From Dumbbells to Data: AI's Role in Next-Gen Workouts

Julien Redelsperger: “And I'm super happy to welcome Andrew Just. He's the founder of Train Fitness. This is an application, a mobile application that uses the motion of users' smartwatches to predict what exercise you are doing at the gym. That's pretty exciting because today we're going to talk about AI, sports, and fitness. Thank you for joining me today. How are you, Andrew?”

Andrew Just: “Doing very well. Thank you for having me, Julien.”

Julien Redelsperger: “Thank you. That's my pleasure to have you today. So we're going to talk a little bit about sports and fitness and AI. So just to get started, what can you tell us about Train Fitness? You started the company in 2021. Where does that come from and what's the story behind Train Fitness?”

Andrew Just: “So Train Fitness is, we're the first application that automatically tracks gym workouts, hands-free using only the motion of your body. In our case, it's the motion of the user's Apple Watch. And so, you know, other applications that track, for example, cardio sports, running, cycling, whether you're thinking of Apple Fitness or Strava, they'll give you rich information on your workout, how far you went, your distance, your time, your splits, your elevation, all of these interesting metrics. But when you actually look at the anaerobic space, there's no way to automatically track, you know, a gym workout or a yoga session or Pilates or martial arts. And that's because you can't use GPS to automatically track anaerobic activity. GPS is not accurate enough to track within, you know, a few centimeters or a few inches. So we have to turn to other technologies and that's what we've built here at Train. We've built and patented the first technology that uses artificial intelligence to track a user's body and then we're able to detect what the user's doing. And so in our specific use case, when you use our application to go to the gym, we'll detect what type of exercise you're doing, count your sets, your reps, your rest time, completely hands-free automatically using only the motion of your wrist. So you kind of, you walk into the gym, you press start, whether you start doing some pushups or bench press or incline bench press or jumping jacks or anything of that nature, you don't need to tell the app what you're doing or how many reps, you just do it and the app will track that automatically.”

Julien Redelsperger: “Okay, wow. And where does the idea come from? Like why did you create that company? Is it like a personal story behind it?”

Andrew Just: “Yeah, definitely. So growing up, I raced triathlons semi-competitively and I was just a huge numbers and data nerd and junkie. So trying to use every single piece of hardware or software or gadgets that I could get my hands on to track every single aspect of sport that I could with the end goal of becoming a better athlete. When I went to university, I started lifting weights more and was just super dismayed and disappointed to realize that all the tech I'd learned to love on the cardio side didn't exist on the strength side or in the strength or anaerobic space. And so here I was trying to become a better athlete in the strength space, but all these tools that I thought I would have at my disposal just didn't exist. And so from that point, I was looking for a way to try and change that and that's what train fitness came out of.”

Julien Redelsperger: “So that means when I'm going to the gym, I have my Apple Watch. The Apple Watch is not tracking accurately what I'm doing. So does that mean that the numbers like they don't make sense? Are they accurate?”

Andrew Just: “So you mean outside of train fitness? Yeah. So the problem is, is any other fitness application you use today at the gym, it'll track the duration of your workout and your heart rate. And that's about it. And so whether you are using Apple Fitness or Strava or any other major Garmin, Whoop, any other hardware or software to track your workouts, the reality is, is you'll log that entire workout and at the end of the workout, it'll say, "Great job, 50 minutes, 200 calories," but that's it. There's no rich data. And the problem with that is lifting weights is far more nuanced than maybe, or at least I would like to think that it's far more nuanced than just going for a run, right? If you want to get healthy and you're going to go for a run, well, you put on a pair of running shoes and you start running. And that's basically it. Now there's, you can get into more details, but at a high level, you can just start running and you're going to make improvements. At the gym, you need to know, you know, what weight did I lift last week? How many days of rest did I have? You know, what was my rest time between sets? How many reps did I do? You know, how was my acceleration in different motions? Is my form correct? It's much harder for a beginner to get into strength training. And if you can't track that data, you know, frankly, you're leaving gains on the table.”

Julien Redelsperger: “Okay. So how do you guys identify, you told me over a hundred exercises, 150 exercises from the gym. About 150 now, yeah. If I'm doing like biceps or triceps, like the move is very close. Definitely. So how does AI come into play here?”

Andrew Just: “Yeah. And so, it's a great question. So when we started Train Fitness, we actually first tried approaches without AI, basically using basic arithmetic and math to try and say, okay, well, the velocities between this range and the watches at this range, and it's moving, you know, this speed, it's probably a bicep curl. The reality is, is there's just far too much variation between individual users for that approach to ever work. And so, you know, the way that you might do a bicep curl is different enough from the way that I do a bicep curl that we can't classify them in one kind of generic, you know, description using math. And so, the way we start to commit that or solve this is using machine learning and deep learning in our case. And so, we've built algorithms that are very, very precise between minor differences in movement. And so, a lot of the exercises that we detect, for example, we will detect the difference between a regular pushup and a close grip pushup, where the only difference is the, you know, the distance your hands are between each other, or even things like an incline bench press versus a flat bench press, where the only difference is the angle of the bench. We're actually able to detect the difference and differentiate between those two exercises. And that's only possible using deep learning.”

Julien Redelsperger: “Okay, wow. So, I understand you are a sports fan and enthusiast. And do you have a technical background? Are you like, how did you create that AI? Did you build it yourself or did you just like hire someone and work with like specialists in the AI field?”

Andrew Just: “Yeah, so I built it myself. And then since then, we've had some team members come on. Our head of AI now, Vivek Verma, has been, you know, an absolute just trailblazer in the space as well. But when the company was started, it was myself building the initial AI. And so, the background there, so I actually don't have a formal education in AI. And I've only been in the space since about 2019, 2020. But then again, most of us have been, it's a pretty new space to begin with. But so I've been a software developer for the last 15 years, though. And so I grew up around data, working with mobile apps and websites and database-driven sites and things of that nature. And I've always been very excited by the concept of big data and now deep learning. But in the initial days, the model was far simpler when we first started. But it was built all in-house.”

Julien Redelsperger: “Okay. And so how many exercises exactly can you detect using your system?”

Andrew Just: “160 right now.”

Julien Redelsperger: “Okay, 160. And how did you define those 160? Was it because it was like the most, I don't know, well-known exercises at the gym?”

Andrew Just: “Yeah, so it's a bit of a song and dance. So when we, our very first version supported five exercises, it was jumping jacks, pushups, a goblet squat, a bicep curl, and a lat raise. And those initial five exercises were deliberately picked because we wanted to show that it would support leg exercises and that we could even track what we call static motion exercises. So like a pushup where your wrist doesn't actually really move that much. And then we kind of grew the list from there. And there was a bit of a method to the madness on picking the exercise list. Initially, we actually tried to pick exercises that looked very different from each other. And so if you think of, for example, the motion of a jumping jack doesn't really look like anything else. And so it's a very easy exercise to add in. You know, the motion of a, as you kind of mentioned, a bicep curl and maybe a tricep extension might look a little similar. Or maybe a cable bicep curl versus a dumbbell bicep curl might look a little similar. And so we initially decided to try to avoid exercises that look too similar because it was just easier to confuse them. But then we got to a point where we were acquiring enough users and they were kind of, you know, shouting loud enough that they wanted some of these key exercises that we ultimately had to find creative ways to solve these previous challenges. And now our list of exercises is almost entirely driven by what our users are asking for.”

Julien Redelsperger: “Okay. So how do you make sure that AI remains adaptive to like new exercises on your fitness? Because you know, sometimes trainers can add like new moves. Do you need the AI to be trained more frequently?”

Andrew Just: “Yeah. So we're actually, so as it stands right now, the models will learn from you as you progress, but they will not, the current version of the app right now will not detect new exercises. That's actually something we're looking at changing in the near future. And so we've been working over the last six months on a completely new architecture that will actually entirely learn from how you move and will also learn new exercises from you as well. And so that's super, super exciting because then every single model is purely customized 100% for each individual user. And so let's say I go to the gym and my trainer introduces a, I'm making this up, but a twisty lunge jump, something, something that's like a completely obscure exercise that no one else would ever do by doing it a few times, you know, three or four times, you could actually teach, train what that exercise is. And the next time you go to the gym, it will auto detect it. And that's something we're super excited about as well.”

Julien Redelsperger: “And today when you go to the gym, you have, I mean, you have basically like two solutions to track your moves. So you have your solution and you have something based on computer vision.”

Andrew Just: “Exactly.”

Julien Redelsperger: “So what are the main differences and the pros and cons of each solution?”

Andrew Just: “Yeah. So we initially actually looked at computer vision. And so the pros of computer vision is it's much easier to build. You know, there's, there's frankly hundreds of startups and medium-sized and even large-sized companies that use computer vision to track motion. And you know, that stems all the way back from, if you think of like the tonal or mirror or some of these other applications where you're working out in front of a camera, they're using that camera to track how you move. The reason we decided not to go with computer vision and maybe some of the drawbacks of computer vision is, you know, if I'm being honest, I think the use case is in the experience, the user experience is far worse. And so it's a little bit invasive, it's cumbersome, it's awkward. It's not very versatile. You know, some can see it as even like a breach of privacy. You know, I don't personally, I don't want to work out in my living room in front of a camera or TV every single time I work out. I don't want to bring a tripod and camera around with me at a public gym, just to be able to count my reps. And there's there's a lot of reasons why I wouldn't necessarily want to be, you know, sweating and, you know, kind of hot and gross in front of a camera every time I work out and not really sure what's happening with that video. And so the benefit with with wearables is none of that exists, right? So you're already wearing your Apple Watch, it's hardware that you already own. You simply download an app and you press start. There's no filming, you can use it very discreetly in a public place. You don't need to like film yourself. You don't need to be like that person with like a tripod and camera out at the gym that's filming themselves. And it's versatile. I can use it at a public gym, I can use it at a home gym, I can use it in the park, at a hotel gym, anywhere I have my Apple Watch with me, I can use it.”

Julien Redelsperger: “Okay. And so it connects automatically with like Apple Fitness and Apple Health. Like you don't need to install your app. And that's it. It works.”

Andrew Just: “Yeah, you download Train Fitness. And so we're very privacy forward. And so you have to give us the permissions for HealthKit if you want us to sync with Apple Health. But once we receive those permissions, we can read your heart rate, calories, things like that. And then we're also writing the workout to Apple Health as well. So it'll save to your Apple Health, it'll close your activity rings, it'll log it as a fitness and workout in Apple Fitness as well.”

Julien Redelsperger: “Okay. And so how difficult was it for you and your team to create an AI model capable of recognizing such a diverse range of exercises?”

Andrew Just: It's challenging. I mean, it took us about four years, and we're still working on it. The hardest part about wearables, or so human activity recognition (HAR), is that there are no pre-built libraries or anything you can build off of. So in the last 14 months, we've seen AI kind of blow up. The vast majority of consumer apps that you see today that are using AI, a good chunk of them, they are a wrapper around someone else's AI. And there's nothing wrong with that. But there's no 'chat GPT' that we could just plug into. Even for computer vision, Google and a bunch of other companies have built base libraries for motion tracking. But none of that exists for wearables. So even everything from the data set that we had to collect to each individual layer and node in the model, we had to construct that ourselves. The challenge of that is it's a very slow time to build. It took us probably 24 months before we had what I would call a functional enough product that a user would actually be delighted to use it. Before that, it was a proof of concept for sure, but I would not describe it as a delightful experience using it. There were just too many mistakes.”

Julien Redelsperger: “So it's like a proprietary AI model that you've built for yourselves?”

Andrew Just: "Yep. Yeah. So we've had a handful of engineers working on it for about four years. We've built, we've patented it, and we're the only ones using it.”

Julien Redelsperger: “I didn't mention that before, but Train Fitness is a Canadian startup. You are based in Toronto, correct? But who are your clients? Who do you target? Is it only Canadian or US as well?”

Andrew Just: “Yeah, so we're global. And I think just given the nature of the global weightlifting demographic and also of our marketing channels and some of the influencers that we work with, actually about 80% of our customers are based in the US, about 10% in Canada, and 10% rest of world. But from a population demographic standpoint, that's pretty proportionate to at least between Canada and the US, the actual populations as well.”

Julien Redelsperger: “So we saw AI is changing a lot of things lately, and the fitness and the healthcare industry, health, well-being industries is obviously part of it. Given this strength and conditioning training, how do you see the future of fitness and AI? What is it going to change for gym trainers and people who actually are going to the gym?”

Andrew Just: “Yeah. So I think a couple of trends that I expect: number one is I expect connectivity of data to be more seamless across platforms. And so what I mean by that is up until today, I have my Strava, for example, I use to track my runs and I use Train Fitness to track my strength workouts, but then I use Apple Health to track my heart rate and steps during the day, and I use a sleep tracker to track my sleep. And I have all of these isolated, independent systems that track different aspects of my life. We're seeing a lot of aggregation and collaboration. And so that stems from Apple Health is a huge driver of that, where it's very easy for everything to kind of come in and sync to Apple Health and then also come out again. And then they really have become the hub, at least in the Apple ecosystem. We're also seeing a lot of private startups trying to do similar things. And so what you'll see, or what I believe we'll see as an outcome of that, is applications using data more broadly across the ecosystem to inform decisions and information. And so as an example, let's say that the sleep tracker I use to track my sleep is now pulling in my steps data and my nutrition data from my nutrition app and my workout data from Strava or Apple Fitness or Train or whatever it might be. It's now going to be better informed on the strain my body's gone through during the day to recommend sleeping. And so conversely, at Train Fitness, we're pulling in stress data and sleep data. And if you're sharing information on, say, your cycles for your menstrual cycle, we're pulling all that data in with users' permission. And so what we can do then is we can use that data to better inform what workout to build for you. And an example of that is, let's say you're coming into the gym and your heart rate variability has been spiking all day, which is a sign of stress. You only slept five hours the night before. And we can see from your nutritional data, maybe you use a nutrition tracking app that you haven't had much protein or something like that. We're going to alter your workout to make it more realistic for your current conditions. And that can all be done automatically using sharing. I think the second aspect of the second major trend that I think we'll see across the industry as well is starting. So we're seeing the sharing of data. I also think we'll see more insights coming from applications. And so particularly with the introduction of LLMs or language models, it's become a lot easier to extract and verbalize or communicate insights to individuals. And we're already starting to see that across different fitness apps. And so where previously you might see something that says, like, hey, your heart rate variability is up 6% today compared to last month, like, great. But if I'm not a physiotherapist or a doctor, I have no idea what that means. But we can actually now use language models to relay that information in a way that users will understand it. And so we might be able to say, hey, Julien, your heart rate variability is up 6%. It seems like you're more stressed. We would recommend meditating today or something of that nature.”

Julien Redelsperger: “So it's like having an AI companion for fitness and well-being that could tell you what to do on a specific day.”

Andrew Just: “Exactly. And I would say up until today, the only people that truly-- a lot of people will track runs and bike rides. But really, it's only more elite athletes that are using that data to their advantage. For the vast majority of us, we're just using it for social and sharing, which is great. But I think for the vast majority of us, we'll actually start seeing more ways to use that data in a way that will help us improve as athletes and improve our lives.”

Julien Redelsperger: “OK. And so do you think in the future, AI could potentially replace, I don't know, like personal trainers, for example? And is Train Fitness a step towards that vision?”

Andrew Just: “Definitely. And I think it's never going to replace them entirely. And so at Train Fitness, we're definitely looking at making fitness more accessible. And so one of the gripes that I have with strength training is there's this huge movement right now of people wanting to lift weights. The challenge is, as I mentioned earlier, it's not just as simple as throwing on a pair of running shoes and starting to run. You can hurt yourself, and you can injure yourself if you're doing something wrong quite seriously. And so you need a good amount of education in order to enter the space. And right now, today, that education comes in the form of personal trainers that cost $100, $120 an hour. You're using them three times a week. And so you're looking at about $1,000 a month, which is just completely not accessible for the average individual. And so one of the things at Train Fitness we're trying to do is, how do we make strength training accessible, where for free or maybe $20 or $30 a month, you can receive this education and know if you're doing the correct form or you're going too fast or too slow or too many reps or not enough reps or not enough rest. And so these are all things that we can build in and teach. Will it replace personal training entirely? Definitely not. I think there will always be some personal trainers that carve out niches. But for the everyday consumer that's just looking to learn about strength training, fast forward five years from now, do I think there will be a lot less personal trainers and a lot more AI in the space? Absolutely.”

Julien Redelsperger: “So what advice could you give to personal trainers? You guys, it's time to focus on AI or to, I don't know, focus on your soft skills, maybe trying to find a niche where you can work on?”

Andrew Just: “So I actually think it's an incredible opportunity for personal trainers who are excited to embrace the technology. So we've worked very closely with a bunch of personal trainers and the biggest problem that a lot of them have is it's a service-based business, so it's not scalable. And so you might have 20, 30 clients, but you can't have really 100 clients. There's just not enough hours in the day to train them all. But AI changes that. And so one of the things, for example, at Train Fitness we're building is what does it look like to have a hybrid personal training solution? So take any individual personal trainer that normally sees their client three times a week. What if they change that to instead seeing their client once a month? And so what they would do is their client would use Train Fitness. And so Train Fitness is tracking acceleration, velocity, range of motion, time under tension, displacement, rest time, reps, all of this super, super granular data, and it's syncing that real time to a personal trainer's dashboard. And so now you only need to see your client once a month because every time they do a workout it pops up in your dashboard. You can see exactly the quality of the rep, the time between reps, acceleration, velocity, all these interesting metrics. And then you can chat with them via SMS or over a phone call and give them training. And that's exciting because now they can scale up their clients. If they're only seeing a client once a month, they could feasibly have 250 clients now instead of 20 or 30. And that's very exciting and a huge way for them to scale up their business as well. And so I think for the personal trainer that's excited to embrace the technology, AI is a very good thing for them.”

Julien Redelsperger: “Okay. And so do you get any feedback from personal trainers or any feedback from users about how Train Fitness has impacted their gym experience?”

Andrew Just: “Yeah, for sure. And so I mean, there's a couple of use cases that come to mind, but we definitely had one individual who somewhat related actually was using a personal trainer for fitness. And what was really interesting is they were using a personal trainer, not necessarily because they needed the help, but they needed the accountability. And so they were an individual that just had a very busy schedule, life going in a million different directions. And they kind of knew that without a personal trainer waiting for them at the gym at 7am, they just would sleep and snooze the alarm. And so they were paying for a personal trainer for that reason. And they, you know, circumstances change can no longer afford the personal trainer and as a result, stop going to the gym. And you know, I think kind of let themselves slide a little bit. And what's really interesting about Train Fitness is so we built a huge social platform around the app as well. And so every time you log a workout, it's posting it to your feed, people can like and comment and share, there's leaderboards and challenges and streaks, and we've really kind of gamified the strength experience. And so when they found Train, Train replaced the accountability factor that the personal trainer used to give them. And so previously, they were leaning on the personal trainer for this, but now they have this feed and friends that are following them and liking their workouts and also noticing when they do miss their workout, right? If I know that, you know, Julian works out five days a week, and then this week, you don't work out at all, I can see that I see you lost your streak, and I can I can, you know, kind of call you out for it. And so with this feed, they were able to get the fitness journey back on track, because we provided a level of accountability through social that they otherwise were missing without their personal trainer.”

Julien Redelsperger: “So you collect a lot of data, you just told us about it. Is there any insights, any trends that you observed about how people work out differently based on their age or region? So demographics, I don't know, do younger people do more push-ups, for example, than older people? Or did you get a chance to take a look at that?”

Andrew Just: “Definitely. So a couple of interesting things. So universally, Tuesday is the busiest day to work out across the gym. What's interesting is, in North America, there's definitely a higher trend towards what I would conversely say is kind of coined as like the bro split or like a traditional muscle group split workout. So like a chest day, usually the most common ones like a chest triceps day, biceps back day, shoulders and core day, and then a legs day. That is actually the most common split in Canada and the US versus in the UK, or Europe and Australia, you see a lot more of like push-pull splits or like upper body lower body splits, which is interesting, and I assume has some cultural roots as well. So bench or chest is the most common on Mondays, which is interesting. And so people start out their week and you know, things that are to be honest, probably not that surprising. But I think the most interesting stat for us was just the split of the workout types between particularly North America and then rest of the world.”

Julien Redelsperger: “And what about the timing? Like do people like go to the gym early in the day or late at night? Do you have any insight on that?”

Andrew Just: “Yeah, so for us, what we see is about 60% of our workouts in each specific time zone happen before 9:30 and after 5 pm, which is not that surprising, I think, around the workday. And that's pretty consistent with what we've seen across demographics as well.”

Julien Redelsperger: “And could it be helpful for the gym to, I don't know, have a better time management or putting more trainers early in the day or late at night or during the weekend or etc., etc.?”

Andrew Just: “Definitely.”

Julien Redelsperger: “Do you ever think of, I don't know, partnering with a gym or with a franchise to work together?”

Andrew Just: “Yeah, for sure. And so the big thing that we've kind of explored is how we could help gyms, even in terms of like laying out gyms around equipment, because we can actually see exactly what exercises people are doing. Now, we're definitely privacy-first. And so any data that we would share would have to be completely anonymized and not in any way tied back to a user. But one of the things that I've always thought about that's interesting is, okay, if you know that chest is most popular on Mondays, if you think of a gym that's maybe in a major city like New York and trying to optimize real estate versus equipment costs, they can actually change the equipment on Monday, maybe they take out two treadmills and put two bench presses in on a Monday. And that's going to yield a much better experience for users. And what we could even start doing is looking at more specific trends on within this gym, does the cohort of people change from different hours of the day, maybe during the middle of the day, it's more common to do cardio exercises, and then at night, people tend to do more like weight training or things like that. But you could actually start looking at each location to try and understand like trends of what people do to better optimize the equipment that's available for users. It's like real-time optimization for gyms.”

Julien Redelsperger: “And so what about you? What's next for Train Fitness? Do you have any upcoming features or any innovations you're very excited to talk about?”

Andrew Just: “Yeah, so always improving on the strength product. I think towards the tail end of 2024, the two major releases that we're bringing out is so we're releasing our own version of an AI personal trainer. And so a way to build workouts based off of your previous experience, your previous exercises, but also as I kind of mentioned data that you would choose to share with us through Apple Health. So we'll be able to see how long you slept or any sort of health data that you give us permissions to receive. We'll take that data and use it to better inform and to better create workouts for you for today, tomorrow, and into the future. So that's something we're very excited about. And then towards the tail end of 2024 and into 2025, we're actually going to start looking at expanding into adjacent sports as well. And so strength training is an area that I'm super passionate about. But the anaerobic space extends much beyond that. And so martial arts, racket sports, pickleball, tennis, badminton, even some team sports like basketball or ice hockey, as an example, are all sports that human activity recognition could provide very valuable insights for. And so we haven't nailed down which sports will come first. But into 2025, we're definitely looking to expand the horizon.”

Julien Redelsperger: “And you're still going to stay on iOS ecosystem and Apple ecosystem. Do you have any plan on moving as well to Android?”

Andrew Just: “It's a good question. We definitely have a lot of users that have requested Android support or support on Garmin or Samsung or other Android watches. For the time being, only iOS. I think we've just pioneering in the space. We learn so much from our users every day, every week. It results in the product changing. And until we're sure that we have absolutely nailed the experience and the product stabilizes a little bit, we likely will not switch to two platforms. You know, we're at a state now where we're making big improvements and changes on a daily basis. And it's super, super exciting, but it's a very volatile development environment. And supporting two different tech staffs would just be a more challenging approach.”

Julien Redelsperger: “Okay, understood. Perfect. Well, thank you so much for that. At the end of each episode, the guest of the day must answer a question posed by the previous guest. After that, you'll have the opportunity to ask a question for the next guest. Are you ready?”

Andrew Just: “I'm ready.”

Julien Redelsperger: “Perfect. So here's your question, courtesy of Amr Awadallah, who is a serial entrepreneur, investor, and founder of Vectora. We can listen to his question right now.”

Amr Awadallah: “The question is, is it really worth it to spend all this money and research to go to Mars? When we have this planet with so many problems that have not been solved yet, can we solve the problems on this planet first and get this planet to be more peaceful, more healing, more unified before we try and go and spend all of this money to move us to another planet? Like let's fix this house first. What's your answer to that? That's my question.”

Andrew Just: “Yeah, my honest answer is I think yes, I think it is worth investing and spending money trying to go to Mars. And it's challenging to say that because particularly now, there is so much political instability, unrest, and major problems going on in the world right now that need to be solved. But I am an absolute huge believer that progress is what drives innovation and going after these challenges and forcing ourselves not to stay stagnant is how as a civilization we will evolve. And so is living on Mars or going to Mars something personally that appeals to me? Probably not. But do I think that the pursuit of that is a worthwhile challenge and one that will teach us a lot? I think absolutely. And at the end of the day, if you want to get really philosophical, we're on this planet for a very short period of time. We have to spend our time doing something. And so I think stretching and reaching for the next great thing is going to yield technology and solutions and innovations that will help the planet. And I think it's always a good idea to be reaching for what's next, the next bigger thing or the next harder challenge, because that's what keeps us moving forward.”

Julien Redelsperger: “Cool. Perfect. I keep that answer. Just one last question for you, Andrew. If you hadn't been an entrepreneur and launched Train Fitness, what would you do today?”

Andrew Just: “That's a really good question. So I hadn't been an entrepreneur. So growing up, to be honest, I for about seven years, I worked at a bike shop and I fixed bikes, I sold bikes, very, you know, a more manual job, but I absolutely loved it. And when I was in recently after I graduated university, the bike shop that I'd worked at for seven years and the owner of that bike shop has become an incredibly close family friend of mine. He was trying to sell it and I thought very seriously about buying it and just running the bike shop. And I think it would have been a completely different life than I have now. But I think I really, really would have enjoyed it. So I think I'd be running a bike shop.”

Julien Redelsperger: “Okay, nice. Well, it's never too late. Maybe in the future.”

Andrew Just: “That's true. That is true.”

Julien Redelsperger: “Exactly. All right, perfect. So now what question would you like to pose for the next guest?”

Andrew Just: “Yeah, so coming from a tech background, I'm going to ask one relevant to the tech space. And so in the last 12 months, we've seen a huge influx of companies building on top of effectively open AI or chat GPT, large language models. And I'm sure we'll see those coming out of Lambda or what was previously Bart or Google's models as well. So the question that I have is, are all of these startups and companies that are building AI this or AI that or basically wrappers around chat GPT, are those real companies? Are they going to be around in 10 years? Or are we in a bubble? And we're going to see similar to what we saw in crypto, a huge rush and then a huge bust and then pop of that bubble as well?”

Julien Redelsperger: “Well, that's a super interesting question. And I'm looking forward to hearing the answer actually to that. It's very interesting. All right. Perfect. Well, thank you very much, Andrew. It's been an absolute pleasure speaking with you today. Thank you for joining me.”

Andrew Just: “Thank you, Julien. And have a good day.”