---

About David:LinkedIn |Twitter/X

Connect with Sub Club:Podcast |YouTube |Newsletter

Connect with RevenueCat:Webpage |LinkedIn

Content Update: Rico is no longer just in beta! You can now access these insights directly in your RevenueCat dashboard.

FULL TRANSCRIPT BELOW

Shamanth Rao:  I’m excited to welcome David Barnard to Intelligent Artifice. David, welcome to the show.

David Barnard: Thanks for having me. I’m looking forward to chatting.

Shamanth Rao: Yeah. David, you’re a

David Barnard: nice to be

Shamanth Rao: on this side of the mic.

Yeah. Yeah, indeed.

Indeed. And you interviewed me on the Sub Club podcast a while ago. And you are quite prolific on the podcast, on Twitter and building your own apps while managing a family and travel. Last year we were just talking about before we hit record. So I’m always in awe of what you do.

I’m always impressed and the last time we spoke, when I interviewed you we did an episode called a Brief History of App Store Monetization. And we did that because you’ve seen the app stores since the app stores launched in 2008, and you’ve seen every major shift firsthand.

So for folks that may not be familiar with you, can you talk about what that journey has looked like over the years?

David Barnard: Sure. So I was a huge Mac nerd. I, that’s really where it all started  and I was working in a recording studio and got familiar with the Mac because of that, and then bought my own Mac in the early two thousands and so I was reading Mac rumors and all the old Mac enthusiast sites.

I was probably reading Daring Fireball when it first launched, and so when the rumors of the iPhone started circulating, I was super excited. I bought the stupid Motorola ROKR iTunes phone that they released before. It was a total flop. I actually returned. It was terrible and then when the iPhone came out, I was in line and first person in San Antonio, Texas. I got my, my picture in the paper. I was

Shamanth Rao: Yeah.

David Barnard: one of the first to get the iPhone in San Antonio. And so immediately after getting the iPhone, I was just awestruck. And it really felt to me like, this is the future, and it sounds like, if you look back at things, you’re like, oh yeah, I called that before it was real or whatever, and you wanna think you’re really smarter or whatever. But genuinely, when I first started using the iPhone, and then especially when I saw my dad first using the iPhone he struggled with  computers as he transitioned into using them in the nineties in his business, and this was now, 2007, he had always struggled with computers. And he took to the iPhone like a duck to water. And that combined with my own experience, left a massive impression on me. This is the future of computing.

This is the future of smartphones. So I dove head first. As soon as rumors were circulating that there’d be an SDK, I started making business plans and, brainstorming apps I could build. I jailbroke an early iPhone so that I could load third-party apps. I actually bought a separate iPhone ’cause I was worried about jailbreaking.

My, my real iPhone. Bought a second iPhone, jailbroke it. Started installing those early third-party apps to like, understand what was going on and then had an app in the App Store. I should have had it day one, but Apple didn’t approve my developer account in time. So yeah, I’ve been building apps since the very beginning and I’m not a coder.

I, I tried to learn how to

Shamanth Rao: code

Yeah.

David Barnard: and I got about a month in and I realized it would take, years to get good at coding. I didn’t have years. I  wanted an app on the App Store day one, and so hired

Shamanth Rao: Yeah.

David Barnard: a contractor to do the coding and never looked back. The one thing that has benefited me a lot over the years is that I’m pretty hands on.

So I’ll read API docs. And because I spent a couple of months trying to learn to code, it’s like I got the fundamentals enough to where for these last 18 years, I can grasp what’s going on and understand the APIs and, look at WWDC Diffs on what’s changed and look for opportunities and stuff. And so for the past 18 years I’ve been building apps with a combination of partners and contractors. I’ve launched, I don’t know, 30 something apps. I’ve sold four different apps. I’ve had multiple apps make over a million dollars. Been a crazy 18 year path.

Shamanth Rao: Yeah.

David Barnard: And then, to be real frank, I did go through a pretty slow spot back in 2018 ish.

One of my business partners moved on to another job and so the app started, revenue started going down and that’s when I started looking for  a side gig. And the side gig that landed was RevenueCat, as a developer advocate there, and that was six years ago that I joined RevenueCat and I joined part-time and then two years in I was doing so much for the company and was going through the process of building a house.

And it’s just so much easier as an employee to get a loan and then I needed healthcare for my four kids and all that stuff. So I finally joined full-time after two years. And, but kept building apps on the side. So I have a weather app that I’ve been tinkering with. And then I think as we’re gonna discuss, I just recently started by coding my own app and it’s pretty exciting.

Shamanth Rao: Yeah. Yeah, that’s quite a journey. And I know we talked about some bits of it when we last spoke, and I also did find, find an old newspaper article about you standing in line waiting for the first iPhone when I was preparing for this. It’s kind of crazy. Yeah. As you said, we are gonna talk about what you actually built, while coding versus what you built with contractors in the past.

So talk to me about the app that you’re building right now. You said,  and when we were preparing for this that, you know, you had the idea a couple of years ago, what stopped you from building it back then? How, how have things changed now?

David Barnard: Yeah, I actually had the idea maybe four or five years ago. I won’t get into the exact idea. It’s funny because for years, people would say it’s not the idea, it’s execution. Like you shouldn’t be precious about not sharing the idea. But I think, the last year has really flipped that on the head that, you know, people can build apps so quickly that if you do have a good idea for an app with a unique angle or whatever, and you do build, I, I think it is the worst time ever to fully build in public.

And once the app is out there, I’m gonna share it and talk about it more publicly. But anyways, the app idea was like five years ago I’ll say it’s a, it’s essentially a self-improvement app. And I was building it for my son and initially it was like a, just a paper version that I was working with him. And, five years ago I thought, oh, this would make a really cool app, but the amount of effort and money it would take to pull off an app  five years ago and pretty much anytime up until the last 12 to 18 months, you, it would’ve cost, 30 to 50,000, maybe 20 if I, did a overseas contractor on the low end.

But then it’s so much more work managing that kind of person. If I went, in, in the US or Western Europe or whatever, and work with somebody more experienced, more senior, it would’ve easily cost 50 to a hundred thousand dollars and it was an, it’s an idea. I think it’s really fun, but, I have no idea how well it’s gonna perform financially.

And it wasn’t the kind of idea like, this is a slam dunk. Let me, convince somebody to partner with me and we’ll build this together. I didn’t wanna waste their time, I didn’t wanna waste my time. But I always wanted to bring it to the world. And that’s what’s so cool about vibe coding and the changes that have happened over the last 12 to 18 months is that, you know, with very little expense.

I am spending $300 a month now combining a hundred dollars for Claude Max, $200 for ChatGPT  Pro. And I’m using my personal accounts, not, I do have access to those via RevenueCat, but, we do have an internal policy, of course, as we should, that you don’t burn company tokens on personal projects. That’s out of my own pocket as I explore this. But other than that, $300 and a few hours here and there, nights and weekends, it’s not really costing me anything. And so I’m gonna be able to put this idea onto the world and see how it does. Maybe it makes, a few thousand dollars and that’s more than enough to compensate me for the time and maybe, you

Go toward a nice little vacation for me and the family and that, that’s just never been possible before. So it’s pretty exciting to be able to do that.

Shamanth Rao: Yeah. You are right. Crazy how cheap and easy it is to put some of this out into the world. I, again I used to work with developers a couple of years ago, and it always felt uphill. Also as somebody that wasn’t as tech savvy. And I don’t see myself as technical, it always felt uphill.

And just, of course now with Claude Code and Codex,  it’s just become crazy right now. I’m curious, you talked about how you’ve been building all of this yourself. I’m curious if you can think about, think of the moment when you realized you could actually build a lot of this yourself.

Was there a point in time or moment when you realized, oh my goodness, I could just do this myself?

David Barnard: Yeah, the, the moment was probably seeing my boss Rick vibe code an app over, I think he did over the holidays and launched it in January. It was really impressive. And I think he actually shared with me the Manus conversation or maybe screenshotted parts of it, of what it took to actually build and just seeing him do it with this very natural language conversation, I was like, okay, like something’s changed.

This is working in a way that it never has before. Because if you go back, even, nine months ago really more experienced coders who were able to move quickly with these vibe coding tools and the, I had experimented with Replit, I experimented with some of the  others, and it was just so rudimentary what it would build that it, it wasn’t interesting to me and I, and it always felt like, oh, I’m gonna need to, I’m gonna need to dig into the code. I’m gonna need to hire somebody to dig into the code. There’s no way I could ship an app at the quality level that I would want to build without some amount of tinkering in the code. But when I saw Rick launch this app, you know, all his natural language stuff. Now Rick is pretty technical, so he may have tinkered a little bit here and there and he actually studied design. So I, it was super impressive, the design of the app, and it was like, this is it. And Rick’s just an incredibly, talented, intelligent person. But it was just a wake up call that, that he could do that mostly just natural language and ship a pretty complete 1.0 with all sorts of features and stuff and, the vibes around the industry I think also influenced me to finally try it. People talking about how good it was.

And I was like, all right, I gotta actually try it and I will say a, a friend of mine Justin, he is actually a developer. I built Launch Center Pro with, we have still  have lunch every once in a while and went to Open Claw meetups in the last few months and stuff like that and he, it really feels like he’s been three to six months ahead of seeing around the corner of what’s gonna happen with these things.

And right before I started, we had lunch and he told me, we’re moving to a place where I don’t think I’m going to ever read a line of code again and when he said that I thought, no, no way. And I still have my doubts, when you get into big complex code bases and like sophisticated backends and like complex data schemas and things like that.

There’s, the expertise of a senior programmer is still gonna be incredibly valuable moving forward to manage those complexities. Even with AI, I could be wrong because, he was telling me he thinks, we’re moving to a place where AI’s gonna be way better at setting up a database, managing that, managing migrations, all that kind of stuff. and so combined with, everything I’ve seen with Rick, the vibes in the industry and him saying that, I was like, alright,  I gotta just pull the bandaid off and make some time.

Shamanth Rao: Yeah.

David Barnard: And it’s tough with four kids in a demanding job and yeah, I will say it’s a little addicting.

And there’s been nights where, I wrap up my RevenueCat work and dive into vibe coding and my wife’s like, “Hey, come to dinner”

Shamanth Rao: yeah.

David Barnard: She has to, like, pry me away from the computer. So it’s tough, but it’s been it’s been so much fun building it.

Shamanth Rao: I like that you used the word addictive. And I would actually echo that, from my experience, I was actually chatting with somebody, and it’s perhaps for the first time in history that you have something being addictive, that’s actually building stuff, making stuff. Because normally a lot of stuff that you would consider addictive would be TikTok or alcohol or whatnot.

Uh, but for the first time, you actually make stuff and it’s actually addictive because it gives you that dopamine rush. It’s a lot of, it’s just from having that instant feedback loop, which again, I still can’t believe as a lifelong non-programmer much, again, made multiple attempts to learn to code and, just had a degree in engineering but just never  ended up learning to code.

And it’s just incredible what we are all able to do now, and I know you spoke briefly, in this conversation about you, you were like reading API documents. You were like getting into the Mac ecosystem, like back in 2008, 2009. So unlike, completely non-technical people, you had a technical background.

David Barnard: Yeah.

Shamanth Rao: How does that translate into your vibe coding experience itself. Does, does, do you feel like you have a better sense of what’s going on? Uh, talk to me about that.

David Barnard: Yeah, I do think that as the tools stand today, it’s pretty good at engineering. It, it feels more junior than the most junior developer I’ve ever worked with. You know, I’ll set it off on a task like, Hey, animate

Shamanth Rao: Yeah.

David Barnard: then, and animations

Shamanth Rao: Yeah.

David Barnard: …are something that it really struggles with because it’s not like taking a video, a screen capture of the simulator to watch the animation happen and introspect the animation.

But that might not be too far away.

Shamanth Rao: Oh yeah.

David Barnard: So from a  development standpoint it’s been as fun as it’s been. There’s also been very frustrating moments where it takes three different prompts to solve one small bug and actually I’ve been bouncing back and forth between Claude Code and Codex and playing ’em off each other a little bit. I don’t know if it brings out the best in them or the worst in them, or if I’m

Shamanth Rao: Yeah.

really

David Barnard: doing the right thing or not. But,

Shamanth Rao: Yeah.

David Barnard: if Claude Code doesn’t nail something in, one or two iterations, I take it into Codex and explain what we’ve been working on. And I have found Codex as of today, all, everything we talk about in some ways is gonna be outdated even by the time this is released.

But as of today, Codex is better at iOS apps. OpenAI created a plugin for building iOS apps and, there’s so much to harness engineering these days that, a lot of the credit goes to the models improving over like December, January timeframe for things getting so good.

But it actually has a lot to do with the harness engineering, where they’ve given it more tools,  more context, more skills internally in Codex and Claude Code to be able to execute on these things. And so Codex with this iOS development skill or plugin, does feel like a more senior programmer than Claude Code and can figure out the more complex code and bugs. Claude Code feels a little more creative and better at some of the more product managery kind of things. But I think that’s where things really still fail today, that I have noticed that it takes a lot of kind of product intuition and you have to really know what to build and why to build it and how to do things to build a really great app. I admittedly, I have not tried, Lovable and Rork and some of these others where you should theoretically be able to more one-shot And maybe they have some skills in some of the way the harness is set up and the tooling. Have some better kind of product management intuition. But so far  with Claude coding and Codex, it does feel like I lean very heavily on those like 18 years of experience where, it’ll tell me, oh, I can’t do this.

I’m like, no, you can! I’ve read the SDK docs, like you could do

Shamanth Rao: Yeah.

David Barnard: So there is a lot of kind of fighting with it and using that, that, that background that I have in order to build something really great. So I’m not, I, I’m not sure about one shotting and, I’m. I’m just really not sure where this all goes from here and, part of me thinks as much progress as we’ve seen even in the past three months you know, maybe ChatGPT or Codex and OpenAI need a product management skill where they.

Shamanth Rao: Yeah.

apply a lot of the kind of intuition

David Barnard: that a product manager would have the same way they’ve done with coding. But I’m not convinced, this is where AI does generally fall flat is in the kind of the creativity and the knowing what to do.

It’s way better at just following instructions and repeating patterns from the past, not like creating new things. So it may be a while  before we can, one shot apps and, I hear a lot of hype about the end of software and maybe we could dig more into that, but at a really high level, I, I think it’s gonna be a long time, if ever that something like Strava or Ladder or Duolingo or something like that is gonna be. Something a a non-technical, non product person can just vibe code a Duolingo. Maybe it’ll get good enough where it can just copy Duolingo, but then what’s it?

Like

Duolingo? And then there’s so much and this is where, again maybe this is coming because if you set up a bot farm with a bunch of iPhones, and then you just have the AI use a bunch of apps over and over again and understand the patterns and the animations and figure out what it feels like to use these apps and the product decisions that went into it. Maybe it can start reverse engineering those kind of things better than it currently has. But I think it’s gonna be a while before you could really one shot a more  sophisticated app. But we’ll see. Things are moving so quickly. And then

Shamanth Rao: Yeah.

David Barnard: …end of the day, and I talked to Eric

Shamanth Rao: Seufert

David Barnard: about this on the podcast, it’s even if you can one shot a clone of Duolingo, you need to get attention for it. Distribution then becomes a

And so

Shamanth Rao: yeah.

David Barnard: a flood of apps, but

Shamanth Rao: Yeah.

David Barnard: apps that have distribution and have network effects and already have an audience will continue to thrive versus these vibe coded apps are gonna put a huge dent in some of those.

Shamanth Rao: Right. And there’s just so much I wanna dig into what you said. But you’re right. You know, you could potentially one shot copying Duolingo. It’s probably theoretically possible and I, I have one shot at Simple Apps, but I what I’ve also come to realize is this, making this into something production ready.

Even just, if it’s just adding onboarding, setting up paywalls that takes so much longer. You know, and it’s a longer journey than just having something that lives on Xcode and in my simulator and I can just run it on TestFlight and that’s, I  think you could certainly one-shot a lot of those and I’ve done that myself, uh, and. Yeah, I, and I certainly think about whether a lot of these systems will have product management intuition, from what you’re seeing, a lot of what you are building builds on your personal experience, personal intuition of knowing the API docs inside out, you know, what’s possible, what isn’t, and you’re able to push back.

And a normal non-programmer isn’t really able to do that. They might just say, they might be able to push back and say, try again, or do this again. But without that kind of contextual nuance, I just, it sounds like it’s not really gonna work. And I know you talked about

your friend who said they’re not gonna review a line of code, read a line of code again and you talked about how you were using, working with programmers in the past, contractors in the past and in preparation for this call, you did mention that for the app that you just vibe coded and launched, you are planning to have a senior programmer review it.

Talk to me  about why you are considering that, given everything else we’ve been talking about.

David Barnard: Yeah, I mean, I’m really waffling on this. So we’ll see if I actually do it or not. I am nervous that if I launch this app and it makes, a few thousand dollars and now I have subscribers and people are depending on it, and as I said, it’s a, it’s a self-improvement app and the goal is to help people improve their lives. And if I get it out there and there’s like a massive bug, and then I try and fix the bug and it breaks the database and the database migration doesn’t work, and then it like kills everybody’s data progress and things like that, then that would be just a terrible experience. You know, to get really bad reviews on the App Store, probably get refund requests and things like that. And so I am a little nervous to put this thing out in the world with code that no experienced developer has ever looked at. But on the flip side, the deeper I get into this and what my friend Justin was talking about is that we’re getting  close and maybe for the level of simplicity that this app is, it’s not a super complex deep app.

There’s not a really complex data schema. It’s

Shamanth Rao: Yeah.

David Barnard: simple, straightforward thing. What have you ever done in the history of software? A programmer writes the code and then they test it and see the output and the code quality varies tremendously. You talk to any programmer who’s worked at multiple companies and seen a lot of code and they’ll complain about code quality and poorly written code and poorly documented code and stuff like that. But like that, like poorly created code is like running important systems and like making money. So there’s part of me that does think that, as good as AI is at the patterns of coding and looking for those kind of things, I bounce back and forth between Codex and Claude Code, asking it to check each other’s work, doing deep QA passes and things like that, and asking it to like refactor and clean things up and look for, dependencies that, that aren’t used anymore and all those kinds of things.

And again,  a lot of that takes, my intuition of having

Shamanth Rao: done QA

David Barnard: on apps for 18 years. I would work with programmers on the code, but then I was the primary interface between the code being completed and then it being good enough to ship. And I got pretty good at breaking stuff. And if that’s what I did for 18 years, as the kind of QA person and I even hired QA people at times, contract, you know, if that was the last line of defense. Anyway, if that is still the last line of defense, then maybe it is good enough. I don’t know, I’m going back and forth. I might at least just have somebody like, peek at it.

But I was talking to a programmer friend recently about that idea and he was like, in, in an hour or two, programmer’s gonna be able to like really

Shamanth Rao: understand

David Barnard: the code well enough to tell you whether it’s good and maintainable or, and that’s again where AI is like you can point Codex at it and say, look for issues, look for bad patterns, look for all these  kinds of things, and it can introspect all, I think I’m at like 10,000 lines of code now. It can look at all of that, in a way and with speed that a human programmer can’t. So yeah, I’m waffling on it. I don’t know how valuable it would be to do it.

I don’t know how much risk I’d be taking. And because this is a side project, there’s a little bit of me that just wants to yolo it out there of, not worry too

Shamanth Rao: much

Yeah,

David Barnard: And I don’t think this is gonna make hundreds of thousands of dollars or anything. I’d be thrilled if it made, a 1000, 2000 bucks a month, it would make a few thousand dollars a month. I do want you to ask me about distribution because I do think this is really important.

Shamanth Rao: Yeah, for sure.

David Barnard: Like how I’m thinking about doing distribution and why I picked this idea to work on and things like that.

Shamanth Rao: Yeah,

David Barnard: Yeah. And so who, who knows? It could make, you know, 3-5 thousand dollars

Shamanth Rao: Yeah,

David Barnard: a month. It

Could do better than I expect. You just don’t

Shamanth Rao: yeah

David Barnard: that’s actually the fun and a little bit of the dopamine of doing

Shamanth Rao: this.

David Barnard: I’m working on it, I go back and forth between, oh, nobody’s gonna care and nobody wants this app to oh no, I see

the App Store making a hundred thousand dollars a month.

That,  I think this is like a better app than that. We’ll see how it turns out.

Shamanth Rao: And you’re right that you know, you could very much drop this into Codex or Claude Code or have them read the code base and say, do a QA pass, do a deep QA, and with your experience, you can just know exactly what to prompt them for, what to look for, probably do a lot of it themselves.

I have considered though, handing some of my code to an experienced programmer who would look at it with AI, someone like yourself that knows what API and parts would be, what would, what lies behind the scenes of the code. I haven’t found somebody who is as tech savvy and AI savvy, maybe that’s something that, I think could be valuable, but maybe not if, you know, Claude Code and Codex are able to look at this directly as well.

David Barnard: Yeah.

Shamanth Rao: And so, yes.

David Barnard: This is where the skills come in where, a good QA skill would encapsulate as much as or

Shamanth Rao: more

Yeah.

David Barnard: than an experienced programmer would think off the top of their head to ask the AI

Uh,

In some ways that’s part of the power of  this is that even for those kind of code review things, it can check on so many different levels that an experienced programmer wouldn’t think to or doesn’t have the IQ or whatever

Shamanth Rao: Yeah,

David Barnard: So yeah.

Shamanth Rao: And you could have the AI come up with a script too. Right? And, uh, something we’ve done is, so a lot of our business, we have about 30 to 40 different pieces of code running at any point of time and just this weekend I was like, oh, things are breaking. Can I have a self fixing mechanism in the code where every time something breaks activate a Claude Code instance that goes in, fixes it, alerts me, sends me a Slack update.

And I was like, holy shit, it’s actually fixing it, and it says, and that was like a Frankenstein moment for me. That’s been incredible. Yeah. And, with your own app that you are building talk to me about how you are thinking about distribution. And I also asked that because you talked about the proliferation of apps that you have seen just with the advent of AI, getting distribution from everything I see and understand is  challenging.

So how are you thinking about this for yourself?

David Barnard: Yeah. You know, if you build it, they will come. Has just never been true on the App Store. I mean, there

Shamanth Rao: Yeah.

David Barnard: are few lucky people like, the I-Beer guy who, you know, so early in the app store, such a fun gimmick that it’s spread virally. But pretty much since the beginning of the App Store, very few apps has it actually been the case where like you just build something cool, tell a few people, and it takes on a life of its own, pretty much always requires some amount of pushing the boulder uphill, to get attention and I will say sometimes part of it is just building something that will get attention.

And so this is something I think a ton about in, building this app specifically in my weather app. is that if you can do something interesting and innovative, that at least buys you, some reason for somebody to pause because it, like we were talking about earlier, if you clone Duolingo, what’s the hook?

Like

Shamanth Rao: Yeah.

David Barnard: Why will people care?  Like they can just use Duolingo and it’s probably better and doesn’t have bugs and is

Shamanth Rao: Yeah.

David Barnard: And even when we were talking about that, we didn’t talk about how much of Duolingo lives on the server that can’t just be copied. You can copy patterns, but it’s not like you can just copy all of their lessons and skills and the AI behind it that’s driving the lessons and understanding the user’s skill level to move you forward or have you repeat lessons, like there’s so much that goes into all of that. But so if you just clone something like why does it need to exist? What’s the angle? And so that’s what I, I’ve been thinking a lot about with this app and in the self-improvement space. It’s the kind of thing where people do want to, this is why health and fitness has been such a big category forever. That’s why we’re seeing a lot of, and headspace and calm and, it’s been one of the bigger, categories if you include all of that kind of in the kind of overall heading of self-improvement. It’s been one of the biggest categories through the app store  for a long time. I would even lump Duolingo into that, learning a language that’s a form of self-improvement. And so I, I think with this app, there’s potential to be able to AI, generate the kind of content. Now I wanna be careful to say I’m a little worried about where we’re headed, where you get fake people doing fake testimonials and fake scenarios and things like that. So what I’m not gonna do is go use the latest video model and create a bunch of fake testimonials of people saying, oh, I used this app, it saved my life. Instead my goal with this and what I’m gonna try and, it may completely flop, we’ll see, but is to create more of that kind of motivational content that, that people do on TikTok and Instagram and stuff, where the hook isn’t as strong. so again, I’m not expecting to make hundreds of thousands of dollars.

I’m not trying to like, build a whole business around this, but I think I can create enough good content, little slideshows and things like that that are motivational and get people  excited about self-improvement and then, see this app as a way to achieve that. I will say for a lot of apps, especially with organic, if you really do wanna build a great business, not just, get a little attention and make a few grand here and there, ideally the app like has a hook that’s exciting and able to be demonstrated, in a TikTok video. So I, I just had Coconote, on the podcast. They built an app from scratch starting two years ago. Got it to $6.7 million in ARR and sold it to Quizlet in the past two years. It’s a crazy story. But what that app does is it, is, it takes notes, it’ll record lectures like at school and college and things like that, and then build flashcards and summarize and help you study based on the lecture. There were so many magic moments in that experience that, they were able to create a lot of viral content that wasn’t just, and the way they described it is like, you don’t want to go viral as a toy, you want to go viral  as a tool. And so is your tool exciting enough to get attention on its own?

The way Coconote did, and they even gave an example on the podcast of, one of the things they went viral for was this, these brain rot, this brain rot feature where you would, it would give you study notes while it had one of those Minecraft Parkour or Subway surfers videos going on. And that content went super viral. People thought, oh, this is so funny. I can

Shamanth Rao: Yeah.

David Barnard: study with, brain rot and those videos went super viral for them, but they didn’t actually convert very well. Compared to the kind of content that positioned it more as a tool that’s really gonna solve a need. And so I think, these are the things you need to be thinking about. This is just one, you know what I’m saying about building some motivational content or what Coconote is doing. That’s just like the tip of the iceberg of the possibilities of getting attention for your app.

If it’s truly  super innovative, you can get press. That’s part of what I’ve done with my weather app and I do have a big update to my weather app coming where we are doing something really unique and innovative in the weather space that I think people are gonna be really excited about.

Shamanth Rao: Part

David Barnard: of the the strategy there is that because it’s really unique and innovative, I think, Apple is likely to feature it ’cause it uses their technologies in an innovative way. And then I think I will be able to get some amount of press and broad attention for it because it does something unexpected and really innovative in a space where pretty much everybody looks at a weather app for, not many people are gonna pay for a third-party weather app. Because the data of weather apps are so good, but it’s such a large space. There’s so many people and it’s something everybody cares about. And we do something unique and innovative. You can get attention. So press is a way to get attention there. There’s a lot of apps where the value prop is so strong that you can do profitable ads. So if you have a really strong value prop in a category  where people are willing to spend money, health and fitness and self-improvement and things like that. You know, you can bootstrap via ads, not just organic TikTok and other stuff like that. So there, there’s just so many ways to get attention.

And thinking about, viral loops, is there a sharing moment in the app that and I’ve put this, I’ve done this in apps in the past, like one of my earlier weather apps. I’ve done two different weather apps now and I’ve been working on weather apps for 12 years. One of these days I’ll crack it, but one of the early ones I had a sharing button and it shared a gif of the radar animation.

And I thought, oh, this is so cool. People are gonna be sharing the animated radar of a storm coming in and, I’m total weather nerd. So I think it’s so cool to look at the radar pattern of the storm coming in and covering me with the, strong red rain and everything. And nobody shared. It was like over the course of years with tens of thousands of users, maybe it was shared less than a hundred times and it was a big feature that we had to put in. So I’m glad I  did it though. ’cause you learn, right? So you should find ways to do stuff like that. But it’s often not the kind of panacea that you hope it would be. But for some apps.

Shamanth Rao: Yeah.

David Barnard: If you need to invite your friends, and by inviting your friends, it does make it more valuable and you can create this sharing moment in onboarding. That’s a way to kinda leverage a little bit of attention into a lot of attention. But the, you’ve gotta, it’s gotta be something that people really care about. First to just even download the app, but then second to bother even sharing it with a friend. But yeah, there’s just, there’s so many ways to get attention. On the podcast we’ve talked about so many different, marketing stunts.

I was even talking to Photoroom recently. They are a much bigger app and they say that they have the money to try stuff like this, but they’re doing in-person stuff now. Where they set up a a photo booth in a popular shopping area of London, and people got to experience the app. And then what was cool about their strategy there was like, they leveraged all of this in-person stuff where people  got Photoroom does photo editing and background removal and things like that. And for people to go into this photo booth, take a quick picture of themselves, and then show them, as if they’re at the pyramids or, something like that. I don’t know exactly what they were doing, but there is this really aha moment of wow, that looks really good. And oh wow, I didn’t know, we’re all so plugged in the app industry that we know these things are possible, but a lot of people don’t even realize these things are possible. And so they took those like surprise moments and created ads and created organic engagement based on it. So yeah, there’s just so many ways to get attention but if you’re building an app, you need to be thinking about how you’re gonna get attention for it. And if you can’t, if there’s not a clear way that you’re gonna get attention for it, then move on to another idea.

Like keep iterating on ideas until you find something of like, okay, it is like distribution product fit, not product market fit. Can you build something valuable that people care about, that you can actually get attention? And I would say those are the  three things that you really need to build an app that’s actually going to make any money at all.

Shamanth Rao: Very cool. Very cool. There’s just so much I wanna take notes on and I’ll certainly revisit this conversation afterward, but I think what really stood out to me was how you’re thinking about product as a hook, how do you make the product itself shareable? And conceptually, I and my team and a lot of folks who work in growth talk about hooks for videos, hooks for ads.

But I think there’s the first time I’ve heard of somebody talk about what’s the hook in the product, what can hopefully, how the product actually be shareable. And I think that’s such a great framing. It’s certainly something that I want to think about as a lot, quite a bit. Yeah. And just to switch gears a bit we’ve talked about a lot of things with AI.

You also mentioned in preparation for this call that you’ve been testing RevenueCat’s Intelligence Copilot. So talk to me about the experience of testing that also contrast that with your vibe coding experience,  Talk to me about what it’s been like.

David Barnard: Yeah, Rico, RevenueCat’s Intelligence Copilot. I’ll be honest, when our CEO is the one who really got excited about this, Jacob. When he first proposed the idea and started tinkering with it, I think he had set up an OpenClaw and was like figuring out like, where do we go as a company with AI and what can we build that’s really useful? so when he first started talking about it in our slack, I was skeptical and I was skeptical because from what I’ve seen over and over and over again, hallucinations, other issues with AI make me very nervous for it to do anything analytical with data. But as they started building the product and I started using the product I started to get more and more excited.

And I think this, this comes down to, how they built it. And I guess I should start by just saying what it is. Rico is a tool currently that lives in Slack as just the Slack bot, but we’ll eventually bring it to the  dashboard, the app, and build it across all the product surfaces inside RevenueCat. But it’s a way to talk to your business particularly is really the best way to frame it, is that it’s not just talking to your data, it’s talking to your business. And the things that are most powerful with it right now are asking it things that would otherwise take a lot of work and thought to surface in our dashboard.

What’s going on with this specific cohort, how is my, trial start rate trending and how does that compare to benchmarks and it gets really good suggestions of like how to improve things. ‘because we’ve trained it on, you know, all the Sub Club episodes, all the blog posts that we’ve done with all sorts of contributors and it’s surprisingly smart about not just getting you the data, but then giving you ideas about how to improve your performance versus the benchmarks. And so it, it also has all this data, subscription apps data in there. so I think why it’s working and why I’ve been so impressed with it is that the team  has done a lot of really thoughtful work around it.

And this is what I was talking about earlier with harness engineering, is that, you know, if you just dump a bunch of data into ChatGPT or cowork or that’s where you’re more likely to get hallucinations it, not fully understand the data. It not be able to give you meaningful responses. But what the team has done is, and what harness engineering is essentially is combination of the AI is the kind of brains, then you need to give it tools and context and what we’ve done is expose all the internal RevenueCat APIs to it and in building this, we’ve actually ended up opening up way more APIs than we had previously because we want it to do more and more.

And so for it to do more and more, we have to expose those APIs to it. So the harness engineering is defining the tooling that it can use, so here’s the data, here’s what the data means. We have all  sorts of specs around the dashboard of why the data is the way it is the structure of the data, the way it relates to each other.

It’s like we have all of that context internally. Can give that to the agent and say, this is a tool to get data. This is a tool to look at benchmarks. This is a tool to understand this and then give it context with a system prompt of how to do it and then right now there’s two buttons.

Good slop and bad slop for every single response you can say, bad

Shamanth Rao: slop

David Barnard: and I’ve done this in my testing where it got something wrong or was a little fuzzy. And then I give very specific feedback. And the team takes that feedback and figures out like, what went wrong here? Is it that the tool didn’t understand, didn’t give enough context around the data that confused the AI?

What, what was going on? And then what’s the additional context? What’s the guardrails? What can we figure out? And so this has been an iterative process with internal testing, and now we have a bunch of people on the beta using it where  we’re hardening the system against the kind of common pitfalls. it’s gotten shockingly good. Like I genuinely didn’t expect it to be as good as it is. Every time I go double-check the data, it’s spot on. I was so worried about hallucinations that every time I’ve used it, I go check the data and it, and they’ve just set up really good safeguards.

Shamanth Rao: Yeah.

David Barnard: It’s like you do it intelligently and do QA passes and verify and checks and balances and all those kinds of things it can get really good. And so that’s what we’re working on and yeah, it’s super exciting to be able to chat with your business and how is this experiment going to, because, we have the paywalls and experiments and like all these products surfaces now that it can talk to.

And so you can ask it like, how’s this experiment performing? It’ll do statistical significance calculations for you and do things like, we haven’t even, implemented into the dashboard yet because it’s so much work to build it out into the

Shamanth Rao:

Yeah.

David Barnard: dashboard. And so now it’s filling out the  product surfaces in a way that’s hard for us to do.

We’re, we’re gonna continue doing it and adding things into the dashboard and other parts of the product. But this is able to like just do so much more. So you ask it like, how’s my experiment going? And I’ve seen some internal threads where it’s so smart about it, and there was one time when I thought it was wrong about something and I realized my intuition was off.

So as much as I

Shamanth Rao: Wow.

David Barnard: worried about hallucination, like humans make data analysis mistakes

Shamanth Rao: Yeah.

David Barnard: and I’m getting to the point where I trust Rico better than my own intuition

Shamanth Rao: Yeah.

David Barnard: it’s been hardened by, people at the company much smarter than me, who understand the data and the relations way better than I do. And so it’s just, it’s really good at being able to pull out those nuances of hey, you’re not at statistical significance yet, but this one looks like it’s going to win. Here’s why. It’s looking into the future of here’s what the conversions are likely to be based on the current data, if we extrapolate that out.

And you  would have to take our experiments data into a spreadsheet and do extrapolations and things like that, because we don’t do some of that by default. And instead you just ask Rico and it does it for you. So it’s super cool. And then it says, it’s like it’ll give recommendations.

You might think about, rolling this out ’cause it does seem like it’s gonna win.

Shamanth Rao: Yeah.

David Barnard: If this wins, you should try this experiment next and yeah, it’s really fun, really great stuff. A really

Shamanth Rao: Nice.

David Barnard: And then, I can see us getting to a place and this, the tooling getting to a place where, I don’t think soon we’ll, like fully automate it where, you know, you just give it an app and say, make me more money. And it’s just figures out what to do. But like we’re heading to a place where, Rico’s gonna suggest, Hey, I think you should experiment with a new CTA on your paywall. I was just listening to the Sub Club podcast where Duolingo said, ‘try for $0 and 0 cents’ outperformed ”try for free.’ And then it just, it generates the paywall, it sets up the test, and then all you do is verify and the human QA, makes sure that  the paywalls look right and that it’s representing your brand and didn’t mess up the colors or anything like that. And then you just hit a button, like start test and then, and again, maybe within soon it’ll send you an email, Hey, the test is winning. Let’s roll it out. And it’ll do the analysis proactively for you versus, I was gonna say, you go back in a few days and ask Rico how it’s performing, but I think, we’re gonna get to a place where it’s proactively gonna tell you how it’s performing and make suggestions based on that. So it’s pretty exciting

In the app world right now

to leverage these tools so powerfully, to help make those decisions, to analyze data, to, just be able to think about your business. More introspectively

Shamanth Rao: Yeah. That sounds so exciting. Remind me, this is not yet out publicly, is that correct? Rico.

David Barnard: it’s it’s in a public beta so people

Shamanth Rao: Okay.

David Barnard: request access to it.

Shamanth Rao: Okay. And that that would be from the RevenueCat dashboard to grant this access.  Okay. Excellent.

David Barnard: I don’t know that we have a button in the dashboard yet to request access. Probably by the time this goes live we’ll be rolling out

Shamanth Rao: Yeah, because,

David Barnard: broadly.

Shamanth Rao: I don’t recollect.

David Barnard: goes out or,

Shamanth Rao: Yeah, I don’t recollect having seen it. I would be very interested to check it out, on a number of accounts you’re looking at. But overall, you know, I can see why it is so exciting because this is an order of magnitude better than vanilla GPT or vanilla Claude because it has the context of your business and, uh, from what I can see and tell a lot of that power comes from you guys giving its tools to access specific knowledge that’s specific to your business, which, even a year ago, was pretty much impossible, unless you built a ton of.

Connective tissue manually. And now it’s just so easy to build those connections. I’ve seen some of that in our own business because we have a Notion and a project management system and our team time tracking, a lot of documentation feed into Claude Code, and it’s just incredible how  much more we can do as a result of this.

So I can certainly understand how incredible it can be to have a tool like Rico.

David Barnard: And part of what’s so exciting to me about it that, any these days, again as we’ve been talking about with vibe coding, like any one app could build versions of this internally. But with RevenueCat, because we’re doing it across so many different apps, with so many different people using it,

Shamanth Rao: yeah.

David Barnard: harden it and QA it and

Shamanth Rao: Yeah.

David Barnard: in a way more sophisticated way than you’re gonna be able to do. internally. And so it’s just exciting and this is how I’ve always felt about Revenuecat since I joined. You know, my little weather app that’s, making a few grand a month is, has more powerful infrastructure than some of the biggest apps in the world. I get access to this tooling with experiments and remote config paywalls and customer center and all this kind of stuff that, some of the biggest apps in the world don’t even have that level of sophistication

Shamanth Rao: Yeah.

David Barnard: in their backend system

Shamanth Rao: Yeah,

David Barnard: And I get it from my  $2,000 weather app. So it’s kind of fun with Rico

yeah, That

I could probably vibe code a little bit of this, it, if I weren’t using RevenueCat, but now I don’t have to do anything. I just get it and I know that,

Shamanth Rao: Yeah.

David Barnard: and I can see my teammates, hardening it and and I see all the people using it and know that it’s going to be the best possible tool for that.

Shamanth Rao: Indeed. We all live in interesting times. It’s an exciting time to build. , David, this has been incredible and, , much like the last time we spoke, there’s just so much I learned, , and there’s so much that I wanna put to use, , just the way you framed what’s changing, but also how you look at the product as a hook.

Just, there’s just so much I wanna go back to, but this is perhaps a good time for us to wrap. Before we do that, could you tell folks how they can find out more about you and everything you do? And we will of course, put links to all of this in the show notes.

David Barnard: Sure. I tweet very randomly. I sometimes think I should like, just embrace my growth advocate role and be like the UP growth guy on Twitter. Tweet all  sorts of random stuff, so I don’t know that’s the best place to follow me. LinkedIn, I’m so hit or miss. I don’t really enjoy spending a lot of time on there, and so I don’t end up posting as much as I should.

But connect with me on LinkedIn. Find me on Twitter. But really the best way to stay up on things is the Sub Club podcast and the Sub Club Live. So we’ve been doing almost weekly live episodes now where, for years I’ve had people say, I was listening to this episode in the car, and I just so wanted to ask this question. And so I’m bringing experts on to Sub Club Live where you can ask questions in the comments and so we’re doing live Q&A’s and topical things. So I’m doing paywall, roasts with Vahe who runs a paywall agency where we look at paywalls and people submit and make suggestions on that. I’ve been doing a regular Sub Club Live with Thomas Petit. He is in demand, very expensive, and doesn’t have room for any new clients and hasn’t for years,  but you can ask him questions, live on YouTube and, help troubleshoot some of what’s going on. And so I’m trying to make the live on YouTube way more interactive and, benefit to the community on top of what I’m doing on the podcast.

So follow the podcast, follow us on YouTube. and then we also have the Sub Club newsletter that you can sign up for that kind of encapsulates everything going on in the sub club world. So we will share links to past live episodes. We’ll share the calendar of what’s coming up and then share really insightful tips.

My colleague Peter actually does a lot of the work on that, and one of the things I love that he does every two weeks is he leads the episode with three recommended tests based on content from the episode. if

Shamanth Rao: Oh.

David Barnard: is to look at that and see, like when the, I, I mentioned earlier the Duolingo podcast episode that I did with their chief product officer. When he said in the podcast, we tried try for $0 and 0 cents and it outperformed ‘try for free.’  So my colleague Peter put that as Hey, try testing that as your CTA. And so it’s really cool. He isolates three different things to test in the newsletter. So it’s super valuable and then we share, what’s going on with YouTube and other things in there.

So that, those are probably the best ways to stay in touch.

Shamanth Rao: Yeah, so many places to link to. We’ll link to all of them and I will perhaps steal that for this episode. But just to list three big takeaways, from everything we have learned, although I think there’ll be a lot more than three to pick from. David. Yeah. Excellent. David, this is a great place for us to wrap.

Uh, thank you so much for your time and I know you are on the verge of traveling out. So thank you for making time and your very busy schedule as well.

David Barnard: Yeah,

Shamanth Rao: Excellent.

David Barnard: It was a lot of fun chatting through all this.

Shamanth Rao: Yeah. Cool. Uh.