He’s building an AI media empire

Andrew Warner: This founder is creating thousands of customized books using ai, but that’s just a start. He visualizes a world where movies, TV shows, every kind of media is gonna be customized to you, and he’s building a company that’s gonna do it. He’ll tell you how he’s doing it, and it’s gonna riff on an idea and a vision for how you can do it too.

Ricardo is the founder of Dream Stories, makers of personalized children’s books.

Intro: Next new thing,

Andrew Warner: presented by Zapier, the AI automation company. Ricardo, how many books have you created that are all AI generated? Would you say

Ricardo Vice Santos: O ma? Many. Right. We, uh, we, um, in, in a sense all of them are, I was gonna try to say how to answer the question.

You know, each one of them is unique, right? So that’s why it’s, it’s hard to, to to be precise. Um. But yeah, but we’ve had, uh, we’ve had, uh, since we launched, we had, uh, about 70,000 characters created on the platform. Uh, and, uh, those are, those essentially, you know, each book, uh, you know, includes up to three characters.

Some people like create book, multiple books with characters. You know, some people get the entire series like, uh, uh, six books in one go. And so, so, so, yeah. So it’s, it’s quite a big number.

Andrew Warner: That’s a lot. And all customized to the actual customer who’s getting it. Um, how much revenue have you produced?

Ricardo Vice Santos: Yeah, we’re not talking about revenue anymore.

I think the, when, when you talk about revenue, the story becomes like, quite quickly about like, sort of, don’t get me wrong, but kind of a get rich quick type of thing. Uh, yes. Where, where, where in the reality I see that what the revenue we’ve had is kind in support of the, the larger vision we’re trying to build.

And so that’s why I would like to focus on that instead.

Andrew Warner: Can we just say get rich quick story here, because I think that’s gonna power people up. Can we make it into a get rich quick story? No. I, I, I think

Ricardo Vice Santos: I, I mean, I think you could if you wanted to. I mean, there, there’s, there’s other products on the market that I, that I think I’ve, I’ve gone for that.

Um, you know, and I suppose you could, uh, you know, we’re, we’re trying to build something. I’m, I’m, I’m teasing.

Andrew Warner: Can I, can I give the revenue that I’d seen, uh, you, you talked about in the past, you’ve obviously got a different number.

Ricardo Vice Santos: Um, yeah, it’s a bit outdated, but, uh, if you want to,

Andrew Warner: you’re saying I could.

Okay. I, what I saw was 3 million a year, but that’s not up to date. Fair. Mm-hmm. Okay. Yes. That’s not up to, all right. I’m gonna get you out of this dis comfortable place. Let’s get to the story of how you did this. And more importantly, I wanna see how other people can create companies like this. From what I understand, you’ve got a family and you decided, you know what?

I wanna create something. What’s the thing that got you started on this story?

Ricardo Vice Santos: Yeah. Uh, so I, I’ve, you know, it’s important to mention that I’ve, I’ve worked in sort of in the media space for a while. I was quite early at, at Spotify. Uh, and so I’ve, uh, I guess I’ve been. Both in consumer and specifically in media for, for a very long time.

And when I saw generative ai, um, I, you know, I, I long ago said this idea of like, I wonder if one day we’re gonna have bespoke content created for each individual. I know it sounds kind of like I’m making that up, you know, with post-talk, you know, now that we have generative ai, but it’s a true, true story.

I’ve actually seen, you know, personalized content on Spotify itself, uh, you know, with, with songs, you know, birthday songs and love songs and so forth to people’s names. And so I had this thought experiment about it. And so, wait, pause for a second.

Andrew Warner: Yeah. You’re saying back then you had, years ago you had seen customized songs on Spotify.

Ricardo Vice Santos: That’s, that’s correct. Uh, this was, this was very early on. Right. And that what, what one of the things Spotify unlocked of course, is this like extremely long tail of content. Yes. You can think that, I mean, it’s very obvious, right? If you go to basically this Walmart, you know, they’ll carry. Say a hundred, 500 CDs, right?

Mm-hmm. That you could buy back then. Uh, and of, so of course, you know, you wouldn’t have this, this sort of long tale of content that, that stuff like the internet and, and obviously Spotify enabled, right? And so there was, there’s a very long tale of content on, on Spotify, of course. And, and I thought it was fascinating.

Even albums of, of songs, like I said, with, with people’s names, which were, it, it was actually fascinating to see because, uh, you know, people ask me when, when I’ve mentioned this, you know, how popular were they? And you know, because, because you know, even publicly you could see. Like which names? Those are more popular, right.

Than which, which songs within that, but I, I thought those were

Andrew Warner: all handmade. I think there was a guy, Matt Farley, who would create thousands of songs and put them in. I thought what he was doing was basically singing a happy birthday song for Happy Birthday, Becky, and then Happy Birthday Andrew. But it was him.

It wasn’t AI at the time, was it? No, exactly. This was before ai. Right. That’s, I see. You’re saying, look, back then you saw customization. People naturally wanted a song just for their name. You said, oh, this is amazing. What if we could have AI created? That’s the thought process.

Ricardo Vice Santos: Yeah, exactly. Right. I mean, but again, I’ll be honest with you, like at the time, obviously you would say AI would be the one to do it, but I, I didn’t go as far as to think how to do it.

I just thought it was interesting that if in a feature would be possible to do it. Right. Yeah. And obviously this was at a time post YouTube and so forth where people were creating videos and uploading, and so it was interesting to think that will people at some point create content. E to, to that point, or to them even to themselves.

The analogy I sometimes tell today is the same way as when Apple put a camera in your phone and in your pocket. It, it didn’t make everyone into a photographer. Now granted more people became photographers because of it, but what it did is it created new behaviors in which people started taking pictures of their food and pictures of their, you know, in addition to their kids.

Right. And that sometimes a bit hard to forecast, right? Yeah. And so I think if you give people the tools to creation, they will do things that are not necessarily an extrapolation of what we do today. Uh, and does I, you know, you can think about, you know, it’s kinda like having a kitchen at home, right?

You’re not a professional chef. You’ll probably cook for yourself, uh, or for your family. And so it’s interesting to think if you give people the means of creating Hollywood level content, you know, will they do so? Right? And will they do so not necessarily with the intent to distribute, but to consume themselves.

So that was, that was kind of what had this seed in, in, in my mind.

Andrew Warner: Okay. And so you said great. I one day this is gonna happen. How did you end up with kids stories?

Ricardo Vice Santos: So, I have a, I have, well, actually, uh, interestingly enough now, two sons, if you, you know, I had a, uh, my second child, uh, literally a month ago.

Um, but with my first son, we, you know, we, we, my wife and I, either one of us attend, reach to him every night, right? And, uh, and I was, as I was reading to him every night, I, I, I realized that there was a disparity between the content that he was reading and the things I remembered from childhood. Uh, you know?

Mm-hmm. I, I was a big fan of tinting. Uh, you know, I wanted to be an adventurer, you know, explore the world. I, I literally have in my office, um, at home, the, um, you know, a big poster of Tintin goes to the moon with a, with a rocket ship, you know, and that was kind, kind of my, my dream. And so I, you know, obviously that’s also the dream I have for my child, right?

I want him to be this adventurous person. And, and when I know, looked at the content that. He was consuming. And granted, he’s a bit younger, I found it, to be honest, a bit more mundane. Uh mm-hmm. And there, there’s content. I, I think it has high production value, but it feels a sort of like, for lack of better expression, lowest common denominator to, you know, it’s kind of like something that’s very agreeable, you know, that that dad in the family is a bumbling kind of funny character, you know, that’s like, I don’t really want my, I want my kid to see me as a fun person, but not necessarily as a clown, you know?

Uh, kind of like as he was. Yes. And, um, it was a bit, you know, to be honest, like, you know, I was like, I think this, this content sucks. And so, and so I was like, okay, I want him to, I want him to be tinting, right? And so. It’s interesting that even who wrote Tintin said, you know, Tintin was modeled by in himself, right?

It was also a, a, a materialization of his dreams. And so in, in a way, I decided, okay, now I can actually do this for, for my son, right? I can create, I can create a book that actually makes him the hero, right? Makes him tinting, right? Um, and so that, that’s kind of how, that’s kind of how I started, right? I wanted to do that essentially for, for myself, for my family, right?

Andrew Warner: And so did you go and create it yourself? I remember when I first saw chat GPT write stories, I said, this is gonna be really interesting for me to create a story for one of my kids. I’ll just tell it what to do and put my kid in the story. Is that what you did? Uh, it’s

Ricardo Vice Santos: a little bit more complex in that, uh, per, uh, you know, because the, the, the challenge with ai, well, there’s different challenges, but particularly with the, if you want to, in this case, if you want to make a picture book, is actually the consistency of images actually kind of complex problem.

That’s actually, let’s pause on that.

Andrew Warner: I’ve heard you say this before. I think that it deserves much more attention than, than other interviewers have given it. Essentially what you’re saying is if I go into Cha GPT, or if I go to Nano Banana or any one of these things and I say, gimme this photo, it’ll eventually get me the thing that I want.

Then I say, put this character doing something else, and it’ll be almost a different character. That’s like a warped version of the original. That’s what you’re saying. It always warps it a little bit, right?

Ricardo Vice Santos: Yeah, it’s a little bit, it’s, it’s, it’s quite interesting actually. And, and to be honest, with different models that, that side the boat, certain ways it gets better and, but you also uncover new problems with, with, with that different models.

Yeah. But, but generally it’s a problem. Right? And, and it’s, I impartial and, and the other thing I like, I also like to say is that it’s not that difficult for you, you know, now none of banana in particular make made things sort of better for the average person to do it. But let’s say two years ago, if you wanted to, to, to create dream stories.

It’s not that difficult for you to do if you’re a technical advanced person to do it for yourself. It’s an entirely different thing to build a product for other people to use it like in a, you know, with sort of some sort of symbols of quality and consistency, right? Uh, right. So again, you actually mentioned this, right?

You go in, you do it, and you eventually you get it right? It takes many tries and, you know, the average consumer is not looking for many tries. The they’re looking for something that kind of works, right? Um, and, uh, you know, it’s similar to, you know, I could have told you, you know, well there was piracy before Spotify and it kind of worked, right?

But, but Spotify made it well, just works. You just, you just search it and you play and it, it starts playing immediately. And, uh, and I think that that’s actually the, that policy is actually, I think what’s actually makes it quite substantially difficult to do.

Andrew Warner: Okay? So the first thing that you did was make it for yourself, for your son.

You had that, what did you use to make it for your son chat? GPT?

Ricardo Vice Santos: We used a, a number of things. Uh, we, we, at the time we used, uh, chat, GPT, we used, uh, uh, stable diffusion. Uh, I forget the first model we used. Um, we used to do it, uh, we used LARAs, which was, uh, we still doing some regards. Uh, but LARAs is something that, you know, that you, you literally use to train models on, on likeness or consistency to actually okay, to actually get, get images to, to kind of stay consistent over time.

Uh, but then there’s a, it’s, it’s a, again, it’s a complex thing. I mean, if you look at the, you know, the, the 70,000 characters I told you, and you asked me like, okay, how many of those really were, you know, problematic this way? There’s, there’s a lot of parts of them that were in part because. If, if the, if the last page, sorry.

If the customer is tech savvy, it’s very easy. But customers in the real world, you know, you ask them for a picture of, you know, show me a picture of the character you would like, you know, single picture. And people upload like a wedding photo with like, you know, 25 people or the, the picture is like potato quality and then, and then, and then, you know, it’s garbage in, garbage out essentially.

So again, there’s a, a lot of it comes from this minutia of like, making something work at scale is actually dealing with these types of things that you know.

Andrew Warner: Okay. But the first one that you did yourself, you made for your son by yourself. Mm-hmm. You did. And that, what did you use? Stable, uh, stable fusion.

Did you use Uh, chatt? Stable diffusion? Yeah. Stable diffusion. Diffusion.

Ricardo Vice Santos: Uh, yeah, chat pt. You know, actually the story, the original story was, was written by me. Uh, you know, I used chat PT as kind of a co-writer, you know mm-hmm. To kind of help me with things. But it generally was, it was written by me. Um, and, and again, civil diffusion, there was some stuff on, on the orchestration of, uh, of, of it.

Yeah. There was like more technical stuff and then just frankly code. Right.

Andrew Warner: Okay. So you did it, you had it, you had the result. You said, now I want to do this at scale. I read an article, uh, about your co-founder, Phil. Mm-hmm. Essentially. Mm-hmm. At that point you said, I think I’ve got an idea. I think you and Phil worked together on something in the past, and you said, will you work together with me to build this?

And that’s when the two of you started cranking and he was taking it from this one to many idea.

Ricardo Vice Santos: Yeah, I, I should correct something that, that Phil was, uh, was, he was playing with Simple Diffusion, right? And I, I was doing these experiments as well, and then he was the one that kind of helped me put together the, the sort of, the, the thing we could literally print, right?

The fi the physical, uh, the physical book, which I’ll talk about in a minute. Okay. Uh, how it became a physical thing. Um, what, what happened though is that, is that Phil, so Phil was a, was a, a data scientist at a company I founded, uh, called Kinko, um, which, uh, which is a, a prior company I, I, I worked on. And, um, and he, you know, at, at some point I decided to step down for my, for my day to day at Kinko.

He, he, later he left as well, and he went to Uber. You know, the, the record is foggy. I, I, you know, my recollection of it is that he went to Uber. Uh, to quote, learn to be an entrepreneur. Uh, and I, I thought it was bullshit. ’cause you know, you don’t, you don’t, that’s that, you know, you at that scale, you don’t learn to be an entrepreneur there.

Mm-hmm. He thinks I’m wrong and, uh, that I’m misremembering. So, you know, so you’ll have to talk to him to see who’s right. Um, but it, but anyways, but lo and behold, he, he was, he was, you know, he learned a lot at Jupyter, but he was also quite bored and he was playing with, with these things. Right. And so he, so he helped put together the, the orchestration of, of these things he was playing with, with, again, with AI quite a lot.

And he helped put together the orchestration and then, yes. And then I told him that I think we build the product around this. Right. Yeah. And I think that this, this is the, there’s a path between where we are and the materialization of that idea I had many years ago of this like, you know, generative media platform, if you will, right.

Where, where content is created just, just for one individual, which, which I should say, I can talk more throughout this, but, but it’s, it’s a bit of, um. Sometimes difficult for people to understand. ’cause, ’cause again, most people think, you know, most people at this point think AI is gonna be a big thing, of course, or some of them.

Mm-hmm. Already is. But most people tend to extrapolate and so they see that, you know, hey, it’s gonna make Hollywood better, right. Because production costs are gonna go, oh, you’re gonna have more content. And, and again, I’m more of a believer that there’s gonna be sort of new behaviors that you can’t see, you can’t see just yet.

Uh, and this is one of, of them, right? Which is if cost near zero, then literally everyone can do it. But if, but if everyone can do it, then of course you know the, who’s gonna watch it. Uh, and the question is yourself, or the answer is yourself, meaning you make it for yourself. That’s correct. Yeah. Which doesn’t mean, by the way, that I should say, it doesn’t mean that you know that it’s the end of broadcast media or anything like that, you know, that that’s another misconception I think people have sometimes she’s.

That, that, you know, one thing will clear cut, you know, replace the other. When in general, like, you know, what happens is that new behaviors unlock right? Granted, some become a bigger slice of the pie. Um, so I do think there we continue to be move ways. I do think there’s gonna continue to be a new tube, but also think that there’s gonna be this immersion behavior of creating content just for your own consumption, such as taking photos for yourself, which is not, it’s not so foreign when I tell you in that way.

Right?

Andrew Warner: Yeah. Yeah. I guess what you’re saying is look, the, when the iPhone came out, it didn’t stop, um, professional photographers from taking those professional photos. It just added more. And most of the photos that we take aren’t even shared with anyone. It’s just for ourselves. And you’re saying the same thing’s gonna happen with ai.

I might just make a story for myself to fall asleep that is customized to me and want to motivate me in the morning to wake up that is customized to me. And all of those things are additive. They’re not replacing. I think that makes a ton of sense. I think that makes a ton of sense. And I could see that happening in a lot of different media.

I have these songs that I’ve made in Suno that are just for me. They’re customized to my little personal experiences, the things that make me happy, the things that make me like fired up or, or sentimental. You’re saying, look, that’s gonna happen for videos. It’s gonna happen for movies, it’s gonna happen for books.

And the reason that you picked children’s books, from what I understand is you said, what are people willing to pay for? What are they so sentimental about that they want customized right now? It doesn’t have to be super long kids’ books. So that’s what you started out with. But your vision is much broader than that.

You would like to get into making video stories for a kid so a parent can go and create a cartoon for their kid, right?

Ricardo Vice Santos: Yeah. Yeah. There’s also an interesting dynamic that I’ve, I’ve I’ve witnessed before, which is I find fascinating products where the, the best way I can describe it is that where the buyer is not necessarily the consumer.

There’s a few products, uh, that are like this. Like, uh, the one I usually tell is like, is Gillette where, you know, usually the, the, uh, the woman in a relationship is the one that buys the Gillette. It’s like, I think about 70% of the purchases of Gillette are by women. You know, the number may not be accurate.

Something to that degree. And, uh, I, the reason I know this, by the way, is because Kinko is a direct to consumer company and we studied, you know, dollar Shift Club and so on, so forth, and, and we were talking, thinking, and talking with other, other brands and, and thinking that. It’s interesting because then if you, if you make that assumption that someone else buys, then you think that the marketing is, for instance, the marketing for Gillette.

It’s not for men to feel I wanna be this guy. It’s more the women saying, I want my men to be like this guy. And so Uhhuh. And so it’s, so there’s a bit of as second degree aspirational, uh, you know, in a way. And, and I think with the, there’s a, I think there’s a big problem in media that everyone acknowledges she’s this desire for dopamine, right?

In putting social media where it’s, it’s very addictive. But if you make with one step removed, you know, you, you know, if you go on the street and ask like everyone, you know, do you think you should spend less time on your phone? You know, everyone’s gonna say yes, but they still keep doing it, right?

Because it’s, it’s a bit like a drug. It’s, it’s difficult to get, get through, but, you know, but again, when you are sort of one step removed and you’re thinking about child or so on, so forth, your judgment is a little bit less, you know, a bit more calculated or, or a bit higher level thinking, right? You’re thinking about their, their wellness beyond what they do.

That’s literally what I do for a child, right? The child can take for themselves, you do it for them, and does that means that you. That parents have this more rational, I think, interpretation about this where, where you can think that, Hey, look, I’m gonna give you the tools. You have a big problem, right?

Which is like, you know, you, you either, you know, I told you mine, you know, where, where, you know, there’s content that I don’t particularly like that my kid consumes, that I would like to get ’em, like other content that aligns more with my belief system, if you will. Maybe Sure. If that’s the right word. But second, I, I, I, you know, addiction is the, the one thing I forbid him for watching completely is like co melon, you know, uh, that one, you know, throwing the, under the bus.

But, you know, it’s like if you, I don’t know if you ever had the pleasure of, of listening to it, but literally drives me insane. The, the, the songs that I just think is playing good. I’m glad we don’t. Yeah. Uh, it’s, uh, yeah, it’s, it’s, it’s mind numbering. The bigger takeaway

Andrew Warner: you’re saying is, look, I’m actually.

If I’m understanding you right, you do have a broad and broad vision for getting a lot of different media. You did start with books because parents are picking it for their kids and it’s important and they’re willing to spend money. Yeah. What’s this point that you’re making about, like, who buys for whom?

Why does that matter?

Ricardo Vice Santos: Oh, okay. I’m sorry. Uh, uh. Because, because again, the, because, uh, okay. So we’re staying, it was taken within kids, right? I’m, I was actually answering why kids in particular, right? Mm-hmm. Why, why, um, why that the family, family products as opposed to, you know, for instance, you have, you have open AI doing soda, you know, which is a, essentially a beyond the model is, um, you know, a social network, you know, essentially, if you will, or, or a content network on, on top of, of an AI model, right?

So they’re trying to get to the actual consumer that’s gonna use it. They’re trying to make a social product in the way, whereas I’m trying to make something that is for the family, right? So something that is focused on, on getting someone, giving someone the tools, you know? So in the sense there’s two users, right?

There’s the consumer that is actually gonna consume the content. Yeah. And there’s the person that kind of. Alters, sort curates the content right, and buys it. Of course. Yeah. So that’s why families, right, why books in specifically? Um, it’s because, you know, there is a, you know, when you’re, when you’re doing consumer, particularly in the entertainment, you know, you’re, you’re suddenly in the detention economy and the one time, which is, which is brutal, you know, like, uh, I probably don’t have to tell you that, but the one thing, the, the one time of, of the, the day I think is kind of sacred for families, it tends to be at bedtime, where literally you don’t, you, you know, the parents are, tend to be very restrictive about what they, uh, what they give their kids, and particularly, um, and, and even the anti screens in, in, in, in many cases.

And so I saw that, but there’s a, a kind of, something I found very fascinating, which is there was a repeatable behavior in it, which is, was like literally every night. Mm-hmm. And when you think about something there, every night, you can start like opening your mind to, okay, I can actually build something that is not a product, something that’s actually a service, right?

Something that’s not a one-time consumption. It has a sort of a ritual to it. Right? And that’s also why Dream Stories was were were written, I told you that we, we sell multiple books, I know to the same person, uh, you know, essentially because they’re episodic, right? It’s a, it’s a sequence. And I see. Yeah.

And, and one other thing that was actually first pioneered probably by Blue Clues, uh, a TV show, which is that they figured out that kids quite like repetition. Uh, I’m sure they didn’t figure this out, but they exploited this where they show the same episode every day of the week and then the next week they show a new one.

And what kids have with that is that they like this. Uh, this repe, this level of repetition. ’cause it, you know, they get excited when they know what’s gonna happen. Yeah. But there’s a threshold where at some point they want the next one. Right. So anyway, so, so I saw all these properties to it where it’s like, oh shit, we can actually build, uh, you can build a service.

Right. Can build something that is epi exotic. That has recurrence. Right.

Andrew Warner: Okay. Lemme take a break and then we’ll come back and I want to talk about the big thing that you’ve got for the future. The break is to just let everyone know. My new sponsor, it’s Zapier, it’s the AI automation company. Now I’ve got examples of how I’ve used it, but you have won, let’s go back to one of your earlier uses of, of Zapier.

What was it?

Ricardo Vice Santos: I use it quite a lot for, for a lot of, uh, both, both big and small things. But one that I can mention is that when, when trying to secure press, uh, there’s all this like mailing list, like helper, reporter out where you get, you get emails like, you know, from journalists or, or blogs requesting like things.

Yeah. Um, and it’s of course, it’s, it’s usually this massive list of lists of things. You, you can pay some services to filter it down, but, uh, but what I realized is that I could essentially create an email inbox in Zapier, so subscribe with that email inbox. Mm-hmm. And essentially have a, you know, filter down, uh, filtered down.

Like things that were, you know, at, you know, when I first say this was based on keywords was very dumb. Mm-hmm. Now it can be built with AI as well. So you have AI filtered, like, listen, out of all these requests, which ones are interesting. And of course with AI now we can even automate, essentially say, you know, pitching back to them.

Right. And so essentially you, you, you have basically a sort of a PR agency, you know, built on Zapier, you know, with, with AI and, and Zapier essentially. So, so, um, so that’s one, uh, that’s pretty cool.

Andrew Warner: That’s killer actually, that you’re right. Today with ai you don’t need a keyword. It will just analyze it. If you tell the Zapier AI option, what you’re looking for, what kind of company you have, it will, it will screen it and then only forward you the ones that matter.

And then, yes, you can’t take it to the next step. All right. I do really love Zapier, and I’m glad to hear your example of it. Alright, let’s talk big vision here. You don’t wanna just be in the book business, which honestly, for many people, that would be a killer business to be in customized books, but you have a bigger vision.

Where do you see this company going Long term?

Ricardo Vice Santos: Yeah. I mean, you’re right. Like you can go very far with books more than, than, than I would say people, uh, would give it credit. Uh, um, also, I, I love books, you know, that, uh, I, I have a large collection, you know, that’s actually my sort of only indulgence, uh, you know, is, uh, my, my big physical book collection, um, at home.

But yeah, but we want to go into other media, right? But, but the, the, the thing is that we’re, we’re kind of, we’re kind of this like. Moore’s lost at a phase where, where the cost per token is going down severely. Uh, um, and so things like what we’re doing now are essentially becoming, uh, you know, becoming or are actually possible, uh, possible.

Now, video is still quite expensive, you know, it’s, again, it’s one thing if you’re doing, you know, video in the old model in which you’re, you know, it’s, it’s, it’s very affordable now for all the wood to use it, right? Because they make, for one, they make one time, you know, and going from, you know, $200 million to, to, you know, a a few thousand dollars is of course very good for them, uh, for if you wanna make it for one on one person and let people, like add it, let people do, you know, try different ideas and so on so forth.

It’s very expensive. This is why it, it’s, it’s kind of known that, you know, soda, you know, open the eyes. Soda is literally burning. It means set to be burning $15 million per day, right? On the, on the, on the app essentially, uh, with people like playing with video. So video is, is, so people ask me this, like, why did you start with video?

It is like one of the answers that the video is actually. It’s still expensive enough. Yes. That, that it’s hard to actually, to actually come up with a, if you’re not a hyperscaler, to come up with a, with a viable business model, arguably it’s not a business model even for them. Um, whereas image is, is becoming, uh, even though, you know, like, like all businesses, we, we, we make money on some, some, and we lose money on some, right?

If, if people add it a lot, and particularly if they add it in the wrong way, we might lose some money on some, some of them. Um, and so getting that balance right is, uh, you know, is important, uh, before you actually scale. Right?

Andrew Warner: So the, the big vision is at some point you would like to create cartoons for kids with I, not just the kids’ photo being in it, but also the beliefs that the parents would like to impose on their kids.

So for me, I might say, I want my kid to understand how to be an entrepreneur, because that’s gonna serve them well. You’re saying, all right, maybe Andrew gets to create a story where his son is an entrepreneur, or the character is an entrepreneur. Someone else might have a religious belief that they want to impose or Yep.

Or bring into their kids. Same thing there. That’s correct. Yeah. That’s the vision that you have, right?

Ricardo Vice Santos: Yeah, exactly. And, and by the way, I should say, that’s already happens today, right? As in that people will create, the stories they create, have multiple characters, uh mm-hmm. We, we actually spent a lot of time on this.

This is surprising is one of the hardest things we, we did, or that we had to solve. Like if you, it’s interesting that if you look at most, not everything, but if you look at, at most comparable products out there, uh, usually they have one character just because mm-hmm. The, the whole thing about having. If it’s already hard to keep consistent with one character, character, it becomes actually exponentially harder with multiple, not because there’s more variables, but literally because AI will mix faces and so on and so forth.

So it’s actually kind of difficult to do, to do things with multiple, multiple characters, but we are doing it nonetheless. Um, and that’s actually a massive unlock ’cause it means people can create content with, with, uh, with the entire family, right? Uh, which is not just one person. So, as such, you know, there’s a lot of parents that put, you know, cousins, siblings, you know, grandparents.

We, we see a lot of, uh, of, uh, uh, you know, disease relatives as well. Uh, you know, sometimes grandparents, sometimes it breaks my heart, you know, when, when the customer tells us that, you know, it’s a, you know, a child, a child that died, you know, it happens quite a lot. Wow. Surprisingly, um, you know, uh, it, it, it’s quite, again, I think the beauty in it is literally that each story is different.

Right. And that enabling these things is kind of a beautiful thing in itself. But, you know, we’ve had customers that are one, one child that, that passed, and then they’re like. Oh, I would like my, my daughter, the, the sibling Right. To have a book with the older brother that they, they kind of didn’t, never met.

Right. And so, right. Um, and, uh, you know, as a parent of a young child, of course, that, uh, uh, that makes me, that’s makes me quite emotional, uh, to, to hear, but you know, it’s also elevates, uh, what we do. Right.

Andrew Warner: Okay. I told you before we got started that I’m a little concerned that people are gonna hear this story and they’re not gonna be able to do this.

It’s not like they can go out and build the next dream stories company. Dream stories is a tough business to, to create. Let’s talk about where the difficulties are. If someone wants to create something like this, and then let’s riff for a little bit about how someone can create a different version of this.

Something that’s a little more attainable. Mm-hmm. What’s the difficulty that I don’t see it on the outside,

Ricardo Vice Santos: particularly most consumer products, right? Or, or most products. You know, the, the, the problem that you really have to solve is distribution, right? Is that, is that we. Particularly in a time where you have all these tools like, uh, you know, vibe, coding, lovable, so on and so forth, where everyone can create a product.

You can, you know, it, it’s a great time in the sense that, you know, if you ever, you know, if you, if you, wherever the person at a party saying that have a, you know, the idea for the next Uber, well, congrats. Now you can do it. Right. You can, I can actually go out and build your next, your next Uber. So I think the problem is actually not so much in it.

There’s some things that are still difficult, but the problem is not so much execution as much as, as, as, uh, as the distribution part is, like getting people to actually, to actually use it. Right? And that, that you can think in, in, in, in different ways, right? And, and no one can give you kind of a, a formula.

Uh, but what I can tell is the way I go about it is essentially, you know, is to figure out the model in which you, a model in which essentially you retain, uh, their retain customers. Again, sounds a bit vague as well in itself, but. Yeah, let’s, let’s say first one thing I do, I do quite a lot, which is paid acquisition, right?

Uh, it’s a, it’s sort of a level playing field in the sense that everyone can do it, right? You can give your credit card to, to Facebook or, or Google and, and ask them to run your ads, um, or TikTok or something like that, you know? So it’s, it’s not difficult to get started, right? What’s d difficult is to actually get the economics to, to work for you, right?

Particularly because it’s an auction model, right? And so you’re gonna have people, uh, uh, bidding against you. So how do you win them? You you need to outbid them. To outbid them, essentially. You need to have a better economic model, uh, than, than they do, right? And so essentially, uh, what I would say is that part is that figuring out, for me specifically, again, depends on, on what you’re doing, but figuring out the economics is probably the hardest, uh, most difficult part.

Andrew Warner: Like what, what’s missing about it? Here’s what I’ve heard you say on Nathan Latkes, uh, podcast, you said. First of all, you’re an experienced ad buyer. This is one of the things that you brought to Spotify growth. Second thing you said is you like Facebook for your product. Re Dream stories does well there.

Third thing you said was that you try to. Get your cost per acquisition to be the same as your average order volume. Those are the basics. What else is there? Let’s go deeper than that clip that I saw, but,

Ricardo Vice Santos: you know, but even that, you know, by the way, I should say if I could get my C to be lower than my OV, that that’s bad.

Right? You’re, you’re, you’re right. Essentially making more money. But what I also said is that there’s this interesting thing in that, in advertising that the, a lot of agencies tend to optimize for this rows of, of one, uh, you know, meaning that you basically, uh, you, you know, you spend $60 to earn $60, right?

But, uh, to, and when I say earn is like literally revenue, but out of that revenue, you need to, you know, if you’re selling a physical product or you have cogs, you know, you need to take them off. And so essentially you’re losing money on every sale, right? If you, if you do that right? So your only hope to recoup is to essentially turn that into a, into a repeat purchase.

But the, the cool thing about what we do in, in our case, and again, this is different for every market, right? For instance, Spotify was different as well, right? We at Spotify, the model was. We have a grade three product, right? We’re gonna acquire people to the, to the, you know, it’s, at the time there was not as much competition for attention either.

So both CPMs were lower. So it was cheaper to acquire, to acquire people to the front door, get them to listen to a few music, and then out of them, you know, I, I’ll say a random number ’cause I actually don’t remember top of my mind, but let’s say 40% of them convert to premium. Right? Something like that.

Mm-hmm. And so then you just make the met and you see if the thing, the if, if the Mets works, particularly, you know, assuming that subscription, you’re gonna have revenue over time. Um, in, in, in our case, no, I should say that the market is far more competitive now than it was when, you know, in, in, in, in online media than it was when I was doing that at Spotify.

But the cool thing I think we have is that we’re, we, we, you know, we. We didn’t try to come up with something that is very exotic, that doesn’t exist in the market. Right. You know, in fact, some people, you know, uh, uh, uh, out of the, the, that podcast, you know, criticize that, you know, oh, but you know, haven’t you seen this other product and this other product that also exists?

Yeah, exactly. That was, that was partially intent, right. That the category exists. Right. You know, so I, and because the category exists, that means there’s, there’s a, uh, the, the, the customer is educated about what they’re about to buy, which is a personalized book. But what you then need to do is to convert that, that novelty purchase into a, into a repeat customer.

Right. That buys many things. And that’s where I said the focus on bedtime, the focus on being episodic, uh, you know, that’s why we’re not doing a, b, c books. You know, people ask me these things, well, what can I get a book about A, B, C? I was like, no, I’m not interested in that. Right. There’s no, there’s no repeatability to that.

Andrew Warner: I see. You know what you mentioned earlier, your previous company, Kenco, what you did, there was smoothies that don’t need a blender. It was like little pouches that you add into water, mix it up. And that was also repeat business. So it seems like that’s what you’re in, in the head space of you’re saying, what can I create that people need to buy multiple times, something like coffee, but it’s gotta be new, it’s gotta be innovative.

And that’s what you’re thinking about with books too.

Ricardo Vice Santos: Exactly. Finding a repeatable pattern, right? I I you’re, you’re, you hit the nail on the head. You know, I’ve, I’ve been doing subscription most my life, right? And the usually thing I tell people is that the, you know, subscription is not necess. I, I tend to be very against this.

I have done it as well, I should confess, but, you know, every time we go to this website and you see this like, you know, save, you know, subscribe and save 30% or something like that, right? Yes. I tend to be quite against that. In principle. I’ve advised other founders on this, and so far I’ve been, I’ve been correct, you know, that it, it improved their, their funnels in which that when you do that, you’re actually confusing your customer because what you’re telling them is that, listen, if I.

I can either buy this one time or I can subscribe. And so and so people, people are not dumb, right? They’ll be like, there must be a catch, right? They’re gonna get my credit card, they’re gonna bill me. I’m not gonna be able to cancel this. You know, so on and so forth. Right? So you’re kind of dangling this 30% discount or whatever with, you know, if you subscribe and that’s not what you wanna do, what you wanna do is you wanna build like a service, something that actually is a service in which subscription is just a feature of that service.

And the example I can tell is that if I came to your, you know, if I basically, if, let’s say at t or something, cut your internet every, every month, right? Right. And the other is, this sucks, man. It’s like, can you please like keep this on? Right. You know, you don’t wanna go to the at t store to buy internet again.

Right? So it’s a service that, that you are, you’ve now grown accustomed to and that you’re willing to pay because you know, you’re just using it. Right? So you’re happy to, to have a, a subscription billing. Uh, and so I think the challenge is creating that, right? And so with, with Kinko, we were trying to solve breakfast, which is highly episodic.

Uh, you know, every day. Most people, it’s the meal of the day. That seems to be more boring in a way, in the sense that there’s less rep change. Boring and mindless.

Andrew Warner: You do the same thing every day.

Ricardo Vice Santos: Exactly. You know, if I tell you, you know, you’re gonna to eat, you know, Thai curry every night, you’re gonna, you know, going call me crazy, right?

But, but, but it’s kind of easy to say, well, you’re gonna have this pro every single day, easier sell, right? And so, and so essentially that, that’s been my goal, right? To turn, I actually remember at Spotify, we, we always quoted this David, David Bowie quote, which was, you know, music’s gonna become like a, like running water, electricity.

So it’s the same thing, right? It’s a service, right? You, you open the tap, you like, you wanna have that tap where you open music, close music, right?

Andrew Warner: But Ricardo, what I’m trying to understand is if you are not immediately getting people to subscribe, instead you’re selling one-offs. In fact, actually, I think what I see.

Uh, you could, the user can buy one at one book or there’s a discount if they buy two. Mm-hmm. But you are not doing it as a subscription, and that’s intentional. You’re saying to the user, look, just buy one or two and then if you want, you come back. Right? Or am I missing the subscription option? You’re

Ricardo Vice Santos: No, that’s correct.

That’s correct. There’s a few reasons for that. I think we, we will probably add the subscription in the future. Uh, there’s a few things I wanna get in place before we get that. Uh, but, but there was a essentially, um, uh, I, I can maybe go into detail in a minute about it, but, uh, but the way you described is correct.

That’s exactly what it is. Right? And, and so what’s the difference

Andrew Warner: between that and the subscribe

Ricardo Vice Santos: and save, which you tell people not to create? Uh, no. I mean, but right now it’s not subscribe and save. Right. You pay the exact same thing, like if you buy another book. So we don’t have a subscribe and save plan.

What I’m trying to say is that you mix when you have the option, buy one time or subscribe and save, you’re, then that’s when you’re confusing your user, right? Because you’re saying, you’re, so, you say pick for them.

Andrew Warner: Let them pick one to buy or a subscription. Tell them how you work and don’t try to confuse them and make them think that you’re tricking them into, into a lifetime purchase.

Ricardo Vice Santos: Yeah, exactly. I mean, and it, it’s, by the way, it’s fine for you to have a subscription and tell people you can cancel any time, all these other things. Right. But when you mix the two things, you’re just adding confusion. And I see, I told you that I, I’ve advised people on this that, you know, there’s a, someone I know that’s, uh, actually a very sort of prominent, uh, influencer that, you know, has a supplement product, uh, where she had these two options, right?

And what I told her is that, listen, you’re telling people that they should consume this every day, but then you give them the option to buy one time, you’re, you’re actually confusing them, right? Mm-hmm. And I said, you know, and if you don’t trust me, they’ll just test it, right? Do an AP test, you know, have a funnel with subscription only, and the funnel where you give both choices and then see what the conversion does.

And lo and behold, the one, the subscription actually increased conversion ’cause it reduced that. Because again, you’re introducing this point of where there people are wondering, wait, what, what’s the catch here? Why is it cheaper to do it? You know, there, there’s something, there must be a catch. People are not dumb.

You know, they’re, they will think there’s a catch, right? And so I see, I think so I think that that’s what I’m a guess, right? You know, in my case, so I, you, you, your, your question is sharp, right? Uh, but, but I think that in this case it would apply if I would’ve both options. Right? Um, and so objectively, I don’t have subscriptions just yet, and, but I can tell you why.

The reason I don’t have it is because for one of the things that we’re, we’re, we’re working on is, and I should probably said this earlier, but right now, you know, the, the books that you can read, like I said, were essentially my stories, right? The stories I created for my child and that you can go and put your family in it.

Yeah. And over time we added, you know, the ability to, to customize it and people heavily customize it. Uh, but it still follows roughly the plot I set, you know, my myself, you know, because that was my. Aspiration for my child, now you’re gonna have a different one. And like you said, you’re gonna have other people have a religious one, uh, moral one, different things.

Right. And I think that the, the, uh, essentially we, you will only, I, I think we can only earn the right to become a service once we give you essentially full, um, full agency in a way. As in I see. Because, because then, you know, that becomes sort of the con the infinite content machine, right? As in that you essentially, you can sell, you can say, listen, you pay this much, and you get like, you know, three, four stories or, or maybe your bill based on how much you usage, but essentially create a service, you know, as opposed to a fixed set of products you can buy.

Right.

Andrew Warner: I see. I actually, that seems counterintuitive. I would’ve thought that it would be the opposite, that you would have to find a way to write the stories for them if it’s an infinite number of stories that are coming at them. Because if they have to every month log in and figure out what next month’s story is, you’re giving them a lot of work.

But if you’re picking it for them and they discover it with their kids, that’s more interesting.

Ricardo Vice Santos: Yeah, but they don’t have to. Right? I mean, the, the, uh, actually, I’m glad you asked that question, right? Or that you went there. Uh, essentially the, um, at the beginning when you come in, it’s, uh, the product is, uh, is what I’ll describe a pool product, right?

You come in and you pull the story out of the system, right? Yeah. And, and by the way, a lot of people say, well, but we’ll, people have the agency to do so. They, they don’t have to. I mean, when you come in and if you, if you tell me I have a 3-year-old, right? And, uh, or let’s say that I ask, uh, I, as in the agent ask you like, you know what, uh, you know, what do you wanna do?

Um, oh, how old is your old, is your kid? You know, do you have a sibling? So and so forth. You know, you can probably make an, an assumption of what the, you know, is it a boy, is it a girl? You can make an assumption that if it’s a boy probably likes firetruck if it’s a girl probably likes these things. Mm-hmm.

And then they just have to ask you like yes or no questions. Right? And then very quickly you kind of build sort of a memory of, of who this family is and what their values are and so on and so forth that then you can basically build on. Right. And you can, so it’s, it very quickly becomes a push thing as opposed to a pull thing.

So you’re right. You know, I, but I don’t expect people to come in to alter every single thing. Um, I, what I expect them is to do that the first time. Um, and then it essentially becomes a, becomes a push thing. You know, as in the episode two, I get three, I don’t need you to necessarily come in. Right. And, and particularly with, with the, sorry, just to say this, you know, it’s the same thing as, you know how people have this thing that, you know, Facebook’s listening to them because, you know, they, they thought about the kitchen knife and now there’s a kitchen knife and ad for a kitchen knife or something like that.

Yeah. So with, with kids, that’s even more predictable, right? I mean, the kids go through sort of, you know, sort of stages of growth. Uh, you know, for instance, my child, you know, just got a sibling, right? His first weeks were, were, um, you know, were very emotional. He dealt very well with this new sibling, but he also had these moments of tantrums, you know, for lack of attention and so on, so forth.

I mean, if you ask Shachi pt, the Shasu PT will tell, well, it’s very normal. It happens to every single kid. Well, no. Right. You know, it’s like, it’s, it’s the, the, so your family will have some sort of different flavors, uh, different, different mind, but they’re not that different. They’re a bit predictable in that, in that manner.

Right. So you can essentially combine I see. The predictability with the data that you provide me.

Andrew Warner: Okay. I get now where you’re going. Let’s take it away from you and go a little broader. What are some ideas that you might have for creating this customized, using AI product to sell people? Mm-hmm. Right.

You’re an investor, you give advice to a lot of other, other entrepreneurs. You’ve created companies yourself. If you wanted to come up with a simpler version of what you’re doing now, customized AI products, service book something, what would it be?

Ricardo Vice Santos: It’s a very, it’s actually, I really like the question. You know, I usually tell, tell friends that if I would’ve, a, a fund would literally be that personalized VC or something like that, you know, like essentially Chase, you know, chase ization opportunities.

I, you know, I can think of, there’s broad categories, right? You know, for instance, I, I think vibe coding in a way is that if I, coding is literally, you know, uh, uh, your own personal software developer building something personalized for you, um, you know, for internal tools or, you know, if you have a cafe, you can have software for your own, your own, uh, your own stuff for, uh, let me think, for consumers.

So it’s actually a hard question in the sense that, you know, the, I, in part I’m exploring the best idea I have in this space, right? So what would be the second best idea, I guess, that we can think?

Andrew Warner: You know, the thing that comes to mind for me is you’re thinking consumer. I wonder if there’s any kind of guidance that could happen in the business space.

For example, we talked about Zapier being the sponsor. Wade, the founder, likes the, what is it? The five dysfunctions of an organization, of a team. Mm-hmm. I forget the name of the book. Anyway, he likes it so much that that at the end of each internal meeting, everyone’s individual contribution to the meeting gets analyzed based on that book, using a Zap.

Mm-hmm. And then DM to them as a coach. Mm-hmm. He’s doing that manually for people using Zapier. I know that a lot of leaders have their own philosophies and books that they study that I think that they would want to somehow impart to their team. This is our approach. I want to, I want to guide you this way.

I want to support you that way. And that kind of customization is not a hard thing to create. Right. And so right now it might be something like a Zap that you sell individual team members that says, look, I’m gonna spend some time with the ceo. That’s, I’m gonna spend some time, right. Assembling all this, and everyone in the company is gonna be do, we’re going to find a way to give you coaching based on your meeting transcripts, which are shared within an organization and can be zapped in based on maybe your email, based on also the, the founder’s vision or the leader’s vision.

Right. There’s some kind of coaching. Yeah. That’s possible that way.

Ricardo Vice Santos: So we spoke about the, the Zapper tool, right? Another tool that I hadn’t contemplated, but I do use is that, are you familiar with Notebook LM from Google? Yes. Yeah. Yeah. So I kind of do similar thing to what you’re describing in, in that, you know, like I said, they have a large collection of, of books, and one thing I do with every book is I take highlights manually.

Uh, I like to read physical books, but then I actually copy the, the highlights. It, it’s, it’s my way of like, of. F you know, memorizing them to kind of copy the notes to, to, to my computer. So I have a big repository of them and I do dump them into Notebook lm to do exactly what you described, which is like, you know, I have this problem, you know, base notebook LM is great in that, you know, the base of truth is literally what the, the, the stuff you put inside, right?

The documents and as you can query, essentially can talk to your books in a way, right? Mm-hmm. It’s a way of, of do a way of thinking about it. So I think you’re right. I mean, another thing that people are doing is, you know, they, they create sort of boards of people they know. Let’s say you admire Charlie Munger, right?

You know, you can have, you can have a bot millet, Charlie Munger and give you advice or, or GPT and so forth. So for sure I think that could be, that could be, uh, that could be a thing, uh, a thing one could productize for sure, right? Um, but, and I think every, any one of these things that you can kind of hack together, uh, you know, with high intent is probably something that you could make as a product for, for a, a broader audience, like in, and essentially take it to 11, right?

So essentially make. Right far, far better than, than what you, what you do as a hack. So, so yes, I would probably user of that and as, uh, uh, an investor as well.

Andrew Warner: Yeah. And you could do it as a service at first. Right now, the, the tools are all available to connect together and customize it for companies, and then eventually the APIs are available.

The cost of doing it is not that much, and businesses would pay for, for coaching that helps bring people, so to speak, into the cult of the company, right? Mm-hmm.

Ricardo Vice Santos: That’s correct. And there’s a, and there’s a, you know, there’s a few tools that are trying to do this publicly, namely, uh, you know, uh, I forget, Delphi I think was one, right?

Where it’s trying to essentially create a chat bot basically

Andrew Warner: based on your approach.

Ricardo Vice Santos: Yeah. And you know what, what’s I think interesting about that, particularly if we think about what you described, which is like data that’s maybe not publicly available, is that. These models are becoming commodities, right?

I, I actually just did this ’cause I, I changed the, I changed from using GPT to Gemini and I was actually asset exercise to see how much I would miss it. And, you know, not much frankly, like, uh, it’s kind of felt like a commodity, right? Right. I used to put gas in one station, I do another one. But what’s, what’s actually valuable is the, is that, that the sort of the ground data, right?

That you, that you actually have, right? So again, the things I put into notebook, LM, for instance, are, are more valuable, right? And so, so managing context is actually a very valuable thing in this context, right? So in my case, you know, the context is the family graph, right? That’s what, that’s what I’m curating.

Um, and uh, and those models will come and go and, and as they come, actually our product tends to get better. Um, but the context is actually figuring out the context management, you know? So in that way you can think that all these companies are sort of memory companies in the way. Right, like a waste organized memory, uh, that then gets, uh, you know, there was actually a time where people said this every, every day, every company is a database company.

And I think, you know, there’s not, it’s not so different right now in that sense. You know,

Andrew Warner: what’s the thing, I, I heard that, uh, from, uh, Jim on your team, that you talk to customers every single day. Do you talk to one today? Um, I do. Yeah. I do talk to them every day. You did? Yes. Why? What are you trying to understand by talking with them?

Ricardo Vice Santos: Yeah. It’s something that, you know, I can’t say it’s always the, the advisable to do, but it’s something I’ve, I’ve, uh, I’ve always liked to do, to kind of feel, uh, you know, essentially if you get a feel for the product is a good way to, to think about how they are, how they’re approaching it, uh, and so on and so forth.

It’s also like, it’s, it’s also helps me be a better leader for, for the, the, the customer support and team, you know, in that function. So at Kinko was the same, was the exact same thing. And, and, uh, I had the. Fairly larger team than I, than I do now, and I will do the same thing, um, a, you know, for at least like an hour of the day or something like that with, with some people.

So what’d you learn today

Andrew Warner: or what’d you learn recently that you didn’t know? It feels like this business is pretty straightforward at this point.

Ricardo Vice Santos: I primarily, I built this product for myself, right. So I think it’s straightforward, right? Um, but, and you know, I’m also somewhat experienced, so I trust my judgment that I’m not disillusional about it.

But, you know, when you’re doing a product for, for families and you know, there’s always gonna be people that, that encounter issues that are not as, as easy as, as, as, as things are, are to you. Right? And so I, I constantly get ideas from, you know, from I can, I can tell I for sure, I had ideas that literally I would say that it, that customers gave it to me.

And, and I don’t say, I don’t mean it in the way that they intentionally suggested, made the suggestion objective that we implemented the suggestion. It’s, it’s more than once than that, right? I just notice a behavior that they were doing. Well, listen, everyone’s doing this, and it sometimes it even feels like a hack, right?

They’re trying to hack something together. It’s like, and oh, wait a minute, is this like something we should actually productize? Right? And, uh, there’s been a bunch of them, you know, that I’ve, that I’ve, uh, that I’ve noticed, uh, people, uh, uh, uh, you know, from doing that. And I think that ideally you would find yourself like head of customer support or something that has that sensibility.

Uh, but it can be difficult, right? And this is why, you know, even it’s known that, you know, even Jeff Bezos, the same thing, right? He would, I’m not saying you would sit and answer emails every day, but notoriously you would respond to customer emails and, and that would, that would prompt like, uh, changes within Amazon, uh, when you saw fit, right?

So, but you can’t think

Andrew Warner: of something specific that you had, that you had learned recently that changed the way you think about the business? Can you?

Ricardo Vice Santos: No. So I can tell you that the, the idea of, of, um, the idea when I started be noticing more. So when I, when I started this, you know, a lot of people told me that, you know, a lot of people, you know, people would not, wouldn’t want to create their own content, right?

And actually, like, I got this pull from customers quite a lot, right? I have a lot of emails to, to show you that the opposite people giving me, you know, from all sorts of agency people that write the entire thing from people to have just small ideas, right? And so that’s a very obvious one, right? You know, people that literally send me an entire script and even drawings of things they wanna do.

Um, but, you know, but other things are more minutia within the product, right? I’ve noticed that, you know, people doing one thing, for instance, people, if people really like one picture in one image, sometimes they would basically upload that picture as a source image is in, instead of uploading a picture of the child, they would upload the AI generated picture of the child, right?

And, and I, and I, you know, initially I thought it was, this is a bit dumb, you know, why are they doing this? But then I realized that. You know, you know what the, the problem I told you about the potato photos, like people putting pictures, and I was realized that, okay, it turns out that the AI generated photos actually high resolution and sort of clean Yes.

Of other things around. So I realized this is what people in AI would call synthetic data, right? So essentially they, they want to use synthetic data to train the models, right? And, and then we decided to try this and it turns out it actually works really well to increase the quality. Um, so essentially use the AI generated image to generate other images, right?

Andrew Warner: You know what I, I to, to my taste, I understand it. Why your co-founder has a photo of him, Phil does on some article somewhere that I saw and I go, this looks really good. I could see how he would want this to be his avatar. You know, and I get that that’s how people feel like you just made the best version of them, the person mm-hmm.

That they are, when they’re on the moon, the hair looks right. I don’t know what you’ve done, but you add some kind of light to them, you know? Mm-hmm. Like to make them radiate a little bit. And so I could see that they say, this is the best picture of my kid. I want you to just use that. That’s what I’m seeing you’re saying It’s also helping you train your model better.

How,

Ricardo Vice Santos: when I say synthetic data, it is actually more used in, in, in the concept of LLMs, right? But it’s like there’s this idea in LL that the, the, that LLMs get better by, essentially by having essentially more data, right? And so, yeah, of course we have the internet data, but at some point, you know, if you believe that that’s a constraint to make a model better, need more data, how do you find more data?

And so it sounds kind of bizarre, but there’s this idea of synthetic data, which is what if we get AI to generate data. That then models used to, to, to study. Exactly. So, so that’s what we call synthetic data. So data that was already created by an AI that is being fed to training an AI model. So in this case, you know what I’m saying is that, you know, the way dream stories works is that you upload the picture.

Um, um, or at, at back in the day, we’d let you upload multiple pictures. Now we only require one. Uh, and then basically we, uh, actually, in fact, this was why we stopped. We, we only, we started being able to only upload one was because we, we noticed people uploaded multiple pictures and then we generated like a preview of what this, this could, could be like.

Right? And then, and then sometimes people like the results, sometimes they don’t. But oftentimes we saw people not liking the, or not oftentimes when they didn’t like the result, we asked them, can you upload a new picture? And that’s when we noticed that a lot of the times people, the picture they uploaded was an AI generated photo that we did, that we generated, right?

And then we’re like, okay, that’s a bit strange, but why would they like to? And then we realized that the, what they were trying to do is that in their mind they thought that that meant I’m gonna, this picture is gonna be the picture used. ’cause you have to understand the customers. Probably think, you know, they don’t understand what AI is, so they probably think that you have someone sitting there cutting the, the photos, right.

And pasting the photo on the book. Right. And so, you know, as like, you know, that’s how they communicate. Can you use this photo on this page? Right? So they think you’re literally cutting. They don’t think that you’re recreating. So what they do, so that’s why they do it, right? They upload the, the AI generated photo, say, I want you to use this one.

So that’s, that didn’t mean, that’s not how we implemented it in the end, but we noticed that holy shit, people do this. And actually the resulting pictures tend to be very good. Yeah. And that’s, and that’s what I mean, right? That’s one phenomenon that they, that I only noticed by virtue of, of watching customers, uh, uh, you know, behave with the product, which ultimately meant that we were able to reduce the number of patients required to literally just one.

Okay. Because then we’re like, holy shit, you know, we can essentially just ask for one picture, generate a synthetic version of it, and then, and then say, do you like this? Great. So we generate a few more synthetic versions based on the same thing. So essentially create four photos that are essentially very, very similar.

We just like the different lighting and that tends to be enough to train the model to get consistent results. Right.

Andrew Warner: Okay. Alright. And then the other thing that I heard that you do is you’ve got some kind of agent for customer service. Can you tell me what that is then?

Ricardo Vice Santos: Uh, I mean, it, it’s quite public, right?

We use, we use Fin, there’s a few fin from Intercom. There’s a few out there. Yeah. Fin I think interesting that Intercom kind of pivoted the company into, into, into AI agents as opposed to, you know, a tool for customer support. It’s customer talk to people

Andrew Warner: about that. Apparently Intercom had a problem where mm-hmm.

People didn’t see them as an AI company. They saw them as like a human chat company and they needed a way to say, we are, we are this new thing. Don’t think of us as an old company, like live person. And so they came up with Thin and that did change the way that. That people perceive them, right? Yeah.

Ricardo Vice Santos: Yeah.

I think they did. Right? I mean, changing the name. Yeah. I think you can see the product that they have remnants. I’ve, I’ve actually used Intercom before, the first full AI pivot. You can see remnants of the old product in it. Right. Um, and, uh, in, but I think they did a good job. It’s, it’s hard to do this as a company, right.

You know, notoriously or famously Netflix did other two times moving their business model entirely to a complete new reality. Mm-hmm. I think Intercom is navigating the same. I think they’ve done a good job now. It’s not amazing, but, uh, you know, but it’s good. Uh, and uh, but it requires a lot of customization.

Right. And I would say that, um, I would, I would still recommend it, uh, but I think the value, you know, if you go out and try fi I think what you’re gonna be very disappointed is that it takes a lot of tailoring to get it to where you want it to be. And that’s actually one of the reasons why I like to do some level of customer support, literally.

Because it gives me a sense also of what. You know, of essentially where, where fin is also not doing such a great job. Mm-hmm. And how I can, you know, where is it hallucinating or why is it, why is it, uh, why is it giving wrong answers or why is it not able to perform something that should be able to perform so and so forth?

Right. So it, okay. So when I, so my point being that when I spend time doing customer support is also a bit with that lens, right? It is with the lens of like, what can I automate? Right? Rather than essentially hiring a lot more people to just cover the, you know, increasing sales. How can we just essentially automate this problem away?

Andrew Warner: Okay. And you’ve got something that you’re using, you would tell me you have an artifact, some document that you use to guide Finn to, to guide your customer support agent.

Ricardo Vice Santos: Yeah. I mean, I think most people have that, right? But it’s like, it’s a combination of escalation guides. It’s a, it’s a, it’s sort of documentation.

We, um, it’s, um, yeah. You know, of course it’s quite elaborate, right. That we spend, but I think that beyond the documentation, I think what’s most valuable to me is the sort of the. The, yeah. The guides, you know, which inform how, you know, oftentimes I will, you know, even when I’m doing it myself, you know, customer support, I ask chat, GPT or something, or Gini to help me answer things.

And oftentimes I don’t like the answers that it gives me as in that, you know, I found, you know, I, I think throughout all the companies I, I worked in basically, I usually, you know, customer support reported it to me. And so I always crafted like a, a specific persona in that I want my agents to, to impersonate, you know, agents mm-hmm.

Or human agents, whether humans or ai. I want them to have this persona. Uh, you know, there is not me. It’s not me at all. So I’m not responding to customer support. Right. I’m, I’m enacting like a customer support agent, right? Yeah. And so, and so I spent quite a lot of time and you know, again, initially I tried to tell that to my team and I have documents and saying how it is, but they kinda learn a bit my, by more by osmosis.

Right? And now I’m basically more trying to, how can I get the AI to believe like this character, you know.

Andrew Warner: Can I see the doc that you give AI to guide it for how to answer your questions?

Ricardo Vice Santos: Uh uh. Unfortunately I don’t have it that time to show you ’cause it’s actually in particular, in the context of fin it’s, it’s kind of broken down in many different days into lots of different things.

Okay. I can, yeah, I can probably send it to you. You know, I would’ve spent time so, uh, not polishing the document itself, but compiling it for you. Yeah. Okay.

Andrew Warner: Alright. I’d love to be able to show that if that’s available. And then, uh, the Zap, I’m assuming you don’t have that anymore because that was an old zap that you did for help a reporter out, or is it

Ricardo Vice Santos: No.

Yeah, it was, but actually when you mentioned, when, when we mentioned this thing, I actually thought, I, I I should do it. Actually, there’s no reason why that so it, so it should give me for not doing it. ’cause I think is particularly with the, with ai, because what I did was exactly how we described it. Right.

But the thing you can do with AI now for sure, is that you can actually get, you know, get it to. Pitch out essentially, right? Yeah. So essentially you, so it’s not just based on keywords, right? You can analyze the content and then essentially can, can compose even the, so I can basically be in Bali, like, or, you know, taking a nap and you, you could essentially, it’ll pitch on your

Andrew Warner: behalf.

Ricardo Vice Santos: Yeah. That’s, that’s usually my running joke in the team is like, you know, can we finally book the trip to Bali? Right? The point being that once it’s fully, like, that day never comes, right? But once it’s fully automated, we can finally, you know, uh, chill. You know, because everything is, is, uh, well oiled.

Andrew Warner: The cool thing about Zapier now is that you can just describe what you said to me, like mm-hmm.

You can talk to it. You say, I’m going to get you, uh, when, when an email comes in from this service, I want you to look for this type of company, and if it’s right, I want you to take this action and just draw, just talk to it. Give it a few minutes. It will create the whole automation for you. Mm-hmm. Then just run with it.

All right. I’d love to see it. So you’re saying I can give you shit if you don’t do it?

Ricardo Vice Santos: Yeah, you should. So you should, we should talk again in a weekend. If I didn’t do it, you know, then, uh, then I, then I’ve, I, I’m out of excuses. Right.

Andrew Warner: All right. Hell yeah. Thank you for doing this interview with me. I’m looking forward to that.

And we’ll publish it and we’ll, uh, we’ll have other people use the same zap. Share it. Share it, right? The nice thing about Zapier is that you can share your thing. Other people obviously are looking for different types of businesses, but it’ll be nice for them to see it. All right? Mm-hmm. Thank you. Thanks everyone.