Why this founder isn't hiring junior employees anymore

Good evening, everybody. My name is Casey Newton. Welcome to Platformer Live. Incredible. Incredible. We are coming to you tonight live from Atlassian headquarters in San Francisco. And believe it or not, gang, this is, Platformer is five years old. This is the first live show we have ever done. Can you believe that? You are here for history, podcast and newsletter history. And, you know, for so long, people say, Casey, when are you going to do a live show? And I said, I'll consider it, but two things have to be true. Number one, it has to take place at Atlassian headquarters in San Francisco. And number two, I want to directly counter-program Microsoft Build. Ladies and gentlemen, I'm proud to tell you, we are here at Atlassian headquarters, and Microsoft Build was today. Yeah, that's what I'm talking about. So if you're here tonight, I imagine maybe you've been following along with this fun project that we're doing. It is a miniseries about AI and jobs and the future of work. We are now, we actually just published our fourth episode that I encourage you to check out. It's with the amazing labor economist, Catherine Anne Edwards. But tonight we have a conversation that I'm also looking forward to, in large part because it's live and we're working without a net. When we published these first few episodes, we've just gotten amazing feedback from readers because it turns out that so many of you, just like we are at Platformer, are curious about what it means that AI capabilities are advancing so quickly, right? What does that mean for the future of work? It seems like it's getting really easy to automate a task. At what point does that mean we can automate a job? So these are the questions that we're interested in, and we're going to have a really amazing conversation about it tonight. A little bit later in the show, Eugenia Quida is here, the founder of Replica and Wobby. Eugenia is somebody I've known for more than a decade. We also have a really fun conversation with Tamar Yehoshua from Atlassian. But to get things started, as we always do here on the Platformer podcast, to tell us about the latest news at the intersection of AI and jobs is Platformer Fellow and Gen Z AI correspondent, Ella Marcianos. Please welcome her. Ella, how are you this evening? I'm doing great. I'm so excited for our first ever live event. Me too. Me too. It is always more fun to work without a net. I imagine that you have, as you always do, sort of gone out and found something for us to talk about at the intersection of AI and jobs. So as you scan the horizon, what has seemed interesting this week? Yeah, so something you may be very familiar with if you've seen your Twitter mutuals, like, embarrassing job change posts, where they, like, have three paragraphs about, like, how they're going to change the world as a, like, B2B AI account manager, etc. Yeah, and be gentle, because some of those people are here tonight. I love you guys. But basically, AI, in addition to, like, creating some doom and gloom around, like, what our jobs will even be once, like, it's superhuman at everything, also has created, like, new job titles and, like, sometimes new job categories. And so I've been, I was interested to see a recent report about, in particular, Box's, like, new categories of AI jobs. So this is Box. This is the sort of, like, AI collaboration company, Aaron Levy, the CEO, was actually our first guest on the Platformer podcast. And so you're saying they have recently introduced some new AI jobs. Yeah. Okay. 13 new types of AI jobs, which almost seems implausible to me. I feel like it's, like, you type a prompt into Claude Code. That's an AI job. But as it turns out, there's, like, more than that. Okay. So, yeah, what's interested me is, like, there are some things that feel, like, a little bit more old hat, where it's, like, an AI forward-deployed engineer, which is, like, something somewhere like Palantir pioneered, where you just, like, basically have an account manager who knows more about the actual technical side of it to, like, talk to people, say, like, why and where you should be using your AI. But there's, like, also stuff that gets, like, a little wackier, a little weirder, like AI business automation engineers, where it's, like, you're doing a new frontier of management science where you're, like, trying to, like, look at, like, a team and see, like, what part of their jobs can be found. Like, uh... That feels like that would be, like, a very unpopular person at the office. You know, I'm just going to, like, sit at your desk and study how quickly I think you can be automated away. Yeah, I did one piece where I tried to automate myself, and I was very nervous the whole time writing this for Platformer. Even though you were the person looking over your own shoulder. Yeah, yeah, yeah. Yeah, it was still unnerving. But, yeah, and then there's also, like, they have, even though, like, AI companies, you know, do their own evaluations, they're completely untrustworthy. And so, like, they also have, like, a B2B company also needs, like, its own internal team to, like, evaluate the AI that it's getting from a third-party vendor to then sell to another vendor via their software. So, yeah, basically, it was just, like, a bunch of new gaps and layers where, like, at least some companies, like, actually are seeing an upswing in people they're wanting to hire. And they have, mostly I've been delighted by the variety of titles they have. All right, give us some more titles. AI architect, AI solutions manager, AI platform leader. What I was really amazed by is Stripe is hiring a forward-deployed AI accelerator. That's just a guy that, like, walks up to your desk and says, do it faster. Yeah, yeah. Okay, so a lot of, like, job titles. Like, what I'm curious about is how many of these jobs do we think are, like, net new versus jobs that are maybe just, like, the, you know, an AI-flavored version of a job that we've already had? Because, like, a lot of these SaaS companies, they have these, you know, like, customer success managers that'll come and check in with you and make sure you're, you know, getting the most value you can out of your, you know, Box.com subscription. But, you know, an AI accelerator, you know, maybe didn't exist before, mercifully. Yeah, so there are obviously some things, like, the quote-unquote AI storyteller, which just seems to be, like, a comms job that boys are more comfortable having. Yeah. Where I'm like, you're really just rebranding. But I also do think there's just, like, a new category here because, like, we don't actually, like, on the one hand, AI is so promising. On the other hand, we don't know, actually know how to use it. Like, I've actually been just, like, scouring the Twitter of Aaron Levy's, like, uh, cracked AI evals intern. And he's always working on something that's, like, very confusing to me. Like, Aaron himself is working on something like that. Aaron is always, and his intern is always working on it. Like, what is his intern working on? Oh, God, I can't remember the guy's name. But he always, uh, he, he's always, like, jumping in these, like, 30 thing threads that I'm seeing where, like, someone who's, like, on Twitter all day is, like, arguing about, like, a new OpenAI benchmark. And he's like, you have to consider X, Y, and Z measurement error that, like, I would never have thought of in my life. This is the AI job I want, just, like, getting involved in stupid AI discussions on Twitter. You know? I feel like I could have a lot of fun with that. That's why I'm a journalist. Exactly. Exactly. Um, alright, so I'm not sure that I actually got your take. Like, what, like, what is your sense of, like, is this, like, at least with the box stuff, does this feel like, okay, like, Box is actually, like, hiring a lot more people, like, because of AI there is more work to do? Or is it, well, we're just sort of replacing some jobs there with it? Or do we just really not know yet? Yeah, very boringly, I feel like in two years, if their clients are like, we can do this stuff X amount faster, then I'm going to be like, whoa, that was a new type of job. They did a new type of thing. And then, like, yeah, it, like, really, I think will depend to me on what sort of outcomes I see. That makes sense. And then on the AI architect, I'm guessing that's not somebody who is, like, using AI to make buildings. Right. What are they architecting? I think they're architecting via scaffolding, which is, like, the whole sort of universe and interface you, like, place your language model in. So, like, basically completely shifts, like, what it can do on your computer, what permissions it has, like, how it views its environment. And, like, very So right there. And so you, there's so much information at our fingertips and how we can use it to create is really interesting. Yeah. So tell us about what the teamwork graph is inside. Thank you for asking about the teamwork graph. So the teamwork graph is our context graph. So you've been hearing about context and why is context important? LLMs only know the generic what's going on in the world. They don't have the context of your organization. So an enterprise one, it's really important if you're using AI in the enterprise that you understand the context of your enterprise. So what is context of your enterprise mean? Like, what have you done in the past? So if your context could be from a tool like Jira, what are the issues you've had? Who worked on them? What documents were related to them? Or your design docs, maybe in Confluence, or they may be in Google Docs or SharePoint. Your conversations in Slack. So it's all of the information in the Atlassian tools and the information in third-party tools. And we build a graph, kind of like the knowledge graph from these are all just graphs of data and the relationships between them. So let's say an incident comes in and I'm an engineer, I can use the teamwork graph. It can tell me, well, there are incidents that were similar to this that happened in the past and here's how they were solved. So then you can solve those incidents much faster. That's just one example of how traversing the teamwork graph makes everything that you do and it surfaces in all of our products. I mean, that's really interesting because I can imagine, you know, somebody who's, you know, like, find a lot of Jira tickets over the years, or I can just imagine a world where maybe the same thing is sort of like happening, like, over and over again. You're sort of like vaguely aware of it. But what you're saying is that you can now apply a new layer of analysis to it and maybe, like, solve the root of the issue. Exactly. And what's really cool is because Atlassian's been around for a long time, all of the information from all of the years you've been using it, even if you used it in our previous data center product before the cloud, all of that is in the teamwork graph. I was meeting with a CIO who has had Atlassian tools for seven years and he's like, oh my God. So that's like really powerful. And then we also let you use the information outside of Atlassian products because we believe in an open ecosystem. And so you can come to Atlassian and use the teamwork graph or you can use it through our command line interface or MCP. So if you're using a coding agent, you can invoke the teamwork graph to get the context that you need for your coding agent to be more effective. I mean, this appeals to me because I feel like I often have the experience of, like, using software for a long time and I just feel like it is getting worse. You know, it's like getting slower. It's like harder to use. Like, they've moved the thing that I used to know. It sounds like what you're saying is if you have used these products for a long time, like, that actually works to your benefit because now you just have this, like, big corpus of things that just knows things about your company. Exactly. Yeah. Right on. All right. Well, so I would love to hear some examples of, like, how are people using this? Like, what are some fun things that you're finding? And, like, is it the case that maybe non-technical people like myself can now do a few things they weren't able to before? Non-technical people can do so much. It's really, it's really opens up a whole new world for people. So I'll give you an example of our marketing team. So we recently had an event a couple weeks ago, our annual event called Team. And I don't know if anybody's ever worked on a keynote, but they can be a lot of people involved. And there's like the text and there's the designs and then there's how are the chairs and where's the light and the sound. And you make one change in the script and then everything changes. And so the marketing team built an agent in Robo Studio that would look at all of the incoming information that was changing and proactively update it everywhere. And then had a punch down list of all of the things that needed to be done. And it was like their personal program manager that automatically tracked everything. And then the punch down list got updated. And what they said is what would take days to get everyone coordinated, they could do in hours. So this was an example that was really effective for them to work in a new way. And then it really reduces all of the toil of the job of just like updating things everywhere, which lets you work on the things that are more interesting. Yeah, yeah. I mean, talk a little bit about the reliability piece. You know, I feel like when I talk to a lot of people who, like, wanted to use something like this, they worry, oh, like, is it going to hallucinate? You know, am I going to like put it in charge of something? It's going to update things with the wrong number and then I'm going to get into trouble. Like, what's your experience of that? You have to be really careful in how you build these agents. So one, the teamwork graph gives you context, which actually makes the results better. Because then you have actual real context going into the LLM instead of the LLM potentially making it up for you. So that's one. And then we give tools to evaluate. So if you're building an agent as part of our platform, you can have tools to evaluate it because you have to make sure that you have some data that you're saying, OK, I'm going to try it and it's going to be OK. And a lot of people, you want human in the loop. So you want to say, especially if it's something that has implications, like I don't want to send emails without reviewing them. So draft all these, but don't send them. So you have to be very careful in the instructions you give. That makes sense. So, you know, in this series, we're really interested in how jobs are changing. I know that you've been using all these tools in your work. You were telling me the other day that you were able to put together a presentation in about 10 minutes that was like live updating with like real time stats from your company. And I imagine that saved you a lot of time. Yes. What I'm curious about is like what you find yourself doing with that time? Like what does it open up for you? Like I find that like AI does like make me more productive, but it doesn't mean that I use computers less, if that makes any sense. So I wanted to know like what your experience of that has been. So it's, there's been a lot of articles written about that lately, about how people are working more in that. But part of that is because it's fun. Like people are enjoying using the tools to do new things. So I'm going to take product management since that's what we do. My org does here. And what you want is that the AI tools enable you to do things that took a long time, like meeting with lots of customers, listening to the Gong transcripts of all the customer calls, writing it up, writing the PRD. So now you can do those in an automated way so that what really matters for a PM job, that you're building the right things. So it takes less time to build things, but you have to make sure you're building the right thing. So then it frees you up to work on more strategic things. Why are we building this? What is the roadmap? Because then you can, the cycle, so the coding agents make the cycle of coding much faster. So these tools make the pre-coding faster of the gathering the requirements, even setting the dependencies and the project plan and who's going to work on what. And so like a little teaser, we've been building those tools internally to make our PMs more productive. And that's something we're going to make available to our customers to enable them to make their products, their PMs more productive. That's the fun part about working at Atlassian. Like the things that help you do your job, you can actually help your customers do that same job. Yeah. You mentioned fun. We were sort of talking about fun a little bit before the show, and you were sort of saying that in your experience, like people have one of two relationships with AI tools at their work, which is either you're scared or they're fun. Talk a little bit about that distinction and like what, what do you say to, to scared people? Like, like, do you really feel like there's nothing to be afraid of or, or do you give them other advice? Well, part of it is scared is change. So people get scared of change and no matter what the change is, no matter what the new product is. And so there are some people, just that's their personality. They're a little bit more hesitant to change. And then there are people are like, Ooh, cool, new tool. Let me, let me try it. And I'm curious. And so you have to what I find is if you have champions in the organization who show that there, that it actually has results and creates value and makes them more productive and it's, they're better at their job, then that's what brings people along. So you'll have these like culture bearers who are the bleeding edge. And then we have them. We have something called the AI builders week where every week, every quarter, all the PMs and designers get together and learn something in AI, whether it's evaling or building prototypes or building agents. And then we have those people giving presentations of like, look at the cool My world and— Which is a Wittgenstein quote, if I remember right. Yeah. And so, I felt like if we figure out how to build language models, that will be probably the closest to understanding the world as well. And so, we started building that. That was really way before any first language models. And then in 2015, Google published a paper where they kind of— published a paper where they talked about the first deep learning model applied for dialect generation. And so then, we just decided to hire every possible NLP researcher we could find that would join us to focus on these language models. And then, of course, after Roman passed away, we built that AI for him and that sparked a lot of I guess media attention. Thanks to your story, wonderful story, a very beautiful story. And then, you know, we just thought, well, we're struggling to find an application because it's a new application. And that was it. We were like, well, maybe we can't build yet a chatbot that would talk eloquently with people, will have meaningful conversations, but maybe we can build one that they can listen and that would be probably enough for many out there. How confident were you when you were doing that? You know, because, like, I talked to some startup founders and they're like, this is a brilliant idea. Like, everything is sort of coming together. And then there's other founders you talk to and they're like, you know, I don't know. Like, we're kind of just, like, trying some things to see what works. Like, do you remember at the time, were you like, this, like, it sounds like from what you're saying, you were like, all of the forces are leading this to happen, that it's actually going to work. We thought it would work at some point. Let's put it this way. The models were so crappy in 2016 when we started Replica that really, we were just thinking, I mean, they would just produce, like, non-sequiturs. It was like sequence-to-sequence models and, of course, there were no models off the shelf or through APIs. So we had to build our own models. And, you know, they were just, whatever, right? Before people moved on to Transformers. That was way before that. And so, it was very hard to say, okay, well, this will become what it is today, but we just felt like we really need to, that it will happen, we just didn't know when. And I think for us, it was more like, okay, well, maybe the tech is lagging, but it's less maybe about tech capabilities, it's going to be more about human vulnerabilities. And there's so many people that want it so much, some connection, someone to listen, someone to hear them out, to accept them, to talk to them, to understand them, that maybe in the beginning, just those people will really react positive to that. But as the tech gets better, we can, you know, maybe increase the range of people that will resonate with that. I think there's something really poignant about the fact that, like, even though the technology, by today's standards, was, like, maybe not that good, like, the human desire and need for, like, support and connection was so powerful that people, like, looked right past it. And, but at the same time, like, I wouldn't downplay the technology either, because I remember, like, when I was talking with Roman's friends and family, the part of the story that will still make me cry when I tell other people about it is how much people learned about their friend and their family member after he passed away and how their relationship with him changed after he passed away because of the conversations that they were having in this app. And that was honestly the moment when I started to take AI more seriously as a thing, because I thought, if even in this sort of very primitive version of the thing that we have today, people can feel that deeply, well, there just has to be something there. I think so. And, you know, to me, just looking at my previous relationships, oftentimes we do have relationships with people where maybe they don't respond that much or it's more of our fantasies. Like, how much do we put in? A good example is, like, talking to God. Like, so many do talk to God and maybe he doesn't really respond. Well, he's sort of famous for leaving you on read. Yeah. Blue wall. Yeah. Yeah. But so many guys that I would fall in love with and I'm like, oh. I mean. Just non-sequiturs. Yeah. Yeah. They're all laughing like it hasn't happened to them too. Yeah. So. And also, I had a lot of experience going on dates and, you know, you would find yourself in some of those and you'd just listen and, like, maybe ask some, oh, tell me more, say some stuff like that. And the guy would be like, oh, that was the best conversation I've ever had. And I'm like, that was easy. And, you know, you just space out half of the time when they're talking. You're just like, think about something else, all the groceries you need to buy, whatever. And they're like, oh, no way. Like, can you? So I'm like, if this is the level of understanding that's required for the most amazing conversation, we can probably build that. Yeah. Yeah. That's it. Once you realized how low the bar was, you thought, there's a unicorn here. Yeah. All right. We'll come back to chatbots a little bit later. I want to, like, zoom out and try to ask you a question about your two companies. Because on the surface, they look quite different, right? Like, one is about an AI that you develop relationships with. The other is a tool for making apps. In your mind, are they completely different? Or, like, do you see a sort of through line there that you're chasing? They're definitely different things. But I think for me, it always was like, the idea was, like, well, how can we make someone's, like, a person's life better or help them unlock their potential through this? And I think with Replica, it's easy. It's like, really, it was always, like, we're going to build an AI to help people flourish and feel better in long term. We had some of the bigger studies published around that with Stanford, with Harvard. Some of them published in Nature, where we proved that we were doing that. And that continues to be the case. And with Gwabi, the idea is, well, like, most of the time today, we spend on our phone, on our phones, using software that's not built by us, you know, and David Foster Wallace' words, like, built by people that don't love us, that want us to just, like, either scroll or click on things. And, you know, just like we build our buildings and then they shape our buildings, then they shape us, same with software. We shape our software and then it shapes us, only that we don't shape it. Someone else does. And then we do what, you know, and in this new era where anyone can really build something in a matter of a few seconds, why not let people take a little bit more agency and maybe not build every app they use, but at least, like, have software be more decoupled from just every app needs to be a business and so you're gonna be, like, just doing what we need for us to make money. So Gwabi is a platform where people can make apps, but can also discover, remix, and use them with their friends, with their families. It's a social platform where you can quickly spin up any app or find any app and start using it with whoever you want. And that allows for, like, in my case, for example, like, really just creating software that's really fitting my life. So really improving my life, whether it's helping me learn more about the art movements I'm into or the language I forgot to play or teach my kids something or find cool events for to take my kids to or even just like a better weightlifting tracker and whatever, personally, I coach and all that. Interesting. So, you know, I'm hearing a lot of ideas that maybe I've seen a version of them somewhere else. I'd be curious to hear, like, what is, like, a feature or, like, a design element of something you've built that made you feel like, this is, like, just truly personally for me and I know no one else would ever, like, build it this way? Or, like, I would not expect to encounter this kind of app anywhere else. I mean, a lot of the software, like, basically when you take away the idea that now you have to make an app and put it on the app store and distribute it and make it for some people, it can just be, like, an off one, right? Like, me. So for me, I have an app that teaches me daily, like, philosophy concepts. And what was today's concept? Epistemic bracketing. Epistemic bracketing? Well, yeah, bracketing. Like, more of the, I guess, hustle, right? Okay, very good. So, whatever. Be present. I'm gonna read up on that. It calls it a way to say just be present. Oh, all right. Anyways. You people nailed it tonight, am I right? You guys are so present right now. Yes. Give yourselves a round of applause. Very good. I interrupted you. Okay, so you built it and it's making an app that lets you explore your philosophical interests. Yeah, and I think, so, this is kind of the simple Well, we love Alex. You know, the weightlifting app thing does sound legitimately cool to me because I, you know, I know that when I'm in the gym, something goes wrong, you know, like I'm supposed to be using this machine, but it's broken. You know, like what's a good alternative? But something that was able to answer me that and also, like, you know, draw on all my previous workouts, you know, I can imagine, like, that getting added to another workout app. But like, if I know I want it right now, like, I don't have to wait, right? And it's become trivial to just kind of snap my fingers and have it. Like, that is like fun. There's something in it where, like, whenever I hit a PR, like, my personal... And that's happening on the regular, by the way. Yeah. I mean, when you set a PR, it's not very hard. But anyways, it's very nice because the agent immediately, the app just talks to me and says, like, oh my God, like, it's a smart app. It's like, oh, like, that's incredible. Congrats on that. That's like, super great progress. You went from, whatever, 10 pounds to 12 and a half on your bench press in the last year. That's what's in the 1%. Look, in these times we live in, we have to celebrate the small wins. You know what I mean? But it's kind of cool when the app, like, knows what's happening, like at all apps. And so, to me, like, the really important thing is that today we have AI that looks separately in the chatbot interface, and then we have apps on our phones. And everyone's debating, okay, MCPs or APIs, or how they're going to communicate. But really, we should not have that distinction. Like, if you think about it, every agent or agent skill should just be an app because a normal, regular person will never understand what an agent skill is, and no one's going to go and read markdown files on GitHub. Instead of that, they can totally understand, oh, it's just an app that, like, goes and looks at your inbox and wherever there is a new email, looks where there's an important one, and just sends you, like, a quick summary if it is. Super easy to understand. Hard to understand if I tell you it's my email agent that triages through your inbox, here's the markdown file, go figure it out. So I think that that kind of vision of ours combines just the regular software, which can be just a simple coin flip app or whatever, a tic-tac-toe app, but still with an agent where you can change everything in it, all the way to this more complicated, like, email, wherever, app. So I want to know what you think about, like, in a world where everyone can make their own software, what does it do to the value of software that other people are selling, like a SaaS company, for example? Well, I think the reason why, again, like, I'm very bad with, like, with the more actual business things. I don't think you're bad at business. I think we have some pretty good evidence you're good at it. Hopefully. We'll see. That's what makes me think, but thank you. But I think that, so, the biggest problem with pipe coding, and no one's gonna use other people's apps if those indie developers own the backend and the data. Because there's just no way, even for consumers, let alone for businesses. Like, if I'm, you know, if I build an AI therapy app and I'm like, Casey, use my fantastic AI therapy app that I built on Replicator. Here you go. And you're like, okay, well, Eugenia can read my all my logs. And even if she, like, it's not maybe, maybe I just, I'm a smart developer, right? I don't know. Maybe I just see them. Like, I don't. And you're never gonna trust me. Or maybe Eugenia forgets to maintain it, so then all my therapy sessions go away. Or she's bad with security, so then all of that is exposed to everyone. And I don't even need to be a bad actor to. So the only way for people to share their personal software is to build it on one platform, like Wabi, where all the backend stays in one place. The platform is responsible for security, the social graph, like privacy, backend maintaining the apps will never go away. And so I think for B2B, it's kind of the same premise. Like, I don't believe that people are gonna start using some, like, indie, nor do I believe that people will build all their apps themselves. Got it. So what's, like, what is the sweet spot? Like, if you project, you know, a couple years into the future, what is the mix on my phone of apps that other people make and apps that I made bespoke for myself? I think only the only big, big apps that will stay are going to be the ones that either have the network effects, like the big social networks, of course, or that have a different, like, basically an offline business behind them, like Instacart, you're not using it for the obviously software or Uber. And those, of course, stay. But everything that's just software, I think will, and especially all the subscription, the long tail of the app store, I think that is going away. Really? Because it's just no need for any of it. Like, it's already barely working. All of the, if you really think about subscription apps, there's only Duolingo that actually ended up, like, going public. If you take out dating apps and, you know, social networks, just the pure, pure software, nothing on the app store. And again, games, there's nothing on the app store that really became a huge, huge business. Interesting. Well, so, I mean, like, would you be willing to, like, name a name of, like, a business, or like, or maybe like a category, like, that you just think, like, is, like, actually in a lot of trouble here? I think subscription apps with low retention, that's just like. Like, what kind of subscription? No one's really. Okay, well, fitness apps. Okay. Like, I don't know, calorie trackers, sport apps, meditation apps, like all this stuff that's just purely software, that's just try to sell you something to, I don't know, like pretty much every app from like that lifestyle and health and fitness categories. Just that they're mostly not necessarily, well, they're not providing a lot of value. If they have low retention, that means they're just selling stuff during onboarding, like, and that's the name of the game for most of these apps, like whatever. And then people just leave and never come back. So instead of that, I think people will want tools that are a lot more agenta smarter and tailored to them, and they can use them with their friends immediately. I have always thought that the meditation apps are in trouble because they're just the people that make them are so calm that they just don't see the threat coming. You know what I mean? There's a certain level of paranoia that you need to survive. And if you've been meditating, you lose it. So maybe think about that before the next time you meditate. What is called spiritual bypass? Well, partly people that spiritual bypass overshoot. This is one of the philosophy concepts of the day, I can tell. All right, so let's know what is it. Okay, so basically, that a lot of people that are, whatever, one of the guys, the guys that coined it, the therapist, John Welwood, said that he spent a lot of time in the ashrams or whatever, and said, the angriest people I've met, I've met at those temples that were there trying to become monks, or, like, living, because a lot of people go to spirituality because they're escaping. Not because they already went past all of their, like, went through all of their traumas and ended up there, but because they just run away from it. And now they're very paranoid and angry and meditating at the same time, so. That makes sense. I guess when I've had, like, a problem in my life, monastery has not been my first thought for where to go, you know. It takes a while to warm up to that. So I used Wabi today. I made a podcast question evaluator. And you may remember if I can take you back to moments ago when I, when I asked Eugenia the question about, you know, whether there was a connection between the two companies that you have founded. I asked it to evaluate that question and it gave it an 82 out of 100. So is that just a bug? Potentially. Wait for a public launch. Okay. I actually, I will give you a bit of product feedback because I think it, you know, it speaks to maybe a challenge that you have. It's gentle feedback. But basically, you know, I could, you know, the app looked very beautiful and it, you know, sort of said, okay, like, tell me, like, who is your guest? And I could type in a little bit about that, you know, submit your question. The keyboard was sort of floating over the like UI element to submit the question. Yeah. And so I, you know, I said, Hey, you know, the element is over it. So, okay, I'm gonna fix that. And then it didn't really fix it. I was sort of able to, you know, like, turn my phone around a few times and was eventually able to hit the button. So, you know to like design everything, like mock everything first and then go develop it and then test it and iterate. Of course, everything should just be built at the same time. So I think we need like slightly different people. I do think that right now if you're building a startup specifically, an applicationally a startup like ours, like you should, you probably mean like 10 to 15 people, but absolutely insane people. And I think you need to, in order to attract them, because so many big companies are trying to get them, you need to also change what you're offering. You're not going to attract them with like 0.1% of equity and whatever the starting salary is. So we're trying to do it differently where we decided we're going to try to, like, I'm a big soccer fan. So like, I'm like, okay, well, let's try to do like a soccer team where there are players on the pitch and there's the back office. So what are the most important roles for us? Let's make them players on the pitch. They're the team. So let's give them the same. Let's give them a lot more ownership that usually like employee number 15 would get. Like all 10 to 15 will get very meaningful sizable like equity grants. They will, it's a pretty relatively flat hierarchy, but they need to be absolute like superstars. And because you're able to give a little more and you're able to pay them a little more, you're able to give them the fame and the ownership in a way that not a lot of other startups can, you kind of create this incredible team on the pitch that everyone who is just doing one thing or like, I don't know, maybe we need, you know, someone to deal with accounting or I don't know, legal or whatever, or motion design, just one thing that does not really, we hire them as contractors, even if they're full time, but they're not part of the team on the pitch and we treat them slightly differently. But they know, and we also want the top people there, but that's like part of the agreement. You guys are coming in to fill in the role. You're not the, but the, the founding team is the founding team. And then, then we would treat like people that put on a jersey with the name of the company. We want them to be active on socials. We want them to be, to put their name out there and we don't want to have to Cook lunch. Cook lunch. Yeah. Yeah. Everything. But that becomes like, that allows to hire like really top tier people, not as a first employee, not as a second employee, but even as employee number 15. And I don't think you need more than 10 to 15 for a, for like to build a billion dollar company. So in this moment is possible to build a billion dollar company and you don't need more than 10 or 15 people. Well, yeah, some people are trying to do it with one, but I think you still need like, I'm trying to do it with one. I guess we got there somewhat. Yeah. So, you know, I imagine some folks sitting here are thinking like, you know, Eugenia, like I, I would love for someone like you to consider me insane and a superstar and worthy of putting on the jersey. But what does that mean in practice? Like, has the skill set changed? Like, like what do people actually need to be able to do in a world where maybe there's only ever 10 or like 15 seats, you know, in, in this company? It really depends because we, we have, you know, designers like, or I guess product people, generalists, engineers. When it comes to engineers, like really just like really either incredible generalists or just super good at what, at that particular, like for example, Swift is very hard to find a great, a fantastic Swift, like iOS engineer. And we went, I think through 120 people recruiting one for like our second iOS person. So there is just really incredible writing code because there the question is always, well, will GPT-6 replace the person I'm hiring? And if the question, if the answer is maybe yes, then it's an extremely expensive hire for us. It's better maybe not to do that because we'll spend more time coaching, rewriting, whatever that, that is. I think for everyone on the team, it's agency for designers and for us, it's very important agency, product intuition, design taste, like whatever your specialization, but across all, everyone, I think that agency and product intuition and not being an asshole, I think are very, three very important qualities. These huge skills. I think I like understand what you mean by every one of those, but agency, just like agency could mean a lot of things. What is, what does a high agency person look like at your company? Like what are they doing? Well, Alex is one. I think we can all agree. Alex is extremely high agency. That pokeballs are good agency. But frankly, the one other thing is like, I really, I really think management or any management in that stage is like almost counterproductive. So people need to figure out what they do. We also build a lot of agents internally that are actually managing our company. But yeah, people that can decide what needs to be done, go and get it done and push in production. That is what, what's needed. Ideally you need a few generalists that can do it all the way. You need some specialized people to be very good at backend, very good at front end, like, you know, figuring out polishing the stuff that, that you're building. But really you just need people to know, okay, well, this needs to be done quickly, agree on it and just go get shit done. And that's like a, like if I'm not seeing constant commit, commit, go ahead. Like if someone needs to go somewhere and agree and then mock something up and then develop, like you already lost because the stuff moves, everything's moving so quickly. There's no time for that. Got it. So it sounds like, you know, being able to like initiate a project and like get it done and like, if you didn't need much help along the way, like this is a core skill that you are hiring for. Yeah. Get shit done. Yeah. Also known as agency. Right. So, you know, there's this set of apps that, you know, tools like yours may someday soon disrupt, uh, meditation apps, for example. Um, but it strikes me that there may be some risk here of the AI labs themselves, like somehow maybe without even trying disrupting something like lobby, right? Like as their capabilities grow, that could become the place where people go in and type in, Hey, make me this sort of app. And, you know, it seems like with the coding tools that are available, some folks are already doing that. So how do you think about like your relationship with the, the big model developers? I think the main idea that we have is like, can we build an interface that's Google first, app first and not chat first. And the, the idea is like, yeah, we do have app creation as part of the product, but uh, we're more YouTube than uh final cut pro. Is that right? Yeah. Yeah. Yeah. So yeah, you can create stuff, but really it's more about discovering apps with lobby. We don't think, we think 90% will not build anything first. They will most likely use other people's apps, maybe then tweak some of those, then maybe eventually start building something. And of course the platform where you use apps. So I come back to lobby every day, not to build apps to use my apps. Uh, sometimes I'll build something new or find something new, but it's not really about, about that. So it's really about using apps, using apps with friends, which is very different from what the models of uh model providers are building. That makes sense. Um, let me bring back uh uh a few uh sort of replica flavored questions because one thing that I was thinking about as I was getting ready to talk to you tonight is while replica has never had like an enterprise, uh, you know, component to it, it strikes me that at some point maybe an agent turns into a companion or like maybe the lines between those two things start to blur, right? Where it's like, you know, today you may have your Clodswarm, you know, uh doing a bunch of tasks for you, but like over time it becomes something that you're talking to about what's going on in the office and about, you know, the gossip on Slack and that sort of thing. So curious if you think that we're heading, you know, into a world where people do have sort of work relationships with uh AI bots. Work relationships with AI bots. I love that. I think Wabi probably has a better chance at like becoming an enterprise thing because just, it's so, we already have tons of apps that we've built as team Wabi that we use on Wabi to manage Wabi. And ideally we want to move on from Slack to full communications on Wabi, like by at least like, you know, end of June, July. July this summer, we promise ourselves we're gonna move on from Slack and try to use Wabi as our main uh messenger. So I think that's easier there. With replica it's different because it's actually our whole premise is like we want to be the, the, you're gonna have an app and AI for work and AI for life pretty much. We want to be that one for life. I don't know. That's a hard question. I don't know. I mean, like, does it seem like good or useful to have an AI companion at work or like, can you imagine a world where that would be useful? I think Actually, you know, my fiancé helped with this, you know. But like, when Opus 4.5 comes out and all of a sudden Clawed Code, which used to be like kind of hard and tricky to set up on a computer, all of a sudden is pretty easy. And also, you really can't just type what you want into a box and that turns into an app. That is a moment where like some light bulbs start to go off. So in a way, like, I'm sympathetic and I think, yes, it is challenging to, like, stay at the absolute bleeding edge of what these things can do. But I think that there's also a version of it that is just, like, sort of like being 80% at the frontier, and like, that probably will be good enough for a lot of things. So in conclusion, subscribe to Platformer, because at platformer.news, we will always tell you what you need to know to stay at that bleeding edge. Thank you, AJ. Yes, and maybe this will be the last one. Yeah. Thanks. I’m super on board with the idea of democratization of the ability to make software, and I think it's going to allow a lot of software that just didn't make sense to build before to be built. I'm curious if you have any thoughts on whether craft or, like, more intentional or thoughtful design is gonna offer any sort of barrier to this onslaught of folks building their own apps. I say this as someone who's a big fan of a lot of like really small indie developers who make really nice software that I'm not sure you would get to just describing it to an LLM. Completely. You know, I mean, I came up in San Francisco in the 2010s where like, you know, design was a priesthood that people went into, you know. And I mean, like, people would celebrate, you know, like, you know, like Lauren Brichter who like invented Tweety and like all of the different, like, you know, UI, like pull to refresh, all of these, you know, like novel like UI elements that came out at that time was really celebrated. And now it's just like, you know, who cares what it looks like, get it out the door. But, you know, as I was sort of sharing with Eugenia, like, Wobby's like a very cool app. And it's also true that the first app I made, I could not hit the one UI element that I needed to get to. So I think the question is just like, how long does that state of affairs persist, right? Is it really the case that we're like a point upgrade away from there being, like, a revolution in design and all of a sudden it can look as pretty as you want? Or do some of these things just like turn out to be a little bit trickier to unpack? You know? But I, you know, I don't know, like setting aside like the business implications of it, like, it's fun to make stuff and it's like fun to make beautiful stuff. And like I, you know, a random story, I've started to read comic books for the first time in 30 years, and I swear you can tell the anxiety of AI's slop is leading them to draw weirder, cooler stuff. And so it's actually been fun for me to just go see Batman drawn in a way that looks completely insane to me. So I think that there, there's like something there. There's like the, like, you know, little green shoots of some kind of new weird aesthetic movement that might be fun. Awesome. Thank you. Awesome. All right. Thanks again. Let's have some drinks and snacks. All right.