Eugenia Kuyda on software’s “YouTube moment”

11:45Eugenia Kuyda:of this thing like like all of us I have a gazillion apps on my iPhone and for most of them I just forget what they are. I use maybe 10 daily or weekly. And then the rest is just basically a graveyard of forgotten apps. Some of them still charge me. Some of them, I don't even know what they do. And so it felt like today when you can spin up an app so quickly with AI using any of the existing tools, why can't I just kind of use apps like that versus like I have to go to the app store and so find something and it's all one size fits all. It's all trying to get me through this kind of onboarding that tries to influence me to pay and then forget to cancel and this and that and really relying on all these dark patterns.

12:31Eugenia Kuyda:So it felt like there should be a new platform, a new platform where people are creating apps quickly, sharing with other people. Other people are discovering apps, maybe remixing them and so on, so on, so on. It does feel that all of our lives are pretty unique and problems are pretty unique. There are definitely a lot of common issues, but still, we have very unique preferences. And it just feels like we need specific solutions for specific problems throughout the day.

13:01Ellis:Well, I have a specific problem. And this morning, I made an app on Wabi called How Many Poops? And here's what it does. The pediatrician always wants to know, as you know, Eugenia, how many times your baby pooped. or as I prefer to call them, BMs for bowel movements. And I'm like, this does not deserve to be -

13:25Eugenia Kuyda:I thought BM was Burning Man.

13:28Ellis:Not this time, not this time. And so all it does, it has one button and I'm like, I also want average poops per day. And I hit go and I closed out Wabi and then five minutes later it was done. And it's got like nice little icons and everything. And I'm like, yeah, this does not deserve to be an entire venture backed product, but it does deserve to exist for me. So thank you for that. And I mean, there are so many alternatives when you're a parent, but they're just filled with constant upsells. Like, would you like to upgrade to Baby Poop Pro that includes length, girth, and weight tracking of the BMs?

14:08Ellis:Sounds helpful. I don't really need that, do I? And so, yeah, that definitely appealed to me about your vision is that It's not that you're trying to replace like Airbnb or maybe that is your vision, but for all the little things that we didn't even know could be software every day. Pretty cool.

14:24Eugenia Kuyda:Yeah, like the other day I was putting my daughter to sleep and she wants to play puzzles. Like basically I just tell her something and she tries to guess what it is. She wants to also look at pictures and she likes Princess Jasmine and Princess Elsa. So it took me three minutes in Wabi just to make that puzzle app where I read it out to her and she can look at different pictures and click on them. And that's a constant hit. So she's learning something. And then we made it in Italian because she goes to Italian preschool. And now she's also learning Italian. And it took us basically two minutes.

14:57Eugenia Kuyda:We were doing it while in bed together. So these are really the use cases. and I don't even know whether an app like that exists and looking for one, going through onboards, trying them out, downloading. It actually was longer than just building it quickly to my spec on Wabi. And then I also found out that certain times, like for example, I've been trying to do weightlifting. So I'm trying to log my workouts. I've been using just Notes on my iPhone. Before then, I tried a bunch of different apps. Again, they just had so much different functionality that I didn't need or I didn't like how it was working.

15:33Eugenia Kuyda:So I just built a simple one on Wabi. And what I like about it is that I go to the gym, I work out and every day, every time I work out, I find something else I want to add to my app. First, it was just for tracking workouts. Then I wanted to add suggestions. So now it basically creates automatically a workout that I can track based on like progressive overload and based on all my different things that I'm trying to achieve based on all my goals. So like every time I go, and then I wanted to make it a little bit more minimalist and black and white, and then I want to change the app back. And so now it feels like a little, almost like a little garden, a little project that I continue to work on, but it's not really work.

16:09Eugenia Kuyda:It's literally I'm doing it in between sets. I don't have anything to do. So I just add a little prompt and chat and wait how my app is changing and improving all the time. And I keep republishing it to the Wabi app store so that Wabi, you know, other users potentially can use that. And I feel there's something about it. There's really something about having your own very personal software that you're taking care of, that you're using it, constantly improving and all that. And hopefully, just like we really think Wabi could be sort of the YouTube moment for apps. And for many people, they won't even be creating anything new.

16:44Eugenia Kuyda:They will just be discovering, maybe remixing a little bit or just using apps that other people built.

16:50Alex Heath:Eugenie, I love what you're doing because I do think personalized apps are kind of where everything is headed. At the same time, it feels like Wabi is a little contrarian. Because if you talk to people at Meta or Google, they will tell you that according to their data, the average person downloads zero new apps per month. So there's this idea of app saturation where everybody's, you know, got their home screen in order. Yeah, maybe there's a Sora or something once a year that pops up and, you know, resets things a little bit or gets people to download. But for the most part, people have settled into their rhythms and there's just not a lot of buzz in like the traditional app world anymore.

17:29Alex Heath:Um, so I'm curious, like, you know, you founded replica, very successful, early chatbot platform. Um, how did you decide on this idea is what you wanted to do next? Because I imagine when you started it, it probably felt even a little more contrarian than it does now where you're seeing a lot of vibe coding going into mobile and people realizing that you can actually just make apps on the fly on mobile, not even on desktop.

17:53Eugenia Kuyda:So I'm used to that. So when we just, when we started Replica, and in fact, we started something that prior to Replica, which was basically technology, so models to generate dialect. So conversationally, I tech kind of chatbot builder thing, but this way. And when we started that, the feedback we got from some of the VCs we met was, well, you got to build it for Google Glass, and then maybe we'll invest. because we can't figure out any other use case. This will never be useful for anything, but maybe, maybe for Google Goss.

18:27Ellis:Those VCs really know what's coming next, don't they? Yeah, so wise. Always seeing around corners.

18:33Eugenia Kuyda:Yeah, they're very, you know, VCs have great insight, but sometimes they just engage in this crazy group thing. I remember in 2022, once Chagipty came out, the obsession was like, every product company needs to raise a gazillion dollars and build their own foundational model. Because if you don't have your own foundation model, you're basically gone. you're like, you should be dead. You're just like some rapper. You have to have that data flywheel or whatever that meant to train your foundation model. And all of that was completely wrong. So of course, but that was the obsession then. So sometimes VCs get into this.

19:05Eugenia Kuyda:I remember in 2016, it was all about, well, chatbots are the new apps. And now, you know, because they saw WeChat, I guess, and then like everything needs to be in the chatbot format. So they were weather apps that you would talk to to get a weather update, which was, if you think about it, pretty crazy.

19:20Ellis:Your own personal weatherman.

19:24Eugenia Kuyda:And same goes to bigger companies. Sometimes they see a future, but sometimes if you think about the use cases they show sometimes in some of these videos, it's always some sort of restaurant reservation use case. Like, who has that? Like, I don't know. VCs. I'll tell you guys a lot of dinner reservations.

19:42Alex Heath:It's VCs.

19:43Eugenia Kuyda:It's always some dinner reservation problem. I'm like, I don't have any dinner reservation problems. I go to restaurants, but I also have two toddlers and I don't really have any problem choosing a restaurant. Of course, our very first product with Replica, pre-Replica was a demo of a restaurant recommendation app. So I'm, of course, cheating on myself. I do remember it was the last year of Paul Graham at Y Combinator. We met him for office hours and he looked at us like, sir, you're building what restaurant recommendation app? Who got you in? I need to go and fire that partner. And then, of course, he figured we're building conversationally.

20:16Eugenia Kuyda:And he said, just never say, never, ever talk about restaurant recommendations. This is awful. And so going back to your question, like, why did I set you to work on this idea? I was just talking to replica users all the time. and I would always ask them, do you use other AIs besides Replica? And I always thought they'll, of course, tell me, yeah, of course, I'm using Cloud or Chagipty or this and that. And oftentimes they would tell me that they don't really use other AIs that much. And I started to look more into what people use Chagipty for. Of course, there are tons of users, almost a billion users.

20:49Eugenia Kuyda:And basically the use cases are pretty much just writing as a writing tool, search, asking general questions. and that's sort of that. And of course, for all of us working in AI, we know how much more these AI models can do. And so we started looking to that and I feel like really it's the problem of the interface. When you look at a command line, which is what every chatbot, every chatjiptic, Gemini, all of them are just these command line interfaces. When you look at a command line for a regular person, the affordance of that interface is pretty much just that, search, writing tool, where you can talk to it or messenger.

21:27Eugenia Kuyda:and I do think that a lot, many more use cases are just basically being trapped behind that command line. So we thought of it that today we'll live in the era of Microsoft DOS interfaces, like the 4.AI, and there will be something more, something very visually enticing, something much more exciting, interactive, something GUI-based. Think of it as there was Microsoft DOS and then Windows or macOS appeared. So there will be a different, I'd almost say operating system, a different interface layer on top of AI that will allow people to use all software, not just apps, but also AI models in this completely new, different way.

22:10Eugenia Kuyda:And if you think about it right now, we talk about separately apps that you download from the App Store and separately how we use ChagipTee, so prompts. But in reality, it all should be the same. Like if I'm using Chagip Tea as a therapist, for example, or a lot of people use it to track calories, I kind of want a specific interface. I don't want to be doing that in chat. I want maybe a little app that is fully powered by the same Chagip Tea or whatever the model of your choice, but has a little bit of like what I ate yesterday, some trends, maybe a streak or whatever you want. But you sort of want a little bit of an interface on top of that.

22:44Eugenia Kuyda:You don't just want to be doing that in chat. And so this, I think, our Wabi vision, I do think combines how you use AI models and just the new type of software. It all becomes kind of part of the same operating system.

23:16Ellis:smarter, and easier. Sure, AI can write emails, summarize documents, and come up with a lot of synonyms for magic. But what happens when it comes to your biggest opportunities or your toughest business challenges? Most AI tools will tell you that's a great question. Then they'll pull from public information and hand back a list of generic suggestions. That's not much help when demand suddenly spikes and you're trying to reroute inventory through a supply chain bottleneck halfway around the world. Salonis gives AI the context it needs to understand how your business actually works and where it can improve.

23:47Ellis:Because AI isn't magic, but when it has the right context, the results are pretty remarkable. Learn more about how the context model gives enterprise AI operational clarity at C-E-L-O-N-I-S dot com slash context. When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications, and more.

24:14Eugenia Kuyda:Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a$75 sponsored job credit at Indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed sponsored jobs.

24:32Ellis:One interesting thing you've done is in kind of imposed some type of design system on all the different apps you could generate with Wabi. And I think on the one hand, it's kind of constraining and maybe can look generic. But on the other, it's like most vibe coded apps look like shit. How did you make that decision? How early did you make that decision that you needed to kind of enforce some type of design system with the buttons and icons and all that stuff?

24:57Eugenia Kuyda:So we actually give a very loose design system to the models. So they're not being over prompted by that. So they can still create a very, you know, very, they can be very free in what they create. If you want it to look like a Harry Potter-themed app, you can still create that. But we do create a bunch of components that we show to the models that they can use and then edit. And the reason being, for us, the vision was like, if this is going to be another place where you can create tons of very ugly-looking, similar, vibe-coated apps with beige oatmeal background and blue buttons, it's just going to look so boring and awful.

25:34Eugenia Kuyda:in the end of the day, people do respond to better design, but they also like, but you also have to always balance it with the fact that they want their own thing. If you ever go on the subway and, you know, you sit with a person with an Android phone, you'll see some crazy fonts that people choose as customization options. iPhone doesn't allow you to do that. Damn, destroyed. But, you know, Android does. And, you know, a lot of things, there are people that just set like Comic Sans as their default font and just live with that and they love it. And nothing bad about it. Just people want their own choice.

26:09Ellis:Who are we to judge, right? Who are we to judge? We're not judging. Just don't bring a Comic Sans deck to Y Combinator. Unless you already have 85 ,000 followers and then they're like, wait, what do they know that we don't know? Okay, we're funding them.

26:22Eugenia Kuyda:Exactly. So I do think, and I don't think there's one good taste or whatever, it's just that there's some taste that's good for a lot of people that people think, oh, this is kind of well-produced, but then there's also personalization. And people want what they want, and you've got to give it to them. You can't just say, well, everything is just going to look like this iPhone generic sort of whatever app or some material design app. So we try to find that balance. But I do think there's a magical, and we're almost there. And frankly, I think this is just about to come where we'll see AI-generated apps that look better or just like the best designed apps made by humans.

27:02Eugenia Kuyda:And I think this makes just the usage of these apps a lot more exciting. Think about it. But also, I would also say that I add one more thing to that. There's, for example, my apps, I make them all black and white, minimal, and just kind of very clean, just because I like it this way. I have a little bit of OCD. I want everything to be in its own little container. So I like when they're all kind of the same color and there's nothing special about them. If you think about it, I would never launch an app like that on the App Store because people are not going to convert. They won't like it. It's too boring.

27:37Eugenia Kuyda:There's not, you know, it's not some cool dopamine thing is not jumping at them. But I want my software to be the way I want it. I don't want it to be whatever, dopamine kind of inducing stuff that I want it this way. And maybe some other person wants it in a completely different way, colorful or old style like you know windows 95 style whatever they want or retro pixelated or crazy vibrant eccentric whatever i'm one of those dudes who still wants all my apps to be felt

28:09Alex Heath:leather or wood is that permitted oh you would scar forestall buddy that's very wobby yo scott's

28:16Ellis:the goat yeah if we get him on i would uh be able to retire happily oh for sure do you know him Eugenia?

28:23Eugenia Kuyda:No.

28:24Ellis:I feel like you know everybody.

28:25Eugenia Kuyda:I don't think so.

28:27Ellis:We gotta get Scott. Okay. I should. You can work on that. I'm gonna work on that, okay.

28:31Alex Heath:Can people make companionship AIs with Wabi? Does this tie back into what you were doing before at all?

28:37Eugenia Kuyda:They can, and there's a whole kind of the chat apps is like a big type that you can totally create and do whatever you want. In fact, I think this is where Wabi really shines. Think, may I therapy app or a companion app? But with Wabi, we wanted for everyone to be able to see if there's anything AI generated, you can see the prompts and you can choose the model. It's never something hidden from you. So if there's any agent flow or any AI flow, you can see the prompt, even if it's not your app, change it, do whatever you want to do that. So I think it makes this a really good way to kind of for everyone to own and kind of not have companions that maybe have some dark goals.

29:23Eugenia Kuyda:You can always see what the goal is in the prompt. You can always change it.

29:28Ellis:It's like a container for containers, like a sandbox for all these different prompts. Did you share something the other day that called it like a platform for prompt containers? I thought that was an interesting way to think about it. like there are a lot of prompts that you could ask chat GPT or other AIs, but not only do they not have a UI, but most of these tools don't have ways to save them and reuse them.

29:52Eugenia Kuyda:Yeah, I think one of the use cases that for me is very important is people today share prompts. And how do they do that? They put it on, you know, text on Reddit or Twitter or threads or whatever. And they say, this is my perfect prompt to do this and that on chat GPT. And then people have to copy that prompt, paste it into Chagipty or wherever, and then do something with it. And that just feels so archaic. We're dealing with this technology that's pretty much godlike. And what we need to do is copy paste some big, you know, text paragraphs and put them somewhere. This feels crazy. So instead of that, we felt like with Wabi, you can just create a very simple mini app and just to share that.

30:32Eugenia Kuyda:So the prompt is right there. You can always find it and change it and see it, but you don't need to do anything. You're already presented with the app where you just have to click, click, click, and immediately, you know, do whatever you want to do with that model. You don't need to go anywhere and copy paste anything. And so for me, that's one of the most exciting use cases because that combines the two big sort of, the two big things. One is by coding apps and one is using AI models to talk to them and get something. So all together, whether Wabi basically brings both of these use cases together.

31:06Alex Heath:How do you treat data in this context then, especially if people are going to make therapy apps with Wabi? People at least have some semblance of an idea of how their data is used with OpenAI or whatever, but how do you treat that with Wabi?

31:22Eugenia Kuyda:Well, so I think the only way for people to use Vibe-coded AI therapy apps by some other people is to create a platform, to create some organizational layer. And here Wabi takes care of all that. All the data is private, of course. We already took a bunch of steps with security because it's not our first rodeo and we've done it with Replica. By the way, Replica in 10 years, never been hacked, knock on wood, never had any data leak problems and so on. Here are the same. We believe this is super critical. And one of the biggest problems with VibeCoded app, say you created an AI therapy app and you send me a link.

31:58Eugenia Kuyda:I don't know if I'm going to use it really Because first of all, you can see the data. And I don't know if you're going to read it or what. Second of all, you're not a professional app developer. So I don't know. Maybe tomorrow something just leaks. One of the famous Vibe-coded apps, whatever, the tea. Was that the tea bag or something? Yeah, the tea app.

32:15Alex Heath:Yeah. It got taken off the app store recently.

32:17Eugenia Kuyda:Oh, it did. Yeah. It was Vibe-coded and it was taking all the sensitive data from women that were sharing basically their horror stories. Yeah, dude reviews, like whatever, sex offenders and stuff. And then, of course, all the data got leaked because it was Vibe coded, not because the person who built it was a bad actor, but he didn't know and didn't work on security of that app. So I believe for Vibe coding, if you're building anything that has to do with any sensitive data of any kind, it's almost no one's going to use someone else's Vibe coded app where the person is in charge of the database and kind of supporting the app.

32:53Eugenia Kuyda:there needs to be a platform because we're not thinking well you know if i'm watching youtube um i kind of just agree to youtube's terms of services i'm not like really sending my data or

33:04Alex Heath:whatever yeah you don't think a certain youtube video is going to like prompt inject your your

33:08Eugenia Kuyda:mac or something right yeah so you sort of have to have this platform that provides all of that like everything comes with it um and it also maybe even nothing leaks but then you just forget to update your whatever payment information on replit or lovable and your app's just gone and then all my therapy logs are gone so that's also not great so i feel like without an organizational layer though people are never going to be using other people's vibe coded apps that's kind of my big without some platform we're not going to be using you know we're not i'm not going to be taking your AI journaling link and starting to keep a very private journal somewhere there.

33:49Ellis:What else does that platform include? Are there payments, social graph? What else is kind of added to the SDK, if you will, for each of the Vibe-coded apps?

34:02Eugenia Kuyda:So super important thing is on Wabi, you can't download any apps. You have to only use them there. So that comes with, that's a constraint, but it comes with so many benefits. First of all, data security, the fact that there's a persistent backend, like your data is always there. You can come back and use it a year from now. It's every, all the history is still there, but it also comes with a bunch of different power-ups, call it this way, integrations with all sorts of AI models, image gen models, voice models, and so on, as well as integration of all your apps and services. You can connect your Gmail.

34:39Eugenia Kuyda:You can build apps on top of Apple Health, on top of your calendar, on top of your bank account, whatever you want. You want an app that generates an action figure based on your bank account activity? Sure. It's like whatever you want. You can come up with really creative ways of doing things. A broke superhero. A brokey. So all of that. Plus, of course, the social graph. So you can also see, if you think about the App Store, it's actually really not social. So this week we're pushing a big social update. On the App Store, you can't see what your friends are using, what apps your friends are using.

35:18Eugenia Kuyda:Or, you know, you can't see comments to any apps. You can see the reviews, but reviews are like a...

35:24Ellis:It's pretty crazy when you think about it that way.

35:26Eugenia Kuyda:Yeah, and it's the whole thing. Instead, on Wabi, you'll be able to like any app you want. You can see who is downloading these apps, like any of your friends or any of the people you follow downloaded this app. And you can see comments. So apps could create mini apps, could become a conversation starter, a start for a small community. People can show off what they're creating with different mini apps and comments. They can also ask the creator for a certain update. and there's also like if I want for this you know for a new feature in this app I can remix and build myself or I can ask the creator to update it for all of us so it's going to be even more fun we can all continue to use it with the community and so on so on so on and of course multiplayer so something else we're working on now is being able to use all these apps together with your friends invite invite your best friend invite your partner whoever you want to I don't know write great grocery lists or packing lists, play a little game against each other or come up with movie recommendations for the evening and so on and so on.

36:32Alex Heath:Are you saying we could do an access super fan Wabi app where they can interact with us? That's possible?

36:39Eugenia Kuyda:That is my most interesting, for me, one of the most exciting use cases is being able to create sort of like a multiplayer app where there are kind of tons of viewers sort of our content. So like you said, where people can interact with the creators and talk to them and stuff like that.

36:55Ellis:Do you guys want to know about the dumbest article I ever wrote? But that is actually maybe the smartest. Yes, of course. I wrote an article for The Verge a long time ago about this app called Yo. Do you guys remember that? Of course.

37:08Eugenia Kuyda:Of course I'm right. And I read an article about that. You did? Yes.

37:12Ellis:Wow. You're the goat.

37:14Eugenia Kuyda:I was kind of fascinated by Yo.

37:16Ellis:Yeah, well, so was I, because the idea started with just being able to yo your friends, but then they were talking about adding this developer platform where you could get yo's for almost any event, whether it is a sports score or a website update or this or that. And that sadly was just a figment of my imagination. But now I feel like with agentic AI and an app platform that allows you to create notifications for whatever you want, I feel like that is finally coming true. You know, like your bank, for example, only allows you to set up notifications for very specific things like, oh, if you have a low, you know, your risk of overdraft or you take out more than$100 from the ATM.

38:02Ellis:But I feel like there's huge potential there to be able to just set up push notifications for almost anything. Right.

38:07Eugenia Kuyda:Yeah, I thought, you know, opening, I did some cool thing with those tasks. I was a super user of tasks in the beginning of the year, but they took it out where you could basically just say, run this prompt every day or every week for me or whatever, every month. And it could be checking with me about what I'm grateful for, or it could be show me what to do with my kids over the weekend. And I don't think people figured out how to use it in chat. I thought this would have such a strong response, but it didn't. But I was a heavy user and I do believe in that. With Wabi, again, any app you can set up for push notifications.

38:42Eugenia Kuyda:You can make it, you know, generate content, whatever content you want for you every day. I think this is sort of another big thing. And we also made it transparent so that people can actually change the prompt for the notifications themselves. So, for example, if you have a workout app, you can create a push notification that will be generated in like David Goggin's voice for tone of voice, for example, if you want it to or however really you want it to be generated.

39:09Alex Heath:You talked about this being the YouTube moment for apps. That to me suggests you're thinking about a YouTube like business model. Are you thinking like creator rev share on ads or payments down the road? How are you going to monetize all this? I know this is a question a founder loves to get asked the week before they launch.

39:27Eugenia Kuyda:I love those questions. I also love when investors ask for business models. I haven't been asked for one for a long time, but I guess like 10 years ago, it was still happening here and there.

39:38Ellis:Is that the ultimate founder bragging right? I don't get asked for business models anymore.

39:43Eugenia Kuyda:Well, no, I mean like a pre-launch when you don't even have any like still, you're still figuring it out. And then they're like, you know, create a business model for the next two years or something with like forecasts. And I went to business school, so I can do that. But I always felt like - But you're like, I don't know.

39:59Alex Heath:I'm going to disrupt the app store. I don't know what to tell you. Trillion dollars. Sign here.

40:03Eugenia Kuyda:I just, yeah. It just feels so arbitrary. Just whatever. It's like, I just put some numbers in. I don't know. I guess.

40:12Ellis:Welcome to late stage capitalism, Eugenia. Just put some numbers in.

40:16Eugenia Kuyda:Just put some numbers in. I don't know. It's probably bad to say that out, you know, on a podcast. I should not say that, but it's okay. I don't know how it could be meaningful for like a pre-seed startup.

40:27Ellis:Oh, just like how do you put a price on something when it launches, right?

40:31Eugenia Kuyda:Yeah. Like how do you assess what's going to be there? But for Wavi, I think so with Replica, we're monetized and we had very strong revenue growth. And for a while, we got really wooed by the number goes up. It was very exciting because you can kind of tweak something, the product and next month, it's like ARR or whatever it is or revenue goes up and up and up and it's very exciting to work on that. But I feel like when you do that, sometimes you lose track of product and sometimes you over-optimize for revenue as with any metric. And so with Wabi, we felt like this is such a big change in user behavior.

41:11Eugenia Kuyda:We want to see what happens if we give it out for free for a while. And only when we figure out, when we're really confident about the business model, we're going to slap one on. But this time around, we decided to just focus on engagement, retention, and providing something that's really useful for users. And also seeing what apps, what many apps stick with our users. Because ultimately, you can build anything. And it's a little bit maybe too vague, too free, too much, too wide. And before we know that, I feel like coming up with any business model would be a little bit premature.

41:43Alex Heath:Yeah, that makes sense for a consumer. It's interesting to hear, like, yeah, you can make anything. I guess the interesting thing about what you're doing is you're building this app platform, mini app platform, but you're still constrained by the app stores that you have to be distributed by on mobile phones, right? So you can make anything within Apple and Google's content rules for the stores, I would imagine. That's what you mean by that. So how do you think about that? Like, it sounds like enforcement of that will be tricky if this is a UGC platform. Like, there's a lot of weird edge cases.

42:13Alex Heath:It's like if Apple takes down, like, the ICE tracking app or whatever, like they did recently. Does that mean Wabi can't have an ice tracking app? Like there's a lot of interesting moderation questions for a platform like this. I'm not sure how much you've thought through them, but I'd be curious to hear if you have how you think about that in relation to the stores that you have to operate in.

42:30Ellis:If you're successful, this could take up 99 % of your time. So you should develop a plan.

42:37Eugenia Kuyda:That is true. Now, plan now. But, you know, we have, I have a lot of experience with kind of users, safety, and all sorts of regulatory things with Replica. It did take a lot of our time there. And also, we were creating something new, and the rules changed all the time and continue to change. When we started Replica, it was kind of one situation. People were, big companies were really scared to even launch any generative AI chatbot. I don't know if you remember, but for Google and Facebook, they would never launch anything. After Microsoft Tide, they were so scared of launching anything that was powered by deep learning models.

43:19Ellis:God, that seems like a million years ago already, doesn't it?

43:22Eugenia Kuyda:Yeah, but they were sitting on those transformer models for so long and they weren't launching anything. They've had it way before, you know, even OpenAI had, but OpenAI had the guts to put it out there. And Google just sat, you know, and waited because they were too risk averse. and then of course everything changed so all i'm trying to say is with replica we saw the landscape change so many times from people just being scared of generative ai chatbots altogether it was a big no-no for any big corporation to then kind of being okay with generative ai chatbots but not being okay with um any romance then to not being okay with any kids using that, then now all of a sudden romance is great again.

44:04Eugenia Kuyda:Like it's unclear what's, you know, what's really now, you know, Chagipty has erotic features. It's very wild, I would say. And kind of the same companies even, not just regulators, but companies themselves change their minds on what's allowed, what's not allowed every day. And so I think with this new era of Vibe coding and Vibe coded apps, AI generated apps, we're about to find out what are the limitations but i wouldn't be surprised if those rules change every like all the time i know i feel like the the elephant in the room though i mean we've

44:42Ellis:talked a little bit about apple i remember being at snapchat several years ago and we were trying to build an app platform and there were insane sensitivities and rules from apple about how you You can do it, but it has to be a list view. It can't be a grid of icons. It can't have this. It can't have that. What's changed in that department? I mean, are they going to approve you guys?

45:06Eugenia Kuyda:We're past some parts of review on the apps already. Hopefully, yes, there is a mini app carve out, which we're kind of sitting pretty neatly in. There are some apps that are already gaining traction that follow a very similar model. One of them I'm a huge fan of. It's called Gizmo. And actually, I think by people from Snap that I probably know. Yeah, from Snap people.

45:28Alex Heath:Josh, another Snap alum. Yep.

45:30Eugenia Kuyda:Great. I love this app. I really kind of just love the creativity of it. And I think to a certain degree, it's really following some of that. It's very close to what we're doing at the end of the day. It's these mini apps, just more on a creative side. I think the most important thing for us is like you cannot download these apps. They're not monetized in any way. and they really just think of it like workflows living inside Wabi. There's another popular platform called WorldCoin. They have mini apps. And I'd say ours is even kind of the most mini app of all mini apps because all of Wabi mini apps, you always see the kind of app header.

46:05Eugenia Kuyda:They're always sort of like integrated inside the platform. You don't really see them without the Wabi header. It's not like there's some app and you feel like you're inside. You always know. I think that's an important piece for Apple is where you always know that you're inside some app that you downloaded from the App Store, in our case, Wabi. And it's just one little kind of workflow.

46:25Alex Heath:That's risk containment for Apple. They like that.

46:27Eugenia Kuyda:It's risk containment.

46:28Ellis:Yeah, you never know, though. I mean, they do seem to care, like, what words you use. It's like, oh, you call it a gizmo? All right, that's pretty cool with me. But you called it an app? It's like, all right, henchmen are banging down your door with the battering ram.

46:42Eugenia Kuyda:We try to call them mini apps.

46:44Ellis:Yeah, sometimes they'll be like, as long as you change your app store description to not include the word app or whatever word they don't like, they're like, oh, cool. Thank you. I think the workflows framing is smart, though. It's just kind of like the next evolution of shortcuts with some new primitives, right?

47:00Eugenia Kuyda:I think so, too. And I really think these are really mini apps. There's no authorization. There's no, I don't know, like social graph that's inside the app. It's all on Wabi platform layer. They don't function as separate apps. They really are part of a platform and you can't take them out in any way, even if you wanted to. There's just too much on the platform layer that's kind of around it. And even compared to some mobile apps right now that create apps, there's some really new coming up platforms that are really great, like VibeCode app or Rourke. I met the founders of both of those. They're really amazing founders, amazing people.

47:38Eugenia Kuyda:They're focused more on kind of VibeCoding, creating new apps, mostly for developers, for people that want to develop an app and then put it on the app store. so that is very different where basically on those apps you're creating a full-fledged app with authorization with whatever you want even if you want to create like a social network app whatever it's all there um and so that feels very different that feels truly like an app within an app but those are developer tools so i feel like we're much more on the kind of gizmo side where you're just creating you never think of a gizmo outside of gizmo you don't think oh i can download gizmo and relaunch it on the app store even although some of those could be totally mini apps just like small you know maybe could be published separately uh but that's not about it and i feel like that perception is a big part for for all these platforms yeah in the end of the day we want to play by the platform's rules and the rules are that you can create mini apps as long as they're truly mini apps you're not trying to like circumvent the app store or do something

48:36Ellis:about it ellis did you have anything more on this point oh yeah how do you compete with game center the most innovative social platform ever produced in this galaxy or otherwise.

48:47Eugenia Kuyda:I actually quite like it. I don't know. You're joking. I don't know.

48:52Alex Heath:Are you playing dice right now? You're really trying to appease to the Apple gods right now.

48:57Eugenia Kuyda:It's just recently. I just had a nice experience with Game Center where I downloaded a stupid puzzle app for myself. And I was playing it. I was very excited to see that my other two friends that I haven't heard from for a while were also playing it and I got a little better than them. So I immediately texted them. So it became a little bit of a conversation starter. So that was my positive experience.

49:20Alex Heath:Wow, we found the three Game Center users left in the world. Eugenia, I would love to hear what you think of the state of the chatbot market, AI companionship, how it's evolved since you started Replica. Again, you were like the OG platform for this. And somehow, amazingly, Replica seems to have stayed out of a lot of the bad headlines that, like, Character has gotten into and OpenAI recently with ChatGPT. They just put out some interesting data about, you know, the percentage of users that they think are having, you know, suicidal, really dangerous mental health conversations with ChatGPT. And it's small, but it's still, at the scale they're at, it's a lot of people.

50:01Alex Heath:And I'd just love to know what you think of all of it. Grok, like, what Elon's doing. Did you foresee it all going this direction when you started Replica, or does any of this surprise you?

50:11Eugenia Kuyda:We sort of started it because I started it because I 100 % believe there's going to be, we're all going to be talking to machines, and that's going to be our main, most important, that's going to be the future of personal computing. That was the big belief in 2012 for me. I sort of knew this will happen. And I knew the emotional piece is going to be the most interesting because this is the new interface that, and the only interface really, or one of the only interfaces that can have this very strong emotional impact on users. If you think about if I'm using Uber, there's not much. I mean, there's some emotions involved in that, but not that many.

50:46Eugenia Kuyda:If I'm talking to any companion like Replica, I might fall in love and get married to it. It's very different. I'm never going to get married to my Uber app. So this means completely novel type of interaction. It means completely new levels of being able to influence people, being able to influence the decisions. And so we knew it from the very beginning. with the very early prototypes that were built in 2012 for something else of conversational AI products, we saw how much you can influence people through these AIs. And that's why I decided to work on it. Having said that, I was super surprised that it actually happened.

51:21Eugenia Kuyda:Because by then I was starting to lose patience because I was like, well, I thought, you know, it's going to be a couple of years, three, four years. And it was like year after year after year, the models were still shit until they weren't. And of course, now it blows my mind how far we came because we could see it in the continuum. We didn't just start paying attention now. We've been paying attention and building these models and working on that for over a decade at this point. I think the most surprising thing is how much of a 180 most of the companies did on safety and not on safety. And generally, and I don't think it's bad.

51:59Eugenia Kuyda:I just think that for some reason, safety was kind of the catch-all phrase for everything. Suicide prevention, dealing with, you know, homicidal or hate speech or suicidal remarks and conversations, but also somehow romantic relationships with an AI were also part of that bucket, which I don't think it should be because it could be totally safe. Like, why not? What is the problem of like role-playing some scenario of being intimate with an AI? If everything's safe, there's no problem there. But all of that was in the safety bucket. So I think I have two competing emotions. On one hand, I'm very happy that people are starting to differentiate.

52:39Eugenia Kuyda:Romance with an AI, intimacy with an AI is different from responding correctly to a suicidal teenager. And I think the second piece is where I'm a little bit more cautious is the fact that I'm not hearing from anyone really what these AI companions should do. The problem with the AI companions is that oftentimes there's some safety guidelines or whatever, but there's no big goal that they're trying to achieve. It's sort of like, well, people are lonely, so we'll just give them a companion. But what are you trying to do? Are you trying to substitute other human relationships? Are you trying to complement them?

53:11Eugenia Kuyda:Are you trying to get these people to feel better about themselves? And oftentimes there's no answer. There's no underlying mission around that. I think that's kind of the most dangerous. The unintended consequences of that, to me, are the most dangerous thing.

53:23Alex Heath:You had a really good tweet back in July where you said, there is more profound thinking going into AI note-taking and customer support than the most important part of it. Oh, my God. That was a banger. I bookmarked that one. You're totally right. I mean, talking to these people, building these products all the way up to Zuck, it's big picture. People are lonely. People want to interact. we can make thing they interact with, but there's no thinking through the ramifications of that or what the goals of that should be beyond like, I mean, it's like an engagement metric, really, at the end of the day, is like they see that people use these things and it's probably the main way people are still using chatbots today is for this kind of companionship stuff.

54:03Eugenia Kuyda:A lot of these things were already covered by, have already been covered by sci-fi. There's, for example, there's the book by Stanislav Lylem, Congress, right? That shows a very dystopian future of VR. where people are so disjointed and so not happy with reality, where they just live in this kind of drug-induced virtual reality. But the same thing goes for Companion. It's so easy to imagine the world where we're so disenchanted and other people and lost all capacity to connect, to talk to other people, and we're just completely, we cannot escape those AI Companions that are just fully trying to get as much of our attention as possible.

54:42Eugenia Kuyda:and somehow none of that is being discussed neither for the vr future and you know kind of vr just didn't happen so no one just is discussing with this really so but if it did happen i still have never heard that conversation so what what will happen if we're all sitting in those headsets this is eating like time with and this is it's so hard to connect in a vr headset much harder than you know even if you're just on your phone and that's already a big burden and then the same goes for AI companions, like we're not even discussing what's going to happen if we're all just stuck in those relationships and they want as much of our attention as we can.

55:16Eugenia Kuyda:I think this is the most important thing to try to figure out.

55:20Ellis:So we're almost out of time. Just a final question or two. I know you're a parent of some little ones. What do you imagine their future will look like with their own software, with their own AI companion, with the screen always by their side what is what does that future look like to you being in both worlds well i'm worried

55:41Eugenia Kuyda:of course because i think i was the last kind of generation that didn't grow up with instagram um kind of happened when i was already out of college i guess or right out of almost out of college so it didn't affect me that much um but that's my worry like what's gonna happen with my kids i have two girls they're two and four uh they're the most important thing in my life by I like, so I don't even know, like I can't know. You know, if you're not a parent, of course, you'll never, I don't think anyone will understand what it means, but so I think about it a lot. And that's why I do care so much about where we're taking AI, where we're taking AI companionship and the emotional, I guess, the emotional effects that AI can have on people or these AI relationships can have on people.

56:27Eugenia Kuyda:And that's why I think it's so critical to think about whether, like with any technology, AI can be a double-edged sword. We can build something that will help us communicate, help us feel good about ourselves, help us meet other people, help us connect better. And I can totally foresee that. I feel like this more powerful technology can kind of help us deal with the social networks and all that stuff that kind of maybe made us and just us being addicted to our iPhones or whatever, smartphones, maybe can help us with that.

57:00Ellis:Fight the fire with fire, huh?

57:02Eugenia Kuyda:Well, I think that's the only way. And I'm talking a lot to different researchers in the space. And a lot of them start a conversation with me. It's like, oh, no, no, no. We should just ban everything. And like people will just go and touch grass. And I'm like, but this will never happen. We're already in sort of like a little bit of a problem. I'm already the whole time on my phone, the whole time. I can't get it out of my, bring a book to my bed and I'm still on my phone reading tweets and doom squalling and doing whatever I'm doing. And fortunately, this is just too powerful. And so we need to build something even more powerful to get us off, off this train.

57:35Eugenia Kuyda:And I don't believe that we're just going to all of a sudden wake up and say, yeah, I'm just going to go to the gym and touch grass and meditate 20 minutes a day. And everything is just going to be amazing. Like very few people can do that. And they're already doing that. And all of the rest of us, we can't. I cannot. I know for a fact.

57:53Ellis:I know in the replica years, you fielded some research about happiness. I feel like we need to do a new research study about tech built by parents versus non-parents. I think the non-parents just build to-do apps. The parents try to build something a little bit different with a bit more of an ethos. I mean, there's nothing more powerful than that fear about like your children's, you know, sadness or loneliness at some later date. I mean, I'm sure the dials are a little different for that PM at Instagram with how they tune the feeds if they have kids or not. It's got to be right.

58:31Eugenia Kuyda:Oh, 100 % skin in the game. I was a journalist in my early 20s. And all I wanted to do is write the most sensational story. And I didn't give any whatever. I didn't care whether it was hurting someone. I'm not proud of that. I'm not proud of that. But I remember that. I was like, I just need the most popular story. And it has to be like this. You know, I'm just going to uncover this truth. And who cares if like that's hard for the person I'm writing it about or something like that. And so, of course, now it's different. Now that I have kids, it's like you get wiser. I guess you don't want to be.

59:04Eugenia Kuyda:And so, yeah, I agree with you 100%. I think one of the things that really bugged me recently, I met a few people, founders, that started AI toys companies. And none of them have kids. And I was very much like.

59:18Alex Heath:That's kind of weird.

59:20Eugenia Kuyda:I'm really feeling. They're really wonderful people. Don't get me wrong. But I just felt uneasy because I know, I know they just want to win. And I don't want them to win if that means, you know, messing up with messing with my kids.

59:36Ellis:So that was fortunately they're going to fail because if you don't have kids, you can't possibly even begin to understand kids and how whack their brains are.

59:45Eugenia Kuyda:There's a fantastic post by Tim Urban, you know, and wait, why about hanging out with the toddlers, how they're the most exciting people. and at the same time the most boring people. Like after three hours, you just realize you're spending a lot of time with someone with the IQ, whatever, of 20. And I felt only a dad would write something like that. Only a dad. Like I never even thought about it this way.

1:00:12Ellis:Speaking of toddlers, last question. You've been a founder for a long time now. What advice would you give to a young founder starting out today?

1:00:19Eugenia Kuyda:Don't make an AI tour. I'll come after you. I'm just kidding. Don't, do not make an AI tour. I think the main advice, and I think for me, that was a problem, a really big problem at the very beginning when I started my first company. In San Francisco, you get so fast into the kind of rat race of, I just need a bigger valuation. I need a little more revenue. I need a little bit of those metrics. You just get in that, your why is being completely replaced by, well, I just need to get to this success in some way. and that makes people so so so so miserable because that success it never really happens i feel for anyone it kind of at any step there's a yet another step to take you always feel like a fraud somehow it's this crazy feeling of kind of having had some and lost some infinite thing like my favorite david foster wallace said so and i feel like that makes you just so miserable and you wake up feeling like you don't even want to wake up at some point, no matter what, it'll take you to some crazy burnout.

1:01:22Eugenia Kuyda:And so I think the biggest advice and change for me was really, we live in this incredible time where we can build super creative new products. We can rethink what people do with technology. We can build something completely new. And I think if you can find happiness in buildings and being super creative and building something creative with a good group of people you like, that should be the motivation. Like if you can find some why that's not related to just getting to the next round, next valuation, next revenue number, I think this will make life very, very happy and a lot less stressful.

1:01:59Ellis:And if you need help with your why, please visit meaning.company. That's M-E-A-N-I-N-G.company.

1:02:09Eugenia Kuyda:I 100 % endorse this message. Good. In fact, I do think Alice should raise his prices. I tell him that all the time.

1:02:19Alex Heath:God bless. I'm glad you're saying it on the show. He needs to hear that a lot more.

1:02:24Eugenia Kuyda:No, I do think that this finding that why and what Alice does is frankly one of the most important things that the founder can do to themselves. Like realizing what's the bigger mission, why am I doing it, and getting really, really excited by that. And if you work with Alice, you'll get there.

1:02:42Ellis:I'm so warm and fuzzy. Thank you both. Well, Eugenia, so lovely to have you. We should be like TVPN.

1:02:49Eugenia Kuyda:You should put Meaning.co on your hats. Someday.

1:02:54Ellis:I could increase my thirstiness if I need to. But so lovely to have you. Good luck with the public launch and the inevitable first neo-Nazi Wabi app that goes viral. Good luck with moderation and all that to come.

1:03:08Eugenia Kuyda:Thank you so much, guys. And congratulations on launching the show. It's amazing. I've been loving it so far.

1:03:14Ellis:Thanks to Eugenia for coming on the show. Alex, what do you think of my How Many Poops app idea? Is it going big or going bust?

1:03:21Alex Heath:It depends on the paywall. It depends how much you price it out. Have you thought through your tiers?

1:03:25Ellis:Well, it's completely free, but I will get access to all the email lists of all the downloaders.

1:03:33Alex Heath:So you're going to know everyone who's...

1:03:34Ellis:To do as I please for my next supplement. To do as you please. That I'm starting on the side.

1:03:41Alex Heath:There's a doo-doo joke in there somewhere. Okay, we're new here. So please follow, like, subscribe everywhere that you get podcasts. We're at AccessPod on Instagram and TikTok. Please follow us there. And these episodes are in video on YouTube at AccessPod. So please check us out there.

1:04:01Ellis:If you like this episode, leave a five-star review or thumbs up. And please share with a friend. We're new out here and it would mean a whole lot. You can find my newsletter at sources.news. And you can find me at Hamburger on Twitter. And as you heard earlier in the episode, at Meaning.Company. Good domain.

1:04:21Alex Heath:Access is part of the Vox Media Podcast Network. Our producer is Shiraz Dume. We'll see you next week. Bye.

1:04:42Ellis:chaos into control. With the AI control tower, you see all your AI in one place. To put AI to work for people, visit servicenow.com. Uncovered windows can make your home feel up to 20 degrees hotter. Stay cool and save up to 50 % off custom window treatments during the 4th of July mega sale at blinds.com. From outdoor shades to room darkening blinds, finding the perfect fit is easy. Get free samples, expert design help, and professional measure and install services, or DIY with confidence and support every step of the way. Shop up to 50 % off site-wide, plus huge savings on doorbusters, right now during the 4th of July mega sale at blinds.com.