#5: Garrett McCurrach - Underground Hyperlogistics Delivery Drones

Alright, welcome. Welcome back to another episode of the First Principles podcast. Today we're joined by Garrett. Garrett is the CEO of PipeDream. Why don't you tell us about what PipeDream Yeah, so PipeDream is the easiest, fastest, and

highest volume way to move things from one spot to another. So whether that's underground or through buildings or whatever it is like it's really easy to move from one spot to another spot to another spot to another spot so kind of like a train so if you thought you think about like a really small subway for things that's Heck yeah so it's like the simple like the way to get people's minds around why it's useful is basically like there are tons of

things that you want to move from place to place you want to move groceries from the grocery store into your house you want to move package from the reception desk up to your office. I think people don't really think about that a lot, don't think about that logistics stuff and how things actually move around in the world, but actually this just

occurred to me. I just remembered that one of the biggest things that helped skyscrapers come to be, skyscrapers really couldn't have existed either before the elevator, which everybody knows, or the phone. It was really hard to move literal messages up and down a large distance. So, until they figured out the logistics of moving a message up and down, they couldn't build skyscrapers. So, I don't know. Oh, that's so cool. I didn't know that. What's the next type of logistics? What else is it

going to bring about? Why don't you talk a little bit about why this is helpful? What sorts of things people want to move around and where you're Um, so Pipe Tree was really born out of, you know, we were looking for something that had 10 years of innovation to go in it, something that, Jeff Bezos has this great quote, uh, my quote is not as good, but

essentially like, you gotta know where the puck is going. Um, in like AI, you could have been working on the old AI paradigm for like, five, 10 years, and all of a sudden Transformers comes out, and you're at the same starting place everyone else is, you're zagging with everybody else. And so we wanted to find something that had 10 years of innovation left in it, meaningfully changed everyone's lives, and was really easy to know where the

puck is going. Logistics is easy, we just need to move more stuff around faster for less money. And that is always true. It's always gonna be that. So we really just had this frustration with how long it was taking for autonomous logistics to scale. And so just dug into it and realized that There are a few reasons, but one of the main reasons was how hard it was to get something in and out of the building. If you think about humans delivering things, it is

expensive to deliver things for a variety of reasons. Main one being that for last mile delivery, we're usually using this giant car to deliver this super small package, but also because the world is super complicated. And where you are, in what building, and then if you're in an apartment, if you're in a house, trying to figure out, okay, where do I actually put this thing for this person is super hard. Same thing with going from warehouse out

into the world. That's super complicated. And then going longer distances, having to wait for batched payloads to fill up an 18-wheeler or fill up a delivery truck. Everything is really locked by having to have a human be

part of that. And autonomous logistics is just, the amount it's going to change the way that we live our lives is drastic. And we're so close, like drones are there. Sidewalk robots are there. Self-driving cars are there. We're so close. We're just like, okay, what is it gonna take? What is gonna be needed to take it across the finish line? And that is the thing that we should be building. And then we can get into

it. Our goal is to make autonomous logistics happen as fast as possible and then make this next state of logistics happen right after that. But that was ultimately why So what, how much of this was informed by like the crazy stuff that's happening in, you know, factory automation? Like if you look at, or maybe even just warehouse automation, like if you look at Amazon, there's not, they're not picking

and packing with people anymore. Like they're, I'm sure that that's all like robots and you see those cool like vertical moving robots that go get something off the shelf or whatever. Like, so how much of what you guys were doing was informed by that versus by, you know, Yeah. I mean, we really had to come at it from

the other way. I think if you look at, especially in robotics, if you try to use too much of something to inform how to build an automated system, then you don't realize just how much minutiae and nuance is put into that robot. Like the Kiva system that Amazon uses, it's So sick. And you look at it and you're like, oh, so simple. It's such a good idea. You have this thing that lifts a shelf and it carries that shelf over to somewhere else and it drops it and then they can, you know, kind of like

drive around underneath. It's such a simple concept, but there's so much that goes into that. Like those floors have to be so level that if they're coming to retrofit a warehouse, that floor they already know isn't level enough. And to get that floor level enough takes like, it's this huge engineering effort to make it like smooth and

level enough for the Kiva robots to run. So you push these robots really simple, but you add a little complexity in the building process and making sure that you've reduced variables in the work environment. So usually like, and Canon are my CTOs, so good at this. He's so good at this. It's just, you need to look at the problem set, and you need to look at the variables that go into it, and you need to look at, okay, what can we control? What can we not control? What

are the most important things to optimize for? And then you build the simplest robot that meets all those requirements. And that is the best way, like, if you didn't introduce any other Like trying to make this thing like that thing or this thing like that thing, then you've introduced bias into the creation process. So it's really a from first principles approach, which is ultimately how we got to Pipestream in general. We

didn't ever start the company and be like, hey, we should do underground delivery. That would be... We came at it from first principles and said, what is the best thing to

build? And then after a series of Yeah. So what were those kind of constraints? What were the Yeah, so it was really what we told ourselves is we want to make autonomous delivery work in every city as fast as possible. That's number one. So we want to do that. And then number two is we want to make this next stable logistics we called hyper logistics happen as soon as possible after that. So when we do our monthly meeting, monthly meeting starts with saying, you

know, we want to make hyper logistics possible by 2030. That is our number one goal outside of any tech or anything that we build trying to get to there. So hyper-logistics, just really quick, is the ability to get something delivered in under 10 minutes for less than a quarter, and just as easily as you can receive that object, you can send it back. So bi-directional, up-down,

ability to get something delivered and really easily send it back. So we're looking at that problem set and just started talking to cities and companies in the space, just different stakeholders. Originally thought drones. I'm gotta be the biggest drone fan in the world. I love drones, always been obsessed with them, done several

drone projects. So kind of like went in with that bias. And then after talking to people, realized that there was this missing space that no one was building in, that if we went and built it, that was going to be the multiplier that would allow autonomous logistics to scale throughout cities. So ultimately it was that, plus we saw it as this really, really, really important piece to making Hyperlogistics happen by 2030. And without it, we would be so far behind in making

that happen. So for us, it was that spot in that we knew no one was building in right now. It was so big that we knew if we got three competitors, we probably should be friends with them because it would take all of us to actually make it actually happen. And What were the reasons, were there like physics reasons or I mean, drones are a super important part of the autonomous delivery. I mean, they're gonna be, I think they're gonna be the lion's share

of deliveries are gonna be done by drones. They're like perfect. There's so much space up there. There's just so much. They're quick. I think people overestimate how much noise there is. I mean, cars make a ton of noise, and we ultimately just kind of like, let that one go. The safety protocols are great. If you look at, I mean, that's ultimately what is taking the longest on them, is making sure everything's safe. But like, the FAA does a great job of making sure That things

are crazy. I mean, Boeing is, I mean, they've lost a door, they've been on fire, and there's been no deaths. It's because they just have an insane amount of safety protocols. They can handle these crazy events happening and everyone is safe, which is wild. But they're gonna be a huge part. It is just a really tough space to be in. It's one of those things that when you get to a service that is that cheap, it really starts to be a commodity unless you have something crazy unique that

you're providing. So if you think about like the internet of goods, At times, vehicles, as they get more and more efficient, ultimately just become a bit. And it's kind of, you know, whatever is carrying that bit doesn't really matter. It's going to be sufficiently fast and sufficiently good. You've got to really, really have an edge to really compete. I think that when we were looking at it, it was like, well, we were getting such a late start. It

just didn't see us catching up with anyone else. It's also, like, and then we were looking at it, and there are so many other problems that aren't drone-related that we're gonna take, it was like, a ton of people are working in the drone space.

Not a lot of people are working on the rest of Is the problem so big that you'll just have to have as many possible solutions working on it as possible? Or do you feel like there's a specific niche that you guys can do that maybe drones Yeah, I think it's both. If you look at global logistics, there's just so many different things to

deliver, and there's so many different geographies that it needs to be delivered for. There's so many different climate and weather conditions and entry and exit points. It's just a really complicated space. I mean, like, sidewalk robots have a spot in that. Self-driving cars have a spot in that. Drones. Underground, I think, has a spot in that. And then there's probably like two or three other modalities that just we haven't even heard of yet. They're going to be super important to

doing that. I think when you look at our space, it's what we really focus on is how do you move around a building really well. So if you kind of think of like plumbing doesn't end at a building, you have the main line and then you have the auxiliary lines and then you have It's ultimately got

to get to the sink, right? You need that point of delivery and then you need a really efficient way, you know, horrible example here, but like if I'm using a toilet, you think a toilet is like a really efficient poop delivery method, right? It's delivering it away from you to a Yeah, yeah. But like you need, your poop start originates in like one point at

a toilet, and then it needs to make it to the main line. It's got to make it efficiently, and then it's got to like be able to go on the main line all the way to the processing Where does poop go? Answering the big questions on the first I'm this close to flushing a GoPro. We're going to crack this wide open. But yeah, that and then same thing with warehouse. If you think about, okay, you have right now 10,000 deliveries that

need to go out this hour. Being able to get all those deliveries into the vehicle that needs to take them to where they need to go is about half of the time it takes to actually get that thing to the end user. So it's a huge coordination problem. And then same thing on the other side. Accepting a delivery in an apartment building, getting it to the end user, going from warehouse to in a city, moving around district. I think there's a lot of different problems and I don't think right now We know

how big each of those is going to be in what places. We know there will be problems. And so our goal is to take the

ones that are going to be the biggest problems, focus on those, and then let Sweet. All right, let's hop into the product. Let's talk about

what you're actually building, like how you're building it and what led you there. So do you want to lead us through the idea maze a little bit of like, you know, how you started, how you got, how you kind of like got closer to a prototype or like a point design, really, and then moved on to a prototype, like Yeah, so once we kind of landed on that idea space, there were a few truths we needed to come up with. So the number one one was for any type of utility like infrastructure to work, it's

got to be as cheap as possible per linear foot to install. So whether you're whether it's Greenfield or otherwise, like The only utilities that actually scaled were the ones that you were able to retrofit into existing locations. So it had to be able to be retrofitted. It had to be cheap. And the way to think about it is like, yeah, how much does it cost per linear foot

to install? So it had to be as cheap as possible to put in per linear foot. It needed to be able to be serviced without being reinstalled. So anything you put in the ground or you put into a building that is locked, you need to be able to fix it without going down there and fixing it. A good example would be plumbing. I think plumbing is

a great example. When you put plumbing in the walls, it's not very often that you're actually having to tear up the floor of your house and tear out the walls to fix the pipes. There's so many ways to fix pipes, whether you're using a rotoscope, Great way to unclog them. If they break, they break in a way

that you just have to open up one little thing and put them back together. There's like a whole, whole industries have been created to like make plumbing be super simple and super cheap to put in and then make sure that it lasts a long time. So we had to mirror that. And three, a lot of this is really complex. The way you would put this in to, you know, going 30 miles across the city versus the way that you would put this in an apartment building, the way you put this into a

retail space. are going to be super different. And so we needed the simplest way to be able to work in all those different environments. And so the way that we looked at doing it was reducing the core actions to individual components that could all work on their own and didn't need to be tied together. So the idea made there was a lot. We tried to make it too simple, and then we added complexity back in, made it

a little too complex. We kind of ended up in this middle ground. Originally, we were gonna use the pipe walls as the rigid body, and so the wheels were pushing out on the wall. This was like super, super early on, which is like a great idea, and it's out there. I don't know It's like right. Dude, no, it's blurry. Go get it. Is it like a little drone that has the wall, like on the walls or whatever? Like is it? Yeah, yeah. Is it like a triangle? Hold

Hell yeah. All right, this is what's left of it. The engineering team loves taking parts out of it, and I think it's because they hated existing. But this is like the very first one, maybe a couple in. There's three wheels or two wheels? It's three, we lost the top wheel. That was one of the things that got cannibalized. But yeah, so it would flex in and then each one was loaded so that it would keep itself in the center. But trying to do that, it caused, it was so cool

how we made it work. way too complex. The package, because the body of the robot was so close to the wall of the pipe, then the package had to be pooped out. So there's a whole mechanism to poop it out and handle it after that. But yeah, it was not the right answer. And so that's when we introduced a rail. So instead of using the walls as that modular, or

the rigid body, then you just have the rail as a rigid body. So the robot doesn't actually touch the pipe, which is great for a few reasons, but it means that the amount of variables you're dealing with are really reduced. You just have the robot on the rail. The open cavity that it's using doesn't matter. So this is great. It means that the robot that works for underground works the same way in the utility space, in buildings, in the attic of buildings, on

the roof. All of it works the same way and you can fit the enclosure to fit the environment. So that is a great example of Caden and the team doing a good job making things more simple. So we have the robot that handles horizontal movement, it handles all the fast movement, and it is fully autonomous and does all its own wayfinding, navigation, and then it's just riding the rail. And then what we call the portal, handles vertical movement and also handles storage and deployment

into the robots. So it picks stuff up in the robots. We have them in here. So everything works with the Tout. It picks up the Tout. and stores it and then either deploys it to the end Is that the actual size of it? Yeah. Nice. That's big. It's like a huge shopping thing that you pick up at a grocery store. It'd be like a huge one Like a double size. That's exactly what it is. It's a standardized tote. Yeah, Cannon knows so much about totes now. He's got to have

every single tote that exists. Wow. They call them totes. It's a huge industry. It's massive. I think he could look at any tote and rip off the serial number. He just got really into totes. So we didn't mean to, but ultimately what we ended up making was really similar to a port where you have a modular piece that handles stacking, it's able to be grasped really easily, manipulated really easily, that's the

shipping container for it, it's the tote. And then you have the thing that's handling vertical movement and storage, That's the cranes, they grab them off the ships and then they put them in stacks and then they deploy the stacks on the trains or 18 wheelers. And then you have the world's arm movement with those trains or 18 wheelers. And so it's like, did it mean to? But we just kept, Canon team just kept taking complexity out, taking complexity out,

making it more reliable, making it cheaper. And ultimately, that's where we got. And it's just like, it's kind of funny that it mirrors global logistics. Yeah, seriously. Maybe part of me is always worried that there's some inherent bias there. Yeah, but we have gut checked that one about as much as humanly possible. Heck yeah. Well, I'm curious to learn more about the tubes, randomly enough. If you are going to lay a new tube in a new development or a new store

or something, if you're going underground, how do you do that? Do you just trench and then drop? Do you tunnel Yeah, I mean, if it's a new build, it's super easy. You either trench, or when other utilities are being put in, it's already exposed, and you just go drop the pipe in. That's super easy. And then, if you are doing, the retrofit is where all the complexity comes in, but it's also where 90% of the value is. Sure. So, if we're doing a parking lot install into

a building, then you would do a cut and cover. So it's like a super standardized method for putting pipes in, where you cut the asphalt or concrete, you pull the blocks up that you cut, you excavate the ground, pop your pipes in, you put the soil back on, and then you either put the blocks that you took off back on, or you do an asphalt patch, which is just put asphalt back over it. It could be as quick as a two-day process, But

doing trenching or cutting covers are super easy. Once you know where the utilities are and then you have the space to do it, they're pretty quick and painless. Where it gets more complicated is going underneath roads or underneath other infrastructure, and that's where you do a horizontal directional drill, which is, and I think this is where sometimes people think about putting underground infrastructure as being a lot harder, Because you only see it when it interrupts

your life. And you don't see how much is put in around you that's not interrupting your life. So Don't think about a dancing bear? What's that? Oh, so, uh, it may not be a bite. Maybe it's called something else. Uh, this is where staying up of material last night is killing me. But, um, there's these videos you can find online where, There's like the gorilla that walks across and it's like, okay. Yeah, Yeah, yeah, yeah, yeah. So when that dancing bear's in

front of the thing you're watching, you're like, oh, this dancing bear's in my way. But then when it's not interrupting your life, you just don't even see it. But once you know what those machines look like, like I probably crossed two or three of them away in the office this morning. They happen all over the place. They are a little more costly and complicated to work, but they're still pretty simple, a lot simpler

than people think. So the horizontal directional drill goes in like this, so it's kind of like, if you've ever seen a video of a non-invasive surgery, where you come in at an angle, and then you go down, and then you come back up. So it does that in a small bore, and then you grab either the pipe or you grab a bigger drill head, and you go back through the path you already made. So once you have that big enough, then ultimately you grab the pipe and pull the

pipe through. So you can do that as long as a mile. More expensive than trenching, but it's still pretty easy and still doesn't, I think people have this idea of when you put it in a ground structure, you have to like, tear up the roads and it stops traffic in Europe, but the underground infrastructure industry has been away from that for a long time. Every once in a while, you have to actually excavate the ground. For the most part, you're staying off the roads in the median. So

that's how you would get all the way up in there. And then in buildings, we try to stay in the air conditioning space. So wherever you put your air conditioning layer, use that same infrastructure. So like if you're in like a Walmart and you look up and you see these big like stainless steel, like they look like square tubes. And then they're put into the ceiling through this common protocol called thread-all. So you put the thread-all and

it screws into the roof. We just do the same thing. So if you thread-all into the roof and then you have the stainless What I'm envisioning is basically just like, you know, in every action movie, there's a scene where like the protagonist is like crawling through an air vent. I'm just, I'm imagining that's your robots, except they're like delivering packages.

Yeah. That's actually, man, I'm stealing that. Yeah. That's a perfect example. So. Yeah, how it works is that robot scrawls through the space and then ultimately the container drops down Very cool. So how did you choose the bore of, is it called the bore or whatever, the diameter of the Yeah, I think the balance we try to find is we think there needs to be a common protocol. That way, no matter where you're at, if you're going into like, imagine you're putting this into an

apartment building that's gonna be there for like 20 years. You're defining the amount of things that can go in, right? So it needs to be big enough that it's capturing this big standard deviation so that you're never like, I'm trying to think of an analogy, you're never like putting in coaxial and then ultimately you gotta put fiber in. So you want to make sure that it's big enough to handle the stuff that people want to have delivered in

the future. But you don't want to go too big, because you can start to chase that tail, the long tail of all the objects and all the different orientations and the sizes and the everything. And ultimately you get to something that's too big to actually put into cities and put into buildings. So, you know, we've, 18 inches is pretty much there. It's, there's like a few sizes that cities are used to putting in there, the standardized sizes.

So you have 12, 16, 18, 24. Ultimately we landed on 24. Just there were too many grocery and other items and other like standardized ways things are done in logistics that being at 24 just allowed us to kind of like slip into those. And it means that we can come down in the future so we can go back down to 18. 18 is where it's most efficient for the future. At 18, you've captured 99.2% of all Amazon SKUs by volume, by the amount of things that people are

ordering per day. So I think it's the most efficient long term. But I think to fit into today's world, 24, is what gets you enough to fit Okay, let's talk about the drone and the track. So I think one thing that people might be wondering is, okay, what do people know of like a tube that delivers you stuff? Oh, it's the thing at the bank, you know, that you put in the thing, and then you close the door, and it like pneumatic tubes, like poof, and like goes up and delivers to the person

on the other side. I'm sure maybe you thought about pneumatic or maybe you just immediately dismissed it, but why not? Why Yeah, we never thought about it. We didn't think about it until much, like, embarrassingly later. It was because we were coming at this from first principles, and we probably thought about it as like a check down at some point, but it just didn't fit that really tight, set of rules that we were building towards. It is not something that is really cheap per linear foot to

build. It is not something where it is super easy to break and cause a problem throughout the system. So your uptime is dependent on the infrastructure. That is the worst. And ultimately, people tried to put in pneumatics in the 1800s, and there was this big push by the Postmaster General for that to be the way that cities operated, was through pneumatic mail tubes, and they did a huge test.

But ultimately, when they would break, when they would lose pressure, you lost that whole route until someone could dig up the road to go and fix the air pressure leak. And so it is not resilient. And so it broke a lot of our rules. And so I'm sure at some point it crossed someone's mind, but it didn't fit in that set of rules that we're building towards. But it is a really good proxy for thinking about ways

that we can fit into buildings and why it's helpful. I think in like if you look at like a Home Depot or, like, stores that use, like, facility-wide pneumatic tubes, they move a lot of stuff around through those pneumatic tubes, and usually the thing that's holding them back is the size of the payload. So you look at, like, hospitals. Hospitals have smaller payloads for the most part, and so they

get all this use out of their pneumatic tube system. They're able to really quickly send things to, like, oncology or get medicine through the tubes. But in just other industries, you just need a much bigger payload. But it's a really good proxy of like, okay, here's how it would feel to interact with the system. And anyone who's been at a bank, if someone's been to a bank and gone through a bank drive-through, it's a really good Yeah. Um, okay. So tell us about the drone. So why,

um, I'm curious why you decided to make it autonomous for one. I mean, it, it definitely sounds cooler. So I get that for sure. But like, why, why does it need to be able to wifi by itself? Like, why can't you just have an operator that Yeah, well, I mean, if you think about our level of autonomy, it's not that much. I would even call it a controls problem more than an autonomy problem, just

because there's so few variables that we're actually dealing with. We call it autonomous because there's not a better way to say it, but it's a little much. So if you think about having an operator, Our goal is for the system to be essentially free. It's automated end-to-end. There is a drawer where you put something in, and then the cost from getting that thing to the next place is the cost of electricity. the maintenance schedule of the

robot and the cost of installing the original infrastructure. And so we've way overspecced the robots. They are absolute units. So the only thing that is really on a maintenance schedule are the tires and the motors. So those are the two places that we've put, those are our two wear items, just because they're easy to replace, easy to swap. So if you have an operator anywhere in that loop, that's a huge cost. And the effort to get it to be autonomous is

pretty simple. I mean, you have a really defined flow space where there aren't that many decisions to make. The main one is what to do when something goes wrong. So being able to identify that something has gone wrong, to change routes. You don't always have signal. So those robots need to be able to make those decisions on their own. Ultimately, end of the day, they're going forward or backward. Every once in a while, they turn left or right.

So it's not like, it's not a crazy car. Yeah, it's very different than an autonomous car where it's like you have to worry about rain and then like dumb humans that are getting in your way and like, I don't know, the infinite number of things that can happen on a road. If it's a closed environment and literally, I guess, yeah, you are turning, well, okay, here's a question. What is

doing the like load balancing across the network? So like, let's say that we're in the future, there's like thousands of these things flying throughout a city, flying in and out of a particular warehouse. Will the individual robots decide among themselves, basically? Like, you go first? No, I'm taking this path. Don't run into Uh, yeah, so the junctions would be the only place where that's happening.

Um, and then junctions have their own protocol, uh, that operates at a junction because any place we have a junction, um, it is connected, uh, to the cloud. So we'll be able to do, um, uh, real time load balancing. Junction meaning like place where you could turn or like Yeah, or like an intersection. So you have one crossing this Does that imply, then, that all of the tubes are one-ways? There's

never going to be a time when you go back and forth? Yes. Okay, that makes sense. I didn't actually know that, but that makes sense that you wouldn't ever want to run into a time when there Yeah, totally. The amount of space you would need to handle that is at minimum 30% more diameter, and that is super costly diameter. It would mean that we would get 18-inch payloads out of 24-inch pipe. or you have to use 36-inch diameter pipe to get 24-inch payloads. Very tough to handle

there and back. It is cheaper just to put it in a second smaller See, okay, this is my own fault for not thinking of this or whatever, but my mental model was, okay, if I wanna connect Amazon warehouse to my house, there's gonna be one tube that's going that way, and I'll put it down, it'll drop down on Amazon, it'll go to my house, and then it'll literally come back through the same tube. It'll just go the opposite direction

through that tube. That's the way that I thought about it. But so what you're saying is that everything is one way and you're just creating, you know, you have to have a forward link and a return link basically from each house or Yeah, I mean, if you think about it, in practice, you already have the mental model for building this out. It's just like a train. So you want to build a loop as much as possible. You want everything to have kind of like this common flow. It's the most efficient,

least amount of feet of track. It is to do a loop. Only if you have to, do you actually do forward and back right next to each other. So if you think about trains, usually you just see one track. In high volume places, you'll see multiple. But for the most part, when you were to go the long distances, it's just one. And then when you go the longer distances, having an ability to pull off for a part of it, so that you can do bi-directional without putting in extra

pipe. So there's like, you know, it's pretty much like a train. Only difference is you can go top and bottom as well. So you can have, you can put pipes on top of each other. So just one trench, you put those pipes up and down, and then you have the bottom one goes one way, the top one goes the other way. Other That does make sense. So when you're building out those networks, you do probably, like you have to introduce turns eventually.

I mean, you're not just doing straight shots all the way there. Can you talk about that? Like how you would plan a network basically and like the ability for the things to do turns. Can you talk about how that whole planning Yeah, so in longer stretches, we're just following the roads. Because you put in all utilities right next to a road. That's where the easements are. So when you think about a road, every once in a while, you want to turn a hard 90. But it's

usually at an intersection. So for us, if you're building out a larger network, we would say, at a turn, do we need to build out a network to the left? Only, does it need to be a bank and we just put a corner here? Does it need to be able to go straight, left, or right? That would be a junction, so we'd put the ability for it to lane switch and go front, left, or right. And then where that's, neither of those are trail, we just follow the curvature of the road. So the

pipes, if you look at them, they're crazy rigid. They seem super rigid, but in practice they're super malleable, so they'll, both take the form of the cut that you make. So when you're going through it, it looks like it's doing this, because it is. It's super windy. So you can make pretty good bank turns, but there's actual discrete 90 degree corners, we should put in a junction and put in just a hard corner, or you put in a

Interesting. That seems like it would make the track a lot harder. If the thing's moving around a lot and going up, down, left, right, or whatever, what considerations do you have to think about for the track then? Because you want to keep it level, obviously. Yeah. This is where I absolutely, I think Canon is one of the best engineers. Canon and then Thomas is on that team making this decision. I think Mac, who does all our software, was part of this. They are so good at thinking

about problems from first principles. Because I think it would have been really easy to over-engineer into that problem set, which is what a lot of younger engineers do. You see this problem, you go, okay, I need this highly dynamic track that needs to affix to the pipe. And I've got to build this way to then go install it. And it's so fun to think about that. It's so interesting. And it's so much more fun to show people that solution than the actual solution. And so

many engineers get tripped up there. While we came in, Thomas and the rest of the team, they are about solving the problem, and they are so absolved of their own ego, they will just make the simplest thing that works, and they will fight for the simplest thing that works.

So on the track side, you need that flexibility to be able to handle a bunch of turns and swivels, and you also need to be able to, as a train goes, or a robot goes and it hits a trap, it's going to hit it with some momentum as it's doing that swirly thing. So if you try to fix it to the pipe, you're creating a brand new wear item and a brand new place where the track can get dislodged away from that attachment. You're

creating just all these variables. So the way they designed the track, is the track can pivot against itself, so it can flex up and down, left and right, and snake along with the pipe. And then it's not affixed to the pipe at all. So we just slide it through any cavity. There's nothing holding it down on the bottom. the robot and the way that the rail's designed, it naturally finds its level, so if it gets out of whack, it will just, next time a robot comes by, it

will just shift it back over to the middle. So whether there's like, you know, shifting in the ground, and there's a new type of snake, it's dynamic to be able to handle that. So it's just like, that's another example of them taking complexity out, so

there are less things that can go wrong, less things that can break. It is like, if you look at our first prototypes and our current one, it's so much more fun to show people the old ones, because they have all these cool little, cute little engineering things that are so impressive, and then I'm impressed by it, but now if you look at the current system, people almost go, oh, that's It's pretty easy. Yeah, it's like, yeah. It

is cool. It is cool how that works though. It's like, I think, I think that is something that people don't appreciate if they've not been around. A lot of engineers is like, you know, or maybe they've only been around some than like, you know, like kind of like, as you mentioned, people who are relatively early in their career, like building things, like it's so fun to work on that, like 99 to a hundred percentile, you know, like just like getting every single

corner case that could possibly exist, like building the perfect system. Like, especially I think like people are graded that way sometimes in school too, where it's like, okay, did you think through everything? everything. And well, that's not actually what matters. Like when an engineer is trying to put something into the real world and they're trying to do it, like you said, cheaply, quickly,

you know, just solving the problem. Like that's when, okay, it seems like you're cutting corners or something to do something really simple and like really Yeah, totally. I feel like an engineer's work doesn't shine as much. It's almost underappreciated when it is really simple. It's like, well, that didn't take long to figure out how simple that thing is. It's not very complex. I think end of the day, There's two camps

of engineers. There's ones who have actually had to manufacture their designs at scale, and then the ones who've done all their designs on 3D printers. And those are just like, it puts the fear of God in you in putting complexity into Okay, one more chunk of the solution which you haven't talked about yet, which is the portals. So tell me about the portals. How do they work, and Yeah, so a portal, if you think about the train analogy, that's the station. So that is a thing that is onboarding

and offloading goods. That's where storage happens. That's where people are actually interacting with the system. Or that's where a drone is dropping off or picking up from a system or self-driven car, cyber robot. They're kind of like the USB port of the system. The internal mechanisms need to be really similar. So all the portals work on the same internal infrastructure. And so all the parts on the inside are pretty much the same. But we have several portals that all work differently

on the outside. So in some scenarios, someone needs to be able to drive up next to a portal and get their thing in 15 seconds and keep driving, not getting out of the car. So that portal is very different from a high-volume portal where you have a delivery driver who's loading the truck, and he needs to be able to get out of his car and really quickly put in 40 boxes into his truck. So that portal looks different than the other one. And both

of those are handing off to humans. Now you have portals that hand off and pick up from drones. You have ones that hand off and pick up from self-driving cars. You have some that need to be space efficient and exist in someone's apartment. You need to have some that are easy to interact with like in a, like outdoors and the weather and the like. So all these different things are very different about the portal, but they all work on the same common

framework. So it's kind of like you have the, be able to have the similar internals of an iPhone and then adapt them to different sizes. This one has a third camera, this one is just a little longer, this one has a bigger battery. You add those external features that are helpful for people, but for the most part, 90% of the build That's actually a pretty good segue to talk about the different types of products or whatever, like ultimate

solutions that you delivered to customers. And so do you want to tell us about what those are and where you're starting and what sorts of customers you're actually Yeah. So right now we're really focused on what we call instant pickup, which is automating curbside delivery. So if you think about the bank tube, It's basically BingTube for everything. So when you get your groceries ordered and you've done it for pickup, making that a faster experience than going through drive-thru. Getting

that under 30 seconds, you pull up. You get your groceries in 30 seconds and you keep going. Both being able to make that happen really quickly, make it super easy for the user, and then be able to automate that whole workflow so that the store ultimately becomes kind of like a vending machine where you

need a lot less people to maintain it. It's ultimately like if you think about where logistics is going, stores that exist in cities, today they're kind of like, I think if we talk to our kids, in 10, 20 years, we're gonna tell them about grocery stores, and they're gonna be like, they made you work the warehouse to get your stuff? And you'd be like, yeah, I guess so. I guess that's kinda what it was, you know? It's like you're doing

your own pick and pack. You get your little card, and you go around the shelves, and you make your order, and then you have to sign them out of the warehouse, and all that happy, you do all that work. But ultimately, those stores just become really It's kind of like edge computing. Those become really dynamic edge warehouses to get stuff to users really quickly. And

for now, that means we're going to pick them up. And then in five to 10 years, that means those are just going to be interfacing with self-driving cars and drones or going to other places. through trucks or pipes or they're ultimately going to be just this node on a network that have

a good amount of volume close to users. And so anyways, this is the part of that is, you know, as they make that transition, being able to handle their in-house automation, you know, our goal is we have this way that we really help them out today. We help them out as they make some transition to autonomous logistics, and then those become all our supply-side nodes to make hyperlogistics happen. It's a crazy, complicated process, but... It's

super interesting. So that's what we're focused on right now is instant pickup. And then we're also working on the bigger in-city network. So going longer distances in cities from warehouse districts to neighborhoods. creating kind of like a, if you think about like a subway system for a city, but for goods. So being able to put goods a lot closer to the end user so that a human can make that last drive to

the end user or put in a drone or put in a self-driving car. Sidewalk Robot, just making those networks more efficient. And then ultimately, the third step of that is creating a retrofit system for apartment buildings and office buildings and homes so that you can accept deliveries straight up into your living area, into your office, into your apartment. I think that is the one that is the hardest to see the value. And I think the value is low today. So that's

why we're not super focused on it. But

ultimately, that is going to be the thing that changes your life. If you think about if a building had one toilet at the front door, And everyone who was in that building had to go up to the front door and use the toilet or it had one sink in Or you had like one light switch that controlled all the lights of the house That is is you know when autonomous logistics starts to scale that's gonna be the biggest choke point is is being able to make the inner workings of the house, be

able to bring those orders, apartment building or office building to the end user really efficiently. And then that is going to be the most important thing to making hyperlogistics works. If you can, that drawer that you're accepting your delivery from, if you can put something back in it and know that that thing is on its own, going to find its way back to the warehouse through a variety of methods, That's the part that's just gonna change the way commerce works, the way we

live. And so that's the, not to go too much into detail, there's some other pieces in between those, but ultimately that's our current strategy for capture, we call it capturing nodes. Capturing nodes, that's Capturing nodes, yeah. That's like, I mean it's really interesting because So here's a little bit of like an aside or story or whatever. So I used to work at Target headquarters. And when I was at- No way, in Minneapolis? Yeah, in Minneapolis. So I'm from Minneapolis. So

it's like the greatest place to work in downtown Minneapolis. It's awesome. That's awesome. So when I was there though, this was like, I guess this was probably like 10 or more than 10 years ago now. But it was when Amazon was still sort of like, they weren't super Not for you, but when they started that partnership. Target at that time was really interesting because the internal discussions that they were having were not about them as a disperser

of goods. It was not about them as a, we're just going to get people things as cheap as possible. That was not how they talked about it. But it was how Walmart was talking about themselves. At the time, Walmart was the biggest competitor in Target's mind to Target. And Target would talk them down. They would be like, hey, Walmart, you go in the store and there's pallets just sitting there. And they don't even take things out of the shipping containers that they get them in and whatever.

I find that there's a super cool parallel to what you're doing because, you know, the Target mindset is like, okay, we think of the store as like an experience or something, but Walmart totally ate Target's lunch. Like, Target totally failed to go into Canada and expand, for instance. And I think the reason for that is that Walmart was really logistics-oriented. Walmart's thing was,

who cares how it looks or whatever on the shelves of the store? We're just gonna take it in, the thing which ships in, and plop that thing directly down on the shelf. And so that's sort of what you're doing. You're kind of taking it the next step beyond that. The target to Walmart to now Pipetream is that you don't even, who even cares about the store as a place you walk it. It's like a cloud kitchen. It's like you don't care to sit down at the table and see the chef

back there making only one thing. You just want the food out of it. You go to a grocery store, you just want the groceries out of the store. And so it's like, I love that you say capturing the nodes. That's really interesting because you're clearly, obviously, you're already thinking about this as a big logistics challenge and it's like, Okay, what are the highest value nodes? It's like grocery stores and restaurants

That's fascinating. Yep. And then, yeah, so that's in kind of like that first bucket of nodes is high supply, really, really close to end users already. And then your second tier of nodes are warehouses that are outside the city, they're farther away, but super high volumes. a really, really high storage. And then yeah, you just do a bunch of other nodes from there. But yeah,

it is super interesting. I think like the future, and you can already see this now, is like what The real store experience is how good your app is and how good and how easy it is to to get stuff from you. I think there's like five years of that then ultimately I think where we get is is all of that changes because right now. Walmart is beating some places because it is the closest store to you. And Target beats some people because of their selection

and how good the experience is, and that they're close to you. And ultimately, they have about the same SKUs. It's ultimately the experience is putting a modifier on top of the convenience. But convenience is most the reason people are going to Walmart to get their pint of Cherry Garcia Ben & Jerry's ice cream versus Target to get the exact same pint. And I think once things become more automated in delivery, Where you get that from isn't going to be a location. It's going to be an

app experience. It's going to be an API that AI plugs into. And the person who gets to win getting that pint of Cherry Garcia to the end user is the person who can do it the quickest. And so it's a mix of who has it in stock close by, who can quickly deploy that to a self-driving car or drone, what drone can get it there the quickest. All that, it would become a lot more commodified, who

is storing that thing near the user. So it would just become who can get the Pilot of the Chariot of Garcia on the edge of the network, right near the user and then who can quickly deploy it and get to the What do you think is like the hardest problem that you guys are solving right now? I mean, the reason I asked to just maybe motivate the question is like, from the outside perspective, like if you're thinking about that speed problem, like how do you get things

there as fast as possible? The most naive way to answer that question would be build a faster robot. Like, just get it there, just make it like 100 miles an hour underground. Just have that thing whip through those pipes. But I imagine that that's wrong, and I'm curious why it's wrong. And maybe it's right, but I don't think it is. What are the sort of hard problems that you guys are Yeah, right now it's a problem of like, you can get

things fast, but it gets prohibitively expensive. And so the most important thing is, you need to have that kind of cherry Garcia. I've never had cherry Garcia. I don't I appreciate it. My dad used to work at Haagen-Dazs and Ben & Jerry's. So it's like, we Yeah, yeah, yeah. You know what? Do you know Ben & Jerry's is the biggest manufacturer of like DoorDash and GoPuff and like all of them, no matter what company gets you under like 30 minute

delivery, number one SKU has been Jerry's. Brilliant. They Whenever I get DoorDash, they always will recommend ice cream. They'll say, hey, you ordered like, you know, whatever Korean food. Do you also want a pint of ice cream? Like every time. A lot of people do. But if you think about it, it's like kind of the example of where everything goes, where it's like, okay, we used to have ice cream in this big tub, and we get it when we want it.

Now we can get whatever flavor we want in an itemized, smaller thing, and because you're already paying for dinner, it's But no, how we got there was by saying, what are the hardest things in making things go I gotcha. Yeah, I think the hardest thing is when you order that pint of Ben & Jerry's, it needs to already be

in the automated system. So right now, when you order it, it's on the shelf, and then a human has to go and pick that item, and either they're part of DoorDash and they're going into a store, they're picking the item, they're going through the line, they are scanning out the cashier, they're paying with the DoorDash thing, they're getting back in their car, they're pulling out of their parking lot, they are going to your house. So

much time and so much cost. Same thing if you're ordering it from like a Walmart, like, okay, someone's now got to go out and into the aisles and pick that thing and, you know, put it on the shelf and then the driver comes up, someone takes it up to the driver, like, and then your other option is like, okay, maybe there's a microfilament place like, I don't know, wherever there's a warehouse that all of this is like there are humans in the loop to make sure that that thing gets

into the autonomous system. So that's the number one speed and cost thing is that you could already have that in an automated back of house. automated system, then as soon as it's ordered, you can deploy it to be able to be deployed to someone who just drives up and grabs it and drives off. You could have it ready for a drone to come in and pick it up. But the number one thing is make sure that you're able to deploy that object without a human

in the loop. When you're thinking about kind of like five to 10 years from now, that's going to be the number one most important thing. in cost. And so all that is, there's storage density, and the cheaper you can make the Autonomous System, the bigger you can make it in a store and still make the ROI, not 30 years.

There's a lot that goes into that, and really, ultimately, it comes down to how cheaply can you automate the system, and how simple can it be, because the simpler it is, the more dense it can be, then it's also the cheaper it is, and you can put it in more places. Interesting. So I think that's the number one thing. I think robot speed doesn't matter a ton. Yeah. I think just having it be able to deploy quickly, the latency of that So what do you think are the biggest challenges that you

are going to face technically moving forward? Putting aside trying to get people to let you dig up the tunnels and selling to customers, which are probably things that occupy a lot of your attention, but assuming that that is all taken care of, basically, what sort of technical problems do you still see yourself needing to solve before this thing is fully

Yeah, there's a lot. I think that's kind of where we are, not to have a little plug, but the number one thing, this is so obnoxious to say, but I think it's like finding great talent is our number one problem right

now. there we could be doing so much more if you look at like a customer pipeline and the things we could do that we just we can't get to because we're trying to fulfill our customer pipeline like all this looks back down into having great engineers who can build and manufacture those specialty portals who can help us accelerate manufacturing be able to deliver more people be able to scale But there, AutoStore is 20 years old and Ocado is 25 years old, if I remember that number correctly.

Like, huge companies, but then you look at the amount of deployments they have, and it's like a few hundred. And so it's like, oh, these are huge companies. that have massive teams and they've been around for forever. And their deployments are, it's just not that much. Automated systems are really, really hard to solve all the edge cases, to deploy, to make it easy to deploy, to scale up your deployment, to scale up your, you know, you have to have third party

technicians. Like, if you have third-party technicians, now everything has to be designed with third-party technicians in mind, you've got to build out. So there's a lot of, a lot of that comes down to how simply and reliably and creatively engineer things so you're solving those problems ahead of time to allow for that scale. And ultimately that's, That's our bottleneck, which is, we in Canada always say it's like the best bottleneck to have, because working with great

engineers is our favorite. And that's, you know, if you have a company, the more creative engineers you have, the better the company is. So it's a good bottleneck to have, but ultimately it's still like super tough to find that talent. So currently And so it sounds, I mean, it sounds like the drone, the tube, the, what'd you call the basket? You had a fancy name for the basket. Oh,

the tote. The tote. I gotta remember the tote. I love the tote. So you have the tote, you got the drone, you got the pipes, you got the track, you got the, like all of that honestly seems like, you know, it's like sufficient or whatever. It's good. Like you guys have figured out those problems and they're like ready. And it sounds like, The hardest part now is just those terminals, basically like the portals that get it up and down. And getting that integrated into these businesses,

it sounds like the SKU count is way higher for that. These are more custom things for individual customers. It's very interesting to me that that's the constraint. It seems from the ad, if you think about, oh, it's hard about making a underground delivery robot. It's like, oh, it's making the robot and making the track. But no, it actually turns out that it's the interface. It's like the interfaces with Yeah, totally. And you're doing it, you know, we have like pretty aggressive COG

limits. So it needs to be like great. It needs to be a great experience for the end user. And it needs to fit under COGs and it's got to be weatherproof. And someone's got to be able to drive a car into it. And it's gotta be a handful, like all these specifications that companies have for different SKU sets and some need to be like, there's a temperature control, like there are a lot of, it's a super fun problem to solve. But

yeah, that is like a lot of the complexities in that. And then on-edge storage, so being able to, if I'm a picker at Walmart, I need to be able to put what I've just picked into a tote. Right now, they can't store more than an hour's worth of orders because the orders stack up too big and they can't handle that with humans. And they've got to schedule people for the hour. The more you're scheduling people for doing it in the hour, the more

people don't want to use pickup. They want to be able to just pick up any time. Well, you could do that. If you had more automated storage, now you think like, okay, what if drones start really scaling? Well, how are you gonna manage that? That's gonna be way more volume. So it's like all these little things about, and it all comes down to, yeah, like managing a store-to-person or a store-to-robot and then on the other side, robot

So the theme of the show is first principles. Are there like other sort of like fundamental design things that you guys had to account for or like, you know, the single most important piece of the, you know, getting it less expensive? Like, are there any sort of storylines like So originally I went to school for mechanical engineering just because I didn't have any exposure to entrepreneurship growing up. So I knew that I wanted to make things and sell things. I loved

that. That was my religion growing up. But no one told me that you should be in business to do that. So I went to engineering because I thought, well, if you want to make things so good, that they sell, which is make them just so good they sell themselves. So I just need to figure out how to make things super good. So I'm like, yeah, classic. No one gave me zero to one. So I went to school. Mechanical engineering like halfway through found Paul Grimm's essays and that was like the religious

literature that converted me. I was like oh my gosh there's just a whole other way I gotta like dig into this and I started talking to people in engineering and I was like oh man People do not let engineers on their own go and solve new things. There's some jobs that exist, but they're so rare. The people who have the most mobility are the people who can put a business model around an innovation. They're the ones who actually get to go. do whatever, build whatever thing they want. And

so I was like, okay, well I messed up, clearly. So coming out of college I was like, well, if I wanna go and learn how to learn business and learn how to sell things, I'm just gonna not get an engineering job. That seems like the worst thing I can do. So I decided not to get an engineering job and instead I was like, well, if I need to, if I have to build businesses, enough to make enough money to pay rent every month, I'm probably just going to

figure it out because I have to. So it's kind of like a little bit of a push a big bird out of a nest that's going to learn to fly on its way down. That was my goal. So I figured out how to code. And like early, early on, I like had these dumb service-based businesses that just like made money off Craigslist ads, like converting people's CLA photos. There's Oh, it's like scanning. Like scanning them into digital. Scanning. Yeah, yeah, yeah. Making them digital. Yeah, that was... Yeah, you

can make a lot of money from that. But then also, just a slurred decode, and then I would just go around to small businesses and ask what their biggest problems were, and just stay up all night building a SaaS platform So I had a bunch of little ones. You were ChatTPT before ChatTPT, man. With ChatTPT if it took 24 hours and the outcome was pretty bad. But yeah, so I was doing that and learned business that way. But

nothing was like scalable. I learned a lot about what it takes, but I wasn't building anything that really had scalability. So I was like, okay, I need to like, Take time. So there was a prosthetic startup that asked me to come on full time to run BizDev. I was doing a little engineering help for them. So I was like, okay, great opportunity. I'm gonna go do that. I'm gonna learn how to operate a startup at someone else's startup. I want to take those two years to

not do anything other than find that one opportunity. that I'll do for the next 10 years of my life, even if it's unsuccessful. So I just wanted to, you know, I felt like I have a big shot in me. I'm in my twenties. I need to just like find that thing. And then I just took a lock in and just like close my eyes and just go for it. Even if it's nine years of failing, like I was still, I'm still waking up that 10th year. Like, yeah, I'm ready to go on this. Like let's keep going. So I

looked at a lot of things. Um, and, and Kelly's talking about earlier, what I was looking for was something, um, where there's a lot of engineering that went into making that thing better, but it was pretty obvious what the route was. So on logistics, you need to make something that is faster. It works in cities. It's simple. It's reliable. It's cost efficient. There's not like There might be, but there's no like new paradigm,

right? Like objects are going to start floating and you're gonna be like, oh my gosh, now I gotta figure out how to handle floating packages, you know? And then also wanted to be something where even if I failed at it entirely, but maybe something I did helps someone else figure it out. that's 10 years of my life well spent. So, yeah, I looked at a lot of things and Lost Mile Logistics was just like, that was the thing. I've been obsessed with it for my whole life. It's

like right on the edge of this huge breakthrough. And so it just felt like the right time to go into that industry. And then I figured that out probably like eight months into the two years and then spent the next like year and a half trying to figure out what the problems were and what the best space to be in, both for where would anyone position themselves and then where should I position myself. And I still think a couple years later, there should be... 30% of

the people working in AI should just shift over to last mile logistics. There's so much to build. I think once we reach automated logistics and then eventually hyper logistics, I think that will be bigger than the internet from the amount of money being made. It's a perfect time to get in. Things are shifting, they're moving really quickly. It's kind of where AI was when Dolly, GPT-3 came out. Yeah,

maybe a little farther. I don't think we're quite at the Delhi moment. It was a good time to get started on the underlying infrastructure. But ultimately, we decided we're going to lock into that. And then through interviewing a bunch of people, we realized that the biggest problems were getting goods in and out of buildings efficiently into a variety of different people, and then the bigger, longer stretches in cities, between cities, doing that efficiently.

And so we were kind of like banging our heads a little, trying to figure out like how to, like which one of these problems to solve, and started with drones, and then checked down to ground-based methods, because we really felt like there needed to be this really good partner to drones that can solve the edge cases and help with managing goods in and out of buildings. And then we're like banging our heads on the wall and then ultimately we're like, okay, there's gotta be like a

comparable here. Cities already handle mass distribution of things super well. Is there a model that we can use and maybe borrow from that? And then ultimately we got to If you think about, like, water and sewage are these huge distribution problems. Yes, you can fit them down into smaller tubes, but for the main part of the routes, they're in these giant tubes. Can we borrow from that and use that to build out. Is that the right medium for this? And at first we were like, absolutely not. That's

so dumb. That's the dumbest thing. But we had two, you know, we still had two years of diligence. We're like, all right, let's just go check it out. Let's figure out why it'd be too hard. And from talking to people, we realized like city regulations are super easy compared to other types of permitting, ultimately because the city owns those easements and because of franchise contracts and other methods, they make

money off those easements. So if you give them a new type of infrastructure to make money on, It's a lot easier than pulling a permit for like an above ground building where the upside for a city is pretty low, but the downside is like someone gets hurt. For an agrarian infrastructure, it's almost flipped. The upside is pretty high. Makes up a big part of their city budget. There's not a whole bunch of new utilities going in. And then the downside is pretty low.

You're, you know, six feet underground. Not much can happen that would negatively affect citizens. So I realized that was a misconception, that construction methods have gotten so much better in the last two decades in putting in this underground infrastructure. So if we could stay really close to the way that utilities are retrofitted into buildings and are installed and managed and regulated. Today, this was

super doable. We just needed to build a robotic system that used that same infrastructure, cheaply and efficiently, and bingo bango bongo. That's the play. So it still felt like really hard, but it felt like one of those things where it was too crazy that other people weren't going to go down that road yet. There was like the sliver of like, yeah, if we like nail this perfectly, there's a big business. So for us to like, it was perfect. It was like, okay, let's go after that sliver. It

was fun. Hell yeah. A lot of people told me not to though. I'd never, I told people, a lot of people about that two year, like, hey, I'm taking two years, I'm going to find this thing. And people had seen what I was doing before, and they're like, oh, this is going to be sick. Like, I'm so excited to hear about what you're going to do. And then I tell people, and they'd be like, don't do that. Don't do that.

That's so dumb. Yeah, people I've never thought would were like, hey, like, I'm your friend, but like, But who's laughing now? I mean, probably still