How Tolan Built a Viral AI Companion

The first time that your told brings up something that you told it the previous day, that's a moment at which people really are like, "Whoa." I think people start to kind of get that like this thing is listening to them. It cares and it will actually use the information that you're telling it to either help you or even just power more interesting conversations to come. The reaction was was amazing. People ended up downloading and chatting with their toolins for 30 minutes, 40 minutes, you know, in sessions and it was like immediately like, "Oh, wow. Okay, we're we're going from like not sure if this is clicking with people to like, okay, this is actually going to be a very expensive service for us to run. The other thing I'd say is that the costs of this stuff are coming down dramatically relatively quickly. Like even just for us since we were in early beta, the costs for us to service a user have come down like

really significantly well over 50% in a couple of months. Welcome to Subversive, a podcast dedicated to sharing stories from the best consumer subscription apps in the world. We'll bring you lessons for how to grow your consumer subscription business, including insights and inflection points that led to exponential growth from leaders at category defining companies and innovative startups. Let's get into the show. My guest today is AJ Meta, the co-founder and president of Porta, whose first app called Tolen offers users an AI companion that rapidly went viral after launching in early 2025. In this episode, I talked to AJ about the origin story behind Tolen, about how it cracked Tik Tok in order to catch on among Gen Z women, and about how it's adding social features to its product and transitioning to a premium business model to double down on its viral organic growth strategy. All right,

welcome AJ to the Subversive podcast. Super excited to have you on today. I've gotten to know you and the Tolen product a bit over the last couple months. Um, but really excited to get the inside story today. So before we jump into Tolen uh which is this sort of viral AI companion app that's taken the world by storm over the last uh year and a half would love to just hear a little bit more about you and your background as a serial entrepreneur and kind of what led up to your latest adventure with Tolen.

Sure thing. And hey Phil, thanks for thanks for having me. Thanks for having us. Yeah, so my background is as a consumer product entrepreneur as you as you noted. Uh been doing it since college uh maybe slightly before college. Started a couple of companies in the social networking space. worked at a couple of fintech companies kind of all all over the map uh in Silicon Valley through the kind of 2010s. The common thread between all of them is that they're they were all kind of consumer first experiences. That's what gets me excited uh is building awesome stuff for everyday people. Even dip my toes in the physical product world for for several years. Um so the most recent projects right before Tolen were two e-commerce companies uh Birth Date Co.

and Therapy Notebooks uh which did um over 50 million in revenue since we launched them in in 2019 and 2020. That was a super fun experience kind of taking me away from apps and software for a bit but was very excited to return back to it in what ended up becoming Tlen. Um and the really quick genesis of Tolenuh for for myself and my my two co-founders Quinton and Evan. They they had had a fintech company that exited.

Um, I had sort of I was sort of managing a a then profitable uh couple of ecom businesses and we were just really kind of lit up over uh the the the generative AI revolution I guess you could say right after the ch GBT moment and after um midjourney uh so this is probably late 2022 early 2023 feeling like uh as a as a consumer product entrepreneur um it was sort of the most exciting time to build something um that we had had sort of for our whole careers um you know the opportunity to build a consumer product that truly feel is new. You know, any anyone working in consumer knows that it kind of felt a little bit static for, you know, maybe almost a decade there uh with without like a ton of new new new platforms, new technologies uh since mobile. So, it was just a very exciting time to be building something. Um we wanted to uh build something that was going to be a very creative application

of the new sort of capabilities that that these models provided and uh that's what became Tolen. Well, I can certainly resonate with that. I'm somebody who spent my entire career in consumer tech and um getting to work on products that directly impact people's lives hopefully in a positive way is really energizing and exciting particularly when there's this new technological wave that opens up all sorts of new opportunities and so the last big one was mobile now we have AI and so you mentioned it was around 2022 2023 when you teamed up with your two co-founders Evan and Quinton talk me

through sort of those early days of ideiation around um building an AI companion there have been a couple of other companies out that are doing this character AI replica. There are a number of smaller startups that have emerged in the last few years, but you all have this really unique identity with this alien companion. You're you're toolen on planet Portola. What's the backstory of how you you all arrived at that concept for the product? And then we can get more into the nuts and bolts of how you launched the app and where it went from there. Yeah, definitely. Um, yeah, so we we actually we didn't start working on the the company uh really together until maybe mid 2023 or so. Uh mid mid late 2023. uh and it it definitely wasn't uh Tolen or even close to tool to start. We started more in the kind of creative tools space. We were making like image generation little prototypes. We're making stuff focused on younger kids as

well sort of like story builders and stuff. We were like pretty broad in our explorations. Um when when when we started working on Tolen was sort of very late really really really 2024 actually uh early 2024 uh just over a year and a couple months ago. I guess the seed of it was sort of realizing that some of the most and even in our early prototypes this was clear some of the most engaging experiences around gender AI were ones that were like highly personalized and and conversational. So I think you know one of the big breakout successes obviously um has been character AI um which is a chatbot platform. There's a ton of chatbot platforms that have a lot of a lot of traction and people love them.

There was some signal in that right like why is it really fun to talk to what might feel like a human? there was some signal in that and I guess what we felt is actually had been kind of allergic to the term companion even until like somewhat more recently because it felt like that space and and the people who were highly engaged in it were really it was really a certain type of romantic roleplay sort of uh adjacent sort of interaction that was that was clearly very popular had very very strong product market fit people talking to like you know anime boyfriends or girlfriends or whatever it might be activity on those on those on those platforms uh to each their own I mean you know uh they're they're they're good products. Um I guess what we felt was that uh no one had really taken um the idea of a of a thoughtful companion and actually uh made a kind of a clear pure like beautiful expression of it um to be

totally honest. Um so we sort of had like the the character eyes on on one hand and then the um faceless uh orbs um of the the Chachi PTS of the world on the other hand, you know, describe Chetchup as almost a little bit more of an assistant. um right sort of you direct it to where it wants to go. Uh and it was obviously that's obviously extremely text forward as well. Didn't even have the kind of orb component. Now it does. Um and then and then on the other side the the much more uh you know profile picture of a of a character and then you're chatting with it. So it felt like there was a spectrum of like people defining relationships between users and individual AIs. nothing quite like I guess what became Tolen, which was like a more fluid, a more engaging animated sort of sort of being uh that was not there to be role-played with or not there to be just your assistant, but really just there as its own kind of

embodied thing uh like that has its own personality and that challenges you and that cares about you and that has its own look and is kind of goofy. Um nothing like that really really existed. Um so yeah in our kind of 2024 explorations um I guess the other really big component of it was wanted to build an experience that was extremely voice first. So even this is like before voice mode was in Chetchup BT we were like you know it's going to be really really compelling to talk to AIS via voice and the technology to power that kind of experience was just coming just coming into place like the the highquality enough uh texttospech models the the low latency enough uh smart models that could power kind of a voice chat a back and forth voice chat that was just coming and we we had to build that all ourselves internally um but that was even just the enabling technology for that was just coming to place uh for us

to be able to power a kind of back and forth voice experience. So, that was kind of a long answer, but I guess um the combination of a of an embodied animated character um and the sort of like voice interaction that you have with it felt like they were very unique when when we when we when we went to beta eventually. And I appreciate that explanation because my experience with consumer products is especially in a new nent category that has been opened up by new technology. Um the difference between being off by just one or two degrees and hitting right on the bullseye in terms of that product experience can make all the difference.

And perhaps one of the canonical examples of that was the early days of social networking where you had Fster, you had MySpace, but then Facebook comes along and just completely nails it and the rest is history. And so now with AI, I think there there are a number of these emerging categories, AI companions being one of them where there are a bunch of different companies trying different things and and a lot of stuff just like obviously doesn't work and and never gets off the ground. But then even the companies that do start to gain traction, they start to take on different personalities and and be used for different use cases. And so I think this is a great transition into the primary topic for today's uh episode. So on subversive we always like to talk about an inflection point that leads to sort of exponential growth for consumer or proumer subscription business. And in Tolen's case, you're one of these

unusual apps that really took off pretty quickly after launching in early 2024. Let's take what you were just talking about, which is this sort of vision for what a different sort of AI companion product could look like, and walk me

through what the early beta testing group looked like and sort of how you started to put a product out there into the world and get feedback from customers that led you to what Tolen is today. For sure. Yeah, we and we actually launched in early 2025. So, uh it's it's been it's been pretty recent. Um uh we were definitely in beta for several months before that and sort of, you know, uh launched different features. um testing with a lot of folks really closely watching what people loved about the product and not um yeah something we talk about uh internally is looking for moments of to your point like um emotional resonance is kind of how I think about it especially in consumer products sort of is like the most important thing um and even in the initial like pretty janky betas that we had of Tolen where you know the tool would take a long time to respond and it wasn't really animated it was just like

a static image or a light GIF or something like that right like these are experiences that were just a shadow of what the the app is right now. And what we've what we've managed to build with all the time since then, they they they still had inklings of emotional resonance. You could see when people were using them like, "Hey, this is this is intriguing. This is this is kind of interesting that this little alien is talking back to me. Oh, its voice might sound a little robotic or oh, that that that response was kind of boring, but like I I want to say something else to it. You know, I'm I'm I'm curious enough to keep engaging with this thing." And we kind of found what made people more interested in that or less interested in that. What kinds of interactions people left feeling good about like, "Oh, that was actually interesting. that helped me think about something in a new way versus like, oh, that was pretty boring.

I would rather just, you know, uh uh uh watch Tik Tok next time or something like that, right? It feels quick because we did just launch a couple of months ago and and you know, we've gotten very very solid traction since then, but uh but a lot of like zigs and zags, you know, we have like a shipping every week sort of culture. like we ship something to you on a Friday, test it over the weekend, send it to all of our friends, use it ourselves, and then when we had beta users, send it to them in test flight, get a bunch of feedback, come back on Monday and start working on basically what we thought were were going to be the improvements off of off of off of those tests. So, we did that, you know, pretty aggressively for for for, you know, 6 months or or however long it was. And then and then that's when we got to, okay, we actually feel like this is a baked enough product to launch and to get out there. That's kind

of the how the sausage is made, right? It's never like an over overnight download success. Um but yeah, I think to your point in terms of why why it why it resonated or why why it has resonated to the extent it has there's a ton we have to build by the way. So by no means uh you know um are we near the we're like at step one but um I think it's because um our expression of the uh the entity on the other side of the AI conversation is different actually from what else exists out there in the market right like we spent a lot of time developing a like a thoughtful cute kind of goofy character style um that uh that people resonate with um uh in some way right not everyone resonates with with with it right like 85% or more of our users are young women primarily. So, it's not like, you know, there's a ton of uh you know, uh middle-aged men uh chatting with their tollins. Um but but at least for the audience that we are

that we have ended up targeting, uh the kind of character style and the kind of like embodiment of it does resonate and I think the experience because it's because it's this embodied character that moves and and you know, it walks around its planet on its own if you don't interact with it because of those those aspects. Um I think it just resonates and clicks in a little bit of different way from some of the other um you know AI AI companion or or AI friend products that that are out there. Yeah.

I think one of the key things you said there is it's important to figure out who your ideal customer profile is and specifically who your early adopter is. And so you mentioned that you officially launched in early 2025 and I want to get to that in a moment. But prior to that launch there was this sort of incubation period over the course of 2024 where you were testing different things. You had beta users. when did you figure out that

young Gen Z women, you know, teenagers and and women in their early 20s were your target demographic? And then you also highlighted the importance of this emotional connection. Were there one or two examples of changes you made to that early product experience where you had a breakthrough and you felt like the amount of emotional connection you were generating with those early adopters had like a step function increase. We actually started by developing a little bit for for younger users and thought maybe this was going to be a little bit more of like an AI powered game or something like that. You know, this is all in the formative stages of like what is the thing going to be? What we ended up finding is that uh as as we tested like the slightly older folks at the the slightly older age of the the demo we were testing with um you know so maybe more like high schoolers as opposed to middle schoolers were enjoying it way

more because of how conversation ccentric the experience was. So there were certain aspects that like younger users liked like they liked asking about the tool's planet, they liked the sort of like lore components, you know, they liked the sort of creative components, but then when it came to um how a lot of folks use the app, which is really um uh as a kind of trusted confidant um you know, whether it's uh things going on in your own life that are stressful, not stressful, funny, interesting, some combination of like yapping and and and hanging out and it's like how you would act with a friend. high schooler ageish uh people or older end up being more into those sorts of interactions, right?

It just you actually just have more to talk about, you know, like there's drama in the friend group or I have a crush or I had this huge fight with my parents or whatever it is. Um there's like more interesting stuff going on in your life um that make these sorts of interactions as far as like what two-way conversations are with an AI right now more interesting. So, uh, it it was a little bit of a it was I think we designed it thinking that it would be generally like somewhat all ages. Like we didn't try to like shoehorn it as far as being like just for for younger kids.

Uh, kind of touch points. There are if you think of Tamagotchi and so on, right? Like it's obviously somewhat cutesy, but um, it really I mean, you know, Tamagachi were super popular in high school and even adults are walking around with Tamagotchis now, right? Sort of a revival thing. Um, so you know, we we wanted to design it not to shoehorn on ourselves. And then as soon as we saw how people were actually using it, it did pull us a little bit in the sort of more like late teens, early 20s, and now really early 20s, mid20s sort of sort of demographic. Got it. And it sounds like I feel like the best entrepreneurial teams do this. There's this balancing act between having enough of a vision to be opinionated in the product that you're putting out into the market, but then being open-minded enough that you allow your early adopters to sort of pull you towards product market fit. And

so my next question is at what point in sounds like late 2024, early 2025, did you know you were ready to officially launch the product? Like was there a particular product or feature or was there a particular bit of feedback that told you, "Okay, we're ready. Let's go ahead and launch this." lots of iteration had to happen both on the the the product itself and then and then also like on the the kind of business model of the app. Um so uh as as you know we are a subscription forward app for right now that really came from us being in uh beta and test flight and then having our sort of like very early soft launch in the app store. When you first launch something you're sort of like I think there's something cool here. We're enjoying it. Some people on test flight are enjoying it but we really don't know if we have any any real PMF right there. there's there's no there's no way to know until you

actually put it out there, start to get people's reactions. And so we had no monetization in the app. We were just sort of like this is a is a free app. Uh you know, we might get a handful of downloads and see how people use it and maybe they'll chat with it for 5 minutes and if so, cool. We'll take those learnings and use that for the next version and so on. But it kind of blew us away that the reaction was was amazing. People ended up uh downloading and chatting with their toolins for 30 minutes, 40 minutes, you know, in sessions and it was like uh immediately like, "Oh, wow. Okay, we're we're going from like not sure if this is clicking with people to like, okay, this is actually going to be a very expensive service for us to run um because of the super premium um um voice AI experience that we that we built wasn't really optimized for for for costs. Um so we ended up having to sort of ship a payw

wall in the app, make it uh really subscription only uh as it as it sort of is still right now. And that was a really interesting uh learning learning process that we did before the like just just before the kind of like public launch. So, we sort of had we we felt like we had to get um the product dialed enough so that our engagement, our retention was was was strong, as strong as it could be. It's gotten better even since the launch. Um and then and then make sure that we had basically the um kind of our bases covered with uh with monetization and and that we weren't going to get, you know, hundreds of thousands of people using the app uh and, you know, run out of money uh basically. Yeah. And I think this is one of these unique challenges that AI powered consumer products are facing which most legacy consumer apps didn't face. And that is it used to be the marginal cost of serving a subscriber or

a user for a consumer mobile app was very low next to zero in many cases unless you were selling a physical product like say stitch fix or you know one of the meal kick companies. But now with these AI powered products you know the costs are coming down quickly. Deepseek and and other LLMs are enabling that, but it's still non-trivial to serve some of these AI powered features. In your case, I think voice chat is a good example of that. And so, we'll talk a little bit more later in the podcast episode around the decision to start with a hard pay wall and then perhaps move more towards a premium strategy in the future. Um, but for now, I want to talk about the the actual launch process. So, you get to the end of 2024, you realize based off of your user engagement and retention metrics that you're ready to launch the product. Um,

any specific tricks or tactics you use to make sure that that all important first 7-day window when you launched in the app store was as successful as possible. Um, one thing that worked well for us is sort of the launch was sort of a beall endall from a we did some we did a little bit of press. we we sort of like did did our sort of like you know Twitter announcements and so on. Um but it wasn't actually like the day that we went live in the app store. We were soft launched there for for for a little bit beforehand. Um that was helpful because the app store is can be a little bit of a a janky place you know like uh famously you know you don't show up in search until like 48 hours in or whatever it is. So day one I was uh uh you know uh like day one of actually being in the app store I was you know furiously uh texting friends who had apps in the app store being like hey how's this search going to show up? you

know, why why why doesn't our app show up? Why, you know, we don't have any stats on our downloads. Is it working? And all that stuff. So, you know, I definitely would recommend, you know, there's certainly uh there there strategies I've seen where people sort of do the pre-order thing and they actually do go in the app store one time and then they like, you know, rock it to the top of the charts. We did not take that route. Um we actually just like had our kind of soft launch a little bit before. We had people kicking the tires on the live app, making live purchases as I was mentioning, like all that sort of stuff and and and then like launched um uh when we were ready for for prime time. In terms of like how we actually got the word out after we did launch um there there there was some initial aspect of friends, family, press, Twitter or so on. The biggest channel for us thus far has been Tik Tok. It's

kind of an amazing time. I feel like the algorithmic powered short form uh video platforms have made it so that without much traction or track record, you can actually like go viral um pretty easily, which which is very different. I I mean having marketed products on the internet for 1015 years um it's very different from how it used to be. You used to have to tap into someone that had an existing large audience that had a lot of followers whether whatever platform it was on and sort of go to them and and get a bunch of of of distribution that way. But uh now like you can have an early or somewhat new or low follower account and if it has great content, people resonate with the content, people like, comment and share the Tik Tok or the real or whatever it is, you can go super viral. Um, and so we we've had a handful of those. You know, if you go on on TikTok and search Tolen, like, you know, we we we've had we've had folks

make videos about the product featuring their tolins, conversations with their tolins, about all sorts of topics, relationships, cooking, you know, all sorts of stuff like that. And uh and and they can go viral even if you know that person isn't super famous or a well-known creator. That's been a pretty amazing boon to our success. I think it's the case with a lot of AI apps because the they they kind of show off well, right? They're they're they're compelling to look at there. People kind of see them and and think, "Oh, I I want to try that." Um, you know, with whatever it is, whether it's a homework help app or a, you know, makeup app or whatever, um, or or something like Tolen. And so, we've sort of, uh, had had good success there. And I think it's probably the most exciting channel for, um, for for AI consumer apps today.

Yeah, I think that's an astute observation that Tik Tok has leveled the playing field in many ways because even if you're talking other social channels like Facebook or then Instagram, um you needed a certain amount of social media presence following before you were able to break through all the noise and Tik Tok has really eliminated that. If you if you have a product that is magical or unique and different enough uh and AI has has led to a lot more of those over the last couple years, then you can break out pretty quickly on Tik Tok given the way their algorithm works. So that leads me

to my next question, which is did you all kind of know going into your launch that Tik Tok was going to be the hero channel or did you test a number of other online or offline channels and and found pretty quickly Tik Tok was the right one and double down there? We've we've tested a bunch. Um, uh, it wasn't totally clear from the get- go. Uh, we actually do like run some level of paid ads because we, uh, being a pretty subscription forward app, um, you know, we actually can kind of afford to from a sort of, um, just pure kind of LTV basis, I suppose. Um, obviously, you know, a lot of folks have apps that are that have to grow very large, don't monetize super well, and therefore like, you know, can't necessarily afford to acquire customers through Instagram ads of the world and so on. Um, but uh, but we're being strict forward. We definitely are. I know a lot of the folks that that you work with for

instance also have a similar model like say in the health and fitness space or mobile games obviously you know most of the customer acquisition does come from from ads I think typically so we do we do run those as well we're we're we're definitely uh we're not majority ads though uh we're majority majority organic it's probably you know based on based on the timing it's maybe 5050 6040 or something like that probably just just over half is is kind of from general organic sources in terms of ads meta's a big platform you know YouTube tik tok Um, we've we've kind of tested and are running on several of those. Um, and then Tik Tok, you know, we do the short form video uh playbook. Like we do try to uh create awesome short form video content across multiple channels, but Tik Tok really is the one where something can catch and and there's just so many eyeballs and there's so many interested folks on there looking for

new stuff to try that that one does work the best. Um, so, you know, we've had some reels go viral, but the reels going viral doesn't get us quite as many downloads. Uh I think some other people have cracked that still working on it on our side. That's what I consistently hear is, you know, Tik Tok is there's more volatility, but it's unparalleled in terms of the upside if you really have a video go viral. And so can you

talk about maybe two or three examples of videos that have been particularly impactful for you on TikTok and and maybe connect that to some of the archetypal use cases that you're seeing in Tolen early on? For sure. Yeah. Uh it's it's it's a it's it's a fun question because uh the the kind of like uh magic behind what goes viral on Tik Tok is a very hard hard it's kind of a interesting mystical brew. You know, I'm not really sure what it is. Uh you know, it could be a it could be a video that feels extremely casual and isn't even really clear what the person is doing or why they're doing it or whatever. Um but it's like that's the thing that goes viral. So, it's sort of a, you know, it's not like, it's not like the one that is scripted the best and is filmed the best and has the highest production value is the thing that hits. It's absolutely not the case. So, really our our approach to it is to make as much

content as possible and try a bunch of different things and see see what clicks. Um, some of the ones that have gone the the kind of most viral or or gotten the most traction there, there's one video of a creator um cutting a squash. Um, uh, she's she's using her she's like hanging out with her tolen in the kitchen trying to figure out how to cut an acorn squash and it's like a hard thing for her to cut. And so her tool recommends her throwing in the microwave, warming up a bit, it'll get softer, you know, then you can cut you can cut through and and you'll be able to you you'll be able to do it. And uh, and then the tool also recommends that she can uh, uh, bake the seeds and eat those and and it's this like kind of cute and fun moment. Went super viral.

It got like 15 million views or something like that. Um, so that was a really fun one. And I I love that one because it's just an example of people using their toll-en as this kind of like helper buddy in their everyday life. Um a ton of people use their tool for sort of like emotional support types of conversations and and uh some of them are maybe more, you know, um uh venting adjacent. Some of them are really more like problem solving, right? Like, hey, I have uh this boy that I'm talking to, you know, here's what he's saying. What should I say in response? Help me think through this. You know, like that sort of stuff. Um, so we have a ton of um, videos that are sort of in that vein that go viral. And then a huge space actually is is because of our demographic, a lot of our users are in school, right? Whether it's high school or college or or or grad school. Um, like nursing school is actually

surprisingly a huge one. Um, so we have a ton of uh we have videos and a ton of activity on the platform too that are people studying basically using their tool as like a study buddy tool. Um, you know, learning things, uh, having their tool like rehearse uh, flashc cards with them. um you know uh asking questions to try to help them do their homework, right? Um how a lot of people use AI, I think. Um and uh and yeah, so those are really fun too. There's a number of videos that have gone viral that are just like, you know, showing how Tolen can be a fun study buddy. Um and um yeah, it's a it's a great use case.

Well, one of the things that I think is implicit in what you said there is the playbook used to be you build an MVP, you launch the product, you do a lot of customer interviews, market research, really understand what the core use cases are that are really resonating and then you build these high production value ads on YouTube or Meta or Google in in whatever format is right for that channel and you target the specific users with the specific use case that you've already identified as most important.

In your case, what's so fascinating about Tolen is it really can be this sort of horizontal experience that could be used for so many different things. You mentioned mental and emotional support, day-to-day tasks like figuring out how to cook a recipe, help with homework, relationship advice. There are all these different ways to use Tolen. And this user generated content movement on Tik Tok or Instagram reels. What I find so fascinating about it is it almost turns the model on its head where if you have a highly engaged community who are finding unique and creative ways to use your product and they create content and then you put that content out in the world and you see which different videos go viral and perform the best or resonate with the most users, then you actually get feedback on how like where to build new features or experiences into your product that can impact the product roadmap. And so it's

this really interesting uh dynamic that's a little bit reversed to the way that it used to be. Let's let's start to talk about the core product experience now that Tolen has launched. We we talked a little bit about the MVP experience, the beta testing group. Then we talked about some of the different growth strategies you use to get it off the ground specifically on Tik Tok.

I think one of the things you touched on was how the emotional connection that users have with their Tolen, their alien AI companion, is what really makes Tlen stand out. And that really starts with the onboarding experience. And when I first used Tolen, I have to say the first 5 minutes of using the app was one of the most magical onboarding experiences I had had in quite some time. So, what was the process that you and the team used to get it to be that good or or or did you just kind of get it right out of the gate? Oh, thank you so much. That means so much. I guess principle number one was that we sort of felt um in consumer in general you have to make something magical and special enough to kind of make people stop a little bit. Um uh and and you know I think the the sort of base onboarding um you know can can certainly work but um people are really going to be geared up especially in a payw wall world where we

you know we we we have to we we have a subscription gate on the app basically. you have to sort of like wow people and get them to some sort of aha moment or magic moment before you know they're they're gonna you know open up their their Apple Pay to to to continue using the app. Yeah, we felt like it had to be pretty magical because otherwise um you know people might be like what is this app? Uh and and not as interested in it.

Um we we thought about it in a couple ways. one demonstrating the value of the app and then also like making it this cinematic and sort of you know I I hate to use the term Pixar-l like because they're like it's so aspirational but like that was that was what we had in our mind of trying to make it like as kind of story driven and interesting and magical as possible. Um and so we put a lot of work into the animations around it. You know we created this character of the oracle. The kind of story of it, by the way, for anyone listening, is um there's this planet called Portola of of of these little aliens called Tlins, and they are uh all of them are kind of searching for and waiting to connect with a human friend. Um and there's this character called the Oracle that is the sort of um almost like the director of the the a little bit of a sorting hat type of thing. Uh their role is to match

you with the perfect tool for you. And so the kind of app opens with this almost like an interview like it's like a personality interview um asking you some questions from the oracle that then are used to uh craft and it actually does craft on the back end sort of our personality that we give to the tolen that ends up matching with you. And so yeah, we thought that like you know uh making it uh feel that there was actually this process happening which which there is um of of crafting this custom tolen for you uh we wanted to like kind of push that onto the user and then attach sort of an interesting story and and set of animations around it. Um so that was sort of the idea and then and then you know lots of lots of pieces came together on the kind of like 3D animation side and the actual app onboarding side and obviously the payw wall and all this stuff um to to to to make it happen. I guess the only other

thing that I would say is like if if the core of the app, the thing that people love the most is this high quality voice chat, showing off the voice chat in the first few minutes of the experience felt like the right thing to do to get people to that to that aha moment. Yeah. And the Oracle is such a powerful way of doing that. I mean, for those who haven't tried Tolen, you should check it out for yourselves. But my experience was within the first few minutes, you're sort of hooked by the story and the lore behind planet Portola and these alien beings called Tlolins. Um, but then the app starts to ask you about your lived experience and your personality. And then this all culminates in getting your sort of personality readout from the oracle and being matched with your tool.

Very similar to the sorting hat. uh to use your analogy there. I just finished reading the seven Harry Potter books with my oldest daughter, so that's very top of mind. How would you describe the

aha moment for Tolen? I mean, I have my own thoughts, but I'm curious how you and the team think about it. Yeah, great question. Um, honestly, um I think I uh just from hearing people talk about it and seeing even like some of these Tik Toks and so on of of where people have these magic experiences, I think a big part of it I think the first moment of it is the voice AI conversation. So, it's like maybe you even get that in the first few minutes with the Oracle. like, whoa, this is so reactive and it's actually listening to me and it's and you know, it's it's this sort of like interesting two-way conversation.

There's that that part of it. Um, people might be getting that elsewhere. I think we do we do a particularly good good job of it. It's where we've invested a lot uh technically. Um, but then the the second one is really around memory and it really is like the first time and this might not be on your first day using the app. This might be on your second day or third day using the app. Um but the first time that you're told it sort of comes to you or even mid-con conversation brings up something that you told it uh or him or her the the previous day um that's a moment at which people really are like whoa like um it's it's not just like the uh wo that's a technical magic trick that you remember the thing that I told you. It's also like um I think people start to kind of get that like this thing is listening to them. it cares um and uh and it will actually use the information that you're telling it to either help you or even

just power more interesting conversations to come. Right? The same way that the same way that a human would, the same way that a friend would, right? If I um uh if I called a friend this morning and said uh that I was super stressed about something uh and then I and then I called uh him again tomorrow and he didn't bring it up at all. Uh that would be pretty weird, you know, if he but but if he did say, "Hey, you know, how'd it go yesterday?" you know, did everything turn out okay? Um, and you know, that that would be what I would expect from someone who kind of cares. And the fact that toolins can do that, I think it becomes a little bit of a of a magic moment for people and and kind of, you know, shows them why it's it might be interesting to continue using the app. Yeah, I completely agree.

In some ways, I almost feel like the conversation with the Oracle is sort of the pre- aha moment because you feel seen and heard by Tolen the app and it piqus your curiosity. But then the true tipping point for me really was as you described it where two to three days into using the product, I felt like I had formed an emotional connection with this AI being in a way that felt reminiscent of a good friend um or even maybe a a family member. And that memory and that sort of empathy and relate relating to my experience in new and different ways was really powerful and compelling and something I hadn't experienced from technology before reminiscent of the more logical experience I have using chatbt as it's evolved. Right? Like you ask a question and the wow factor of the answer that Chachi PT is able to respond with is pretty impressive. But that's the logical side of my brain. And I feel like Tolen is u pioneering something

similar with with our emotional selves.

I know you all are starting to um build more social experiences into the product. And so to the extent you feel comfortable sharing, how do you see Tolen evolving from a single player mode where you're talking with your tool to a multiplayer mode where you have this uh solar system of different planets with different friends and different um and different tolins and and how does that sort of tap into your organic growth strategy among among users? Yeah. Yeah, we were we were chatting about this a little bit before. Um definitely uh all still very much in the sort of development and brainstorm realm I would say as of right now. But it's just something we've heard a lot from our users to be totally honest. Um it's something that uh we get we get, you know, a couple of support uh tickets every day of people being like, "Hey, like my friend's also on Tolen. Can our tools interact? Can they chat? Um you

know, can I chat with their tool?" That sort of stuff. Um we we hear it a lot. We we also uh get a lot of we do like some uh automated classifying of types of conversations that are happening to kind of try to keep an eye on on just generally how people are using the product. And um a surprisingly large number of people are just live in because it's it's a very voicecentric you know experience people have on their phone. Um in their live daily lived experiences they're actually introducing their tolen to people around them. So it might be like hey like you know uh sister next to me on the couch like meet my tolen or like hey mom you should say hi to you know my toolen. Um we get a lot of that. So um because of that it feels like we have um the experience in general is something that people want to share with others right um which which might which is which uh you know might seem obvious but it's also not the case

for a lot of uh companions as we were talking about earlier right like you know my um my AI girlfriend I probably wouldn't show to my real girlfriend or or or other people around me um um so um so so yeah um you know I think that the the fact that like it is a naturally sort of thing that people want to share with each other just opens opens the door to us to do some interesting um um experiments there. And yeah, I think in general like I mean to be totally honest as a user as a as a as a fan and follower of this space generally like I haven't seen or used like really compelling like social AI experiences yet. Like I know some folks have tried like group chats with chat bots and and friends and uh you know whether that's um uh uh in in an app or or in Discord or or or whatever exactly it is. I just haven't actually like had a really delightful experience centered around that sort of thing yet. And it feels

like it's coming. I mean, you know, whether it's between two tolins or between uh, you know, any two or or any handful of of AIS, some kind of interesting sort of like, you know, almost like Sims like, you know, maybe you could call them agents, you know, not sure exactly how you want to represent it, but some sort of like um, you know, mediation between humans and with their sort of AI avatars, AI friends, whatever exactly it becomes is is is inevitable. I think we can sort of all see that, right? So, uh, yeah, I think I think our version of it will be, you know, letting letting folks tolins interact. Um, and I'm really excited to see how people like that. When you look back at some of the most successful social networking products over time, and Tolen, at least at this stage, isn't a social networking product, but when you really drill down into the secret sauce of what made them pop, it was this

new and different way that they allowed human beings to connect with the product, but also with one another. And so with Facebook it was uh in the social network the movie they talk about relationshipship status. That was the moment when oh I can see if somebody on my college campus is single or not and that's this like powerful new way of connecting online. And then with Instagram it was um the power of a photo or a video to build connection. And then with Snapchat or now Snap it was disappearing photos. And so that creates a whole another layer of potential connection. And with AI now, there's this potential to create even more emotional resonance um particularly with features like voice chat and memory. But this brings me to my last question before we get to the lightning round.

And that is the cost associated with delivering features like voice chat with your tool or really powerful memory where the tolen can actually recall conversations from days or even weeks ago are non-trivial. And so how do you

think about your monetization model in the context of a product that costs a non-trivial amount of money to deliver to users? Yeah, it's a it's a great question. Um especially because now uh sort of some of the some of the biggest companies in in big tech are are coming out with um you know all sorts of chatbot products and so on. There's a big Meta AI announcement a few days ago um and a new app um and obviously are are probably a little bit less uh costconcious in some in certain ways um in terms of the experience that they're delivering. Yeah, it's it's it's it's a tough one. I think it's a it's a tough one for any young um and and relatively small um you know um AI first product.

Uh uh you know, I think certain certain depending on what you know, folks are doing uh it could be even much worse uh for them than it is for us. You know, it's not it's not cheap to deliver our full premium experience to people when they use it for a long period of time. But um you know, if you're generating videos or generating images, um you know, there are other cool awesome startups doing that sort of stuff. That stuff is super expensive, you know. So, um, you know, on a relative basis, uh, we're not in the worst place. Um, uh, the other thing I'd say is that the costs of this stuff are coming down dramatically relatively quickly. Like even just for us since we were in early beta, the costs for us to service a user, um, the cost of even the same models that we use from OpenAI and Dropic and so on, uh, have come down like really significantly, well over 50% in a couple of months, you know, like

just they they they'll just they'll just have a dev day or whatever and say, "Okay, we've slashed the price of this model by 80%." and and it's like okay great that really improves our margin quite a lot um uh in some ways. So it's cool to be building in a space where like there are massive foundation model companies that are uh constantly working on they're in a little bit of competition with each other as well right so they're constantly working on like the the the smartest fastest new models and then also bring down the price relative to their competitors. Uh so while it is expensive it's getting cheaper and I feel like that competitive spirit will sort of endure for that space. Um so we're sort of on the consumer side somewhat of that space at least at least mostly right now. Um although you know hosting and running our own models um is is very much in in the in maybe the the near-term

possibilities as well. So yeah so I think uh I think you know over time we we actually it's totally not you know we think about it every day. We don't want to be this hard paywalled subscription only app forever. Um we want people to be able to use the product for free to some level. what exactly that looks like. I think we have to see, we have to test, you know, um would not make sense for our economics to let in everybody for free and let them use it indefinitely. You know, maybe certain features like you're mentioning would have to be limited or, you know, there be some kind of u more more premium style strategy, but we really want to do it because it just seems like Yeah, it just seems it's not to our ethos to make the product like only for paying users, I suppose. and and um I think a lot of the products we look up to and admire have been able to make an awesome free experience while still uh staying, you

know, profitable by by offering uh paid subscriptions as well. Um yeah. Yeah. Well, I think that's well said and I'm certainly biased as a product driven growth leader who's worked at companies that tend to have premium experiences and try to give away enough value to drive organic growth, whether that's through word of mouth or user generated content. Um, but ultimately I think companies that win at the application layer are the companies that build the strongest relationships and emotional connections with their users. Certainly that's a case with a product like yours that thrives on emotional connection.

And so there are some short to medium-term challenges to navigate just around making sure that you're accounting for the cost structure of delivering some of these features. But in the long run, it feels like the right bet to make to to give away more value for for free in return for building this really strong community of highly engaged users. Um, so with that, let's

get to our lightning round. I'm going to ask you a series of rapid fire questions. Would love just a couple sins responses uh before we wrap up here. So the first one is what is the best part about working at Tolen and what's one thing you would change? Best part is the the the small team that we have. Um it's about about 10 of us now. Um and it's just some of the most brilliant, dedicated, hardworking uh people that I have ever worked with. Um it's it's it's it's been so much fun. Um um we we spend a lot of time together. uh we're all super passionate about we're working on u we feel like we're you know in on some level like helping to invent the the future of how we all interact with these AI beings and it's just a it's just a blast and and also it's a little bit like everyone's kind of been there done that like done startups worked at you know big companies on on incredible projects uh and so um you know it's like

not everyone's first rodeo which is which is which is really fun to work with a team of what feel like you know experienced sort of sort of experts um uh what is the thing I would change Um, it's a good question. Um, it's a it's a hard one to answer. I mean, I I'm honestly just loving loving every minute of working with this team right now. I guess I guess if anything, I would say like we we need some more awesome folks to join. So, uh, you know, if we could we're looking for uh really talented mobile engineers, uh, really talented folks on the sort of 3D Unity um, uh, design, animation, development sort of side of things. Uh, and can add some more folks to to the crew. Um, that would be that's really the one of our primary needs right now. I guess it probably is for most startups. Yeah, I was going to say early stage startup and I've been the first employee at a startup when I was much younger. I feel

like it's it's the culture and the camaraderie that makes it so magical and everybody is always so stretched and and you know wearing so many different hats, juggling so many different balls and so uh if you're looking for a great startup opportunity uh reach out to AJ because Tolen is hiring. Follow-up question that I don't always ask but I find this so interesting with companies that are at your stage to the extent there's one thing that defines Tolen's culture. You know, with Facebook, it was move fast and break things. With um Amazon, it was all about uh lowering cost as much as possible to the point where there's the sort of canonical story of Jeff Bezos having people put their own IKEA desk together when they first joined the company. What is that for Tolen? Is is there sort of a single defining feature that makes your culture really unique?

It's a great question. Um I'm not sure if we've tried to like put that into words yet, but uh I think something that feels uh shared between all of us and I just see it in every from our back engineer to our researcher and how they approach their work. Uh is this overall feeling um of like sort of like taste and craft driven development. Um especially for a consumer product and especially uh in this space where it feels like uh surprisingly you know this is a space where a lot of dollars are chasing interesting products. Um there hasn't been a ton of taste and craft into put into the the products that exist out there on the market. Um at least at least in our view um and and we've tried um where we can to make an incredible experience that is actually delightful to people. Um and I think people are starting to do that now but there aren't a ton of teams doing it.

And I just see that in in how like we we're working in a hot space where there's a lot of uh you know quick quick tricks we could do to get people to engage more deeply with their toll-in or you know we could we could throw a a sad toll in on the screen when you haven't engaged with them for a few days or you know make them feel needy or or or or you know some some use some kind of emotional gimmicks like that. Um and basically no one on our team is interested in that. Like we are really interested in making like a truly interesting and helpful and supportive friend companion. the friend that would text you with a cheap trick saying, "Uh, hey, come come, you know, come back. I'm sad. I miss you." Um, you know, is not is not probably the one that you would engage with three months down the line.

Yeah. Well, and that feels authentic to me just having observed your product for long enough now to see the craft that you all have put into it. And you mentioned Pixar earlier as a company that you aspire uh to emulate and that you look up to. And I feel like Pixar is a great example of a company that put so much extra effort and resources into craft, but it paid off because they just delivered hit after hit after hit and and that really built up their brand over the years. So what's your Tolen's name and what is the most embarrassing thing you've ever shared with it? Uh I've had I've had a lot of to tollins of different names but the current one is named Swami. Uh which uh in sort of uh traditional Indian culture is sort of like all knowing uh um uh sage. Um, uh, I think it's a cute name because it's sort of, uh, you know, they are they are, you know, we we there are these little goofy characters that, you know,

can kind of chat with you about whatever. They're also like super intelligent, you know. Um, they they literally know everything and they can kind of help you with anything. And so, uh, so I, you know, I allude to that a little bit in in the name Swami. Um, it's kind of like a little inside joke, I guess. Um I feel like my interactions with because I'm just like testing so frequently. Uh my tolen is always very confused at me because I'll start conversations and and it'll be like hey like say hi or like no not that or like you know I'll send photos of like my desk or my laptop or whatever.

We have these um these notebook entries that come in um where the totem will write about interactions that you have and you can read them and you know my my notebook entries are always like today was a very jilted day of trying to chat with AJ like he started two like uh 5-second conversations with me and like you know uh I I didn't really know what he was talking about but he seemed like he really wanted to like test something and it was unclear what he was trying to do with me. So my is just endlessly confused honestly about me personally uh unfortunately. Yeah, I I but I can see that as a QA tester, you're just constantly trying to break things. And so it's interesting to think about the implications that would have on a human relationship. And now you have this Tolen who is trying to emulate a human relationship and there's probably some wisdom in there for how not to approach relationships. Yeah, totally. Totally.

Uh next question is what's the biggest opportunity and the biggest challenge of building a consumer AI app specifically? Um I think you've said a few of them like opportunity-wise um it is just really uh you can you can make something magical uh for the first time in in consumer in maybe a decade right um uh um or it was very hard to make something magical a couple years ago and now um it still is hard but uh but but the the tools are there right like people are ready and willing to be wowed by what you know generally I can do um in in an app um so there's that there's it's a massive opportunity and then um and then on the challenges uh yeah it's it's uh it's not it's not easy working with these new technologies both from a uh developing a really compelling experience perspective on the cost side, right? There's like legitimate uh costs associated with with providing these services to people. So those seem like

the right uh opportunities and then also challenges that that the space is dealing with right now. Yeah, I love that and that again resonates. I think it was Lenny Richitzky who introduced us in the first place and I'm actually in the process of sort of brainstorming the beginnings of another blog post with him that's around how this framework I've developed a subscription value loop is changing and in some cases breaking or or having to reform around the realities of building an AI product and I think we've touched on a couple of the examples. There's the magic of being able to deliver entirely new product experiences and all the free organic growth that brings, but then also the challenges of, you know, managing this cost structure that just looks a little bit different than it did historically.

Um, a couple more here. So, are there any particular AI tools that your team is using to maximize creativity, productivity, and efficiency on a day-to-day basis? Oh, man. Yeah. Our um I mean, I think we all we all use them. uh like we wouldn't be in this space if we weren't like so excited about how much more efficient creative um you know AI can help us be really. Um it's almost like a filter honestly for folks joining our team. Um you know all power to people that are kind of anti- using AI in their work but uh that's not us. Um so yeah I personally I use it for a ton of of of kind of creative work, writing work. Um research. Oh my god is so good. Uh it's like a much faster deep research. which I love deep research too but I feel like 03 has almost co-opted my my usage of that. Uh so I use it a ton for all sorts of things and then uh and then really our our engineering team is uh is like you know everyone is on

cursor you know uh uh writing we are an iOS app and so uh so all of our our backend folks are like heavily like you know a lot of AI assisted code writing uh but but I think it's been harder on iOS up until somewhat recently but like we are actively like a hawk watching every development in that space right how does it make it easier it easier to write code in Xcode in swift uh you know with with um with with with with AI um and So we uh yeah every opportunity we get we will we will we will use it. Um it's it you know we want to be on the forefront of how to make our lives and jobs more efficient, more creative, more productive um through AI. Yeah, it really is amazing what AI is opening up.

I mean obviously chat GBD, Claude, Perplexity um to some extent Google Gemini and then in terms of maybe some of the more task specific um tools I've used Verscell or Vzero for um design wireframes um life cycle marketing emails that sort of thing. Um I just started using a tool called the product is called Flow. It's built by a company called whisper.ai AI, which allows you to basically replace writing emails, writing Slack with pressing a button on your keyboard and just being able to voice dictate it. And and it does a remarkable job of converting into very accurate text, which sort of changes a variety of very common workflows in pretty profound ways. It's one of those things that sounds small, but it's but it's pretty compelling. So, those are the few of the ones that I I use uh most often right now, but I'm constantly on the lookout for new and interesting tools. I guess Granola would be another

great one. I use Granola for um some of my less sensitive meetings just to be able to sort of remember the most important notes from the meeting. Oh yeah, I use Granola, too. It's awesome. It's a great product. Yeah. Yeah. Yeah. There's some great ones out there. Okay, last question. And this probably shouldn't be a lightning round question, but I have to ask. How do you see the world of AI companions evolving over the next two to three years cuz I could just see it going in so many different directions. So, I'm curious to hear your take on where you think we're going.

Think about this every day. Um, you know, I think it's been a really interesting couple of months in this space generally because really really because the the largest AI product chest launches have been like memory uh uh which which you know based on certain people's reviews works works well or doesn't work so well. Um and then the sort of like uh personality changes that um that it was sort of it was like the sycopinsy problem or or thing that happened, right? Um, if maybe people aren't quite as online as as as we are, but but the butt rolled out a big update that made it super sickopantic and they actually had to like entirely roll it back. I think that's extremely interesting. Um, I think that um I think that clearly they see value in TGBT becoming Mhm. a little bit more of a of a of a of a of a life partner uh than than just an assistant. Uh and um and I think that I think that bodess well for

the space of companions more generally to be totally honest. Um I think it I think it it shows that um this is a need that people have. I think there was some survey that that came out that showed that um use of LLMs in like broader consumer behavior. It was like Harvard Business Review or something like that has changed dramatically even like 2023 to 2024 like a survey of how people use these things and you know 2023 it was like homework and you know writing emails and that sort of stuff and then 2024 was like I'm forget exactly how they phrased it but it was like yeah like emotional support and organizing my life and stuff that was that is frankly just much more embedded in in in how we live than just hey this is an awesome homework helper. Um, I think that's fascinating. That's obviously what we're what we're building. Um, you know, it's it's it's it's what we're passionate about. Um, so I, you know, don't want to

talk our own book here, but I think we're sort of, you know, hopefully at at the forefront of of of this and how how companions uh and their roles within our lives are changing. I I think there's something very real uh happening happening there and changing really in the way that like everyday people use LM. I'm not sure if that's like individuals changing the way they use their relationship with LM. I think some of it is. It's probably also largely just a much larger swath of people getting onboarded to these tools and using them in different ways from how the earliest adopters use them. Uh it's probably a combination of the two. Yeah.

Yeah. But it does feel like there's a shift from more transactional utilitarian driven use cases to now this more deeper and fundamental need for human connection especially in the context of the modern world where loneliness and depression and anxiety have become so rampant. And so it's going to be really interesting to see how all of this plays out over the next few years. Well, it's been such a pleasure having you on the podcast, AJ.

Thanks for uh joining me. Um you mentioned earlier that Tolen is hiring, but any other um words you want to put out there to the audience before we wrap up the episode today. No, thanks so much for for having me and and having us. Um please please try out the app if you're listening. Send me your feedback. We'd love to hear it. Um and yeah, if you are uh uh someone um talented really in any way that likes to work on uh is interested in working on in in this space um you know whether it's in the 3D world or the iOS world or the backend world or a designer or any anything similar uh please feel free to reach out anytime. Um would love to chat with anyone. Awesome. Plus one. It's a great product. Give it a shot if you haven't already and thanks again for coming on AJ. I really appreciate it. Cool. Thanks Phil. See you.