and grab it. Cell phones, drugs, guns. [00:27:41.07] John Collison We had Keller from Zipline here. It's Zipline for criminals. [00:27:44.12] Garrett Langley And these guys have built a comparable product. I was with a sheriff that I know really well, and he was like, "If I shoot that down with a shotgun, I am technically breaking the law. I'm breaking federal law if I shoot this down with a shotgun." And so you're seeing states now pass, like Louisiana has a state bill, Georgia's working on a state bill that says, sorry FAA, we're going to do what we want. And that creates a problem, because most law enforcement officers just want to follow the law. [00:28:11.01] John Collison Yes, and presumably that won't stand up from a federal preemption point of view. [00:28:14.16] Garrett Langley No, and the other challenge is, I think you and I were discussing, who's going to enforce this? FAA is not an enforcement body, but these people want to follow the law. And the law says clearly, you can't shoot a drone down. You can detect it, you can't mitigate. [00:28:28.12] John Collison I had no idea. [00:28:29.17] Garrett Langley It's a totally crazy law. [00:28:31.11] John Collison Yeah. What else has gotten worse from an offense perspective? [00:28:38.08] Garrett Langley This is one that's like a… I don't know if I have strong feelings on this topic, but this concept that we appropriately hold local law enforcement to a very high standard of accountability and audibility. I think that's very good. The downside is we don't with criminals. And so as a citizen, you could be the victim of a crime. You get really frustrated that they're not working hard enough. It's very rarely that. It's that they don't have the tools, they don't have the data, or they're not legally allowed to get to the data. [00:29:08.15] Garrett Langley The warrant system is a very, very good thing. It's a very effective tool. But there's a debate of should law enforcement have more ability to solve crime faster. The example of the framework that we use is, I don't know how you land this, is that the severity of a crime should be commiserated with the sophistication of technology. And I'll give you an example. Facial recognition, hot topic. There are thousands of cities in America that have banned law enforcement from using facial recognition. John, facial recognition is not bad as a technology. It's not good either. [00:29:44.19] Garrett Langley It's just technology. I think a way more effective measure would be to say, "Hey, look, facial rec has its pros and cons. You can't use it for shoplifting, but for homicides, crimes against children..." Go to your list of things that we as a society care deeply about. Law enforcement should do everything in their power to solve those cases. And then the petty stuff, we should say, "Look, we're going to make progress there. You stole a pack of Skittles, we probably shouldn't deploy a drone to find you. If you kill someone, I think we should work really, really hard as a society to hold you accountable." But there's no nuance. [00:30:20.18] John Collison That nuance does not come through. [00:30:22.00] Garrett Langley No, they're just like facial rec is bad. And we don't do facial recognition for that reason, because it's too controversial, but that's counterintuitive to me when the technology gets better every day. [00:30:31.07] John Collison Yes. Another interesting thing that's happening amongst these trends is the prevalence of body cameras, which I think maybe was, in some corners, promoted by law enforcement skeptics but I think now has— [00:30:47.06] Garrett Langley It's backfired. [00:30:47.22] John Collison I wouldn't say backfired but just like it's, in a lot of cases, been somewhat exonerative, right? [00:30:54.14] Garrett Langley No, I think in many cases police were really hesitant to get body cameras. Just like, I mean, imagine if you wore a body camera all day at work, you'd be like, I'm not sure if I love that idea. It's a little bit invasive, but they did. And I would say in the majority of cases that I see where foul play is called, it is almost always the inverse, where the law enforcement officer was just trying to do their job. And on the other side, there was either a mental health issue, there was something gone wrong, and it exonerates the officer, which is interesting. [00:31:26.11] Garrett Langley I think the same thing is true. What at least we're seeing on the camera side on the streets is historical policing, sadly, is quite prejudiced. We all have our biases, whether it's conscious or unconscious. So if you have bad data in, you get bad data out. And so the traditional way of policing is you go to dangerous neighborhoods, look for suspicious people, and arrest them. It's just like you're perpetuating a trend against a certain community. And when you look at Oakland as a good example, who's quite a big supporter of Flock, they're like, we need a more objective way to police. Let's just focus on stolen cars. We don't care who's inside. We will find out eventually that car was stolen. [00:32:07.17] Garrett Langley And then you don't wind up policing where crime has happened historically, you wind up policing where crime is happening right now in real time. And that is a fairly fundamental shift in how policing works. And for many cities we work with, changes the perception of this community with law enforcement. Because they're no longer felt like they're being targeted. It feels like it's wherever crime happens, we're going to go chase it. And if there's not been crime in the last couple of days, then we're good. [00:32:35.07] John Collison Yes. Is crime up or down in the US? [00:32:39.07] Garrett Langley It is down. I think it'll continue to go down. COVID was like really bad. [00:32:47.01] John Collison So it surged during COVID? [00:32:49.21] Garrett Langley Crazy surge. People lost their minds. I think we tested society of what happens if you keep people inside for too long, and people got really violent. [00:32:58.02] John Collison Yeah, so do you have an explanation? Is it that it was a mental health issue or was it a crime of opportunity that things were less well-guarded, and there are more opportunities for crime, but you think it was just people went a bit batty being locked up? [00:33:12.08] Garrett Langley I think so. [00:33:12.19] John Collison And results in crime. [00:33:13.19] Garrett Langley But I'm obviously not a clinical psychologist by any means, but you look at the data and there's no other way to articulate why homicides went up three to 3–4x and then plummeted back down to let's call it somewhat reasonable levels. There's a very few number of cities that are left at COVID levels. Most cities had this massive spike. [00:33:35.03] John Collison Yes. [00:33:35.08] Garrett Langley And the way to… If you look at the people who were committing that violence, they tended to be 16 to 22, male, and very online. [00:33:49.07] John Collison Very online? [00:33:50.02] Garrett Langley Yeah, and so this is really sad. [00:33:52.23] John Collison How do you measure very online? [00:33:54.04] Garrett Langley Let me give you an example. I was in a major city, not too long ago, and I was asking the chief to tell me kind of what's going on. This is a couple of years ago during COVID. And she was like, "Garrett, these kids are just killing each other." I was like, "What?" She's like, "Yeah, they're literally getting in their car and just shooting each other." I'm like, "Why?" I'm like, "Because this guy posted a picture of him with that guy's girl on Instagram." And I think in a normal situation you might have called that person and been like, "Hey, bro, that's my girl." And in other cases, they're getting in a car and shooting someone. And that's not normal, that's not normal behavior. And it wasn't normal before COVID, and it happened a lot during COVID, and then it's largely gone away. [00:34:36.11] Garrett Langley But it was a very specific social phenomena where the race to violence was so dramatic, it was so scary. And I don't think society's necessarily gone fully back to normal, but if you look at the data, it's hard to articulate. That's the most obvious explanation to me. [00:34:53.11] John Collison And I guess there are other empirical measures like air rage incidents, road rage incidents, all that kind of stuff went up during COVID? [00:34:58.21] Garrett Langley Everything went up. [00:35:00.12] John Collison Yeah. [00:35:00.22] Garrett Langley Like everything. People just literally lost their minds. I think we need to be outside. [00:35:04.23] John Collison Yeah. But how are we doing now on crime? You say it's down. Is that across all categories? Because you read Twitter, and it's like everything's locked up at CVS and San Francisco is full of... Like, there's all these memes which are… Maybe do some debunking and confirming of all the memes. [00:35:20.04] Garrett Langley Every city is a little bit different. Major cities still have a major crime problems. I would argue... I live in Atlanta and the mayor, he's made it very clear, anything greater than zero homicides is a tragedy, and therefore we're on the race to zero still. It's an unacceptable level. We're down historic lows over the last decade, but it's still pretty hard to wake up and go, only 52 people died in Atlanta last year. You can't pat yourself on the back for that. I think where we're focused on is more so clearance rates, and that's getting better, but only better in certain cities. [00:36:00.17] Garrett Langley We track it across Flock, not to overly promote, tends to be that if you have a Flock product, your city has a much higher clearance rate. You look at San Francisco as a great example. The new chief, the prior chief, Mayor Lurie, they are making crime a focus and the focus is on solving crime. And when they solve it, as I'm sure you follow on the news, they talk about it. And they should, because it's really cool, they're working very hard, and what you'll see is back to the online point I made, all these people are online. And so when San Francisco PD is dominating X— [00:36:34.19] John Collison It has a deterrent effect. [00:36:35.22] Garrett Langley Gosh, yeah, you don't want to get caught. And that's the unintuitive or the counterintuitive point on crime is that most people would intuitively say, "If the punishment is really, really bad, people will do less of it." But you were 16, 18 at one point. You operate on a Boolean mindset. I will get away with this, therefore I will do it. I'm going to sneak out of my house because I'm not going to get caught. You don't care if the punishment is being grounded for a week or a month, you're going to get away with it. [00:37:02.08] John Collison It's just will I get caught or not, yeah. [00:37:03.20] Garrett Langley Yeah, and so when you flip— [00:37:04.15] John Collison I didn't realize criminals are so online. [00:37:05.23] Garrett Langley So if you flip that... It's a subculture, you got to get there, but I mean, you'll see— [00:37:11.23] John Collison Sorry, is this like Reddit, Twitter? [00:37:13.18] Garrett Langley They're everywhere. [00:37:16.06] John Collison Pinterest? [00:37:16.15] Garrett Langley Probably less Pinterest. But TikTok's really big. Instagram's really big. Snapchat's really big. And it's how they recruit though because the whole recruitment effort is, I'm going to show a lifestyle online that seems dream-worthy to recruit these people. And then they're in, and they're going to perpetuate that. And the data shows it's actually not a very good job. You actually don't make very much money being a criminal. It's like the medium income for your average criminal... It's very, very low. But they promote a lifestyle of wealth. So if you look at those cities, so you're taking San Francisco. Crime's coming down, clearance rates are going up. I would say as soon as we can get somewhere like San Francisco to 100% clearance rate, which we have in other cities, crime goes down. [00:38:00.13] John Collison 100% clearance rate is like 100% solved? [00:38:03.08] Garrett Langley Yes. And it's doable. We have it in major cities. Cobb County would be one, which is like the second-safest county. [00:38:10.04] John Collison Cobb County has 100% clearance rate. [00:38:11.01] Garrett Langley Yes. If you commit violence in Cobb County, you will get arrested. [00:38:14.20] John Collison And that's across what sample size? [00:38:17.04] Garrett Langley That's the second-largest county in Georgia, so let's call it a couple million people, almost a million people. [00:38:22.16] John Collison So a decent number of violent crimes? [00:38:24.03] Garrett Langley Oh, yeah. [00:38:24.15] John Collison Yeah. [00:38:25.11] Garrett Langley But it's going down every year. [00:38:28.06] John Collison Sure. [00:38:28.12] Garrett Langley So look at places like that— [00:38:29.05] John Collison Because you will get caught. [00:38:30.01] Garrett Langley Yeah, and so we have this new concept that we've developed called Safe City, and we'll go into a town and say, "Look, this is the platform you need to solve all the crime in your city. It's your choice. It tends to be, like about 20 bucks a citizen a year." [00:38:43.23] John Collison Yeah. [00:38:44.17] Garrett Langley It's like, it's your choice. And what's fun to watch is we find these mayors with really strong backbones, and they're like, "I want to be a safe city." There's this awesome town, Greenville, Mississippi. You've probably never heard of it. 26,000 people. They have our drones, they have our cameras, they have our AI, they have everything we do, and they are lighting it up. And kudos to the mayor and chief, like every crime they solve, they're on the five o'clock news. [00:39:09.23] John Collison So these highly-effective cities. What effect does it… What happens to the criminals? Does it shift it to other localities? And you need to… It's like disease eradication, you ultimately need to do kind of enough to blanket it, to snuff it out. Because if just one municipality does it, it'll shift to a local municipality. Does it within that municipality, people just go to, something non-criminal. I'm curious, what are the effects of one municipality getting really good? [00:39:38.08] Garrett Langley There's two phases. The first phase, you're spot on. People, they just change cities. It's like when San Francisco started adopting Flock, Oakland crime went up. So then Oakland needed to adopt Flock, and now a couple of false starts, now that it's fully going. They will have the same, but first is the shift. And then you're right. People don't… The majority of criminals, let's call it 99% of criminals, are not evil people. They're not. Evil is a random act of violence, and that is exceptionally rare. [00:40:11.07] Garrett Langley It's all opportunism. You and I were fortunate enough to be born in a family, in a social construct that we could go build great companies, but not everyone had that chance. Some people get pushed in the wrong direction, and that's what we have to stop. It's like, as a business, we don't… While we're proud of the impact, making a million arrests, it's actually quite disheartening because that means a million people will now have to go through the criminal justice system, which doesn't work that well. There's a million victims. It's bad in every single way. It's also crazy expensive as a society to jail that many people. It's a double negative bottom line where it's costly and prohibitive that they'll probably never reenter society, so we're much more focused on the preventative mechanisms of how do you convince someone this is not a good lifestyle, this is not a long-term plan. Eventually, what you will see is they will go get jobs. These are functional members of society. They can be functional members. We should push them in that direction. [00:41:06.07] John Collison What kind of crime is on the increase? [00:41:08.01] Garrett Langley I'll tell you one that's pretty fascinating. If you look in the enterprise community, organized retail crime in stores was really hot during COVID and right after COVID. That's why you saw CVS lock their stuff up. You saw places like In-N-Out leave Oakland. This was a huge problem, and it's still… For someone like Walmart, I think they reported just shy of a billion dollars of theft last year. It was a lot of money, and their peers aren't doing dramatically better. Now, Flock customers are doing really well. [00:41:38.21] John Collison And Walmart aren't neophytes at this stuff. [00:41:41.15] Garrett Langley Yeah, we'll get there with Walmart. But you look at some of our partners like Lowe's, their shrink has gone down order of magnitude. It's not really a problem anymore. But where the criminals have moved is to the distribution facilities. It's safer and bigger loads. Probably the most sophisticated one that I've heard of is an Eastern European group. They went and bought a legal, well-running freight broker. They now own this asset, and they bid on all these… Low bid on all these projects, show up with real paperwork, fill up a 20-foot with product, drive away, dissolve the company. $7 million in a single day. That's pretty good business. That's very sophisticated though. That's harder to solve. That's much harder to solve. [00:42:36.19] John Collison Because every car, phone, everything will be disposable used in that. [00:42:42.12] Garrett Langley Everything's going to leave the country as fast as possible and just show up on a street somewhere, and then the other interesting challenge is you have to unpack... This is a little bit complicated. Who's liable? As soon as the product goes into a store, that retailer is liable for the theft. It hits their bottom line. For a lot of these companies, they negotiated such that there's insurance coverage from a different broker that owns that asset until final delivery, and so it's not super clear who actually cares, and it's going to take— [00:43:18.03] John Collison The insurer cares. [00:43:19.01] Garrett Langley The insurer cares eventually. But when you talk about the scale of what's called tens of billions of dollars, hundreds of billions of dollars of product being moved, you're like, "Oh, we lost a couple of hundred million." [00:43:29.12] John Collison I see. [00:43:30.17] Garrett Langley "We'll be fine. We have insurance." It will catch up eventually, but that's probably the most interesting type of crime we're tracking is the tacking on the distribution. We have a lot of partners on the distribution side where we're now deploying our product to try to prevent this from becoming an epidemic. But it's much harder to solve. It's a real company. [00:43:49.22] John Collison Yes. I want to ask you more about this, but I'm going to get another Guinness. [00:43:56.09] John Collison This is a good segue into your corporate business, because again, people associate you with selling to municipalities. But obviously, you sell a lot to corporates as well. [00:44:06.03] Garrett Langley Yeah. It's a big part of our business. [00:44:09.17] John Collison Can you say how big? [00:44:11.23] Garrett Langley North of 100 million of ARR. [00:44:13.10] John Collison Okay. [00:44:14.03] Garrett Langley It's real. You told me to go all in. [00:44:17.06] John Collison Oh, sorry, you had some dregs left. I'll get to that. [00:44:19.15] Garrett Langley That's my fault. [00:44:21.06] John Collison You want to just go full up, and you'll get more of a head that way. It's perfectly, it's the correct… [00:44:25.15] Garrett Langley It is. [00:44:26.01] John Collison You can trust it. Yeah. Okay, so the corporate business. [00:44:29.01] Garrett Langley Yeah, north of 100 million of ARR. Probably the fastest growing segment. There's some wild stories. We help businesses solve two problems: how do you keep your employees safe, and how do you keep your assets or your stuff safe? So it's kind of funny, because we have a Fortune 5 company that's a customer. [00:44:49.03] John Collison You probably don't want to talk about specifics, but we should just imagine big box stores that kind of— [00:44:53.09] Garrett Langley Well, this wasn't— [00:44:53.15] John Collison These Walmart stores— [00:44:56.06] Garrett Langley We have a Fortune 5 company as a customer, and they spend like $100,000 a year. Because they have like three locations in the country. That's it. They have like three really big campuses, so they're actually like, we don't proxy towards necessarily like market cap or revenue. We proxy to physical locations, like how many locations? Someone like a Dollar General is a way better prospect for us because Dollar General has 7,000 stores. [00:45:18.18] John Collison Yeah. You want Subway. [00:45:20.15] Garrett Langley Subway is a good example. We tend to focus on retail, healthcare, and logistics. They tend to have big physical footprints, a lot of employees, and a lot of challenges. [00:45:33.18] John Collison Is their main challenge theft? That's what they're worried about, that they want to prevent stolen cars coming to steal stuff? [00:45:43.09] Garrett Langley I'd say employees is… I'd say three years ago when we started the business unit, it was assets. It was like, "Oh, my gosh. Coming out of COVID, everyone's stealing everything." It is much more shifted now to keeping employees safe. You think about some of these businesses, they might terminate half a million people a year, a quarter million people a year. It takes one angry employee to come back, and so in our system, you can have it automated with HRS that when the employee is terminated, they're added to a localized hot list. If that employee ever comes back on campus—it's not illegal, but it notifies security. This is why people have security teams. Don't let them in the building. It's about kind of moving that layer of safety farther out. The other example is we do quite a bit of work in executive protection. I'll give you a good example. If you think like a Fortune 2,000 CEO, our stuff's deployed at her house. We also deploy at corporate. It's not illegal, but it's quite weird if the same vehicle on the same day goes to both locations. [00:46:43.07] John Collison That is not that person. [00:46:44.06] Garrett Langley Yeah, it's not that person or their EP team. I think keeping people safe is the number one thing. The assets are like, it's good, but worse case, you'll just raise prices, which is not good. But we had this crazy case with one of our healthcare partners where this group, pretty smart actually, would show up to the hospital dressed in a certain company's uniform, and be like, "Oh, the robotic surgical arm needs to be repaired. Can you help me grab it?" and some clinical surgeons were like, "Oh, yeah, let me show you where it is. No worries, no worries, John." They walk in there. This is a multi-million dollar piece of equipment, and they literally walk out with it. This happened in— [00:47:29.16] John Collison It's some Thomas Crown Affair kind of stuff. [00:47:31.11] Garrett Langley But it's like, I mean, you're a physician. You're not thinking like, "Oh, is John actually working at this company? Is the product… I think I used it yesterday." You're like, "Oh, yeah, that sounds right. It's a super expensive product." This is an example to your conversation of where the business gets interesting is we then connected that healthcare provider with federal authorities. This isn't going to be a local authority to solve this kind of crime. It's a federal crime at this point. They were taking the product, exporting it, and selling it in a different country for kind of clinical work. We think that one is pretty interesting. That's an asset one, where actually the healthcare system is like, "This is $20-something-million of product stolen." That's a problem. [00:48:08.18] John Collison Is your business going international? Which countries is this best suited to? Is this a universally applicable product? Do those differences in law enforcement agency structure make it... I'm just curious. [00:48:24.23] Garrett Langley We're dabbling, but we think payments is probably a very global, everyone wants to process payments, I think. [00:48:35.17] John Collison Yes, but there's a lot of local nuance. [00:48:37.08] Garrett Langley Yeah. I think for us, maybe our ambitions aren't big enough, but when I look at the domestic opportunity, we should be able to get to five, 10, 15 billion of revenue in America, and we're not there. The difference of, to your point on the nuance in payments, it's probably comparable to the nuance of working with local government. But then you overlay hardware, and that just gets maybe two steps too difficult for my stomach today. [00:49:07.07] John Collison Got it. [00:49:07.21] Garrett Langley But there's, I mean, we get an inbound... I saw an email this morning from an Australian police department. It's like, "We'd love to do a demo," and it's like… [00:49:16.16] John Collison Why not just do… I mean, I agree there's a lot of nuance. Is there some aspect of your tech… You've put a lot of work into the cameras, vehicle recognition, and stuff like that. I know, again, from painful experience that localization is never easy and never as easy as you think. [00:49:32.11] Garrett Langley Yeah, here's… [00:49:33.12] John Collison But maybe it's easy. [00:49:34.13] Garrett Langley Well, so here's the argument: We pursued one really big deal last year in Mexico, and we had support from one of our investors, like, "We should go after..." And they had connections to the Mexican government. We got all the way to the finish line, and it was us versus Hikvision. Hikvision is a Chinese camera manufacturer, arguably a subsidiary of the CCP. We were almost 10 times the price, and it came down to like, well, is the Department of Homeland Security going to subsidize this for the Mexican government? Because we know their listed price is way more than this. It's very clear this is being subsidized for the Mexican government to make this purchase decision, of which case they will most likely be giving away real-time feeds to another government, and we didn't win that deal. I think that if you look at... There's a great book on this topic. If you look at what China was able to do in Africa with their infrastructure deployments on connectivity, they own those countries. [00:50:42.12] John Collison What's the book? [00:50:44.18] Garrett Langley I'll find it for you. [00:50:45.11] John Collison Okay, we'll pop it in. [00:50:46.23] Garrett Langley I'll find the book. But the book pretty much shows that one of the maybe smartest moves that China did was going to these developing countries and saying, "We will give you 5G, really low cost. We'll bond it for 100 years, so we're not giving away for free," but guess what now? Chinese government has access to your pipes. Maybe that's not scary for some countries. For me, I'd be a little afraid. That was our pitch to the Mexican government: "You don't look at just cost; look at the sovereignty of the data, who has access, and we'll domicile this data in your country. We won't domicile it in America." They were like, "Sorry, it's too expensive," and we've seen that. We've done a couple other projects like that, not just picking on Mexico, but other countries where we're not competitive with China, not because our products aren't better. I think they're much better, but we're not being subsidized by the federal government. [00:51:36.09] John Collison Presumably NATO countries, Australia? [00:51:40.07] Garrett Langley Should be better, but I mean, no. I mean, we should go walk around a strip mall and I could probably point out 80% of the cameras are Chinese-manufactured, and your average would-be business owner is like, "Yeah, I got it for 15 bucks on Amazon." I don't know if you know that story on the vacuum cleaners, but someone hacked into it, it's like… You should care about where your data is stored and who's storing it, and that kind of stuff matters to me, but not everyone. [00:52:12.20] John Collison That makes sense. [00:52:16.08] John Collison As we've learned from Flock Safety, fighting crime is most effective with network data on your side. The same is true for payment fraud. Most businesses only see their own transactions, which makes it very hard to spot patterns and prevent future attacks. But Stripe Radar tackles fraud using the power of the Stripe network. Our machine learning models train on hundreds of billions of data points across the $2 trillion of payments we see on the Stripe network each year. When new fraud patterns emerge, our models can quickly learn and start blocking similar attacks immediately. So if you want to fight fraud like Flock Safety fights crime, check out Stripe Radar. [00:52:54.07] John Collison As I think about your competitive landscape in the US and companies that sell to law enforcement, I think about Axon—which people would associate with making the body cameras, but they make lots of it and Tasers, I think, but they make lots of other stuff—and then Motorola Solutions, which has kind of the radios and the get-up in the cars and things like that. Are you three the main three players? Are there others I should be thinking about, and how will it... Presumably you guys compete over time? [00:53:22.00] Garrett Langley Yeah. I remember in 2020, trying to raise our Series B, and every investor came to the same conclusion. There were three strikes, you're out. Strike one, you're based in Atlanta. Now, post-COVID, that's become less of a problem. But strike one, you're in Atlanta. Strike two, you're doing hardware. That's really bad. Hardware is really expensive. [00:53:44.02] John Collison But now it's the only business with terminal value. [00:53:46.21] Garrett Langley I would argue that AI is not going to replace cameras or dig holes. It's like, we mean, a third of our employees dig holes for a living. That's just AI is a long ways away from, I think, replacing that. [00:53:57.11] John Collison Sorry, what are they digging the holes for? [00:53:58.13] Garrett Langley Most of the time, you want a camera where you don't have any infrastructure. [00:54:02.08] Garrett Langley We actually show up. [00:54:04.00] Garrett Langley We dig the hole, we trench it, we pour concrete, put our pole up. I mean, I might be— [00:54:08.15] John Collison Okay, so it's on your own pole. I assumed it's just mounted on an existing pole. [00:54:11.18] Garrett Langley I should pull this data. I would not be surprised if I'm the largest general contractor in America. We pulled 77 permits a day last year, and if you've ever built a house, you know how difficult it is to permit something. Seventy-seven permits a day, like crazy scale. [00:54:26.11] John Collison Yes. [00:54:27.00] Garrett Langley AI is not going to replace that. It's like a very safe asset. But when we were trying to raise that Series B, the third strike was... Second strike was hardware. Third strike was, and you're trying to sell to the government? The last company to go public was Axon. They went public in 2005, which would have been 15 years prior. In the VC speak, it's like, "Well, if you can't get big enough to go public, you're not worth investing in." It's like, well, crap. Luckily, Garry Tan shared my beliefs. Safety should be a public right, it shouldn't be a privilege. It should be a right if you live in America. He did our Series B which is, I think, the only Series B he's ever done. [00:55:03.21] John Collison That's awesome. [00:55:04.18] Garrett Langley He was right, and I was right, so it's the three of us now. It's Motorola, who's the biggest. They're about a $90 billion market cap, 120 years old. Invented the radio, still are the… I think they have 80% market share globally of landmass radio. It's a crazy scale. [00:55:18.22] John Collison It's funny how durable some of these businesses, Garmin and other, like you just start doing GPS, you start doing radio, and you just stay doing radio. [00:55:25.20] Garrett Langley Yeah. Own it. And to Motorola's credit, they've driven a ton of innovation on the radio itself, similar to Garmin, which you'd thought Garmin would have been crushed, and it's like they just… [00:55:34.09] John Collison They've been so creative in creating verticals, you know. Aviation, all this stuff. [00:55:37.20] Garrett Langley Awesome products. [00:55:38.11] John Collison Sports. [00:55:39.12] Garrett Langley And no one would debate whether Motorola's radios work really well. When you need them, which would be like a natural disaster, they work especially well. That's like pretty compelling when your job is to respond to natural disasters. [00:55:53.16] John Collison Totally. [00:55:53.19] Garrett Langley So you have Motorola. Every single thing we do, we compete with Motorola. That's fun. They're big. Second one would be Axon. You know, I said public in 2005, about a $40 billion market cap. [00:56:05.18] John Collison Where did they come out of? I understand where Motorola came out of. [00:56:08.12] Garrett Langley Yeah, so Rick Smith is the founder. Graduated Harvard. He plays it off, he's a very smart guy. Graduated Harvard, found this Taser company. It wasn't called Taser at the time. [00:56:22.05] John Collison Okay, so they started with Tasers. [00:56:23.12] Garrett Langley Yeah, found the IP, the product, and then tried to commercialize it. If you love a good founding story, it's a good one where they almost went out of business. His dad mortgaged the house. He mortgaged the house, went public because they couldn't finance it. I think they went out at like a $50 million market cap. This is like when people went public much earlier, and just has done a really good job of growing the business in the public markets. [00:56:51.03] John Collison And Taser is a brand. It's like Kleenex. They own it… They started as Taser. [00:56:55.03] Garrett Langley They started as Taser and rebranded to Axon in 2012 or 2015? [00:57:01.18] John Collison Yeah. [00:57:01.21] Garrett Langley And got into body cams via acquisition. They bought a company that had just won the contract for the New York Police Department to do body cams. So bought that. They have a dash camera, and they'd recently launched a competitive product with me. Now we compete. Everything we do we compete with Axon too, which is fun. [00:57:19.12] John Collison So it's just a Mexican standoff? [00:57:21.03] Garrett Langley Every single city. And then what's interesting… But your question was also in the future. There's this really interesting phenomena happening now where because Flock's been somewhat successful, VCs have poured in. When I used to call a police department 7 years ago— [00:57:38.17] John Collison Ah, now you have a bunch of competitors from all over the world. [00:57:41.09] Garrett Langley No one would… You could literally just walk into a police department and be like, "I'm going to meet the chief." They're like, "Oh, that's great." Now chiefs are like, "I have 17 different people calling me for the exact same price." Now they don't compete with us, thankfully, but you think about trying to deploy AI to monitor body cam footage. There's seven different companies with VC backing doing that. It's not a big enough market. I think your annual report was great, by the way. You all are at what, 1% of GDP or something, probably? Okay, I'm in 60 to 70% of the cities that matter. [00:58:15.14] John Collison Yes. [00:58:16.13] Garrett Langley There's just not enough space for seven competitors. The market's not big enough, and we're not creating new cities. There are new businesses created every single day. [00:58:22.23] John Collison I think you tend to see structurally in these smaller… I mean, it's a large market in total, but where there's a finite universe of buyers, the distributional advantages are very powerful. [00:58:35.21] Garrett Langley I think so. I think what we'll see is… I see a new VC-backed company every day. I'm like, there's just all these things are going out of business, so they're just going to be consolidation. You see, I think Motorola has done 40 acquisitions in the last two years. Axon did five last year. We did one last year. I just think you're going to see… [00:58:55.02] John Collison Everyone in this space is going to consolidate. [00:58:56.02] Garrett Langley A ton of consolidation. Because once someone builds a good product, you go, "Great, I already have a sales force. I have the customer base. We'll just pick you up." [00:59:02.16] John Collison Yeah. I had no idea about the… You're talking about where the tech goes. I had no idea about two things actually, that I learned here. One is that you're doing real-time analysis of 911 calls. Is that being fed to an LLM basically, or like 911 calls now scored by an LLM? [00:59:20.01] Garrett Langley The way I think about it is we are trying to build an orchestration layer for a city's safety. The majority of police departments are understaffed. I think the worst I've heard recently, I was at a higher ed college campus, a large college campus, a football school, they're at 40% of staffing. Could you imagine if tomorrow you woke up and Stripe had 40% staff? [00:59:45.02] John Collison It was 40% staff. Yeah, it'd be tough. [00:59:47.14] Garrett Langley You would figure it out, but a lot of things that you would deem core today would just stop. [00:59:52.04] John Collison Yes. [00:59:52.23] Garrett Langley That's how most of our customers operate, which is not sustainable. [00:59:57.10] John Collison Yes. [00:59:57.19] Garrett Langley We view our job as being that force multiplier, or I like to think of this orchestration layer where every manual thing that's done by a human should be automated. The difference in public safety versus maybe payments is like, I think you want humans at the start and finish still. If I call 911, still want a person to pick up. [01:00:17.13] John Collison Those dispatchers are incredibly well-trained. [01:00:19.16] Garrett Langley They're very well-trained, they're very calm. Where I think it gets valuable is like, but let's say there's a mass event. AI is way better because if you have a surge of demand, and you go from only one call a minute to ten calls a minute, AI is better than a busy tone. That makes sense. Then I think about for us, we call it amplified intelligence. When that 911 call comes in, LLM is able to pick up the call, is able to determine what are the characteristics. [01:00:43.04] Garrett Langley Can I go start to build an investigation? Can I go pull this up so that if you're the detective... Whereas historically you walk in and go, "All right, I'm going to go get the call transcript, I'm going to go look up my record system, I'm going to go do a bunch of analysis." It's the equivalent of having... Whether you use something like a Glean or another product that's like, if you're going to do a lot of the hard work, so that as a sales rep, you show up, and you're like, "Got it." We do that for investigators. They show up to... They get delivered a case and a lot of the busy work has been done for them so they can use the human part of their job. [01:01:11.10] John Collison Aggregating all the data they need that— [01:01:13.05] Garrett Langley All the different data, like let's go do this. Like we solved, one of the cooler cases we've solved was armored trucks, the Brink's, they move money around. It's one of the best, it's called a jugging case. It's very popular in Texas, where you follow these armored trucks, and you typically have a shooter on a nearby building. When the person walks out, the shooter shoots him, the other guy runs in, grabs all the money and runs away. It's a very, it's obviously sad, but it's a very, very profitable thing. We built an agent that tracks armored cars. At all times, it is literally saying, "Okay, here's an armored truck. If we ever see other vehicles tracking this car, flag them automatically." Now, it's not illegal. [01:01:58.05] John Collison Yeah, sure. [01:01:58.18] Garrett Langley Right? But it is enough that if you're a city like Houston who has a jugging problem, you want to get a subscription to… You let me know whenever you see someone doing that because I want to know. [01:02:07.18] John Collison Yeah. [01:02:08.03] Garrett Langley I want to get a step ahead. [01:02:09.17] John Collison Yes. [01:02:09.20] Garrett Langley There's always types of crimes that you can start to say, "Okay, this is helping officers not do something new. It is just doing things that when you're at 50%, 40% staffing..." Sure, at 100% you would have been doing this yourself. You'd be checking the cars, you'd be in Flock, you'd be running searches, but you can't do that anymore. [01:02:25.03] John Collison What you're describing is again, having this data to be able to do anomaly detection just allows for new police work that otherwise wouldn't have gotten done. [01:02:37.11] Garrett Langley No, it's people's lives, you know. [01:02:41.14] John Collison Then the other thing I hadn't realized you guys were doing is the drone assistance where, again, this sounds to me, obviously already it's the case that many police departments have a helicopter and then for very serious crimes, they will task it. But that's expensive, very limited resource. I think what you're doing is allowing for some manner of air support to just be much more cost-effective and available to more officers. [01:03:06.21] Garrett Langley That's right, yeah. I think it's another example where there's just a lot of tasks we ask law enforcement officers to conduct that a drone could do faster, more effectively, and cheaper. If it's calling out a hit-and-run, great, send the drone. I was working with a town in Tennessee, it's a good city, their average response time to 911 calls seven and a half minutes. Their drone from us gets there in 68 seconds. It's like a better quality of service. The incremental cost is less than the cost of one single officer. It's really nice when you see technology deliver both a 10x better product at a dramatically lower cost. That's how it's supposed to work, and it's working for them. [01:03:54.11] John Collison What kind of task or intervention are the drones best suited for? [01:04:00.19] Garrett Langley There's three primary use cases. The first would be vehicular pursuits. It's like one of the most dangerous things cities conduct is high-speed pursuits. Typically, it's not the suspect that dies, it's some random person. For most of our towns that have adopted our drone program, they end pursuits, they don't pursue anymore. Send the drone. [01:04:20.07] John Collison Oh, wow. [01:04:20.14] Garrett Langley Send the drone. Way better. It's way safer. That's a— [01:04:25.08] John Collison S you don't have like a police car running at 80 miles an hour down a residential street. [01:04:28.17] Garrett Langley No, you just have a drone 400 feet up in the air. Quietly, safely, waiting for the car to pull into a gas station, pull into their home, pull in somewhere safe, get to a red light. Then meanwhile that video feed's being broadcasted to the entire police department. It's like, great, John's over here, block him in. [01:04:45.05] John Collison Every pursuit goes to GTA level five immediately. [01:04:49.03] Garrett Langley Yeah, if you're smart. That's a big one. Second is 911 calls. What's been interesting is that, we'll send the drone first. It takes something like Elk Grove, up in Northern California. They've got a bunch of our drones... [01:05:03.21] John Collison Where is Elk Grove? [01:05:05.14] Garrett Langley Elk Grove? It's a suburb of Sacramento. [01:05:09.08] John Collison Oh, okay. [01:05:09.12] Garrett Langley But maybe 75,000 people. I say suburb, it's a pretty big town. But so they'll dispatch for 911. Majority of the calls actually never need a human to show up. The example I'd give you is like, you see a fist fight, you call 911, and then what happens is historically like seven minutes later someone shows up, guess what? They're no longer fighting. But now we've dispatched this officer, he's like, "Well, I'm in the area, maybe I'll grab a Gatorade, walk around, check it out." 30 minutes later, like we've wasted a bunch of time. [01:05:40.06] Garrett Langley You send the drone. In the best case, you see the guy's fighting, and you're like, great, I'm gonna keep an eye on them and allow the officer to do his job. Or there's not a fight anymore, you dismiss the call. You actually decrease the response time for situations you really do need to go to by removing the junk in the system. That's the second one. The third one is search and rescues. Not every city has a helicopter. Cities that do is a very expensive proposition. Sadly, people go missing all the time, and you can pop up the drone, throw on thermals, surely it's at night, and find the person. [01:06:16.01] John Collison Actually, I haven't thought of that. Drone plus thermal camera is very transformative for SAR. [01:06:23.13] Garrett Langley No, yeah, we had an interesting case in a cold state right now who's flying our drone. There was a, sorry, a lot of my cases involve homicides. I'm going to find a better topic than homicides at some point. [01:06:38.03] John Collison I have a better topic for you in a second. [01:06:39.07] Garrett Langley Okay, great. There's this 911. Car on the side of the street, launch the drone, throw in thermals, actually see where the person went. Sensitive enough to see the heat pattern, like there's like a dip in the snow. Want to find the guy, find him and everything. But it's interesting, to your point, I think we're still very early on in the use case exploration. I think it's just the type of technology that until it's fully proliferated, which give it two or three more years, I think we'll continue to find more ways to augment how we respond. I think, I guess the last example I'll give you is we typically launch the drone for all first responders. It's not just law enforcement, it's fire, it's EMS. The whole community gets advantage of it. [01:07:23.02] John Collison Okay, my final example, because I agree a lot of the crimes are quite heavy topics. I feel like one of the most satisfying genres of YouTube video to watch is people who laser aircraft, but they are mistakenly lasering a police helicopter. This is an insane crime that happens. [01:07:44.03] Garrett Langley Wait, why do they do that? [01:07:44.23] John Collison Okay, so there's this insane crime that happens. [01:07:47.06] Garrett Langley Isn't that bad? It's, like, very dangerous. [01:07:48.08] John Collison It's very bad. It's very dangerous. But bored people, to your point of just maybe people going mad during COVID and being cooped up, people just for fun, laser aircrafts. It's very dangerous. It's dangerous for anyone's eyes, turning a laser into them. But, like, a pilot who is flying a plane full of people at that moment, it's especially bad. But it's a real problem that happened. You listen to ATC recordings all the time, like "Airliner going into LAX getting lasered by someone on the ground." People buy these lasers on Amazon. [01:08:11.18] Garrett Langley Yeah, they're very powerful lasers. [01:08:13.18] John Collison Exactly. They're very powerful, and they're just lasering the cockpits of aircraft. But occasionally what happens is they're lasering an aircraft, but it turns out that aircraft is the police helicopter. [01:08:22.14] Garrett Langley They're just like, "Gotcha!" [01:08:23.09] John Collison They got the thermal, and they're just… You get to watch the whole thing unfold. There's a bunch of these on YouTube, where you'd see the squad cars coming up, and it's, yeah. [01:08:30.14] Garrett Langley I'll have to check that out. [01:08:31.09] John Collison But it's such an odd crime, and it's very satisfying to see them caught in real time. [01:08:38.10] Garrett Langley Yes. That's a really weird... The other interesting thing, because I know you like aviation stuff. We're seeing more and more police helicopters have to turn off ADS-B as well. Because criminals have the same data that you have. [01:08:51.09] John Collison Yeah, sure. [01:08:52.06] Garrett Langley Most of the police helicopters actually fly without ADS-B now, which is like a whole challenge. [01:08:58.01] John Collison Sure. [01:08:58.04] Garrett Langley From a separation point of view. [01:09:00.16] John Collison I have so many more things to go into or jumping around, but I like this. How's the business evolved? You're now, you said around 500 million in ARR selling to both law enforcement agencies and corporates. Have there been interesting changes in how you monetize? Is it just a question of scaling up? [01:09:22.11] Garrett Langley I mean, I'd say the biggest challenge is two, three years ago we were single product, single customer. We had our neighborhood business. It was growing 20-30% year over year, but it was just operating. Law enforcement was going really, really fast. We had one product and then maybe made a mistake. I know RJ from Rivian was here. Similar story, probably built too many products for too many customers really quickly. In hardware, that's really expensive. Hardware tends to follow this J-curve of huge CapEx investment upfront to get the thing going, and then you monetize, and it actually winds up being— [01:10:01.06] John Collison And there's very significant scale economies at very high orders of magnitude. [01:10:05.16] Garrett Langley Yes. We went from one camera that tracks cars to a camera that's focused for people, a drone, trailer, multiple customer segments, and that's really hard. [01:10:20.11] John Collison Looking back on it, you think you went too broad too quickly? [01:10:22.23] Garrett Langley Yeah. I would have better— [01:10:25.03] John Collison Did you discontinue products and stuff? [01:10:27.20] Garrett Langley No, we muscled through it. [01:10:31.05] John Collison But you slowed down the rate of adding products. [01:10:33.14] Garrett Langley We were like, we cannot do any new hardware products this year. We need to take a year or two off of hardware products. We can debate software products because there's no incremental burn or cash outlay for it. But we know what the J-curve looks like. We saw it in the core business. The core business now is profitable, which is great. We'll generate free cash flow, hundreds of millions of dollars of operating cash flow this year. But those new businesses are effectively Flock five years ago. We know how painful that is, but we were just like, "Oh, it'll be great, it'll be fine." [01:11:04.14] Garrett Langley I think that's been challenging for us. Is how do you balance from a product roadmap servicing two customers with just very different use cases? Like your average Amazon distribution facility is 30 acres, 50 acres. That's like a neighborhood block. [01:11:22.00] John Collison Yeah, yeah, yeah. [01:11:23.01] Garrett Langley Then you've got, like, San Francisco that's got hundreds of square miles. It's the problems are different. That is probably like the hardest thing that we're still trying to muscle through that. Like how do you organize your company to service these two different customers without having redundancies? I haven't figured that out. [01:11:39.03] John Collison What have you learned other than to don't make too many products? What have you learned by building hardware? Or what have you learned about building hardware? [01:11:48.06] Garrett Langley Yeah, I mean, a couple of things that I would jump to, the first would be, I don't know how you guys think about forecasting demand. It's a full-time profession at Flock. [01:12:01.22] John Collison Sure. [01:12:02.12] Garrett Langley It typically needs to be 12 to 18 months out. [01:12:04.14] John Collison Yeah, yeah. [01:12:05.18] Garrett Langley If you look at the hardware companies that don't make it, that's actually where they fail. [01:12:09.06] John Collison Yeah. No, we ended up with, I mean, our hardware business is a very small part of our overall business but we wildly overproduced at one point. [01:12:16.15] Garrett Langley You're just sitting in a warehouse like, there's all my money. [01:12:19.04] John Collison Exactly. [01:12:19.17] Garrett Langley Sitting in a warehouse. [01:12:20.10] John Collison Sitting around. [01:12:21.00] Garrett Langley Luckily, it's not... It's not going to go bad. It's not like bananas. It's like, "Oh, gosh. We have like a week to move this product." [01:12:26.11] John Collison Yes. But it kind of goes bad. [01:12:28.20] Garrett Langley It does go bad at some point. I think I remember, during YC many, many, years ago, talking to Eric at Pebble of how it was like the irony that in the best year ever of the company in terms of revenue is the year they went out of business. I'm like, that's crazy. [01:12:46.10] John Collison Yes. [01:12:47.03] Garrett Langley But they just, they overproduced in Q4, even though it was still a record quarter, it just still wasn't enough. I think that has been amplified by also, our distribution process, which we're full first-party. We not only design the stuff, build the stuff, we install the stuff. I have to have forecasting— [01:13:07.20] John Collison Maintain it, presumably. [01:13:08.16] Garrett Langley Maintain it as well. I have to have forecasting not just at the product level, but at the geographic level by product, with some level of discrepancy. I don't need to know this city versus that city, but at least the general area. I think forecasting, that's been pretty hard. The second is, every decision you make in hardware is millions of dollars at a minimal and often tens of millions of dollars. When you grow up in this Silicon Valley mindset of like, everything's a two-way door. [01:13:35.18] John Collison Yeah. [01:13:36.10] Garrett Langley Bullshit. Hardware, everything's a one-way door. [01:13:38.11] John Collison Yeah. [01:13:39.13] Garrett Langley You want to pick that part? [01:13:40.09] John Collison Yeah. [01:13:40.23] Garrett Langley One-way door. [01:13:41.14] John Collison We're gonna live with it for the next five to ten years. [01:13:43.18] Garrett Langley Yeah, and like, I don't know, you probably don't track this as much, but probably the dumbest financial mistake I've made in the last year is our supply-chain team came to us like six months ago, and they were like, solid state memory is getting really expensive. We had this one part that was in our cameras and the price had gone up 4x in near time, which indicates that typically one much, much larger company placed a massive order. We've seen this time and time again where an Apple or a Sony or some consumer, Samsung, will pick a part for a new product that's coming out in 12 months. The supply globally disappears. [01:14:19.22] John Collison You're feeling the AI data center buildout in your supply chain? [01:14:24.06] Garrett Langley Yes, and so our supply chain leader was like, "This is getting crazy." I was like, "Well, who do we use?" I go, you use SanDisk. I'm looking at their stock, and I'm like, should we be buying like SanDisk? I started my CFO, I was like, "Should we be buying SanDisk stock?" I was like, their prices are going up 3x. It means that like a lot of people are buying a lot of products. Of course, the stock's up like 1200%, and I feel like an idiot for not like following that conviction. But no, yeah, I mean, our BOM, now luckily it's a small part of our BOM. [01:14:46.14] John Collison Yeah. [01:14:47.04] Garrett Langley But we have a full team of people who all they do is mitigate global supply chain risk. Because parts just disappear. [01:14:55.08] John Collison Just getting the products into the hands of customers as already specced is nontrivial. [01:15:01.08] Garrett Langley What we wind up doing, which I think a lot of companies do, I'm not sure if we're special, but it's just like you don't think about this is when we look at a BOM of a product, we'll risk purchase not necessarily the whole thing, but the cheapest, highest risk things. We should have bought... We did, we bought a ton of memory so that we didn't run out because we can't— [01:15:19.08] John Collison But I would have thought, often these supply chain crunches come on the leading edge where everyone's fighting over TSMC, three nanometer node production capability. But the auto chips, at much larger gate sizes are not, or larger nodes, are not as contended. I would have thought that you guys are not using tippity-top end cameras and lenses and memories and things like that. But that's not a panacea. [01:15:51.16] Garrett Langley It's like capacitors. The point I give you is, let's say we ship 100,000 units. That's tiny. All it takes is, if we need this capacitor from Texas Instruments, all it takes is Apple saying, well, the iPhone 27 doesn't have that capacitor, and Apple's like, "Cool, we'll buy everything for the next three years." Then we go, "Oh, crap, we've got to completely change our strategy." [01:16:17.17] John Collison But I'm surprised you're competing with Apple for components. I would have thought their capacitor— [01:16:23.17] Garrett Langley Well, capacitor is a capacitor. But so it's actually know the capacitor you want. You have to have a supply chain to buy it. You have to have enough inventory with enough lead time. What you wind up doing is early on, your designs are very simple. You buy everything from out of fruit or whatever. You buy everything from easy places where you can buy 50 at a time. Then, when you move to tens of thousands at a time, you start having to have designs that have four different derivatives for every part so that your supply chain doesn't have to call engineering being like, this part's out, they know these parts are all sub-ins. [01:16:55.01] John Collison You're describing the scale diseconomies of manufacturing where it's easier to buy three of something than it is to buy 30,000. [01:17:00.10] Garrett Langley 100%. I never would have thought we would need a team of people who all they do is spend money for a living. Just buy stuff. [01:17:07.04] John Collison What's hard about operating the hardware? How do you keep the lenses clean on the cameras? Just how do you… I don't know what the other— [01:17:13.21] Garrett Langley It's like on the one hand, we're some of the best weather forecasters. It's like we keep track of every major storm. It's like a storm, we need to be back up. We do, we have a pretty cool, we call it our flight team, and there are technicians that only fly. It's like that big storm in New York and Boston, they were flying into the storm. We had surge demand to fix stuff. [01:17:37.19] John Collison To be on site, when they get a call that something is down, go repair it. [01:17:42.09] Garrett Langley Yeah, well so everything's, we have really good telemetry on the equipment, so if a customer calls, something's gone, we've really screwed up. [01:17:49.03] John Collison Yeah, you know before the customer knows. [01:17:50.04] Garrett Langley We know before the customer knows, and most of the maintenance at this point is fairly predictive. We know this mechanical part malfunctions after between 100 and 200,000 uses. If you're in the area, it's ironic that the largest cost structure in replacing equipment is the driving. It's driving. If we think a part needs a replacement in six months, and we're nearby, it's cheaper just to replace it, refurb it, and get it back in the field. That part's been like, we had to build a software company, a hardware company, and a field services business. [01:18:22.00] John Collison Do you have parts that wear out? I would have thought the whole thing is fairly solid state. [01:18:26.06] Garrett Langley Most of it's… Well, drones obviously— [01:18:28.02] John Collison Sure. Different kettle of fish. [01:18:29.18] Garrett Langley But on the camera side, there's one part, which is the IR-cut filter. When we operate at night, we operate on infrared, and you need a different filter so that you don't have pink images during the day. That literally changes twice a day. [01:18:45.03] John Collison Oh, like something mechanically— [01:18:47.12] Garrett Langley Put a lens over it. [01:18:48.19] John Collison I see. [01:18:49.18] Garrett Langley After a couple of hundred thousand uses... [01:18:51.14] John Collison That's the one moving part. [01:18:52.19] Garrett Langley It is the one, and it is the only part that breaks. [01:18:55.11] John Collison Why not just have two lenses? [01:18:57.20] Garrett Langley With two different image sensors too? [01:18:59.04] John Collison Yeah. [01:18:59.22] Garrett Langley BOM. [01:19:00.13] John Collison Okay. [01:19:01.03] Garrett Langley No good reason. [01:19:03.16] John Collison Okay. [01:19:03.20] Garrett Langley Other than it would just be more expensive. [01:19:05.09] John Collison Yeah. [01:19:06.00] Garrett Langley But yeah, that part's, I think a third of the company, like I said, digs holes, drives bucket trucks. [01:19:13.00] John Collison Yeah, sure. [01:19:13.16] Garrett Langley Keeps track of all the inventory. But if you look at it, I didn't want to build that business. But early on in the company, there was this horrific case in Atlanta where a woman was just running in our version of Mission Dolores, or pick your nice park in the city, random act of violence. There's cameras everywhere in the park. [01:19:34.19] John Collison Yep. [01:19:35.19] Garrett Langley None of them were working, and the city got blasted appropriately for it. They're like, "Well, but also, do we really want our police department being in charge of camera uptime?" That's a dumb idea. We should just pay a company, like Flock in this case, so just make sure the stuff always works. It's not my favorite part of the business in that sense of like, it's a lot of stress. It's very operationally intense, but I think it's valuable to our customers. [01:20:01.09] John Collison Yeah. How do you think about the right... When you talk about cameras in the park, many movies center on this idea of universal surveillance in the Bourne Identity or something. They have cameras on absolutely everything, or I guess the Bourne Supremacy, I think, is more the Waterloo Station scene and same with lots of other things. How do you create the right guardrails once you move off roads, and into parks and public spaces and kind of creating access controls around that? [01:20:33.11] Garrett Langley Yeah. The cop-out answer is like, I don't want to be in charge of deciding that. Thankfully, we have elected officials who we vote to make that decision. [01:20:43.07] John Collison But what do you think a sensible place for them to land is? [01:20:45.18] Garrett Langley I think it's way higher than we have now. I think that for me at least, for every crime that occurs that doesn't get solved, it means we didn't have enough cameras. That's, to me, the easiest rubric. Now, I think to your question though, it's a question of where they are, who has access to them, and I think it's one of the few cases where the disparity between maybe my knowledge and your knowledge of how the technology works and how much official work is pretty big, and their dreams of how it might work versus actually how it works, it's much less sophisticated than they think or dream up. It's just a camera. [01:21:29.17] Garrett Langley For me at least, we should have an abundance of cameras and have incredibly restrictive controls of how and when they're used. And the example I'd give you is like today, everything in Flock just you can generally do. There's data retention where we protect how long something is stored. For live video, it's typically 7 days. For LPR data, it's typically 30 days. You can go longer or shorter if a democratically elected body votes on it. But I would challenge, why not do more cameras and have a warrant restriction? Like why not do 30 days of LPR data, but if you have a warrant, you have a year. You will need less cameras. [01:22:08.18] Garrett Langley I do think there are some nuanced ways to do it, and thankfully, we have a really good government affairs team that is lobbying for that kind of legislation to say like, "There's a way for us to both be safe and maintain civil liberties, and it needs to be legislated though." I can't… Those ideas are not in line with what my customers want. They're law enforcement customers, they just want to go catch bad guys. They want to follow the constitution, catch bad guys. But that nuance in the middle of like, but what is societally acceptable today really belongs in your elected officials to make that decision so that it can change, and so we pushed them to be like, let's legislate this now before it becomes a problem. [01:22:51.19] John Collison Where do you think has passed sensible rules? [01:22:55.21] Garrett Langley Sensible. I think Virginia's bill last year was pretty good. It defined... It did a few things well, and one thing I don't agree with. What it did well is it defined a modest data retention period of 21 days. I think that's fine. I like 30, but tomato, tomato, it's fine. I think the ACLU was lobbying for 3 minutes. It's a little tough. It's hard to swallow. I think seven, 14, 20-something days is enough. There's a trade-off there. [01:23:29.04] Garrett Langley They mandated formal auditing, which I think is great. Not enough of our customers audit themselves on a regular basis. We can build software to make that easier, but we need to be pushed to do that. Customers don't want it. They need to be told to do it, so I think that was good. It also validated that this could only be used for criminal investigations, which I think is really good. While that's obvious, it's helpful to write it in law. [01:23:51.23] Garrett Langley I think the only thing that I disagree with is they did say effectively there's no participation with the federal government. I think that's just… It's their choice. I think it's their choice, and that's the beauty of the country, is like Virginia should do what feels right for Virginia. But I worry about the types of cases that you don't want to read about on the news that tend to get solved by the US Marshals, the DEA, the ATF, the FBI, and like they can't use our technology in Virginia. Which is like I said, I live in Georgia, so thankfully it doesn't really impact me. But as a business, I'm like, I don't know if that's right, but I'm more than anything just happy they passed some legislation. [01:24:34.22] Garrett Langley New Mexico passed a similar bill this year. California has a similar-ish bill. I think a couple other states have... I think the worst type of bill is actually not whether it's 14 days or 30 days of data retention. To me, the worst bill is an unenforceable bill. So you can imagine a bill that's like, this product cannot be used for possession of marijuana. I'm like, "Who's going to enforce it?" [01:25:05.12] John Collison It's unknowable. [01:25:06.07] Garrett Langley You might believe that, and that's great. But someone has to enforce this, and actually what's going to happen is no one's going to enforce it now, and that's, I think, really bad law in those cases. [01:25:15.08] John Collison Yeah. That makes sense. I'm curious what your view is on police department procurement. What do they do? What do they buy not enough of? What do they buy too much of? They don't feel like they're swimming in procurement dollars. But yeah, I'm curious. [01:25:36.07] Garrett Langley Yeah. I mean, the thing they buy too much of is, and not to pick on Motorola, things that only matter in the 0.001% case. It's like, guys, why don't you just use cell phones? Like, well, in a natural disaster, and it's like, "Got it, okay." How often, so you look at a landmass radio contract, and it's going to cost something like San Francisco County $200 million. [01:26:03.23] John Collison You're talking about that kind of money? [01:26:05.21] Garrett Langley On a TCV basis, oh yeah. [01:26:08.02] John Collison Wow. [01:26:08.12] Garrett Langley I think... Motorola is a really big company. [01:26:11.04] John Collison Sure. Yeah. That's how? [01:26:13.13] Garrett Langley Seriously, you don't get there... And you go, "Well, I get it. If there's an earthquake and every single cell tower goes down, law enforcement definitely needs a way to communicate." Man, is that really the only way? [01:26:25.23] John Collison Yeah. [01:26:27.07] Garrett Langley Could you guys just have your own tower? There's gotta be something. But you look into it— [01:26:31.22] John Collison Is this a "your margin is my opportunity" situation for you guys? [01:26:35.09] Garrett Langley Maybe, because law enforcement's unique in this case in which they really also like that they own the infrastructure. They own that bandwidth, and so they can do whatever they want with it. I don't think they do anything interesting with it, but they can. Versus if they're riding on Verizon or AT&T, they have no control over that. [01:26:52.06] John Collison Why don't you guys just have the exact same products like radio in the spectrum? [01:26:56.03] Garrett Langley We probably should. Yeah. But we got to— [01:26:58.11] John Collison That's my… It's like a VC in the boardroom— [01:27:02.23] Garrett Langley You should go to an $80 billion company that's been in business for 130 years. Like, I mean, I guess you did it in payments, and it worked out pretty well. But yeah, so I think like that part, I think, is really, really kind of crazy. At least in the business world, I feel like I buy for my majority case, and I deal with the ramifications of the edge case or build around it, and they do that. [01:27:21.08] Garrett Langley I think procurement is exceptionally slow and exceptionally laborious. Everything goes to RFP, but the RFP is written for one vendor, so all it does is wind up taking an extra year, six months. I get why RFPs exist, but they're not actually RFPs. I've never seen an RFP that's not written for one vendor. I'm sure it exists somewhere. [01:27:44.10] John Collison Yeah. [01:27:46.05] Garrett Langley I haven't seen it. But what I think they do well is one of the things that I think the government did figure out is maybe you need to have an RFP for a $200 million contract. What about a $10,000 contract? Great. So they do have spin levels where if you're a police chief, you can go spend $25,000 or $50,000, and not have to go through the entire process. You know, it's tough. I would not wish upon anyone selling to local government. It's more negatives than positives in a lot of ways. [01:28:14.13] John Collison Yeah, and again, that procurement process has evolved for a reason, and to protect against certain other failure modes. But yeah— [01:28:21.05] Garrett Langley Well, I mean, you think about it in business, and I know you. We use Stripe. I'm sure you have like six competitors, I don't know. But I know you, so I'm just going to buy Stripe. That in the business world is considered very normal. In the government world, that is called illegal. Which is really... It's really interesting. [01:28:41.08] John Collison But again, it's evolved for a reason. The rules are written in blood. [01:28:44.18] Garrett Langley Well, no, and then like you have plenty of cases where it's gone sideways, but it is just interesting dynamic. For so many things that we take for granted as normal business practice are definitively illegal when procuring with government. [01:28:56.08] John Collison How do you guys use Stripe? [01:29:00.12] Garrett Langley A lot of our private-sector customers pay either via credit card or check. We have a lot of checks. [01:29:05.09] John Collison Okay, so for the public sector stuff, that'll generally be via a torturous RFP process, NPO and something like that, but for your private sector— [01:29:18.14] Garrett Langley But even, I mean, we don't want to be in the check deposit business. [01:29:20.16] John Collison Sure. So you use Stripe for the check functionality? No one knows about our check functionality that Stripe can accept checks for you. [01:29:26.23] Garrett Langley Am I allowed to talk about it? [01:29:28.23] John Collison No, no one knows because we're bad at marketing. [01:29:30.19] Garrett Langley I was just like, oh. [01:29:31.16] John Collison Not because it's a secret. [01:29:32.19] Garrett Langley No, yeah, no. It's like we don't want to be in that business. So it's a remote deposit box, or whatever it's called. But yeah, that's right. [01:29:39.14] John Collison We can give you an address that you can give to your customers, and they can mail checks to it, and we will turn it into digital money, and the fact that a bunch of atoms and an envelope going through the postal system were involved, you can forget about those details. [01:29:53.06] Garrett Langley Not a single customer of ours pays via ACH. It's either check or credit card, and we do a lot of checks. I'd probably say 80%, 90% are checks, maybe higher. [01:30:01.10] John Collison You're one of the few tech companies to mostly use Stripe for checks. [01:30:06.06] Garrett Langley Just trying to buck the trend every day of the week. [01:30:09.08] John Collison That's really funny. I love that. You were talking about police departments maybe being over fixated on kind of the 0.1% cases. Does that apply to... A common critique that you hear of police department procurement is the sort of militarization of police departments, and you know what we really need is a BearCat for this town of 20,000 people, which is like an armored personnel carrier. [01:30:34.21] Garrett Langley It's like a $3 million vehicle, and it's like, why couldn't we just get an F-150? [01:30:37.22] John Collison So do you think that applies also to kind of the shiny stuff? [01:30:40.16] Garrett Langley Oh, yeah. I mean you look at like... Early on when we were building the company, we'd get the question of, can your license plate reader work on a car going 175 miles an hour? I was like, "Probably not. I've never driven that fast. I'd have to rent a runway to go test this." They're like, "It's really important." I'm like, "Really? How often does that happen?" They're like, "In a high-speed pursuit, people drive very fast." I'm like, in a high-speed pursuit, you know who they are. We were criticized, and had to build a product. Until we got to like 120, 150, it was a major blocker to sales. [01:31:16.05] John Collison Really? [01:31:18.07] Garrett Langley I mean, it's like we built a camera that, I mean, we tested it on roads that we drove on, so we'd get up to like 80 or 90 and it worked fine. But we had to eventually, we rented an amateur racing track and just drove around in circles at 120 miles. We emailed the employee base and were like, who owns a car that goes really fast? That was actually kind of funny because a lot of people were like— [01:31:38.18] John Collison What car do you use to test this? [01:31:40.22] Garrett Langley There's some fast cars. Plenty of Teslas, Rivians, some other nicer cars. They drive really fast, and that was like a fun day. It was like, put the cameras up and do it. But that's not the performance. [01:31:50.20] John Collison Did the cameras work out of the box or did you have to tune the model performance? [01:31:53.13] Garrett Langley It's fine. Yeah, the only place where we had to make one modification, which is kind of interesting. We deploy a radar. So on really busy roads, where our angle of incident is particularly tight, we can't shoot super far down. We have to shoot at a sharp angle. We have a radar attachment that we tilt backwards to notify the cars on the way, like get ready, which helps. But that's a very far edge case. But the camera doesn't actually— [01:32:19.22] John Collison You prime the camera essentially? [01:32:21.11] Garrett Langley Yes, to get ready for a car coming. [01:32:23.23] John Collison I'm sorry. [01:32:24.21] Garrett Langley Because the camera is offline unless there's a vehicle. [01:32:28.21] John Collison I see so you booted it up, and that's a power-saving measure? [01:32:34.05] Garrett Langley Yeah. The most expensive thing we do is take a picture. The second most expensive thing we do is send stuff to the cloud. Third is just like being a computer turned on, and so it's similar to your iPhone turning your screen off. [01:32:46.20] John Collison Okay. [01:32:47.15] Garrett Langley So we added the button to take a photo right away. [01:32:49.17] John Collison The continuously running radar is very low power. [01:32:52.20] Garrett Langley It's negligible. [01:32:54.08] John Collison I see. That's cool. [01:32:55.15] Garrett Langley Yeah, that was a fun one. [01:32:56.18] John Collison We were talking about hardware previously, but what's building your own drones been like? [01:33:02.18] Garrett Langley A lot of fun. [01:33:04.14] John Collison Sounds fun. [01:33:05.05] Garrett Langley No, it's fun because like… I have kids and you have a kid. It's really fun to build a product that your kids understand. It's like we drive around Atlanta and my son will count the cameras from home to school, or we'll go to the airport, or we'll go to the park. I love that he can actually understand what dad does. [01:33:29.15] John Collison Yes. [01:33:30.16] Garrett Langley I'm not sure if you've taught your son like how to— [01:33:32.08] John Collison I'm screwed on that. [01:33:33.10] Garrett Langley Yeah. It's like, yeah. "We process the world's payments." And so when you show him a drone, he's like, "Oh, this is so cool." Obviously, he has like a little miniature drone. [01:33:43.15] John Collison Well, also, you don't only do drones, but you do drones to catch bad guys. That's the very early age. [01:33:50.06] Garrett Langley Yeah, you get like, yeah, to be clear, he'll watch PAW Patrol, and be like, "That's what dad builds. That's the helicopter." Well, we don't put any people in it. But I think what's been fun is we made a hypothesis that if you've studied planes, I think the most interesting plane, military-wise, is the Warthog. Because whereas traditionally, you built a plane, what kind of missiles can we add? They were like, "No, no, let's design the best missile. It's very precise, very big, and then we'll figure out how to fly it." [01:34:18.20] Garrett Langley Our thesis was, "Well, we're really good at cameras. Drone is really just a camera that flies. Let's build the best payload and then figure out how to fly it." I could show you the payload the next time you're in Atlanta, or we could fly one out here. It's like the coolest camera ever. It's huge, it's like this big, and it's got four different image sensors, maybe six, different optical lenses. We can read a license plate almost a mile away. Like crazy specs, great thermal, then we're going to get to fly. [01:34:51.01] Garrett Langley I don't know, I was an electrical engineer, one of my co-founders was a mechanical engineer. It's fun to build things that fly. So it's like, all right, we have a payload, we have these airframes, we have aeronautical engineers now. It's been fun to grow the engineering team. [01:35:03.09] Garrett Langley An imaging team, you think about the dock that it lives in, it's effectively a commercial grade HVAC system. The drone lands, it needs to be cold, it needs to be hot, it has to charge. If you know anything about lithium-ion, Lithium doesn't like to be too hot or too cold, so you've got to keep that well-conditioned. It's like this huge compressor, it's this massive thing that opens and closes, and if it's snowing, if it's frozen, all of these engineering problems, and that's the fun part of this stuff. [01:35:26.17] John Collison Yes. [01:35:27.14] Garrett Langley Everything else is... Selling is good, but building stuff's fun. [01:35:30.21] John Collison How many drones do you have out there in the world? [01:35:32.23] Garrett Langley We don't disclose that. [01:35:34.12] John Collison But decent number? [01:35:35.10] Garrett Langley Yeah, no, it's hundreds of cities that are flying. [01:35:37.22] John Collison They are flying drones. It's funny when you talk about the A-10 Warthog, you're reminding me of the Boyd book by Robert Coram, which I only read recently, but it's like one of those kind of Silicon Valley, and everyone talks about it, and everyone talks about him in the context of his OODA Loop. You know, orient, something, decide, act. But I think that's actually kind of overrated. [01:36:04.06] John Collison Actually, the main reason Boyd is interesting is helping the Pentagon procure better planes. I'm just reminded of that with the Warthog and also, you know, your description of needing to run at 170 miles an hour. Because basically he came into a Pentagon that had a bunch of bad planes because all the generals were just obsessed with specs. [01:36:21.15] Garrett Langley Yes. [01:36:22.02] John Collison They wanted, a high top speed and, like, they really judged planes on specs, and actually, you know, the fighter pilot joke is that there are only two throttle settings in a dogfight: maximum full military power or throttles idle. Those are the only two energy states that you're in, and what matters is maneuverability and the ability to add energy or lose energy quickly. Anyway, he was involved in basically all of the good planes that were produced, including the A-10 Warthog, because he got them out of the mindset of just kind of speeds and feeds. [01:36:54.14] Garrett Langley Well, that's for us. The spec we track is time on scene. And so, one of the reasons why we care so much for the payload is like, if you have a payload that can see really far away, you don't have to fly there. Actually, you get there faster virtually. Which is what, if this was a drone that was carrying a payload, like a Zipline or something else, like actually physically in there matters, but for us, it's just about time on virtual scene. [01:37:15.18] John Collison Yes. [01:37:16.08] Garrett Langley That's what we measure ourselves to. That's why we fly high, because physics allows you to see farther. That's why we have these huge payloads. You don't have to actually fly there, and then you don't have to fly as fast, which means you can conserve battery life, which means you're in the air longer. All of these designs were around that use case versus like, I'm sure Zipline went through a whole separate use case of like, you know, what's the max payload and all that kind of stuff. [01:37:36.04] John Collison How many drones does a city the size of San Francisco need? [01:37:40.03] Garrett Langley Twelve. [01:37:40.22] John Collison Okay, so that's a very small number. [01:37:42.16] Garrett Langley No, because our average drone can cover a 30 square mile radius and get there within under a minute, and thankfully, you know, knock on wood, there actually aren't that many 911 calls that merit a response, so we look at it both ways. We look at it in terms of geography and then 911 density. So in more rural parts, you need less drones because there's less call for service, and then in more dense urban areas, you need more drones, mostly from a volume of service. But even in our most dense customers, it's pretty rare that they fly two drones at once. It happens, but it's like... I think our busiest drone is being flown 90 hours a week. That's a lot of flight time. [01:38:22.22] John Collison Sorry, do you dispatch the drone when it's needed from its charging dock, or is the idea that it's out there flying already and you just task it? [01:38:28.15] Garrett Langley Yeah, it lives in the dock. There's debate of whether a drone should be in the air at all times. [01:38:36.20] John Collison Does that actually save you meaningful time? [01:38:39.09] Garrett Langley Being in the air? [01:38:40.05] John Collison Yeah. [01:38:40.11] Garrett Langley It would save a lot of time. [01:38:41.04] John Collison Oh, really? [01:38:42.01] Garrett Langley Yeah, but you look at the Carpenter case in Baltimore, and they had an airplane with a very powerful camera, 24 hours a day, and that was deemed like an unwarranted search, so it got killed. And so we're very conscious, like we have other part of our business that is interesting now at scales, like we have a full team of constitutional attorneys, and I'm sure you have like a regulatory team that when you want to build something, they're like, let's check it before we ship. We have a constitutional team that's like, cool idea, let's actually make sure this doesn't violate the constitution. [01:39:13.23] John Collison Let's just look up the Fourth Amendment here real quick. [01:39:15.10] Garrett Langley Yeah, let's just like double-check this thing, and so when you look at the drone, like we believe, and this hasn't been tested in court, but these are smart people. They're like, look, it's unclear how that would end in court. But if you call 911, there is a reason to fly the drone. If there's a gunshot, if there is a gunshot detection, if there is a stolen car, that is a reason to dispatch versus just flying around looking for stuff. It's not unconstitutional, but we would not push that as a use case. [01:39:44.18] John Collison I'm sorry, like a lot of this precedent, as new technologies come along, people reason by analogy of like, a car is somewhat like your house, but somewhat different, whatever. Isn't a drone flying around just like a police cruiser on its patrol? [01:40:01.10] Garrett Langley So the... And I'm not an attorney. Hasn't been the case yet. [01:40:05.02] John Collison Neither am I. But, it's never stopped me yet. [01:40:06.16] Garrett Langley The other analogy that would be the butterfly effect, which is like when things are much, much cheaper and much, much easier, historical precedent gets thrown away, and so you take the helicopter example. You could say, "Oh, well, helicopters fly sometimes." Yeah, but helicopters are so expensive, it's not practical to have 24/7 aerial coverage. [01:40:23.15] Garrett Langley With a drone, it's not impractical. It's the same reason why when we launch our drone, we want to go from the launch location to the in location, the camera's pointed at the horizon the whole time. We don't want to look in your backyard. That's like... We feel really strongly, and that's not a law, that's just our point of view of where the law should be, so we'll build a product for that. Then, if the operator wants to tilt down, that's their control. But as a default, we're flying straight on. [01:40:51.03] John Collison That's very interesting. Last question. You guys have grown with cameras out there in cities, now getting into drones, building the software OS to help law enforcement agencies, and others kind of synthesize all the information they have. Just what comes next? What future product ideas are you playing with? Where do you want to go? [01:41:11.13] Garrett Langley So, if I think about it, we talked about this earlier, failure for Flock is prison population goes up. It's actually like really bad. We look at, the products today are very much focused in the middle of a crime. A crime has already happened, and therefore we should solve it, and that's really good, and I think we're definitely not done, but we've done a lot of work in that category. I get pretty interested in expanding that and going, "Well, what about what can we be doing from a product perspective to prevent crime from happening?" [01:41:45.12] Garrett Langley That actually doesn't necessarily look like software. It's like one of the interesting things that we started last year, so we call our Thriving Cities Fund. It's probably in an analogy similar to your Stripe Press. Which is like it's never going to be the core of your business, but you feel really good that it's a part of your business, and so when we go in places like Greenville, Mississippi, we also commit to deploy capital as growth partners to those businesses. Because if we want to convince that 16-year-old to not be a criminal, there does need to be jobs. Jobs that like a 16-year-old can get, and so we deploy capital in restaurants, nail salons, like pick your business that you can be 16 and work at easily, and we want more of those to exist. [01:42:26.02] Garrett Langley Last year, I think we were at 21% IRR, so it's not a bad business. I think Flock's stocks done a lot better than 21%, but I feel really strongly we could deploy hundreds of millions, if not billions of dollars of capital in the cities that also choose to be safe. I don't want to go deploy capital in a place that doesn't want to be safe. But for that, I want to do more there. [01:42:46.06] Garrett Langley Then, I think to our conversation on the other half, the majority of the crime we solve is not violent. It's non-violent, and today, the discrepancy for a juvenile and non-juvenile, you still wind up in some type of penitentiary or prison system. I think that's crazy. All the data shows, as soon as you wind up in prison, you're going to get violent, and you're going to come back. [01:43:12.01] Garrett Langley So, I would articulate there is an opportunity, don't know yet what it is, to say, "Well, hold on. If this was an opportunistic criminal, is there a product—with a capital P, because it might not be software, it might be hardware or software, I don't know—that allows that person to have a second chance?" And in a way that is not going to increase their likelihood of doing it again. Like, that's bad. But like, prison can't be the answer. It just doesn't work. [01:43:43.07] Garrett Langley The whole concept of prison will never work, and there are some incredibly well-run prisons with really well-intentioned wardens doing the best of their ability, but the concept of putting a bunch of violent people together is by default flawed, and so I question what could Flock be doing to say, "I'm going to prevent kids from becoming criminals, and if you do wind up on that path, how do I get you back on track as fast as possible?" [01:44:07.04] Garrett Langley We're a for-profit business, so I'm not looking to be a non-profit. But I think there is something there. I mean, you're talking about millions and millions of people who really need... It's actually as a society in our best interest to get them back in and productive, and I want to do both of those. [01:44:23.08] John Collison So, fewer crimes, fewer people in prisons? [01:44:25.07] Garrett Langley Yeah. That's the end goal. [01:44:27.19] John Collison Thank you. [01:44:28.08] Garrett Langley Thank you. It was fun. [01:44:29.04] John Collison It was awesome. [01:44:29.18] Garrett Langley Yeah.