Latest / Talk Commerce / From Dog Poop Detection to Aerial Innovation: How Caleb Olson's Poop Copter Took Flight

Transcript

From Dog Poop Detection to Aerial Innovation: How Caleb Olson's Poop Copter Took Flight · Talk Commerce

• Brent Peterson (00:01.721) Welcome to this episode. Today I have Caleb Olson. Sorry, Caleb, it's been a long day already. He is the inventor of the poop copter and I recently saw his video from MiniDemo in Minneapolis. And we're here to talk about entrepreneurship and inventions and all kinds of other fun stuff. But Caleb, go ahead, do an introduction, tell us your day-to-day role and one of your passions. Caleb (00:29.174) Nice, yeah, thanks for having me. It's really fun to have conversation stuff. Yeah, I'm a traditionally a software engineer. That's at least. Peter science, my undergrad. I've got a machine or a masters in computer science emphasis in machine learning. And I feel like I fit the typical mold in terms of my day job doing bread and butter is just web web development usually flavored with. machine learning in some way, working with data science teams. Yeah, that's my typical, typically just doing bread and butter web development. And then my passions sort of out or bear their face more on my side projects where, I don't know, there's much less hand like rules and things you have to hold to. It's just whatever you can think of, you can do or try to do, which is really exciting for me. And that's... That's where my passions lie. Outside of work is building things that I feel like haven't been done or solve a problem that hasn't fully been solved or made better. It's fun doing that kind of stuff. Brent Peterson (01:44.741) That's great. and I'll make sure I get the link to the video on our show notes. But before we dive into pooping and helicopters, I wanted to tell you a joke. I do this thing called the Free Joke Project. And all you have to do is give me a rating one through five. And it won't be a chat GPT joke, because as we discussed in the green room, they are terrible. So here we go. And actually, they're going to be terrible no matter what. But we'll give it a shot. I actually have two and we'll see which one worked better. So here we go. Did you hear about the whale that swallowed a clown? It felt a little funny. Caleb (02:16.92) What's better? Caleb (02:27.95) Okay, I gotta rate it. It's clever, but it's not. It doesn't really. It just makes me think, that's stupid. So I'd give it like a, two or three, three. I'll give it a, well, two, it's supposed to be really, it's supposed to make me laugh. Not just be clever. Who. Brent Peterson (02:38.735) Good. Brent Peterson (02:45.317) Thank Brent Peterson (02:51.139) This one's gonna really not make you laugh then. Here we go. The past, present, and future walked into a bar. It was tense. Caleb (02:54.126) Ha! Caleb (03:02.926) Feels like a joke you'd tell before like a English English class or something like before your English lesson. I'm like reading like a reading thing. Not to be a don't want to be a bomb, but I'll give it a two again. It's just clever. These are I mean these are clever, but. Not to be start this out with. Brent Peterson (03:17.861) All right. I would like to call them... Okay, I got a poop one for you. One more, that's it, because I keep trying. Why couldn't the toilet paper cross the road? It got stuck in a crack. Caleb (03:28.972) Okay. Caleb (03:33.954) Go. There you go. Let's see. That's more childish. I'll give that one a three or four. That's a four. Yeah, that's a good one. Brent Peterson (03:39.853) All right. I just had to keep bidding. It's like a bidding war here. I tried to get to four. All right. So I love that. I'm a my weekends when I decompress, I like to do a little bash and maybe a little Python coding. But you've gone you've gone both you've gone to the creation level, which I love. Like it's not just creating. Caleb (03:45.698) Yeah. Yeah. There you go. Caleb (03:59.053) Yeah. Brent Peterson (04:06.597) code but it's creating something right so tell us a little bit about tell us why you came up with this idea and kind of that journey Caleb (04:16.28) Yeah, I mean first like yeah my view on just code and people nerd out on. On like good code and clean code that does something useful, but I view it exclusively as a tool of building like I'm a builder and I have these tools and they're just the means to do something with and that's that's where I'm most. I've said in the intro I find most my passion is building something that's. That's what's so magical for me still with like coding and stuff is you can build seemingly anything. I mean, it's digital, but as soon as you bring the digital and merge it with the physical in somewhere or mechanical or electrical, can do really cool things. the Poopcopter specifically, which is what you've stumbled upon, is like the final form of this like poop journey I've been on with. like two or three years ago, I, in with this idea of, of using building a completely computer vision based dog poop detector using my security camera. So security camera in my backyard is just looking at my yard. I was thinking I could probably detect where the poops are and then keep a log of, of where they are. So I don't even have to go and do poop patrol. I can just go to where I know the poop is kind of thing. So that worked and. I believe at the time it's the first, my claim, my arrogant claim is that it's the very first completely computer vision based dog poop detection system. know there's stuff like collars that detect, they have like gyros on them and they can detect like angle and movement and they can assume the dog's pooping. But this is purely from like a stream of footage detecting a poop. So that was a fun project, but I don't know. That got some... some notice online when I just put it up and there's just all sorts of feedback on like, different directions of where to go. And they're just like, well, now pick it up and put it in my neighbor's yard. Throw it over the neighbor's yard is what you should do. So I didn't, I didn't really realize it at the time, but like two years later now, there's one sort of gimmicky project in between the dog poop detector and the poop copter. It's a involving a laser on a, like a robot arm and it would point to where the poop is. So it's kind of. Caleb (06:38.498) this over engineered middle tier project that moves it closer. So I don't even have to look at my phone, but the poop copter takes it to the final version in my mind where it's completely, the vision is completely end to end automated where there is no manual need to go and pick up poop. Cause that's the dirty part. That's the part that people don't want to do. And everyone does it. It's such a... It's such a common the genesis of the original ideas. So many people have dogs and every dog poops as far as I'm I'm aware and you gotta pick up the poop. So it felt like a very common problem. Very common. I don't know manual task that could do with some automation. And I doing research in the space there where there are some, there's one attempt in particular. I don't know if I should say the name, but you can look it up. It's just look up like RC car, like dog poop detection or dog poop pickup system. There's only one other. Automated completely automated attempts to pick up dog poop, but I don't think it. It hit market and it seems prohibitively expensive to build, from what I'm, what I'm gathering from the stuff I found online about it. I think this is what I've come up with. It's technical, but the materials involved are really basic, mostly just it's a very simple drone for the most part. And then the actual mechanism for picking up poop is just 3D-printed plastic. And it comes together in a way where all you got to do is a physical movement to open and close kind of thing. So yeah, I don't remember what the original question was. like how I came up with the poop copter. But the short answer is seeing the problem of needing to pick up, like it's annoying to pick up dog poop. And then the evolution of ideas and attempts over, I guess, like two and a half years now, the poop copter taking almost a year of nights and evenings working on it the last year. The other projects were much smaller in scope and simpler, but led me to this. Caleb (08:56.182) Also probably over engineered solution or attempted solution to automate the problem of dogs pooping in your grass. Brent Peterson (09:05.861) So you mentioned a little bit, you mentioned 3D printing, are you making a lot of these parts yourself aside from the actual drone and then kind of engineering them together and then getting some servos or some motors to get it to work? Caleb (09:19.916) Yeah, I mean I have it. It's in a broken state right now. Most of the mechanisms not here, but the drone itself is a very typical. There's some a lot of small decisions I made along the way for. Or because it's tuned for carrying a payload and like the actual payload, but then also this mechanism underneath it. Most drones aren't carrying things so and it's not a racing drone, right? So I've tuned it. like lots of little things like the type of motors, the length of the arms, where wind is going when it shoots air down. So I had to pick up, like I built this thing and to pick up basic soldering skills. And then you started with just getting the drone working and look like completely entering a hobby or a field that I'm totally unfamiliar with drones and how they work. then immediately kind of going off the beaten path with how drones work by like throwing a a small Raspberry Pi and there's a camera in the actual box. So when it's at its open state, it's receiving or it's recording or it's streaming footage from the camera to the Pi and some processing happens on board. But the current version or the current design is it offloads the heavy machine learning compute tasks to a nearby or a machine in the cloud kind of deal. And then it over a web socket. communicates back with the Pi, which then communicates with flight controller to actually control the fine movements of landing and stuff like that. Brent Peterson (10:59.759) think in your video you had described either a phase one or a phase two where you actually laser pointed or laser, not laser, but you located the poop in advance and then found it. How does the location work now? Is just camera based? Caleb (10:59.915) Caleb (11:15.278) So completely remove the laser and what I mentioned about the dog poop detector. Those are all old versions. This is a totally new system and a way of doing it. The way they're looking, yeah. The dog poop detector was actually not looking for poop. The very first version of what I built a couple of years ago was actually looking for a pooping dog, which is way easier from a cheap camera's perspective, 50 feet away. It's way harder to find poop and distinguish it from dirt. that it's way easier to just detect if like 99 % of the time if your dog is in that pooping position, they're probably pooping, right? Unless they're constipated, I guess. So that's how that one worked. But this one's actually in real time doing, I arrived at a accuracy versus speed trade off in terms of the type of model and what it's actually doing to actually detect poop. on the ground as it's patrolling. So that's kind of the model I suggested in the presentation is it just patrols in like a grid in your backyard or wherever. Some defined GPS bounding box where it can just scan, like it just 10, 15 feet in the air and it just patrols. And as it's going in real time, it's processing and looking for poop. And if it sees one, it will just, it'll move to the precision landing stage of the pipeline basically. And then goes, it does other stuff to land using there's other stuff that's hard to see on here. Like this is a LIDAR for detecting distance. It's factored into the actual precision landing. didn't invent precision landing, but I feel like I invented precision landing on poop with the sort of design of what happens where and how it works. then plugging into the existing use case of a drone doing a landing. But I needed it to be way more accurate. People are doing drone landings, but they're usually on the couple meter basis with GPS. Mine needed to be, it needs to land with the poop in the bounds of the box. So there's a five, six inch error that I have here, or room for error. I've come up with this whole system for how that works. The biggest concern is speed, because the faster it knows. Caleb (13:42.742) where to land, the faster it can adjust. Because if it's really slow, originally it was just landing. It was missing the target all the time. But it needs to be really fast because there's wind and other things that push the drone. It's not a perfect system where it's just like a claw machine. It goes down and it's there. So there's a lot. And there's still a lot to fine tune and make repeatedly perfect, but at a point where it's real. And I'm trying to figure out sort of what's next. Where do I, where do I take it from here? Brent Peterson (14:17.029) Have you worried about the 10 year old neighborhood kid collective where you're paying the neighbors to clean up the poop and now suddenly they don't have a job anymore? Caleb (14:27.662) yeah, I don't know. It depends. It depends on the model. Like someone suggested even just like you could have like a training or like a kid's mode where they could like fly the drone itself, which is kind of dangerous. Even it's not that big of a drone, but there's no real like safety guards on it or anything. It could cause some damage and it, yeah, it's kind of difficult to fly a free form drone like this, but, Brent Peterson (14:29.061) You Brent Peterson (14:40.728) Yeah, there you go. Caleb (14:55.276) Yeah, I don't know. There's depending on how it operationalized or like what the end model is. It could have some broad impacts on like the service. There's service industry of people that actually come out to your house and they'll people will pay to have people come out and pick up dog poop. And I've had some conversations with a couple of people that own those types of companies and they're interested. Like they'd like to, like trial it out with like, buy like 50 of them and trial it out and see if they can, make the process more efficient, I don't know. Or I feel like direct to consumer would be way more fun though for people to have their own hoopcopter flying around. feel like people, people would like that. Brent Peterson (15:37.283) Yeah, so you hinted at like, is next? Where are you going with this and what do you think is next? Caleb (15:46.998) Yeah, so I mean I mentioned it's taking like a year for me to get to this point where I'm talking about it and it's real, but. I'm starting to just take a break and and have conversations like this and and get more eyes on it and think about what's next, because I'm trying to balance. Like. If all I care about is money, then maybe I would try to turn it into a business and I would stand up the business side of things, whatever that looks like. And there'd be a lot of learning and bumps along the way, but I'm trying to balance like, sure. I like money. Everyone likes money, but trying to balance it with like what got me into it in the first place, which was just the passion of working on something fun that I feel like no one's doing. And I'm just kind of like cooking up something new that the world hasn't seen in my basement. That's where I really get ticking and it's way more rewarding to like get something and accomplish something. So I'm still trying to decide whether I buckle down and sink another hundred hours into productionizing this thing and then actually just selling it myself, whatever the model is. Like, do I just let, just lease them out on like a monthly basis that people can pay whatever and have their own like a subscription model or do I just sell them to people? I have lots of concerns like safety if I'm If it like misfires and it just runs into someone's house or like a dog or something, that could be really bad. So I feel like there's, there's a lot of questions that would need to be answered. And I'm not afraid of doing that. I just don't know if I want to do that. And I feel like the dream for me would be to partner up with someone or some company and sort of offload it and honestly move on to some other project. I've got other ideas and things. Leave her not poop related. I've slowly become just the poop man. It's just weird. Brent Peterson (17:42.926) You've been on the news, you've been on VGORD MiniDemo. How are you handling all this exposure and everybody wanting to talk about poop? Caleb (17:52.738) Yeah, I think it's fun. yeah, it's starting to get old. Like at least the really short conversations. I enjoy conversation like this because it's longer form and we go below further than just the surface where people ask like, well, what does it do with the poop after it picks it up? And then I can make my joke about sky dumping because it can drop it from from the sky and stuff and or you drop it on your neighbor's house and then all these like short like two minute. clips where I have jokes and I want it to be funny and entertaining. But I don't know. I think it's just fun. At this point, it's fun to just talk about something that I've been cooking up for a long time. But also the hope is that the right eyes land on the project and the right connection can be made and I can get involved with people. I can fill in the business side and the productionization side of things. to take this thing to the next level. Because I don't want it to just die. But I also don't really want to spend hundreds of hours turning into a business because it doesn't sound fun. yeah. Brent Peterson (19:01.381) What do you think is next? you have other ideas? You had mentioned you had some other things brewing. And in the past, have you had other sort of inventions that you've come up with that maybe you haven't been so vocal about? Caleb (19:11.819) Yeah. Yeah, yeah, I have a YouTube channel where I'll put up a usually that's my model is cook something up and I'll just put a video out as honestly just like a diary for like something I've done. and just like record key thing, but on the note of other projects, there's a, there's a few videos and a couple of them are open source. A couple of them I'm keeping close to my chest in case I want to do something else with them. but. So the two things, two big things in my life and in a lot of people's lives, dog and children. So I've got two kids now and two or three of these other projects involve, well, one of them was sleep tracking. It's like an open source. This is free. Anyone can go. If you have the only requirement is you have a camera that you can tap into like a live stream for the RT real time streaming protocol. That's the only requirement. Otherwise you should be able to run that. basically tracks your child's sleep when they go to sleep, when they wake up and keeps a log of that of, and predicts when or tells you when they should next go down for a nap kind of thing. So automating that, like some people like put it on their fridge or whatever, or like keep a log. There's apps for manual data entry. that the goal of that was to just automate that process. So I can just open an app and see when my kid woke up. So don't have to like think, okay, I put them down at one and they go down in two hours. Automate all that away. And then the other one was Hungry baby detection hungry baby alarms what I called it Where the the pitch in my mind was for that one was In the middle of the night, especially with newborns they they'll typically show signs of hunger with like lip smacking There's a few different heuristics I was looking for in an automated way and basically if enough of them happen in a short enough period of time Caleb (21:10.414) It'll trigger a notification sent to wherever you want it, like your phone or something. and then the idea is if you can get ahead of, like, if you need to go prep a bottle in the middle of night, like the husband could go to the fridge and get milk and prep it and get ahead of the baby eventually or inevitably crying. And then maybe they'll go back down to sleep sooner and it'll just be a more streamlined process. None of this is really backed by too many studies or anything like that. It was just something I'm There's a lot of assumptions baked into that, but that's another baby one. But yeah, I've got other stuff cooking up. Nothing fully formed. I like to host these brainstorming sessions where I'll have my brothers and some buddies come down, grab some beers and pizza or tacos or something, and just wipe board and just throw stupid ideas on the board. It's mostly just a fun hangout, but sometimes it gets ideas flowing. And you think of, I don't know, it's good to get other people's brains involved and just like spitball on things. but yeah, I'm trying to think of, like something I haven't seen. Maybe you can help me brainstorm on this on the fly. Something I haven't seen, really tapped into is there, there are smart callers for dogs, but, and there are some that will record footage of like what your cat's looking at. But all it does is record to an SD card as far as I'm aware. No one's really doing a real time, like you'd probably want to stream it off board, if you could process on the footage off board, but if you could have a live stream of what your dog or your cat is looking at, what kind of data sources would be of value or interesting that your dog is looking at? like one use case is, my dog will just, a lot of people's dogs will just bark. And I'll be like upstairs, it's late at night and she'll just be barking out the door. And I'm like, okay, am about to die or is it just someone walking across the street kind of thing? I was like, maybe what if you could like in real time, see what the dog is looking at or barking at, and then have some kind of processing that's that classifies or tells you what like it would say in real time, you get a notification Twinkie's barking at the person going for a walk or someone's at the door kind of thing. Caleb (23:33.838) I don't know. I'm kind of exploring that realm, like what your dog is doing, like in terms of like what data sources they are generating or tapping into in some way that you could process or work with to create something of value. Not totally bought into that idea of what I just suggested, but it's kind of how I evolve these ideas as well. I like to usually you're supposed to everyone says you're supposed to start with the problem and then find the solution. But I like I like exploring interesting data sources that I have that feel untapped. Like what is my dog looking at? What what could I do? As soon as you have a data source, you could capture it and distill it or transform it in a way where you can work with it. If it's computer vision or audio or some other kind of metric or like sensor, you can transfer that. and work with it in Python or whatever you're using to do something. Brent Peterson (24:35.461) Yeah, I did try putting a GoPro on the back of our dog one time and I got, they make GoPro dog harnesses, but I think the problem is as soon as you start running the, everything starts going like this and it's hard to, it's hard to manage it, manage that look. But yeah, if they're stationary, I suppose our dog has a fee collar, FI, so we can track him where he's at. It doesn't, and it does have real time, but it doesn't have video or audio on. Caleb (24:41.452) Yeah. Caleb (24:45.432) Crazy. Yeah. Caleb (25:02.595) Sure. Caleb (25:05.922) Real time video. VPS. Okay. Brent Peterson (25:08.234) GPS just says GPS yeah it's connects to my Strava which is even more important Caleb (25:15.702) interesting. Yeah. Yeah, I haven't done a full. Yeah, I'm just kind of playing with ideas. I don't know if this is something I'll keep going down. Like, I'm not ready to start working on that, but I feel like there's cameras that are small enough. Like these, I've got one right here, Raspberry Pi camera. Tiny. Like you could put this on a caller. I mean, you got to power it and have it plugged into something, but you could have, you could use those Picos or Pi zero W or whatever they're called. Brent Peterson (25:16.965) You Caleb (25:42.51) You could have a pretty small footprint and that fits on a little collar. But yeah, that's probably the first thing you'd have to address or solve for is motion because they're moving all the time. And is it, I don't know, workable enough that you can do something reliably? I don't know. But yeah, I'm not bought in on the value of the idea totally to be even worth exploring. Sorry to interrupt you. What were you going to say? Brent Peterson (25:59.994) Yes, so for Brent Peterson (26:06.521) No, no, for the geeks that are listening that are interested in sort of the technology part of it, you're using Raspberry Pi and then Raspberry Pi has a bunch of peripherals that you can attach to the unit and put it together. Like on your drone, you got a kit for the drone or is it a commercial drone that you kind of attached a bunch of equipment to? Caleb (26:27.438) So it's a kit, and then you buy, it's just like building a computer, and anyone's built a computer, it's just you have to solder things. And there's a little bit more, you have to research and make sure things are compatible. So the drone itself, I went down, I originally wanted it to be an off-the-shelf thing, that felt like a quicker path forward, but they're all closed systems, the firmware. As far as I'm aware, from my research, they're closed systems, and I've got an open source firmware called ArduPilot. is what's on the actual flight controller. So that's how I can tell it. I can issue it commands programmatically and control it in an automated way. That was a decision I had to make pretty early on. And then I'm like, OK, well, if I'm building this from scratch now in order to use ArduPilot, I got to learn all this stuff. So most of it is a pretty typical drone. This is a very typical quadcopter. But then where it gets weird is this deal here is just Double taped. literally just tape. thought I'd have to come up with a fancier mounting system, but it's solid and it's endured many pretty fatal crashes. But it's just a Raspberry Pi off the shelf. It's like 15 bucks. This if I recall, yeah, the Pi zero is like 15, 10, 15 bucks. And then this camera, like the this is a lens that I had to put on top of the camera underneath it. It basically looks like this. for the most part. This is the typical Raspberry Pi camera. had to switch from, originally I just had this on it, but I was getting gnarly artifacts in the footage that was streaming because of the motion. In the off the shelf Raspberry Pi cam wasn't handling it well because of gets into camera tech, something to do with the shutter of like the way it captures information for the frames. It was creating these weird wavy effects and it was making it really hard to like do anything meaningful with it in terms of a like an image processing perspective. So I've got this separate, slightly more, it's like a $50 camera. the global shutter camera, from raspberry PI and then a lens on top of that. And then that's just wired. can't see it, but there's a, this sort of orange strip is the camera cable. So it's sending footage to the raspberry PI. Caleb (28:47.468) which for the most part immediately does some initial transformation and then immediately sends it or streams it to an off-board machine, which is doing the heavier stuff I mentioned. And then WebSocket back to the Raspberry Pi because latency is extremely important with this because it needs at least 30 to like, ideally you have like 50 frames per second or like a refresh rate or a processing rate of 50 commands being issued to the flight controller saying, land here, land here, land here, as it's moving subtly and moves down. And the height changes as detected from the LIDAR. Thankfully, that landing process is already built out. It exists within our new pilot. But then plugging into it with this extra stuff on top is basically where my creativity was involved. That's basically it. It's just the camera feeding into a Raspberry Pi and then working with a bunch of code that's running off board. I can talk about that if you want, but it's just Python. And then. Mostly off. I did train my own model. I had to have a lot of footage of just poop on my phone where I had to pull frames from and I would basically label. I only needed like 80 to 100 images of labeled poop basically. from my backyard to make it work in my backyard. That said, thinking about that, if you're familiar with machine learning and image classification or object recognition, it's really biased to my dog's poop. I bet it start to struggle with a larger dog's poop or different factors that aren't captured in my backyard. So taking this to the next level, you'd have to take the model itself that it's using and add away more data and make that more resilient for different use cases or different environments. Brent Peterson (30:47.077) So you answered my next question, is you're using artificial intelligence to identify the poop in the yard. You're not doing it manually on a screen, which you probably could do, right? Somebody could probably fly around the yard and land and go and do it. And I know how many machines it takes to run a bigger LLM. Caleb (31:01.942) Yeah. Yeah. Brent Peterson (31:11.753) Are you offloading it to a SaaS provider for your LLM or are you running your own model that you've trained to do this? Caleb (31:17.518) Yeah, it's my own model. It's technically it's not using LLMs at all. It's using a computer vision. Computer vision like it's an object recognition task, so it's looking for like you've everyone's heard of or most people have heard of like Yolo the model Yolo, which is like you only look once it's quick shot. Like it frame by frame like a video it'll recognize little person walking across the street or like. chair or something. So it's that, but it only recognizes poop, not poop. So it's exclusively for, that's all it's looking for. So if it misfires, might, there's a chance it lands on something that looks like poop, but it's not actually poop. But it's only looking for poop. And I didn't like craft a neural network myself. Like I have some data science, I'm like a peasant data scientist. I'm mostly plugging into existing like models that are proven to work for that, for this type of task. And then I just overload it to fit my use case specifically. You might be able to get more precision or more juice from the squeeze if you go completely like custom. But there's trade-offs. All along this whole project, there's trade-offs of like, Do I go down this rabbit hole? it, is it worth exploring this rabbit hole for whatever the perspective gain is for the amount of time I'd have to think into it? You know, I could have spent another a hundred hours like fine tuning things. So it works a little bit better this way. or certain, certain like environments, like if there's a tree or something, what if it, it, the poop is under the tree. How would that work? There's certain things like I are unanswered that I had to just like prune out. For the sake of proving, can it in an automated way move around, see poop, land on it, and remove it? That's the core of the project. And I feel like that's what I've proven. But taking it to production, whatever production looks like, you need to answer a lot of those types of questions. Brent Peterson (33:31.173) What kind of dog do you have? What size is it? Caleb (33:34.574) I a corgi named Twinkie, 30, 35 pound corgi. A little smaller, not tiny, but smaller. So mean, yeah, bigger dog. It's not going to work for the size of a box. It would need to be, I mean, maybe. But it wouldn't be hard to have different versions of the poop copter where it's slightly bigger. Brent Peterson (33:57.039) you'd have to have a Great Dane version and a miniature Chihuahua version. Caleb (33:59.406) Yeah, problem is the bigger it is, the more expensive it is. need bigger, stronger motors, more material and it's more dangerous. it runs into things drones. don't know. The drone's going to be fine. They're pretty, they can take a lot of hits, but it's the risk of it. Like what if it hits a dog or something that could, that would be bad. so that was something I, I was biasing all of my decisions along the way was keeping this. I wanted it to be as small as possible and cheap and simple as possible. That's why this is just, again, this isn't a broken down. This is just the base. is a face and these other components that work together to make the whole box work. This is the spike front face plate. And then these leaves fit onto that track in the base, and it's able to slide and stuff like that. Pull up the video if anyone's watching this, and you can see the actual functioning version. It's taken a big fall recently, and I haven't reassembled it. yeah. Brent Peterson (35:03.653) And last question, did you explore, there's a ton of the robot mowers out there now. Just, you know, the other idea would be to just have something that roams around your yard and kind of picks things up like a combine or something. Caleb (35:20.398) Yeah. Yeah, yeah, I don't know. Part I mean, partially what set me down the drone route was it felt cooler and no one's tackled the problem from the sky, but it makes the problem harder. It's it's there's way more variables involved in terms of like making the landing. That's the hardest part of the whole thing in my opinion. But yeah, you could. My first thought is the the main benefit I see from being in a drones perspective or patrolling from a drones perspective is. Brent Peterson (35:29.978) Yeah. Caleb (35:52.93) you're higher up and you have, can process on much more terrain at a time at a given moment and you can move faster. It can move through the yard pretty quick and, and maybe slow down if it sees, if it thinks it sees something and then once it has more confidence, you can actually make the landing. If you're confident, I don't know how well you'd be able to, mean, maybe there could be like a camera on a stick or something and it's got a little more height or something. don't know. But yeah, the, the, the, the car version of the, Earlier in this discussion, I mentioned an existing prototype that a team of researchers produced that seemed really expensive was basically what you're describing. it has, I question how well it works because it has like a camera in front of it, but it's, can't see that much unless it's really big. But what they had was pretty small. It feels like you'd have to do a many, many more like movements and it would take a lot longer. But I don't know, you could probably come up with an answer to that problem. And then you would remove a lot of the complexity in operating from the drone's perspective and danger of it flying around. So I don't know. There's a million ways to do anything. I just headed down the drone route and it's worked. Maybe re-approaching it from the car perspective is worth a thought. And coming up with a design, maybe it can be even cheaper and simpler. I just haven't sat down and like, thought through more fully, like attempting to solve for that. Brent Peterson (37:26.021) Have you thought about other things you could deliver with the drone, like a beer? Maybe an out to the deck or something like that? I mean, if you wanted to repurpose it for something? Caleb (37:38.242) Yeah, drop off. don't know how. A beer would be probably too heavy, at least for this one. This this version of the drone would have to be bigger to carry a beer. It might be able to get it there, but it'd be overworking itself and probably burning out the motors and it would have like no no battery life flying you your beer. But I guess maybe it's worth it. Brent Peterson (37:58.509) Maybe that's the dog's job. Bring you the beer and the drone goes picks up the poop. Caleb (38:00.94) Yeah, yeah, yeah, that's a good idea. I know it was just a joke, someone suggested, and this is a good idea, but like this version, the version of the poop copters designs for grass. It's got all these torture looking spikes on it. Really gnarly looking, you could, right now it's just glued together basically, but it'd be easy to design. I think it'd be easy to design some pins where you'd be able to pop off the faceplate and put like a a more low profile with a rubber faceplate so it can get traction and pick things up off the ground. So I don't know about a beer, but people have suggested picking up garbage on the ground if it can find trash. And you'd have to adjust the faceplate because the spikes probably wouldn't work if you're picking up stuff on concrete. But you could adopt it to that use case, like picking up crap or trash, not crap. Can't use that term so generically when most of this conversation is about crap. Yeah. Brent Peterson (39:06.417) That's awesome. Caleb, who are you looking for next? Are you looking for investors or what's next on your list? Caleb (39:14.124) Not investors. Honestly, I don't know who, if it's a drone company or a pet company, or maybe this could, this project could be applied in a different way. I'm not seeing, but anyone that sees value and is interested, anyone or one entity that's interested in taking this to the next level and turning it into a product. I think that would be fun and cool to see this thing go to the next level. I just don't know if I have. the will and drive to do it myself. So at this point, I've done the most of the heavy lifting in terms of proving something's possible. And then the question is, is what's possible worth scaling up and worth the value is worth the cost of building and shipping out poop copters. So I think it is, but I'm a little biased a year's worth of sweat equity and definitely going to taint my outlook on the project, but. That's the dream, or the goal is to partner up and sort of offload or productionize it. And then that frees me up to move on to some of these other ideas that I've got cooking, which is that's where my passions lie. I want to be careful with. There's only so much time, so I just want to spend it in the ways that are most valuable, however you define value for yourself. Just trying to make sure I land on the right spot for myself. Brent Peterson (40:46.149) That's awesome, Caleb. It's been a really enjoyable conversation. Caleb Olson is the inventor of the poop copter. Thank you so much for being here today. Caleb (40:52.206) That's right. Caleb (40:57.294) Thanks Brent, thanks for having me, it fun.