Revolutionizing Athlete Coaching with AI and Human Hybrid Models

Kamal Datta: Hey everyone, welcome to What Runs Around Pod and a special series on Tech in 26.2. I'm here with Matt's. welcome Mats. and this is my first time doing a in-person recording. So I'm looking forward to it. And hopefully there are less glitches that I need to take care of it. But it's glad to have you here and meet you in person, Mats. Hey, thanks. Likewise, Kamal. So let's start with a quick introduction, Mats. I think you are the Founder of Hybrd based in Boston. You're also a Y combinator alumni. Let's start with how the idea of hybrid came in, and if you can walk us through your journey of starting the company. Sure, happy to. So you're exactly right. We're building hybrid, which is an AI agent for athletes. And how we came to the idea was my co-founders and I, we were all, and we all are athletes ourselves. And we noticed that the the wearables and the fit tech industry, like that whole ecosystem is getting really, really good. You know, we use Whoop to track our sleep and recovery, our garments to track our runs and our rides. we're using training peaks to get our training plans and we're using Strava to like share our work with friends. All of these are great products that we're really excited for, especially Whoop. We we really love three out of the four on the founding team. are Whoop alumni, so we we feel we've always felt you know that's this great product. And but we noticed that, you know, even with all this this abundance of information, none of the products were really doing a great job of coaching you towards a specific goal. And we felt that the reason for that was that one, they didn't have all the right information in the right space. And they also didn't actually understand fundamentally like what is your goal? Like when is your race? if you're not training for race, like what is your goal and you know, what's actually on the training plan from each of those days. Yeah. So we have this idea that there'd be something in the space. And then fortunately as AI models have gotten better and better, turns out you actually can make really good coaching recommendations with if you have the right combination of data model and some of the world class human coaches that we collaborate with. That's good. And that was the initial thought you had when you were in Whoop that that was a gap. We had the initial idea a little bit after leaving Whoop. a number of us worked at another startup here in Boston in the in the mobile app space. Okay. and but that's when we started the you know getting the idea. And we were super lucky to get into Y Combinator right when we started the company. And pretty impressive, yeah. During the it been off in the race since then. Yeah. Cool. We have Strava, as you mentioned. as you said, you have Whoop and Training Picks. I think I had the opportunity to chat with Dark Freel of Training Picks on the pod as well. I know he mentioned still, I don't know if you had an opportunity to listen to the pod. I think I specifically you have live coaching, in person coaching, and you have the AI based coaching and how do you see the industries going? I think I'm talking about almost two years ago I spoke to him. So since then a lot has changed in AI. He had a thought maybe that you still have to have the human in the loop. I'm curious to see as you solved to come up with a personalized feedback to the athlete itself, how does it play into the role of a human in the loop? Totally, yeah. I think human coaching is really important. And I think human coaches will always play an important role in the coaching industry. Yeah. there's just certain things that NA agent is not good at, like, Deeply understanding your psychology, understanding what motivates you, inspiring you to be the best version of yourself. But there's a lot of stuff that a human coach has to do that's kind of rote or like grunt work. Like you could imagine if you were training for, you know, a marathon and your your coach you know has to shift workouts around on a training plan around you know, let's say your rest day changes or you're traveling for a week or the weather looks bad. Yeah. These are sorts of things that Actually are just kind of grunt work for the coach and don't let them focus on the things that are really, really important to coaching, like understanding you as a person. Yeah. And so we have always thought of this kind of hybrid approach where today all of our training plans are designed by world class human coaches. And then the agent is what's just modifying those training plans depending on the data that we see about the person and their progress. So understanding like what equipment do you have, what what are your time constraints, what progress have you been making. and we expect that to get more and more proactive as we can add additional context, like you know, what's the upcoming weather in your area, those sorts of things. that is interesting. So all the plans are not AI generated. You have human generated the plan and then they're tweaked based on the person who are assigned to that plan too. That's correct. We have we have delineation internally of like, okay, is this something that an AI agent should do or is this something that a human should do? Yeah. And I think we want to keep running with that where you know you could even imagine that future versions of hybrid, perhaps perhaps you can, you know, we have a number we have many athletes on the platform who who love hybrid, but they continue to work with their human coach. and they're asking us like, hey, can't you like it'd be great if you know my human coach could be on hybrid as well. And like I'm, you know, the coach is giving me the Broad strengths of the training plan, then I'm executing the training plan and hybrid is making the modifications as necessary. So I could totally see a vision like that in the future too, where it's you know literally a hybrid between the human and the agent. Cool. Let's talk about because this is very, very fascinating. But this is one of the things I've talked to lot of folks on. Like I I I ran London a couple of weeks ago. Congratulations. End of April. And and I was talking to a sleep Scientist, just before that. I think I just released the conversation last weekend. and then we talk a lot about the role of jet lag or travel on your athlete performance, right? And I think you're alluding to the fact that there are a lot of like weather is one of the factors you mentioned. I'm just curious, like, what are the data sources you're pulling in to say, hey, this is your plan, the Coach said. Yep. And this is the data points I am seeing. So that's the modification you should think about. Like you're supposed to run a 16 mile say for next weekend. Yeah. Based on where you are, the data we are seeing, we recommend X versus Y, right? Totally. What are the those data sources? Yeah. So currently we've been very focused on this initial customer of someone who identifies as being a hybrid athlete. So someone that does curious, yeah. And cardio. And so the data sources for that particular athlete has been yeah, we all want to understand like what are all the Cardio workouts like running and cycling that you're doing, and what sort of weightlifting are you doing? And so to capture that data, we can connect with basically any wearable on the market from Garmin, Koros, Whoop to pull in workouts. and then we built this seamless way to track your weightlifting in the app itself. So you could, for example, just I think we were on the one of the first apps to build something where you could just like basically type into the app, tell our LLM like I did five by five squats at 185. And then that gets categorized and and logged. so those are the primary data sources. And you know, we d we're very interested in adding additional data sources like sleep and recovery, calendar, weather information, that sort of things. We're just kind of going one at a time to make sure that the experience is great all the way through. that's what is hybrid athlete? I'm curious too. So you you you you you coin this phrase as you're building hybrid, right? Yeah. Yeah. So a hybrid athlete, As defined by the person who invented this term, is a guy named Alex Viada, who his him and his company Complete Human Performance are actually those world-class coaches that I mentioned that we partner with. and Alex coined this term in 2014. And he basically means someone who's trying to do two different modalities concurrently. so in the classical definition that's someone who's like, you know, they're running for Half marathon or marathon, but they also want to spend a lot of time in the gym and get strong. And I think if you look at search trends, for example, that the interest in the hybrid athlete world has like 10x over the last few years. I think more and more athletes are realizing you know, if you're you don't want to just be like doing all of one thing. Like if you're a pure runner, you might be very well suited to do a little bit of extra strength training for like injury prevention. frankly, some people like the aesthetics as well. Yeah. Or if you're just a total gym meathead, like, you know, you you're probably best off like incorporating some walking or running or cycling to improve your cardiovascular system. we actually especially see this trend amongst women as well, where there's been a boom in the recent years of of more and more women getting into strength training as well. and still keeping a lot of the other modalities that they're doing before. so it's a really exciting space. So it has to be minimum two modal ADC. Yeah, that's how I would define it. Yeah. It's a bro it's an expansive definition. And so, you know, we see anything from you know, the the runner and lifter that I mentioned, we see a number of triathletes who are also getting into strength training. you know, it's a little depending on the modality, you have to tweak things a little bit. recently we've seen a lot of like more military and tactical users as well. and it turns out I I think I think everyone's kind of a hybrid athlete wherever they're best off doing more than, you know, just one thing. Yeah. I I wonder if you see the data of athletes, do you feel like people are not a hybrid athlete at their segment of folks? I think I think the there's a you know, most people, I think, if you actually ask them, want to be something like a hybrid athlete. Like they say, I want to run and I want like, you know, I think lots of runners or cyclists want to do some weightlifting in the gym. We're like, No, they do should do a little bit, and vice versa. I think most people or many people that are great weightlifters at this point probably realize they should do a little bit. Yeah. balancing those things is really hard. And so we see a lot of users who in the past have struggled with that. especially around things like the what's called the interference effect, where you know one of those modalities can interfere with the other when it comes to making progress. And so and that's kind of why we create hybrid in the first place is to help you through that that struggle. Yeah, I'm I'm very familiar with this. So I enjoyed my tender night is doing polyometrics through my London training too. So it was just a couple of weeks before. So to your point, it like can happen. I'm sure it happens a lot. A lot of other folks do. is injury prevention one of the outputs of or outcomes of hybrid? Totally. Yeah. So you know, to go super tactical with it, for example, one of the analyses that we show all of our users is the ratio between their chronic load and their acute training load, which put differently is basically like how quickly are you ramping up your training volume across different disciplines. And generally like one area where injuries can happen is if you get super motivated to become a hybrid athlete and then you start going from zero to a hundred in the gym or vice versa, you're a lifter Like, okay, I'm gonna you train for this Iron Man and immediately start doing all these disciplines. That's you know, not only but one of the very common cases where someone can injured and hybrid can tell you, like, hey, you know, you're actually on ramping a little too hard here and recommend a different approach. Okay. that that's interesting. And what other feedback an athlete can expect from hybrid? Yeah. So I'd say the number one piece of feedback that you should look to towards hybrid for in the same way that I think this is one of the best functions of a human coach is to help you with your consistency. I see. I've really learned you know, I was mentioning a little bit before, like I I had the honor of being on the US national rowing team for for a while. and what struck me about even at that level was that like each particular workout, yeah, of course there were some hard days, but like each particular workout, you know, wasn't like the hardest thing I ever did. and wasn't like anything that you wouldn't do at a lower level. Like, you know, there's there's no like secret workouts at that at these high level or like magical workouts. You're just doing more consistently. and there's much smaller margins of error for like, you know, missing a session or missing a few sessions. Right. and so that's kind of inspired how we think about Harbor as well, where like, you know. Training plan that we provide you is really important. But a lot of where I think the magic of hybrid is is that we can also adapt things in the cases where, you know, you might know you have something that would cause you to miss a session or you might have missed a session, and we can help you figure out like, you know, how can I get the most out of the week that I'm traveling, or you know, I did end up, you know, life happens. I'm I'm I you know, I miss a couple of workouts, like how can I safely and efficiently get back back on the wagon? So the consistency is the one what High but you are suggesting that it is is helping athletes to get to even if you say life happens kind of scenarios, how to get back to the consistency. I think you know, if if I could wave a magic wand and give any athl any athlete like one particular attribute, I'd if I could give someone consistency, like I think that would be the biggest. The biggest one. Let's talk about you have a background in athlete or you're an athlete yourself. If you mentioned that you're in the rowing national rowing championship, how does it shape you? coming with the idea and actually, you know, bringing it where it is today. Yeah. Hybrd. I mean, I'm super lucky to get to build something that I actually want to see in the world. I think one, it makes my job a little bit easier where if I'm deciding like, hey, should we take the product in direction X or Y and like, well, you what would I want as a as a user? Yeah. That's really fun. I think it also guards against burnout a little bit where you're like, I'm super lucky to get up to work on the thing that I want to work on every day. Yeah. Yeah. I also think that just being an athlete is good preparation for you know many things, but especially being a founder. being a founder is a very uncertain thing. Like you don't necessarily have a guarantee that things are gonna work out. But I think athletes learn that again, back to the point of consistency of you know, you get a lot from just showing up every day and yeah. If you are if you like trust that process, like usually good things will come. You know, there's some things you can't control, but just control what you can control and keep working hard and like that's generally better than the alternative. That's amazing though. That that's good to know. The showing up, I think there is a saying, right? Yeah, if you just show up, things are supposed to move forward and it it helps you. Yeah. The conference is true too. If you don't show up, here is no chance anyway, so you might as well show up. yeah, yeah. That that that is so true. Yeah. That's good to know that your background is shape. you mentioned about data sources and the feedback you're absorbing different inputs from either different apps or different dimensions to give a a constructive feedback to the athlete, right? To you know, either to get to consistency or how to improve or give a feedback that is slow down and either ways, right? we live in a world where pretty much for every data point. You have a device, right? Or you have a device collecting a sense of data sources, right? So I'll take an example of a runner since you know we would they're not running point. So I may have the Apple Watch or the Woo band gives me certain data points. a lot of runners wear hard rate monitors. now we have smart shoes, can give you feedback on shoes. I think I had an opportunity to talk to Roy. I think it's also a possible bit right now in Miami. But some data collecting on shoes. So when you talk about all this, then we may have my favorite watch or I may have some other monitors that and I'm collecting data point. Do you think that there is an overcollection of data compared to what feedback actually athlete is getting there? What what you are seeing as you're building hybrid? Totally. I I think it's a good question to ask. For one, I think the most important thing about using any of these devices is just realizing. They're all just one tool in your toolkit. Yeah. some of the best athletes in the world, historically and even today, have used none of this stuff and gotten really good. So, you know, I think always wondering like, hey, am I still chasing like a particular goal or outcome or habit? Or am I like, you know, living for, you know, not to pick on any one of these companies, but like am I, you know, does my strava fitness score mean more than like yeah, r running the actual race? So first of all, like th they're all just tools in the toolkit. But that being said, I think all of these things can be really useful tools. I've found Whoop over the many years I've been wearing it and was working in the company, like a really, really helpful habit formation tool. you know, it just seeing the impact of like a night out or staying up late or eating a late meal or something like that. I always knew that that was not good for my performance, but when you see it in the data immediately, Maybe not for everyone, but from my personal point of view, it really changes your habits really quickly. And so I think that was a good use of of that tool. if you find yourself stressed out about you know your wearables data and things like that, I'd you I just encourage people to try taking them off for a while and see if it works better. Like these aren't you know, there's no laws here. Yeah. the other point on the over collection of data, I think that's a risk, but I think that particular problem can be solved by. tools like hybrid of like, okay, you know, now we have all this data. Some of that data is really high signal and a tool like hybrid can recognize that. And some of that data is noise and you know hybrid can also discard that data. So I think as the agents that are building this are getting better and better, like we'll be able to use the useful stuff and discard the the noise. I see. i you talked about noise. Do you think there's a noise and there is A meaningful feedback, as you mentioned, right? Yeah, getting people to consider that's the outcome you're trying to dive, right? Right. You talked about Woob, that, you know, helping you to build new habits or change your habits. because with all the data collection, as you say, there are so many data points. And how does a general user who probably doesn't know what this scientific impact of those each data point is, make sense out of it, right? How do we make those data point and meaningful feedback for the end of this like okay I can I understand this what it's saying versus here you go this 20 yeah gadgets or 20 you know data points and I'm trying to figure out what does it mean for me. Like I don't know. So is that a problem you see as you're collecting as I think you're sitting in the intersection of data points right across all these different platforms and sources that's coming in. Yeah. Totally could be a problem. I think that's where think you our message around consistency is so important of like let's really encourage people to do the basics first. Like you should have a certain training volume depending on your goal. You should have the right blend of strength and cardio if you know if you're wanting to be a hybrid athlete. That's where like 80% of the value is like doing those things consistently. And frankly you you can do that without a wearable at all. Like you could or without any tool at all. Yeah. But I think that tools like hybrid or like the other products that we mentioned can make that more fun, more enjoyable, frankly more addicting in a positive way. And so then I think they're really useful tools in the toolkit. you know our users especially like talk about hybrid being a great way to offload some of that like mental load of trying to sift through like, you know, what's the what does the sports science say about this? Or you know what are all my different gadgets telling me. Right. and so I think these tools should be making your life easier and not harder. And that's certainly what we're trying to do at at hybrid. And I think the w if you're an an athlete listening to this, I would just put all of your tools under that microscope as well. Like is this something that is making me more excited to train and making me feel good about my training? Or is this something that's causing me more stress than the actual goal that I have? And like is it if it's not serving you, like it's okay to try stuff and and move on. That's a very good point. because that's that's the lens probably everyone should evaluate these tools and devices that they want to use for sure. you mentioned that you use humans building the training plans. Yeah. And with where we are with AI today, why do you think it's still important to have that? Or why not what Rana does? Sure. Yeah. So I think this is something that's changing really fast. Yeah. I think that whenever you're giving sort of like health-related advice, it's good to have an extra person who can like look over things and make sure that you're going in the right direction. I think that, you know, the exact engineering constraints here might change over time. What I mean sp specifically is like, you know, perhaps we could move away from having A each individual training plan like drafted by the human, but we we'd still want to have humans involved in like the the testing of an agent's output. And so like being able to build tests for our agents that say, like, hey, these 10 principles in training are really important, like don't violate these, or these are situations where you know, if a user asks hybrid about an injury, like we'll always say, you know. Here's our recommendation, but like you should talk to a a real doctor if you know on under under XYZ conditions. And so that like human imprint is is really important, even though I think that like frankly the even just in the in the times that we've been working on this, like the performance of the model and on some of our internal benchmarks is is really blowing us away. So I you know, I I would expect that we'll use more and more agents in the future, but we're gonna d we'll continue to do a good job of Steering them with with human oversight. Interesting. And you mentioned about engineering constraints. What are you talking about here? Yeah, so whenever a hybrid answers a user question, we have additional agents that can kind of like evaluate the agent's response and judge those on various rubrics. And so what I mean by those constraints is like us telling the agent like these are the parameters in which you can you can operate, these are the parameters in which you can't. I think that level of like human control over the agent is important in many different domains. so this are these more on sort of the guardrails you have on the agent output that you're putting on. Okay. Combination of guardrails and then the context that the agent has to to make its decision. We actually Alex Viada, the guy I mentioned before, he he originally wrote the first book about hybrid athletes and 2014, but he actually recently released an updated textbook called The Ultimate Hybrd Athlete. And through our partnership with Complete Human Performance, a lot of the content out of that book is used as context in the model to make sure that like the model's responses are grounded in world class coaching and physiology as opposed to you know who knows what. What AI knows about the shelf. Yeah, cool. No, this is this is interesting. And I think I would guess you mentioned about rubric. So the rubric probably is human input going into either the guardrail or whatever the context you're giving the models. Yeah, we work closely with our with our coaching partners to you know we'll test a bunch of different scenarios and rank how we think the agent did. it's actually a really interesting problem of like how do you evaluate agents output at at scale. I'd imagine there's Other companies doing this in different industries too. Yeah, yeah. Yeah. So I'm just curious now, how long does it take for you to say generate the plan for say half marathon? If you have to consider all this constraint that you're doing with? de depending on the constraints, I mean, it's pretty fast at this point. That's the other thing, like latency has just come down really, really impressively. Like we could probably generate the whole plan, you know, on don't quote me on the exact number, but on the order of seconds. okay. Which is which is exciting a lot. What's been interesting as we've built this is like, for example, like the first time that we built, I briefly mentioned that we have a way to make logging your weightlifting really easy. Remember like a year and a half ago when we released that. it was like the slowest thing ever. Like you tell it like I did five by five squats, I want eighty five, and then you just like sit there for two minutes while while I was processing all that and like You know, there's things we could do to make it faster. And I think these models just keep getting better and better too. Where it's like, you know, something that's really fun about building in this space is that, you know, you could build something that like almost like almost doesn't work right when you build it, and then you just wait like a month and like magic happens. Magic happens. well. Is it a voice prompt that giving like I'm manually dictating? Most of our users right now just type or use the Apple Like voice to text. We are very interested in voice agents as well. just trying to stay focused on like a bunch of things on the roadmap as well. Right. Now you mentioned that it's taking two minutes loads. I like, Hey, I dictated and now it's actually loaded and I I do think that'll be one of the form factors that people will use hybrid, once we are get around to building the the voice agent side of thing, because I agree that'd be such a fun and convenient use case. You know, people have their AirPods in in the gym could just say what they did and dislog. Yeah. Yeah. Yeah, with eleven laps and all that getting so advanced now. I don't know. It's crazy. I mean, I I've started to get on calls where I'm like, I'm not even sure if that's a human or not. yeah, yeah. those calls that you you get. Yeah, yeah. It sometimes you're like having a normal conversion after a couple of minutes you realize, am I really talking to real people? Yeah, someone else. I know I think that just goes to show that really, you know, are I think it's really dangerous to make predictions like too far into the future right now because you know. You might have thought a year ago that an AI could never you know give you reliable like athletics programming advice. And I'm like I'm gonna say that like it's getting pretty good and like it's gonna continue to get better. and so just it's hard to forecast these things in in in the future. Yeah, I mean, so I'm not a Rana subscriber. but I know the folks that you know the run it's it's a great company. Yeah, basically app that built. so the first time on the London marathon, I actually did a AI generated training plan, which I followed. except my priometric messed me up. but it was to your point, I gave the different data points that personal to me, not maybe Strava data, but you know, things, my routine, my kinda Times that I think I work out like, you know, with family and all sometimes I cannot just go out for like two hour strength training. It's probably not that's the luxury for me. but you know, how do you give those constraints? And sometimes I say travel, so give those. And it it did a great job. At least it's not it's more like not trained like the way you're defining hybrid as it has a real coach giving input input. But it was whatever the model had, I I used Gemini. But to the point, yeah, it was and I give an outcome. I went to run under four hours and it did a great job and it was a sixteen week plan. to a lot of action it takes me to where I was pretty close, except the the injury part. But you can rely a lot on those kind of plans, right? What model can generate. And if you are tweaked away with professional inputs and coaches and it especially intentional caliber, it can get you to you know the outcome that probably you desire. Yeah. Yeah. Totally. Maybe we could have even helped you avoid that biometrics injury by you know I would be curious, you know. I mean that's maybe I'm sure runners get injured all the time. Of course, it's part of the sport. Yeah. If you work with runners, you probably see it's you know a couple of them. It's the same ones popping up some other. Yeah, it's a matter of time. Yeah. Mats, so what are the trends you're seeing in this space? Yeah. Well, you know. On the wearable side of things, like I think the wearable boom is just starting. you know, if you look at both Wooper and Aura are like decad coins now, which you know when I was when I started out Whoop, you know, everyone was like, What the what the hell are you doing? Like this joining joining this company, and now it's like you know, the most common thing ever. so there's gonna be more data. We are just talking about how the agents keep and the models keep getting better. and then I think the Strength training is having a moment too. you know, I mentioned you know broadly, but it's also especially women are are getting more and more and more into strength training. in the running world, like you see athletes who are maybe kind of moving away from the idea that you just need to be like as light and skinny as possible. Like it actually helps to avoid injuries to be a little bit more robust. Yeah. and then I think in the broader fitness world, like, you know, we might actually I'll be curious to see and I hope that say what you want about GLP ones and those sorts of things, that there's gonna be a whole new segment of people that previously felt like the fitness world wasn't accessible to them, that now are gonna become, you know, you know, one much more able to work out, but also to realize like, hey, I should do things to counter the side effects of GLP ones, and that might be like strengthening. Strength in cardio. So I wouldn't be surprised if a few years from now we're we're seeing that the fitness market is even bigger than we thought before. that's amazing. and you do see that there is more space for devices? Yeah. I mean we just saw the launch of of the Fitbit Air, for example. Yeah. I think that as markets mature, like we'll see more segmentation around price and brand and and those sorts of things. There's there's a lot of people in the world that have heartbeats to track. So I think I think it's gonna be a pretty big market. cool. And what do you see that the trend is not telling you that, you know, should be there but not there? Interesting, yeah. I mean, I think the million dollar question is just what is the role of apps like hybrid in a world where cloud and chat GPT are really powerful? Like will we will we go to a world where you just do everything in Claude and you have no use for you know any other product. Or will we go to a world where you know what I think is gonna happen is I think people are starting to call it like the bring your own tokens model where you know you can have a lot of the intelligence that you want in, you know, I'm using Claudaton right but could be Gemini or whatever, like in Claude, and then you for that like specific task that you want to do, then like you can connect VMCP to an app like hybrid where you're like, okay, I have this training plan that I built in Cloud, but now I want hybrid to help me execute on it, make it really fun, update as I as I progress, send to my wearable, send that wearable data back to Claude. And so you'll have this hybrid experience between maybe a few of these like super powerful LMs and then the apps that help you actually use that intelligence for really specific Tasks like training. And how does the commercial availability work in that in that model that you mentioned? It's very interesting, makes that you're super interesting. Yeah. I think it just I think if you're someone who gets you want to position yourself to be in a position where as models get more exciting or get more powerful, you get more and more excited. Yeah. Like if a new model release makes you really nervous, I think that's a tough spot to be in. But for us, we're like, hey, new model just got released. Our weightlifting input is getting faster. Maybe like in the future we have voice and like that's getting more realistic and feels more like talking to a coach. the the plans are getting even more reliable and accurate. Like we feel like hybrid gets better as these models gets better. And so we think that's a nice place to be. but yeah, certainly if if a new model release makes you really worried, then it's it's tough right now. It is true, yeah. And with this whole trend is just like okay, they're learning. If I have a special niche, probably they're learning tools that and then maybe it becomes another at release for a model. Yeah. And then pretty much you're talking about okay, what happened to my vertical that I built. Yeah. And I think like as as evidence, I think cursor is a great example here where like, you know, Everyone has access to these models, but Cursor managed to create a great experience of like using those models in the context of software engineering. of course, there's other tools out there like Cloud Code or whatever. But I think that's a great example of like, you know, I think right at the start of all these models coming out, everyone was like, like, you know, these are just wrappers around the cloud, like it's nothing. And then you see Cursor has an amazing exit because like it's actually real important to match. Take this like raw intelligence and put it in the right form factor for something very discrete. And software engineering was a natural first candidate, but I think it's gonna happen in many industries, and I think the fitness world is one of those industries. Matth, have you noticed, and since you are in the Y Combinator ecosystem, and you've probably seen a trackload of AI enabled companies within Y Combinator or outside. as it becomes easier to build something, where do you think is the differentiator to actually have some traction on the product and service? I I mean, you probably talked to a lot of founders or mentors there. For a company like hybrid, I think those differentiators end up being one quality. Like to your point, everyone everyone and and their cousin has like, you know, their the thing they vibe coded overnight. Yeah. And then every day we'll see like some new thing like, this looks a lot like hybrid that someone just vibe coded overnight. And then like a week later you forget about it because like that person didn't continue to improve it. Like it's actually we've been working on this for like two years now, almost two years now. It's actually pretty hard to get like all those moving pieces that I was talking about working together. And so like I think it's gotten much easier to start, but I don't think it's gotten that much easier to like stay in the fight as long as you need to to to make it. actually in some ways thinks it get think it gets a little bit trickier where if every day you can just start some new thing that you're interested in, it's actually a harder and harder. And we face this internally too of like, we have to focus on the thing that actually makes hybrid good. instead of being like, yeah, we could spin up this, you know, XYZ feature and then you think it moves really fast and then like those last twenty percent still take forever. so yes, easier easier than ever to start, but maybe harder than ever to stay focused and to stay focused on something the amount of time required to succeed. Is staying focused is the I I'm trying to get to like what is the mode when the cost of building is like kind of some tokens you're talking about. Sure, sure. So I think focus can help you on the quality aspect of things, which I think is goes goes a really long way in terms of like actually having the product experience that one use. Yeah. And then I think related to that level of quality also goes into random distribution. yeah. getting how do you stand out amongst a sea of like a hundred people trying to build something that you're talking about. Especially the people who are white coding things like overnight you're mentioned. Exactly, yeah. on the one hand, a lot of noise. On the other hand, in some ways it's almost more like the bar comes down, where like people are so used to seeing slop that when they in when they engage with something that feels purposeful and high quality, they let us know about it. as an example, like even two years into the company and with like thousands and thousands of paying customers, we don't have a customer support line. Like users can just press a button in the app and it opens literally an iMessage to me or my co-founder and can text us directly with like any issues that they have, feedback, and that level of feeling like a human cares about the experience as opposed to like, hey, I made this overnight and post on Twitter and like never talked about it again, has really helped us stand out. Build this brand that people are excited to support. it's awesome to see people wearing like hybrid hats in the wild and those sorts of things. nice. How did you come up with the idea that the people can actually directly contact you? Or what I'm sure something driven, some experiments driven. Yeah. so during our YC batch, this was right when Brian Chesky, the founder of Airbnb, was going big on like the founder mode stuff. and he like came by and gave a speech on this and like really fleshed out those ideas. And so like a lot of the times we were just asking yourself, like, okay, like you know, now we've got a bunch of users, like, how should we do support? And then we're like, what's the most like Brauncheski also laid out this like idea of like the 10 star experience? Like, what do you think would be like the the like almost absurdly good experience? of you know, of course, having to talk to support is like never usually a great thing, but I was like, what if they just like open a text message to to us? Like it's kind of goofy, but let's just see what happens. and I've really enjoyed that as a feedback loop and we've tried to take the same principle to like any issue that we're facing. Like what's the thing that would make the user think like, Hey, that was like that wasn't just like fine, that was like absurdly good. I I have listened to brands, I think he talked to Reed Hoffman on masters of scale pod with the tester, you know, what do you do? Yeah. Do things that doesn't scale as a founder. I think it talks about it. is it? Okay. It talks about his New York experience when I think he came to New York and I could take the photos and all. this is good. Is it scaling? Yeah. Yeah. Yeah. We're super pumped with our with our growth. And even that like that text to founder thing, like yeah, in my opinion, so you know, at this point I'll I'll probably have text conversations with like 10, 20 users a day. Wow. Which is Which is, you know, not not nothing. But on the other hand, I'm like on with my role as being the guy on the team most focused on product and design, like the two main things I can do are like build the product and and talk to users. Yeah. and so like I'm happy if I don't think we'll change anything about that system until like my literally entire day is just going back and forth with people. Yeah. And then then that'll tell me like, Okay, well maybe we need to hire someone or do or do something in that space. But until then, like I'm super thankful for anyone to wr that writes in because that's just what's gonna get us closer and closer to to product market fit. And I'm I'm I would guess you've got so many feedbacks probably are putting back into the product to improve it. Yeah, yeah. I mean super exciting when in a world where you can like build these have these agents that can go, you know, build new features for you or fix bugs or whatever. You take a text, you make a plan for how you're gonna adjust it and then like Pretty often like a user can text me in the morning and say, like, hey, I had this issue or this bug or whatever. And by the end of the day, it can be hey, like this is fixed now. and that feedback loop is just really, really tight and something that you know, I w as a startup you have certain disadvantages, but like things like that are how we can stand out at first. you know, bigger companies that are gonna do this. How does the cycle look like? So you're gonna text, yeah. Hey Matt, I or yeah. Somebody text you. Like can you walk through the cycle? 'Cause I'm very fascinating. Sure. Yeah. They'll I they'll they'll text me or my co founder Ben. we have built something where like the start of that message will have basic information like their user ID and their app version and stuff like that. Sure. And then in my head I'm just like, is this like a bug that they're describing or is this like a general feedback thing? if it's a bug, you c I can usually try and like USSL. Yeah, I I just like try and triage it quickly. Yeah. You probably could build something here, but like finale's been fine, like Triage it and then you know, I'm I'm not the most technical guy. I can do like a little bit of damage. So if it's something really small, I can try and like yeah you know, put up a PR. Usually I'll talk to my our founding engineer or CTO. broadly w if it's usually if we see a bug, like, you know, we'll have a couple of users report it. So it's something that's kind of on our radar already, try and squash it out that same day. if it's like a bigger thing, like I wish hybrid could do this or that, then I'm like I'll just put that in my notes and That's really helpful. And I try to give them a little sense of like what does our roadmap look like and when they can expect us to address something in the vein of what they're talking about. cool. And then if it's a feature set that probably it has its own cycle based on a roadmap. Yeah. Yeah, yeah. Okay, okay. this is fascinating that you still address fifteen to twenty personally and you're as a leading product, you're getting the feedback yourself. And then directing it to our sway. I think I'll never stop doing that. Like even, you know, even at you know, hyper scale, like, you know, maybe, maybe I only see like a some percentage of the ones that come through, but I think it's important that like I feel the pain of something being broken. Like it's it's easier. It's you know working at a big company. sometimes it's tempting to like you can just kind of show your laptop at the end of the thing, like that'll be there. Yeah but like when you get a text that like, you know. eleven o'clock at night or about when you're about to go to bed and the user's super upset because like they were so excited to try hybrid and like something dumb wasn't working. Yeah. Like it r really makes you angry and like fires you up and like gets you excited to keep pushing on it. That's true. That's true. Cool. Would you have any other trends you want to share about surrounding industry as you're seeing building hybrid or overall the endurance sports industry. Yeah. You probably have not touched on. Yeah. I mean, one of the trends that I'm interested in, and I think hybrid fits into this vein, that we haven't maybe explicitly followed it yet, is that I think what's exciting about tools like hybrid, tools like Whoop, tools like Strava, is encouraging a mindset around fitness that is not one of like, you know, sure people can have certain goals around their physiology, but like less about like, you know, the aesthetic or looking a certain way or whatever, and much more around. you know, it's just really rewarding to have ambitious goals and to struggle for something and sometimes you make it and that feels good and sometimes you don't, but you know, learning to do hard things and to pursue them. That's certainly things that we want to encourage with with hybrid. And I think or I hope that there's gonna be a broader shift in that direction as you know people think more about like, hey, what does it mean for me to be like a a a human person, like yeah in a world of like overwhelming AI intelligence and like content 247 and like, you know, all sorts of messaging from social media or whatever. I really hope that that will be like a foundational thing that a lot of people can latch onto. and so that's something that I'm excited to to push forward with hybrid. wow. That that how do you how do you see that happening? Yeah. I I guess I don't I don't fully know. I mean, it it's kind of on us to one build a product that can make something like doing your first half marathon feel much more accessible to more people. so I think it's really exciting when we see users who, you know, most of our target customers like in that, you know, intermediary phase. But like we certainly get users who they've done their first five K and they want help to get to their first ten K, or like they've done their first ten K want help to the first Half marathon. Yeah. So again, like making something that makes that experience exciting and enjoyable and addictive is isn't that good. And then you know, we create a good amount of content ourselves ourselves. And I think we try to have a like positive and you know uplifting message around being you know training for whatever it is that you want to set your mind to. Okay. Well that's that's pretty awesome. Thank you. Matt, it is great to have you on the pod. And thanks for sharing what you're doing at hybrid. And I think all the best to the success you have so far. And I'm sure you'll do great. And hybrid will do great, especially the mission you have, right? helping the hybrid athletes who are, you know, everyday athletes who are trying to get themselves better for themselves or whatever reason they're striving towards. So thank you for sharing what you're doing and leveraging technology to the extent you are doing. Yeah, it's my pleasure and thanks for having me on the show.