Andy Liu of Lullaby: Her Boyfriend's Bedtime Story Beat Calm, Headspace & Melatonin
18VC Podcast · Episode 12 · July 4, 2026
Guests: Andy Liu
Source: https://www.18-vc.com/podcast/andy-liu-lullaby

Andy, happy to see you here. Thank you so much for hopping on and joining us on this episode of recording in 18VC podcast. So before we talk about Lullaby, which is exactly what you're building, right? I do want to start with you. Maybe could you walk us back a bit and tell us what you were doing and what was your life like before this company? Yeah, absolutely.

Thank you for having me as well. So I like to track very back. I was born and raised in Chongqing. -Obviously. -And then, yeah, I went to college in New York University, and that is the reason why I moved to New York, for my college education. And after I graduated, I stayed in Manhattan, joining this boutique consulting firm, which is a strategy consulting firm that specializes in event industry.

But I stayed there for a year and a half-ish and realized that there's something that is more interesting that I want to pursue, and that's when I made my -decision to start Lullaby. -Cool. So you mentioned you spent some time doing strategy, right? And it was a consulting role. So what was your day-to-day life like in a consulting environment? Well, I would say we're a little bit different because we specialize in events -industry, but we can kind of separate it to two parts.

-So the first one is the normal consulting -activities. Like meetings on and on and creating pitch decks, drawing the beautiful pitch decks, doing interviews to get insights from the professionals from that industry. That's the most part. And outside of that, since we're doing due diligence for some of the events -we're going to be traveling in person to join some of the large conferences and talk to the attendees, talk to the vendors at that location and ask about their opinions about the conferences and stuff.

And essentially, we're going to integrate that into our decision making, which will -ultimately be outputted and back. I see. I tried to get an idea of what was your work pressure mainly coming from? And is it more so the physical pressure or more mental strain? I think both. As we all know, consulting is a very steep learning curve industry, and people have very high expectation about your -output, about your execution.

So mentally, people when entering this industry kind of expect that they will be doing great. They need to push themselves to excel so that they can get promoted and get seen by -the company and the leadership. And on the other hand, it is very physically stressful as well, because normally we will do -9 to 11. -Oh, wow -as like a normal time when we're on the pipeline.

Yeah, so I think both kind of compoundedly contributed. Oh, that sounds so tough. And you mentioned that you started having sleep problem. Was it right from the beginning, or does it occur sometime during your journey there? I think it really started in the beginning, but I gradually realized how -impactful it is kind of later in. I've always been a light sleeper.

-But back in college, I think this is something that is still manageable because I don't have a very -fixed schedule. But I think once I started my full-time job as a consultant, there is a hard -deadline. A hard deadline, you need to wake up every morning, right? -And for the prior day, you are not exactly -when you can log off. So having uncertain time of falling asleep and have a very set time of waking up, I think that really kind of exacerbates, not exacerbate.

-But that really made my sleep quality even worse. I see. And in your early time experiencing that lack of sleep, did you try to, I guess, normalize it just as being productive, like a lot of young ambitious people would do? Yeah, I think absolutely from the very beginning. Because -especially people in financial industry and other industries, they almost feel like -having less sleep is kind of like- -flex for them excelling at their work.

That's how I used to think about it when I first joined, but then I realized that there is no equal sign between being an ambitious person versus -How so? -lack of sleep, right? Because the less sleep I get, and as it compounds over weeks, I realize my brain just start functioning really slow, and that kind of gives me further to me trying to be excelling at my work. Mm-hmm.

I see. Totally makes sense. Definitely. And I think that experience actually relates to the initial or origin story of why you decided to start up as Lullaby and solve that problem. So, we do observe that you have a post about what's behind the initial incentive of building such a product. You describe as it was like 2:00 a.m., and you are after the consulting deadlines, you have reread your Slack messages again and again.

You already read before, but you still feel so tired, and you feel so worried that you are exhausted to do anything about that, even though you are tired, but you just cannot get to sleep. And at that moment, and you mentioned your boyfriend started making a bedtime story for you, and actually that worked, and you were finally able to fall asleep. Can you tell us a little bit more about that moment? What was happening in your mind at that night, and what do you think is Sorry.

That's it. I will cut it out. Yeah, absolutely. So I think, going back to those days, I think the moment I touch my bed, countless thoughts just come up, and it could be about very different things. But all of those are about maybe a future deadline like, "Tomorrow I have to do XYZ by this deadline." And also "Today what is the things that on my list that I haven't been able to finish but needs to get done by tomorrow morning?" And also, sometimes I even kind of thought about some occasions that happened 10 years ago, and I think I should have said that sentence instead of this sentence for that specific moment.

Yeah, so it was just countless thoughts. And I still remember, I used to dream talk, and it was actually with my friend. We're traveling to Barcelona for a music festival, and we're sharing a bed. And the day before I arrived, they already arrived there, so I joined them, and the next morning when I wake up, they were like, "Oh, do you know you were dream talking?" I was like, "What did I say?" And they were like, "Oh, you were talking about you want to geoclone some events in New York City, Washington, D.C., and Seattle," blah, blah, blah.

I was like, wait, that is exactly the context that I have for my BAT project before I head out for this music festival, and because it has been nonstop in my mind from the back end, it's somehow kind of like I dream talked about it. And I was even, at that point, feeling really stressful because I was like, do I have to check with my team if this violates the NDA we signed for the client? Something like that. Yeah, so those moments are actually quite common, and the point where my boyfriend helped with that was kind of like I tried multiple methods to help me fall asleep and try to not overthink and try to cut off my thoughts, those things that are too wired on my mind.

-But I realized the meditation apps, like Calm and Headspace, they have very structured and a large library, but I try a lot of them, and doesn't seem to be working for me somehow. And then I also try some of the supplements, like melatonin, magnesium stuff, et cetera. But still, I feel like since my sleep is not categorized as sleeping disorder, I think what I need is more a kind of emotional side of -acknowledgement and like have a familiar or comfortable voice that can cut off my thoughts. So that is the point where I realized, okay, actually, I can have my boyfriend telling me bedtime story.

And he actually proposed it once, and we tried it once, and I realized, wow, this is more helpful than any other method I used. So we've been keeping that as kind of our winding down routine for a long time. Yeah, that's pretty sweet. And I do see that you're mentioning that maybe that's not actually a physical disorder, but it's more of like a mental feeling thing, that it's about mental comfort, mental ease.

And then moving on to your boyfriend's telling the bedtime stories to you, do you think it's actually the story itself that matters, or is his voice that matters, or it's something behind that is the kinship, the sense of security, the sense of trust, and the sense of being taken care of that really help you ease your anxiety? Yeah, I think that's a good question, and I think that definitely comes in different layers, right? So I think to start with, the first layer is actually the content. But it doesn't mean all of the content for that bedtime story is really important.

I think what is most important is the starting, first paragraph of the bedtime story is the most important. -A little bit of context is that my boyfriend used to pick the fluffy toys from our room and use that as a character to create the bedtime story. And I think it is because I'm listening to a story about a toy that I'm so familiar to that I get attracted from the first place to engage in the story. But what keeps me continuously engaged in that is my -boyfriend's voice, right?

And because I really like his voice. His voice is not A lot of meditation apps, they give you a kind of a therapy voice, a therapist voice. They're absolutely professional, but kind of just lack a little bit of fun and -intimacy. So I was like, okay, I really like my boyfriend's voice, and I think that is something that has been very helpful for me to fall for the whole kind of bedtime story.

And I think the third layer underneath that is what you mentioned, the familiarity, the security, the trust, right? I think the difference between me listening to Calm, Headspace versus listening to my boyfriend's bedtime story is really like, I know the words he said are staying true to me, -being very honest to me, right? Whereas for the content for the meditation app, sometimes they're very structured But they are very guidance oriented instead of saying, "Oh, I feel you, I see you," and put that integrated into the bedtime story itself. Does that answer your question, hopefully?

So you're trying to recreate that sense of the safety and authenticity through building Lullaby, right? And so my question is this: Will you be competing directly against those AI boyfriends/girlfriends kind of apps in the future? I think my answer is no. Yeah, I think we are trying to position ourself as very sleep focused, right? -And I think our difference between us and other AI boyfriend/girlfriend companion apps is that they're trying to get you engaged more on the content.

They want you to stay on the platform as much as possible. But for us, it is really like we want you from scrolling your phone to turning -off your phone. -Yeah. We want to help with that part. So like I can actually feel that because when I was signing up for the test fly and testing version of it, being a test user myself, I tried to throw a lot of noise around it.

I tried to talk to it by saying, "Tell me something outside that's totally irrelevant to the sleep. Tell me more about the stock market. Tell me about the weather tomorrow." And then I realized that gradually observed that it will always try to nudge me into going to bed. And so I think that's the biggest difference between this and most of other traditional companion apps out there, is they all want you to stay.

But what you're building is trying to make me drift away from the platform. Our North Star metrics is really different from those platforms in a way, also going back to why we put the focus on sleep. Because in order to fall asleep, you really need to -kind of zoom out of the informations that you're kind of always on top of, right? Like the stock market, you know?

Like what's the next to-do list, right? And after we talked to a lot of people who have similar falling asleep issues like us, we realized a lot of times they really like to listen to very nonsense content that make no sense at all. But because it doesn't make any sense, that it helped them to cut off from a very logical daytime experience, like work routine, and drift towards a sleeping state. So I think that is essentially what's the difference underlying us with other companion apps.

Totally. That very much align with my experiences. Is that actually more of a distraction from your daytime work? Is that what the app is trying to achieve? -I think it's part of it. So we started with the bedtime stories, but apparently we're not only focusing on the bedtime stories, -right? So, we're doing the normal kind of content people consume for sleep are bedtime stories, meditations, white noise, et cetera.

So, bedtime story is our first motivation of start doing it, but ultimately we really want to provide the right solutions, or not right solution, the right kind of comfort for the user based on their daytime experience. And the part of cutting off is for those users who tend to overthink and tend to have very wired sleep before going to sleep, right? Whereas for some other users, they might just be like, "Oh, I actually had a very happy day today I want to -share with you." Then that's where the chat function would place in a bit more because if you don't speak it out, you kind of like, "Oh, I feel so excited, I feel so excited." -It keeps replacing your mind. But once you say it out and you feel acknowledged, you like -that also kind of builds towards your wind-down routine.

Right. And going back to how you actually validate and figure out that there's actually demand beyond just yourself that's craving for product like that. So you posted this onto Reddit, right? And then you receive immense public support. What was your first reaction when you saw the top comment, which is, "Protect that man at all cost"? Yeah. My first reaction was I actually showed it to my boyfriend, and he was like, "You see?

People acknowledge" "people cherish me more." I was like, "Yeah, absolutely." Yeah, but I think outside of that, I also see people's need towards the acknowledgement, emotional kind of comfort before going to bed. And a lot of times, people would kind of recall their childhood memory of their either uncle, auntie, grandparents or their mom and dad talking bedtime story to them, but it somehow kind of paused in the middle of the progress them growing up. Right? But then they were like, "Oh, that's so sweet.

I wish I had that." So that's the point where I've realized, okay, first of all, it looks like an interesting topic to explore more. And second of all, I'm really curious what is the mechanism behind that. And that, going back to Philip's topic before is why we think the content, the voice, and the level of trust is really important. I see. And then before you actually set out to founding a company and build your own solution to target this problem, I'm just wondering, have you done comprehensive market research on all those potential competitors out there, like Calm, Headspace, Endel, and Better Sleep, et cetera?

I'm kind of embarrassed to say that despite of coming out of a business background, I indeed started -the business trusting my gut. -That's a way to do it. Yeah. But I think the research and interviews is something we're exploring along the way. -Because I do think, in order to receive feedback from people that actually is surrounded by the actual product you have, that is more helpful than asking them a general question like, "Oh, what is your pain point about going to sleep?" Right?

They can give you a thousand answers. But after you create a demo, which is our first version is a vibe-coded app to them. After they try it out, they can specify the feedback, specify the areas that work and that didn't work. Right. I think, agree, I think doing market research is a very -well-known and a very structured way to do it. But I think right now, in a day where there are so many -products coming out every day, right?

You need to catch their attention first by -putting it out there, and then -gradually build towards it. -Absolutely. I think the speed and how fast you can iterate is actually becoming the moat, and you don't want to spend too much time doing desk work research. Yeah, absolutely. But that is also important, definitely because I realized I started from a founder's perspective, then I talked to a lot of other investors.

I talked to other founders. -Multiple time founders. I think they really value that, investors especially. So I feel like that does have its value in it, but ultimately at the end of the day, it's trying to test, see if there's the actual need in this market, and does this actually contributes to your traction, -right? -For sure. Yeah, I think we can move on a bit to the audiences or the target customers you are actually aiming at.

So from your ideal view or the other, as you just covered actually, investors or founders, these guys are people who have a really quick working pace and also have a busy life, and what is the ideal customer look like, and who do you assume they will be willing to use your product? Yeah. I think the ideal customer profile really started with myself, and now we kind of draw a more clear picture of the ICP being of overstimulated achievers, and these tend to be either students who are preparing for exams or young professionals in very high stress industries like consulting, banking, or -healthcare, et cetera. But at the end of the day, it's really whoever's mind keeps running at that time when they really want to fall asleep.

Yeah. Hmm. Definitely, I can sense that I may resonate with that. I don't really have a problem sleeping, but that would definitely be helpful, especially when sometimes you do want to sleep, but it's not a normal time for sleeping. Oh, I'm pretty sure, Philip, you don't have a problem sleeping because we need to do more work then. I think we need to pack our schedule a bit more, so.

Oh, man. Don't pressure me. Anyway, I think in the future, I would need your product definitely if Lucas keep going this way. And then let's move on, and maybe can you just walk us through the product as it is today? What are its main features? What does it look like? How would you describe it to a user, people who may be interested in your product but have not yet used it before?

Yeah, absolutely. Also, I just want to add one more thing to the ICP that we discussed earlier because since we also talk about the medical supplements that could be potentially used for helping you with sleep. I think we are really not targeting clinical insomnia, which I think a lot of other, the hardwares and -the sleep counseling therapists, they are targeting, and we do think that is something that really needs medical attention when you get into the stage of having sleep disorder instead of want to be seen and want to be filled. Yeah, so I think that is a very clear distinction, and that actually makes our sweet spot for the ICP to people who can sleep but cannot turn their brain off.

Yeah. But hopefully, Lucas, just stop pressuring Philip into adding more work. Yeah. I would rather people not having the pressure of going to sleep. Ultimately, we want people to get rid of the app and kind of the starting point of Hinge is being wishing one day user can -delete their app, and that is also kind of like the same perspective for us. Okay.

But going back to your point of the product. So I think a proper way to describe it is: It is an app that you can open at night that can help you wind down with a conversation before bed that understand how you're feeling and then creates the exact audio experience that you need for that night. Does that make sense for you guys? Yeah, definitely. And because from our understanding, is it more like a personalized content or personalized comfort?

I think there are two layers, right? The content itself is in a more shallow layer, and the comfort or mental support is deeper down. Which one is closer to what you actually want to build in the long run or in your ideal space? Yeah, I think we are really targeting both, but for the personalized content is kind of like the delivery mechanism for the personalized comfort, right?

After we talked to a lot of people, we realized people have very different routines for falling asleep, right? So, -like content is something we can get their attention first, as I earlier mentioned about what works for my boyfriend's bedtime story. But essentially, we really want, by that delivery, create that sense of, say, "This app is really my go-to. I feel kind of trusted, that at night, I can fall asleep.

I can open it, and it will help me to get to a point where I can kind of wind down." Yeah. I will push in a little bit forward on this track that about building the sense of trust with an AI companion. Correct me if I'm wrong in the wording, but I think at the root of the problem is that why your boyfriend's bedtime story really helps, maybe it's, again, the relationship, the kind of connection you are feeling between these two people that let you feel that your anxiety was seen, that you are taken care of, and he will stay with you no matter what happens and always support you. But do you really think an AI companion can achieve the same level or achieve a similar level kind of trust or sense of security that is building between you two?

Yeah, I think it's a great question, but I want to make a distinction first is that the trust that I talked about earlier is not that I trust the app as if it is a real person. It's more like I trust this app to help me fall asleep, -right? So like that is kind of different from the other apps, trying to build a persona and create an image and kind of wants to convince you that it is ultimately a real person from the internet that you cannot see. Yeah.

So, I think there's a huge distinction about that. And for our level of trust, we really wanted to make sure why we want to do personalization is because we really want to shorten the amount of time and energy people use to kind of find the right content for them, right? A lot of people, they will explore three, four months on Calm, trying to find the right piece of meditation or white noise for them. And some people may never realize that a bedtime story is something that could be potentially helpful after they try all these different kind of apps or solutions, et cetera.

Yeah, so what we're trying to personalize for is to get the right content in front of you at the bedtime, and the level of understanding is rooted really in the sleep intelligence that we know your previous behavior patterns, we know what's your previous preferences, and then like, "Okay, on a day like this, what would work best for you?" Yeah, definitely. And, I do remember you mentioned that you are positioning yourself a little bit different from these AI companion, AI boyfriend, AI girlfriend kind of products, because you are not looking at engagement as the most important metrics. But there was another problem I would want to ask is that, because if we want for your app to get more personalized context about what these people is working on, who they met, or what happened during their daytime life, maybe you actually want the user to get more connected to your app or your AI companion that they feel natural, or they feel willing to talk to it and share about their story. And would that actually a little bit contradict to the AI, like your positioning of staying away as a AI companion product?

Yeah, I think that's a good question, but for us, a little mechanism behind our content creation engine is really like we create the content based on a classifier, and that classifier is something we get from that conversation. But kind of not really high level, but something that is a bit more general than who you actually meet, who you actually talk to tonight, right? Because at the end of the day, still as I mentioned, we kind of want to separate the daytime experiences with the audio itself. So, the conversations to share with are really there to help us get the classifier.

For example, we have easy mind, right? We are not trying to be 100% detailed about -what is on your mind. We just know like okay, it looks like several things are hopping on your mind, and you really want -to get out of it. Then, that's the point where we can recommend a -body scan type of meditation for you. And that content, we realized a lot of people mention they don't want anything being talked about in that conversation to be included in the content -like they listen to itself.

I think that is a very kind of natural way for people to -fall asleep, the preference for that. Yeah, so, to answer your questions, we really just need the Not right, but we definitely need less context than those AI companion apps to help personalize the audio companion for the users, if that makes sense. That's -Yeah, that's legith. And obviously, there are all types of sleep-related problems, and you also can't solve all sleep-related problems at once at this stage, right?

So what's the most accurate, I would say, description of the problem you're trying to target now? I guess, is it the loneliness? Is it too much noise, right? Is it always anxiety or something else? Yeah, I think essentially, if you want to fall asleep and cannot fall asleep, just find us. That is what we're trying to be. We're not targeting very accurately on people who cannot fall asleep because of overthinking or because who have a very tense body after doing work.

We are really trying to bridge the gap where people know they want to sleep but cannot fall asleep and do not know what solution may potentially work for them, and that's where we're trying to serve. So what are those people doing today to cope with the solution? What are some incumbent solutions right now existing in the market? Yeah, I think a lot of people are doing, they try Calm before they try Headspace.

Some of them are scrolling social media before going to bed, like doomscrolling. And we also have friends who would read books, who would watch a childhood kind of drama series, soap opera series. Yeah, or just listen to our podcast. It can be pretty good for going to bed. Actually, I did fall asleep one time. I was listening to one of our podcasts.

Oh, you're not supposed to say that right now. This is your own podcast. Lucas's voice has the sense of familiarity and trust for Philip that this exactly would work. No, I think it's we are interviewing a professor, and his voice just reminds me of the day sitting in the classroom and just -And just sleeping through his classes, -right? -falling asleep.

-Yeah, absolutely. -Yeah. So what's the part that none of those current solution that you mentioned solve well? I think if you go ask anyone who have some certain of sleep concern, most of them would be like, "Oh, I want to sleep better, but I don't know how." Right? And I think that is why there is a lot of friction to get started using apps like Calm and Headspace, because they have a comprehensive library, but you need to try each one of them.

You need to explore the library and praying when you open the content for the next 40 minutes and be like, "I hope this is going to help me fall asleep," but you cannot guarantee. Right? That is what we're trying to do, is we want to make it a more -less passive content selection, but more active kind of providing the content to you instead of you having to spend multiple days or months trying to find the content yourself. Yeah, that's a pretty clear distinction.

I think for other apps out there like Calm or like Headspace, those are more like Spotify or Netflix, where you need to browse a library and select content from that. But for you, it's just like right now, the only and the first thing that users like me, the first entry point, the first touch point will be just the chatbot. So it's very straightforward, very clear. And I would like to move on a little bit to actually, I do remember you say you have interviewed a lot of potential users, and what surprised you the most during your talk with them?

And for example, do they have a clear idea of why they are not able to sleep, or are they clear what kind of feature they want, or actually, they don't know? They're there asking for a solution. Yeah, I think they mostly know why they cannot fall asleep, but they don't know how to help them fall asleep. -Yeah. There are so many solutions that they've tried, and again, people have different type of experience before going to sleep, and it's really hard to draw one conclusion, but overall, they've been all like, "I want my sleep to get better.

I don't know how. Help me with that." Yeah, that is the thing we think are very interesting to find out, that kind of align with our expectation. And another interesting thing that we realized is actually people's need for familiarity -to help them fall asleep. Because there's a girl who struggles to fall asleep, and after searching on Calm for like four or five months, and finally find a piece of white noise that works perfect for her.

And it is kind of like it's a white noise for a natural landscape, and there's little insects making -sounds and then having the grass in the wind, something like that. And we were really interested, why this specific scene, why does these specific voices work for her? She was like, "Oh, I think it has to do with my childhood." When she was really little, she used to go to her grandparents' home and in this kind of a similar setting. And whenever she goes to her grandparents' home, she will have super good sleep.

And she was also saying, some people, they are really used to listening to very loud white noises that no one else could accept because the interviewee was raised in an environment where it's super close to a shopping mall. And when the interviewee was really young, the parent would bring him to the shopping mall, and he feels like that is the kind of environment where he feels safe, and that is why the white noise worked for him specifically. Yeah, so we saw these little patterns of what could actually help them fall asleep, and why it has to be that piece of content. And yeah, we just feel like it looks like growing up, since people are changing their sleep behavior sleep pattern a lot because they move to different cities, they start different jobs, but really at the end of the day, it's those things that are very familiar since their childhood that actually help them fall asleep.

Yeah, so that is something we found super interesting. Yeah. And from that observation, did that help you guys kind of redesign or adapt your functionality or the feature you are shipping on your app, so that maybe, for example, we would consider adding in the onboarding procedure that your AI companion will ask actually about the user's childhood experience, if that's a higher chance that you will find something useful for finding the right content for -them. -Yeah, exactly.

That is exactly something we planned out. Because, again, at the end of the day we're trying to help people find the content that could be helpful for them, not since day one, since the first week, I would say. And we're really trying to find these little patterns that those correlational kind of relationship to their better sleep. And we realize now a lot of academic research, they're really heavily focused on the sleeping disorder, but there is less of them that focuses on not having sleep disorder, but having the trouble falling asleep.

So it's really also kind of a similar progress that we're doing our little research ourselves. But yeah, onboarding from the first step is something we want to integrate so that we can compound it to the classifier we already developed to provide better quality -content for the users. I actually tried your product. And Personally, the biggest aha moment for me is how realistic your voice narrator could be, and just how good the quality is.

And it just sounds so good with the natural flow and cadence. And I'm just wondering, is it sort of aligned with your expectation that this would be the biggest aha moment for first-time users? I think so. The first-time users, yes, but for the long-term users, we're trying to build the level of trust on it, as we discussed. Yes, I think most of the users, after using our product and then get back to us, they were like, "Whoa, is it actually real person or is it actually AI?" -And that is actually something we have been kind of put a lot of energy on, is trying to- -orchestrate the voice and the little breaks between the sentences, and it is even a different pattern between the bedtime stories versus the meditation.

There's very nuanced, specific designs inside the content and the voice, and the tone itself to create that very smooth feeling. And I think, yeah, we also mentioned earlier, from my personal experience of using Calm and Headspace, I just don't like the voice. I know. These are very much robotic. I don't know why. Yeah, exactly. And they're even narrated by real persons.

So I was like, if we have a better way to create better audio experience, why not, right? Obviously. And no intention to roast you, though, but I'm sure you also get this question a lot if you've been talking to investors recently. What's going to be your moat if suddenly ChatGPT or even Claude Voice become more and more natural, like human? Yeah, I think that's a good question.

I think the difference down to the very deep is still the sleep intelligence. Like ChatGPT, Claude, these general large language models, they're capable of doing so many things, but at the same time, there are so many apps that seems to be kind of easily replicated by the functions of these general large language models that still exist and getting a lot of attention nowadays. And the real difference is the sleep intelligence. I think GPT, a lot of times, if you talk about your daytime experience to help you fall asleep, it would say, "Oh, I feel you." Like, "Yeah, just share more." Like, "Absolutely.

I feel that way. That must be frustrating." It's kind of like an ongoing and nonstop validation. -But it really never gets you to a point that will help you zoom out from that context. -I see. And also, I think, again, what works for different people is so different. For example, some people with just white noise, right? Have you ever tried using ChatGPT to create a piece of white noise for you?

And also, for those white noise, how do you make sure this is something that works for you? Yeah. So essentially, the sleep intelligence and that kind of level of scrutiny on sleep is something I think essentially differentiates. And that is why we have a sleep and wellness lead on our team who has I see sleep knowledge and expertise behind this. I see.

And since you were just talking about sleep intelligence, I assume there's something like an intelligence brain system in your application. Would your app benefit from something sort of like data flywheel advantage? Would you provide better services for all users as more users come in? Sorry, can you clarify a little bit? So do you mean as -Can you clarify your question?

-Yeah. Are you able to feed and not take advantage, but further refine on your underlying tech or model as more people coming in, and therefore provide better services for future prospective users? Oh, yeah, absolutely. I think my engineer is going to be a better person to answer -this question. But essentially, we have agents that are creating your bedtime kind of experience personalization, and that agent has a learning capability within itself.

So it will learn and adapt to your previous seven-day selection, et cetera, and your sleeping time behavior, et cetera. So what we're trying to do is really to, as you use it for longer term, you really kind of eliminate the need to need to select A, B, C to get -Awesome. -the right content. -Awesome. You got it totally right. That's exactly what I was asking, so yep.

Yeah. I would like to ask, what actually do your users now sharing with the AI companion? Do you know that piece of info? Or, and the further down question is, would there be a kind of a little bit privacy concern? Because I think this is a little bit more personalized context comparing to the other workspace context, right? Yes. Can you clarify a little bit about what is the thing you mentioned about the AI companion for our users?

Like the feedback for that. Yeah. I was just curious, are you able to access basically the chat, right? The chat the user was sending to your companion. What are they actually sharing, or what are they talking to your product? And the second down question is, would people be concerned of a relative privacy issue if -their personalized content is being -used in a certain way?

Yeah. -Got it. Yeah. Absolutely. So, I think to start from a higher level, it has to do with the memory design system, right? So, we deeply value users' privacy, and that's why for the memory, we have short-term memory and long-term memory, right? And as I mentioned earlier, we're not trying to store every single sentence that you shared to the chat to the database and try to create a super nuanced profile of you.

So, the shorter memory, again, my engineer would be better at answering this, but from my not very technical explanation, the shorter term memories are only there to get the classifiers. And the classifiers are the things that are going to be stored in the longer-term memory that helps you predict what the next day content would be more suitable for you. So, we are not trying to store every single little detail about you, but what we really care about is what is your status for your sleep that night, what content you listened to that night, and does that inform us to provide you with a better audio for the next day, right? Because, again, the familiarity is something that works generally very well for people to fall asleep, and that familiarity could be the content itself instead of saying, like, "Oh, I recall yesterday you mentioned you had a breakup." You had a breakup, your heart broke, something like that.

And it's like, oh, yesterday this content, meditation, worked really well for you. And I assume today with this new classifier, I think this is still going to be the best content for you today. And you just mentioned about the classifier. Out of curiosity, can you share with us what are the different classes you have for now? Or is that your secret sauce that you want to rather keep it private?

Yeah, it is a little bit proprietary information. -Because this is developed by our sleep and wellness lead. So again, I think I shared a bit earlier, for example, one of the classifiers is the busy mind, right? And for the busy mind -the simple content would be a meditation with a body imagery that guides you to focus on different parts of your body and then kind of refocus on the physical world instead of getting caught up on all those ideas in your mind.

Yeah, that's just a very simple -kind of example. But for each kind of classifier, you can think of it, the status of you before going to sleep. And for each of them, there is a content that would be most suitable for that status. And it is not the only decision factor for that. As we discussed earlier, your preferences for your childhood memory and other factors are also going to play a role in deciding what is the thing that needs to be played for -you tonight.

-Definitely. And want to looking at the future, so going forward, what do you see Lullaby will be down the line? Would it be a smart sound space, or would it be a bedtime story generator, or just more like a close friend who's willing to listen to you and tell you a story before your bedtime? I think I'm open to possibilities. -We're still very early in our journey, so I think all of the interviews we've done have been super helpful for us to kind of frame a little bit more clear picture of the next step we want to head to.

But again, that's only for a very shorter term thing, right? I've talked to another founder friend of mine, and he was like, "I only make decisions for the next three months. I'm not going to make any decisions that is longer than that." So I think that is something very true for a very fast-paced tech world today, and I kind of agree with him. So, yeah, I'm open to all type of possibilities and also, from our conversation, not from this podcast, but talk about the hardware potentials, right?

We also think that is a very interesting trajectory to pursue, and we're thinking about, say, do we have to collaborate with those brands, or do we just start adding the plugging kits directly to our app? So instead of having to access that through a hardware, we can directly get every information from the app itself, with access to different platforms. You can have WHOOP, you can have Oura at the same time, right? Instead of, say, we can only work with WHOOP or Oura is it possible to get all of those together and then magically create something for you?

Yeah. And how would that be a game changer in terms of from user's side, in terms of user experiences? I think, as we mentioned, we want as little friction as possible, right? We're trying to make the personalization part work well, and for that purpose -You need context. -Yeah, the context. Context from the user's talk, but also from user's physical data points, right?

And we think a lot of these hardwares, they're doing an amazing job at providing the most accurate metrics -of your body, right? ... there is kind of a little gap for the implementation, the suggestion, the solution part, right? And that is the part where we think is a very promising and -very fun perspectives for us to explore. -Right. You could become just a sleep layer -embedded inside those smart hardwares.

I think that could be a good direction going forward. And just a little bit reflection at the end of the episode, what are something you have learned most about yourself, actually, from your own founder's journey so far? Yeah. I feel like it's not about myself, but it's more about my understanding of this world is that -I used to be a little bit perfectionist, especially working in consulting where even something as a comma at a pitch deck would matter a lot.

But as a founder, I realized there's really not a perfect standard. It's all artificially created. And a lot of times, instead of trying to perfect things, I think getting things done and put it out there might get you move a step further. Yeah, so really start doing things, and even when it's not perfect, let people -know, right? That may give you very unexpected -discoveries.

-So you're talking about shifting the mindset from having to achieve all As, having to achieve 100 on exam, to just getting enough, getting a passing grade will be enough. And then also building in public, right? Being able to embrace the raw and all the criticism and feedback out there at the very early on. Yeah, exactly. I see. And for those young founders who are still feeling early uncertain, what's your words?

What's your message to them? Ganbare. It's not an easy journey. I think it takes a lot of self-belief and resilience to be on this journey. And I think a lot of times, you really have to have this kind of self-belief out -of nowhere. backed by no evidence and to keep on going. And I think, again, the road is not as scary as it looks, so you have plenty of chance and rooms to do things wrong and see how you can do your next step after that.

Awesome. That'll be a wrap-up to this episode. Thank you so much, Andy, for your time and for being willing to be with us. We look forward to hosting you next time. And folks, if you want to hear more from us, don't forget to follow us across all of our channels: Spotify, Apple Podcast, and YouTube. And follow us and subscribe, and stay tuned to what has yet to come.

I'm pretty sure now it's bedtime for me, so I need to use Lullaby. And good night, folks. Bye.