Building an iPhone app with zero technical skills | Bryce Rattner Keithley

I built an app called Daily Hundreds. I opened Lovable and Replit actually on the same day and left the simplest prompt. It was incredible to me that I could tell these AI tools, I want this, and it spit out a very basic minimum viable product of it. I asked you the other day, I was like, are you in TestFlight? And you were like, yeah, I was in TestFlight. And so now it's in the App Store. You got it approved. It's ready to go. We have anthropomorphic animal demos. This is my favorite part, and we're going to see some pretty, pretty amazing ones. But can you walk us through how you generated this? I make the animal in Gemini. I film myself doing the exercises and then I mash up the anthropomorphic animal with Bryce exercising to create the videos. Were there any hard skills you felt like you developed as you went through this process? I got really good at copying and pasting, but I think I knew how to do that beforehand. I don't actually really know what Railway does. And yet it's there now. The fact that you, a very non-technical person, are buying or acquiring Railway as your infrastructure without really knowing what it is, it's kind of amazing from a go-to-market perspective for these AI tools. And I'm sure saying what you don't know and then trusting the beep boop robot gods also helps you discover and push further than you would if you knew the boundaries of things. Welcome back to How I AI. I'm Claire Vo, product leader and AI obsessive here on a mission to help you build better with these new tools. Today, I have my friend Bryce Ratner-Keefly, who is decidedly not technical. She's actually spent her entire career in talent and people, but she has somehow beat me to the App Store with a vibe-coded fitness app, including customized animal videos for any workout you want to do. She's going to walk us through her step-by-step and show you how a beginner's mindset can mean you can build anything you want with AI. Let's get to it. This episode is brought to you by WorkOS. AI has already changed how we work. Tools are helping teams write better code, analyze customer data, and even handle support tickets automatically. But there's a catch. These tools only work well when they have deep access to company systems. Your co-pilot needs to see your entire code base. Your chatbot needs to search across internal docs. And for enterprise buyers, that raises serious security concerns. That's why these apps face intense IT scrutiny from day one. To pass, they need secure authentication, access controls, audit logs, the whole suite of enterprise features. Building all that from scratch, it's a massive lift. That's where WorkOS comes in. WorkOS gives you drop-in APIs for enterprise features so your app can become enterprise-ready and scale up market faster. Think of it like Stripe for enterprise features. OpenAI, Perplexity, and Cursor are already using WorkOS to move faster and meet enterprise demands. Join them and hundreds of other industry leaders at workos.com. Start building today. Bryce, to start our podcast, can you remind me, as one of your very good friends, are you technical? Claire, not at all. Perfect. So just to get that out of the way for everybody listening, I have my friend Bryce here. She is talented in many, many ways, but I still have her beat on the software engineer index. And yet, she has been the first of us that has vibe-coded a iPhone app to production into the App Store. So you beat me. You beat me there. So tell me how you turned from your background, which I'll let you talk about a little bit, to a iPhone app developer. Yes, so most of my career has been in talent and recruiting, and I've had the privilege to work with really talented, incredible technical leaders throughout my career. And on one hand, through working with these folks, you come to see how talented they are with what they build. And on the other hand, do you come to see that they're living, breathing humans as well. And without being a computer scientist myself, there was this massive gap and full abstraction between who they are and what they could do. And with these new tools that came out, even going back to last fall, I kind of got an inkling of like, well, maybe I can try something too and see what the magic is all about and see what the high is you get from bebopping into a computer and having what you build show up on the screen. So what did you build? I built an app called Daily Hundreds for consumer number one. The story dates back to the pandemic where I spent 20 hours per day every day in this literal room and wasn't moving and had two young kids at home and I craved exercise. And there was this very short-lived hundred push-up-a-day challenge that I started doing, but felt like it was boring and monotonous doing push-ups every day. And I kind of thought, I wish somebody could just tell me something different to do every day. I would happily do a hundred of them. And it kind of came back to mind. I've been doing it intermittently throughout of just jumping jacks or squats here and there between meetings. And I was like, that seems like a lightweight thing that I can try. And so I opened Lovable and Replit actually on the same day and left the simplest prompt of, hey, build me a tool called Daily Hundred that pushes a different exercise to me to log a hundred reps. And it ended up building something remarkably fast. It's not what you'll see today, but it was incredible to me that I could tell these AI tools, I want this, and it spit out a very basic minimum viable product of it. I ended up sticking with Replit because I could phone a friend. I have a friend who's a director of engineering there and I figured at a worst case, he could unblock me, but that was really how it started. And I built this starting in October and I got it into the App Store earlier this month. So it took a couple months, but walk me through a little bit of how you used Replit. I think this is a really great platform for folks who really want to go zero to a hundred into production. We had one of our early How I AI episodes with John Blackman, our 91-year-old vibe-coding grandpa who used Replit to build an app for his church. Replit is pretty amazing when you're building stuff that has a database on the backend that you really want to be functional. So how are you interacting with Replit on a kind of daily basis and how are you moving forward with the app? Just kind of tell me, tell me your process. I tend to think that a beginner's mindset can be used to your advantage here because I truly don't know what I don't know. A big unlock, somewhat from your pod, is that there's a plan mode. I would sometimes get myself in trouble by asking Replit, you know, how can we change the progress bar? The progress bar is like, you're adding some reps and now it's in a circle, but it started as a line. And initially I would say, let's change the progress bar to a circle. And it would just do something bananas. And I would have to quickly like undo that. We don't like that. And coming to see they have a plan function was really helpful and profoundly changed how I would approach things. And tend to start really big picture of like, okay, robot, here's what your non-technical friend wants to do. How can we collaborate on our new idea? It has a preview panel that y'all can see right here, which makes it really easy to see if you're on the right track or the wrong track. Sometimes it magically happens on the right track right away. Other times it goes on the wrong track many, many times over. But it's really all in one and made things very simple to at least work at a small scale. I have a question. So what concepts, development concepts or software engineering concepts did you find yourself learning as you build this with Replit? Were there any things that you just kind of got like a little, you know, because you've done technical recruiting again, you're friends with me, you're friends with directors of engineering. You've been around technology for a long time, but were there any sort of like hard skills you felt like you developed as you went through this process? Oh, gosh. Are there any hard skills that I felt like I developed? I got really good at copying and pasting, but I think I knew how to do that beforehand. I got way better at labeling things so I could find things on my computer. I still don't think I know, candidly, any more about software than I did when I started to the extent that, like, when I learned that, heaven forbid, I had to take my app from my precious Replit and move it to somewhere else before it can be an app. I moved it to Railway. I don't actually really know what Railway does. And yet it's there now. OK, so I actually really, really love this. Let's just, one, we're going to say I got really good at copy and paste. We're going to save that one for the trailer. But what I reflect on this is like, you really are vibing. You are really the true vibe coder. And it's so interesting to me, you know, something like Railway, just like a hosting kind of platform. My friend works there. We know the CEO. We're good friends with Railway. We love them. But the fact that you, a very non-technical person, are buying or acquiring Railway as your infrastructure without really knowing what it is, it's kind of amazing from a go-to-market perspective for these AI tools. Because I think, you know, they used to be selling. I was in dev tools for a long On the mat, the head should be to the left, and the feet should be to the right, the knees repositioned above hips and feet forward in a tabletop position. I have to pause here as my friend Bryce writes this very precise prompt about accurate positioning on your crunches. I have to call out that she has a secret power here, which is she used to teach bar. And so being able to prompt somebody, knees over hips, feet shoulder-width apart, hands behind your head, who knew that all those skills would come into prompting an AI image gen model to make a leopard doing crunches. It's really true. As I often tell people, like, you never know what lateral moves in your career are going to end up being tools that pay off in dividends later. And I am quite good at physical exercise cueing, and now I use it a lot. Okay, let's see what this gives us. And I know a lot of your guests like to speak. For me, some of the precision, I'm visual, so I like writing it out, but I have found that we want to leave no room for interpretation. And I'll often ask myself, can I be any more literal in what I'm describing? Okay, and we're using Nano Banana here in Gemini. And it's creating your image. I have to call back to one of the earlier things you said, which is you originally, your MVP one was using Sora, and Sora is being sunset. So I'm glad you have a different flow here, because you would have been deprecated out of. RIP. I know, right? Yeah, I think it's going to be interesting. You know, one of the challenges of building stuff in these day and age with all these new products is sometimes they end up being experiments that that go away. And then you're like, wait, I was building my very important animal videos off of this. That's a, it's a really disappointing. So I'm glad you got to this, this new, new flow. And then Gemini is going to take a couple of minutes to load and we will just wait for it to show up and then we'll go to the next step. This episode is brought to you by Metaview, because who says hiring has to be fair? Every founder, hiring manager and recruiter I speak with feels the same pressure. Hire the right people as fast as possible. But recruiting is brutally time consuming. Alignment is hard and the competition for great talent keeps getting tougher. That's why teams like Riot Games, Brex, GitLab and Replit, plus 5,000 other organizations use Metaview, the agentic recruiting platform giving high performance teams an unfair advantage in hiring. It works by giving you a suite of AI agents that behave like recruiting coworkers, finding candidates based on your exact criteria, taking interview notes, reviewing every inbound application, gathering insights across your hiring process and helping you identify the best candidates in your pipeline. Don't let your competitors out hire you. Metaview customers close roles 30% faster. Get started with Metaview today and get your first 100 candidates sourced for free at metaview.ai/howiai. Okay, so it loaded. Let's pop it open and see how good it did. So it's kind of flawed. It didn't do a great job. It didn't listen to the hands behind its head. It didn't listen to the head down. It got one leg in the proper position, but it got one like not in the proper position. We're going to try it again. We are creating an anthropomorphic leopard in a gym setting. The leopard should be wearing exercise clothes. And sometimes, and this is probably me not knowing anything and summoning the AI gods, copying and pasting works when we're on a win and starting over sometimes work when you have an L. So I'm just restarting the karma and writing it all over again. This is to match the video because it doesn't do a good job interpreting anything. Knees should be above hips, knees should be bent, feet forward in tabletop position, both feet off the ground. There should be no other characters in the image. Sometimes AI gets excited and adds friends for the character, and we just don't need that. Okay, so just, we prompted it once. You copied and pasted your kind of original prompt, made a couple edits, did not give you what you wanted. You say, don't copy and paste. That's just welcoming more of the same. Instead, start it over, rephrase. We added a couple more literal things together in terms of like saying both feet off the ground or both hands behind your head or left and right and just being very precise. And then you're using Gemini to create this image. We'll see if it does a little bit better. We'll see if Gemini does a little bit better. It's really fascinating. Sometimes we'll get things on the first try. Sometimes it'll take a while. Yesterday we made a cougar and what it got quickly was the cougar. What it didn't get, and I wanted it to do burpees, I wanted the cougar to be all the way forward on the mat. And I wrote that a couple of different times. The cougar should be standing all the way to the front edge, left side of the yoga mat. It gave it weights the next time. I didn't want weights, but it kept it in the middle. And so what I then did was I took an image of myself and showed it a different version because we just kind of got stuck. But we eventually got the cougar a little bit further forward. Okay, let's see. Okay. Oh, yay. There's your leopard. Its head's not down, but I think we can work with this. Okay, can we pop it up and just show what the improvement we got? All right, so both legs up, tail still looks awkward, but you know, it's a tail. It has a tail. It's hands are behind its head like we want for crunches. It's in exercise gear. There's no other friends. We love it. Okay, so what is your next step? Here is another actual hiccup about Higgs Field. Sometimes the downloads work well and download what you want. And sometimes it downloads something totally random. Like it might give us a different animal than the one that we created, which is fascinating. So we'll see if it gave us the right one. It did. I saved the animals by name. I saved in the starting positions folder, so I could pull them for recall. I give the animal and I give the position that they're in, so it's a little bit easier. And then do you want to try and turn this into a video? Let's turn this into a video. Okay. So then we go to Higgs Field and Higgs Field has, it's like an embarrassment of riches in terms of how many options you have and what you can do. It can be a little bit overwhelming. But in this case, I use motion control. I use cling three. It's under the video option right here. So cling three motion control. I add the motion. So in this case, we're going to pull my video. And this is going to be crunches. We're going to do crunches cut because it's shorter. Fortunately, all of these exercises are not in my living room and that's what you would see if we didn't have this magic tool. And then we start, find the position that we want. And this was our leopard in the supine position. It lets you choose the scene control. So in this case, we're choosing the image. Still loading our video or still loading our image. Excuse me. So just to call out for folks, a couple of things that I noticed if you're not watching on the YouTube. So Bryce took a video of herself doing these crunches with her legs in the tabletop position. Very good form. Excellent work, my friend. And a couple of things I noticed is you're, I see now why you prompted for the head to be at the left because you don't want the video to be mirror image of the image. And so you put these two image and video side by side. They're roughly the same shape of character. You're using this cling 3.0 motion control model. And then you're saying the image is what's going to control the scene versus the video is going to control the scene. Because you don't want the videos in your living room. You want the videos in 24 hour fitness animal edition. Correct. Okay, great. Exactly. And then you click generate. We click generate. And it does require patience. They tend to take quite some time. You can tend to do, I can maybe do three or four before it gets overwhelmed. So I might queue some up if I have my videos and my images ready. But it is, you know, go for a walk, do some of your daily hundreds, get a set in and it'll eventually get you your video. Amazing. So we'll let this process and we'll come back and see what we got. Do you want to see some of the past ones while it's working? Yeah, let's do it while it's loading. Okay, so there have been a lot of friends that we've made. We have animals and we have mythological creatures for making sure they're getting good coverage. There's a goblin, there's a cat, there's a fox, flamingo. Some of them I'm able to get on the first try. Some of them take a while and some of them are just like real, real tricky. I'll show you one that I think demonstrates like how AI can do you wrong. So I wanted to demonstrate swimmers, which is a great postural exercise for us people with desk jobs. Excuse me, supermans. So it's laying down. It's lifting all of its limbs up off the ground. It's using your posture muscles. It's using your glutes, basically your whole back chain. This is what I wanted. This is our genie doing some supermans. But your hands have to be like that when you do supermans. No, you At some point, like, yes, you can do a lot of this in Replit. You can do a lot of this on your own. At some point, you're going to have to have someone technical unblock you and level up to get this in the App Store. And a friend's husband who's technical, I engaged him to basically write me a fairly intensive doc that I could shop around if I wanted to hire a contractor. But as I went through the holidays, as the new models came out in February, and I started I started getting ready to advance the project, it was wild to me because the same technical folks that I was asking, Claire included, it was sort of like, actually, with the current models, like you can probably do this yourself. And so I did. Beginner's mindset, went into Claude and was basically, how do I prepare a Replit app for App Store submission? And kept it really broad and gave it the context that I am not technical. I can use Claude code, where it was starting, some of what was involved, and what do we do next? And it gave me, I think it was like five pretty meaty items that we were going to have to go through, including moving, letting go of my Replit comfort blanket, and moving off of it. And ultimately got it ready for App Store publishing. And so what I would do is I would use Claude, original Claude, as like, my friend in, I don't know, in the cockpit of like, what are we doing? How are we gonna approach this? And he would tell me when to go into Claude code. Claude code would write me code. I would bring that back into Claude and say, this is what you told me to have it do. Here's what it did. It would confirm or give me thoughts on it. And then it would tell me to put it into the terminal directly, which was a fascinating and at first terrifying experience. But that was ultimately what I would do. I'd bought between Claude, Claude code, Claude and terminal, and really continue. And in fact, I hit my limit like 51 minutes before it reset one day, and then I got 90% another day. But I was able to stay within my limits generally and did start to outwork it because on a couple of occasions, it would say, you've done a lot today. You can come back tomorrow. I'm like, absolutely not. We're going on to the next one. And so I was able to go, I took about 25 or 30 hours in one weekend and just cranked through it, but ultimately got it into the App Store on the second try. Yeah. So just for folks and I love this workflow, again, because it's like beginner's mindset, right? You were just like, plain old Claude, OG Claude is very accessible to me and not scary. So I'm going to ask OG Claude to give me a plan. And again, like, let's go back to plan mode. I need a plan. Give me a step-by-step. And I love the honesty of like, I'm not technical, so like, let's not pretend I am. Give me the step-by-step. I'm going to work with Claude Code. And then you would give those steps to Claude Code. So you almost have like a product manager or technical architect in Claude. Then you had like your software engineer in Claude code. And then they both said, you know, we don't have the ultimate permissions here. So you need to step in and do these specific things in the terminal. And then you went to, I asked you the other day, I was like, are you in test flight? And you're like, yeah, I was in test flight. And so now it's in the App Store. You got it approved. It's ready to go. It's ready to go. Yeah. I got some pretty low-hanging fruit from my first submission. That was mostly user error. Yeah. But yeah, we're, we're live and in the App Store. What was the user feedback or the feedback? Do you remember? Yes. So there were three things. One, and of course, I just took it all. I copied it and pasted it from Apple's feedback into Claude and was like, we did not achieve full publishing. Here's what it said. The first one was like, oh, it looks like you checked the wrong box on parental. You're like child proofing something. And so it was like, this is the box to check. I probably went back to it and was like, you need to clarify further where that is and what to do. The second one was definitely user error because it does use your phone number to log in. And because it needs to be compatible with iPad, I needed to add sign in with Apple, which I had done, but I never tested it. And so that was my bad. It didn't work. And so I needed to troubleshoot that and make sure that sign with Apple worked. And then you need to have a way to delete it. So I needed a button where you could delete your app. And so I added a button to be able to delete your account. And she's not lying, folks. It's right here on on the App Store. She truly beat me. This makes me very mad. I am a little bit competitive. I'm a competitive fight goer. This is super impressive. So again, for folks who never thought they could do this and, you know, we've had people on the show who are like pseudo technical, like product backgrounds. And it's just what I love about this moment is really people are unconstrained by execution and they can take a good idea that adds value to them and build it and make it real and not just make it real, make it production ready, which I think is super impressive. So, Bryce, before we get you out of here, I'm gonna ask you a couple lightning round questions and then we'll get you going. So in your day job, you think a lot about talent, teams and leadership. And I'm just curious your thoughts reflecting on this exercise, which has brought you much closer to the builder side of things. What are things that are top of mind for you about how to hire and how to be hireable in this moment? Yeah, it's really it's really interesting. I've personally always gravitated towards and appreciated environments where the best idea wins. Right. It's not only certain people have the best ideas or certain functions can have the best ideas. But I think in this day and age, the attitude that different skills and different ideas are going to come from everywhere. And it's table stakes to cross pollinate and have both like the humility and the curiosity to work with others in ways that you haven't before. I think people that get, dare I say, territorial or constrained by what they used to do or what other people used to do are going to struggle with relevance and playing nicely in the sandbox and instead trying to zoom out and consider people can contribute in ways that they haven't before. And being both ready for that from other people and keen to to do that and bring that attitude to other places as well, I think is is necessary. It requires there's a lot of adaptation. In fact, I I feel like there's an opportunity for me to adapt because for a lot of the things where my job or my function is changing, it's still a nice to have. I think there's other functions where it's it's table stakes. And I have friends who especially do technical recruiting who have seen that when there's resistance to seeing technical problems in a new way is where folks really struggle. So for an engineer to come in to a technical interview and focus only on finding a working solution fastest, it misses the point because the robots can find a working solution faster than they can. And if they're not recognizing, hey, in this equation, the human role has shifted. I need to step back and consider my role as well as the full suite of tools and understand that my role as a technical expert is broader and different than it was six months, a year ago. That gap, I think, is only going to be greater and more challenging. And so that willingness to almost be open to outcome and unattached to what was and recognize that what got me here won't necessarily get me there, but also that at least for now, there's still a lot of opportunity, I think, to preserve humanity and maximize impact when you're willing to see things differently than you did before. I love that. And I think that's a real challenge that needs to be put out to everybody in this moment is your roles are fundamentally changing and the skills you need are changing. And that's okay. I think there's still value to bring into an organization, to a team. But the sooner you face that things are different now, the better. Bryce, my second question is, I know you read or listen a lot to books in this moment in AI. Has there been a book that you've read recently or one that you just have in your pocket from, you know, the before times that you think is really relevant for folks right now as we're navigating this different sort of moment in our careers? If you haven't read what got you here won't get you there or how women rise, which is kind of the sister book to that. It's a good reminder. Not everything is in the title, but it's really the premise that like, yes, you should trade on your strengths and the things that have been positively reinforced to date. And also definition of insanity is doing the same thing and expecting different results. And so your willingness to believe in yourself and scale to the next altitude is really your opportunity. I did listen to A Whole New Mind. It's the book by Dan Pink, which actually came out in 2014. And I think is really ahead of its time, but talks about the benefits of an and thinking of both left brain and right brain thinking. And again, this is it's old now. And the way he's talking about automation and acceleration and abundance was