Stickerbox founder Arun Gupta on building AI-powered screen-free kids toy that sold out in days

and CEO at Stickerbox. Happy Co. Uh the uh Creativity First AI product went viral over the break and I'm excited to talk to him. He is in the reream waiting room. Let's bring him in to the TVPN Ultra down in my room. How are you doing?

Good to have you. [clears throat]

Hey, great to be here.

Uh could you please uh kick us off since this is the first time on the show with an introduction on yourself and the company and just some of the background here?

Yeah, absolutely. So, my name is Arun. I'm the co-founder of Hapico along with my good friend Bob Whitney. Um, previously I was in Y Cominator back in 2009. Uh, same batch of Strike.

Amazing.

Yeah. I actually dropped out of Yale um in the middle of my junior year um and six months later was lucky enough to get in, talk to Paul Graham um on a weekly basis. It was amazing. Okay, let's let's actually just dig in there since we have some time. What were you building in 2009? What was the mood on the ground at YC in 2020 in 2009?

So, we were actually one of the very first hardware startups. Um, a lot of it was, you know, Reddit had just happened, Airbnb right before us. Um, you know, there were a lot of B2B stuff, a lot of consumer things. Uh, we built a personal sleep tracker. So, it was a wristband that you wore while you slept. Um, and it used uh sleep science called to track your sleep cycles and wake you up at the right time in the morning. So back when Fitbit and Draw Bone were like doing so early. So early

so early. [laughter] H I mean it must be had been painful to see Aura and and Whoop and basically have have like the right have the the right idea and and what what

Yeah. Yeah. What was the blocker? Was it supply chain or distribution? Like I mean I I went through YC in 2012 or first company, first idea, same thing where a lot of it like wound up getting built by someone else years later, but uh what was your experience like if you were to do the postmortem?

Honestly, for us it was big data. So back in 2009, everybody who worked on big data worked in biotech or worked on Wall Street. So you couldn't really hire them away unless you had a ton of money. And all the money is in data analysis, right? So like Aura, they're analyzing tons and tons of data points and giving you all this information about how you can live better, eat better, be more healthy, and we just didn't have that. So the business really flies when you have a subscription that works, but without big data analysis, it's really hard to build a subscription that works.

Yeah, that's crazy. I remember 2009 was like the dawn of Hadoop and like distributed systems that you had to orchestrate. It was like it was like doing a multi-raining LLM AI infrastructure. You couldn't just be like put all the data in one database and like Google handles it or AWS handles it or MongoDB handles it like you

databases became a thing.

Yeah. Yeah. All these we take them so much for granted now but back then it was such a pain to orchestrate anything. You're you're you're running like true infrastructure just to do like basic stuff by now by by by today's terms. Uh fascinating.

Yeah. So so so what led you to start this company? When did you start this company? What were you doing in the interim?

Yeah, so uh after my YC company um gracefully uh shut down, um I started another company called gra.com. Um and GRA was basically

You started Gra.

Wait, the fashion company.

That's so that's so cool. I didn't Somehow we do a lot of We don't always have time to do uh we don't always have time to to do research, [laughter] but on your printout Grail didn't make it in. Uh, I've used GRAS uh for a decade now.

That's amazing.

Uh, so

yeah, I started GRA in 2013 in my bedroom in San Francisco. Uh,

yeah.

Were you a collector of uh like fashion items? Yeah.

Big time. Big time.

What were you into back then? What was the first product?

Did you ever wear chrome hearts to an investor pitch? [laughter]

No, I I I never wore it to an investor pitch. Uh when I went to see the investors, I was wearing more Laura Biana, right? You know, that's the vibe.

And that's still like what, five years beforeh starts uh introducing it to the valley via the All-In podcast. That's that's fantastic.

Yeah, exactly. No, back in the day it was really about like work wear and like raw denim. Uh and then it became really big through collabs. So there was like an H&M Bond collab that was really big. There was the Yeezys that dropped. There was all the Nikes that dropped. And then Supreme obviously had their massive moment. Um, but no, I was really into like all the OG Raph Simmons stuff. Um, actually if you remember ASAP Rocky did a video, I forget what the song was called, like please don't touch my raft. And in that video, he's wearing there's like three of my personal collection pieces in that video.

That's crazy. What was so um what uh

uh what what was the what was the like the customer acquisition funnel at that time? And I'm interested to hear how I've heard a lot of founders talk about uh you know if they got started years ago they were maybe doing like uh Minecraft bots. A lot of people were doing like uh sniping for sneaker drops sneaker bots. Uh did that play a role? Was there an automated portion or was it mostly organic? Like walk me through some of the trends that were happening at the time with Grail. And then I promise we will get to sticker box. That's why I want to talk to you. But this is

it's the most exciting. Sticker box is the most exciting part. But yeah, um really it was all about community honestly. From day one it was about community and all the way through until the end. Um because we were like very heavily focused on men's versus women's. So it was about 7030 men's versus women's and at the time in 2013 people were like do men even wear clothes? And I'm like yes they're people obviously they wear clothes. Um but men's fashion was like not a thing at all. Um so we were able to sort of like ride that wave. Um, and see sort of like

riding the Instagram wave in some in some ways. Like that had to be be the biggest catalyst.

Yeah.

Yeah. I talked I used to talk a lot about how people when you were like, "Hey, what are you into?" people would be like, "Oh, I'm into like Nirvana or like Radio Head." And then when you're like, "What are you into?" They're like, "I'm into Noah or like Supreme or like Jill, something like that, right?" So, it's like your personal aesthetic became more defined by the clothes you wore because we became such a visual society through Instagram rather than the music you listen to. um which was more maybe of like a MySpace um you know OG uh type thing

and there and and you guys had uh people building real businesses on the platform. I I have uh my my uh oldest friend uh teamed up with a college buddy of mine un unrelated to me and started selling on GRA ended up building uh their own brand called No Maintenance that uh that that's now you know thriving. Uh but they got their start just reselling. That's how they made generated the revenue that they needed to actually do their own inventory run. So, uh,

made their own boutique. A lot of fashion designers, too. I think Reese Cooper probably uh, you know, started out on there's a there's a viewer in the chat that says, "I bought my first pair of Rick Owens Dunks from Grail. This is so cool." Salute.

That's awesome.

That's amazing. Okay,

that's like seven grand. [laughter]

Probably still holding them. So then, yeah, take me through the the idea to get back in the arena for the third time. Uh, when did you start the company? Was this post chat GPT moment? I mean, I I I I feel like we've talked about this in the show, like there is such an opportunity for interesting businesses here in the age of AI, but doing

and I was always saying everybody wanted to build these AI hardware devices for adults. I'm like, I have a phone. It is an AI hardware device. They don't market it that way, but I mean Apple kind of does, but I was like, I don't need another device that like helps me order Door Dash, right?

But my kid I don't want my kids to have I don't want my have my kids to have an iPad, but creating a product. I always thought like if you made a product that let just let kids ask questions.

Well, the Rabbit R1 the Rabbit R1 came out and they used Teenage Engineering. It's this beautiful orange package. Very cool. But a lot of the reviewers who were adults were like

just comparing it to their

comparing to my iPhone and and and and a and I never got around to actually buying one and giving it to my four-year-old. But I was thinking that like it feels like a great kids toy. And then there is still the narrative of like great technology starters toys. So I always thought it was a great market to go after. But what was your inspiration?

Yeah. So I can take zero credit um for creating the sticker box. Uh it was all my co-founder Robert Whitney. uh he is the original inventor um along with his son um who helped him basically co-develop the device uh and basically the story is pretty classic uh you know son was like hey I want a coloring book of tigers eating ice cream and he was like I don't have that but I can just print it off you know an image generator and then you can color it and then he's like okay great I'm going to print that and then color it and then he's like now I want lizards riding skateboards and then he prints lizards riding skateboards u and then you see this magic moment go off in the kid's head and they're like Wait, I can say anything. And like for us, like we're a little bit more jaded, I think. You know, we're like, "Oh, yeah, AI, LLM, it all works." For a kid, they're like, "I just came up with this idea, and not only do I see it, now I can hold it in my hand and color it."

Yeah. I was I was I was uh w with my three-year-old yesterday using an image model, and he said in his words, he was like, "It's like magic."

Yeah. No. And and and I think the physical aspect is is just a it's so important. uh when I I I use image generators with my son and I have AirPrint and I will print the photo even though I'm running out of ink left and right uh because the you know full page color photo it just goes out really quickly but it's amazing and it's way different than just being like oh here's a Gibli of your toys uh to be like here's a thing that you can put on your wall you can enjoy

physical representation is or or color it in like it's a coloring book now like there's so many cool things you can do on your water bottle But I think that's what my co-founder really understood is that this physical representation, these physical stickers are just what's so powerful for children. And it really is

such an imagination accelerant. You know, it's like these ideas, they they're in my head, but then once they come out of my head onto paper, he was like, "Wow, this really transforms what they're able to do."

And there's two there's two sides to this. There's like the cynical be a good parent, no screen time, uh like print stuff out. that's better from like fighting the iPad cake thing. And that's true. There's a lot of evidence of that. That's a good thing. But it come you come to it with like a negativity of like I'm avoiding a bad outcome. So I feel like I'm taking my vegetables. But I just think that like there there are there's like a lindy like beautiful thing about tech texture and physical things.

Yeah. We we have like an old uh like it's not a Polaroid camera, but it's this really

bad digital camera that then will print out kind of pictures on it. So I what one thing I like about sticker box is it's not like these kind of products haven't existed. It's just like integrating new technology into it and it's a form factor

and that's the right like my co-founder brought the product to my house. He had already built it. He had in a cardboard box.

It looked awesome and I look at it in one second and I'm like it's all about how you put this together, right? It's like not like these things didn't exist. It's all about your product sense and basically like how you intuitively were like this is the right kind of product for a kid. Um autonomous self, you know, serve stickers, lasting physical pieces. Uh it's pretty magical stuff. But we like to say, you know, that the um it's an AI product and like you said, AI for kids and AI hardware for kids is amazing, more magical. But we like to say that the AI itself disappears behind the magic of the device. Because it's really not about the AI, it's about the idea. Like the idea is the hero, the sticker is the hero. It's not the device or the screen that are the hero. It's the kid's imagination that becomes the hero of the story.

So talk about the prototype. It sounds like your co-founder did all the hard work. He's the workhorse. You're the show pony apparently. [laughter] We'll give him all the credit in the world.

But it but but it it seems like he was a tinkerer, hacker, was able to whip something up pretty quickly, I imagine. Like Arduino, Raspberry Pi, like are there wires everywhere? And then and then do you go to a co-acker or a manufacturer? Uh is this something that's like pretty turnkey and there's a solid ecosystem around or was it a long uh multi-year process to actually get the first one made with a high quality? It was difficult, you know, because like these things are thermal printers, right? So, he came with like the idea fully formed and like a lot of it already baked, but to turn from a prototype to a production product that, you know,

our Christmas load was insane. You know, just the amount of kids unboxing on Christmas and then hitting the servers like he was stressed

like, you know, rightly so. Um, but to get there, it's difficult because when you think about it, this thermal printer, it's meant to print receipts. It's meant to print text on receipts. It's not meant to print images. And if you look at the fidelity of the images that's being happening, it's like I think something like 300 dpi like on the image, it looks beautiful, you know, and like that is sort of like a little bit of our secret sauce is, hey, we're able to make these images look really great through a combination of many different things. Um, but that's something that he was able to do really, really well. Um, and something that we had to work with uh custom honestly with the factories um to try to get them to to make it happen. So,

And you're sold out. You're sold out. I I just bought one while you were uh uh here with us. Thank thank uh thanks to Shopify. It makes it really easy to check out, but uh it's going to get delivered in February. It's been selling like crazy. You guys are just trying to keep this thing in in stock.

Yeah. Um so we had a ton of units available for Christmas shipping. Those went in like two weeks or less. Um and then we just kept selling. Like we had the December 28th ship date. Those sold out in like a few days. The January 6th ship date sold out in a few days. Then we had a February 17th trip date and I was like February 17th. It's like not even um it's like like a month and a half away. Those sold those all sold out. Now we're in February 24th. So can't keep this thing in stock. Um but we're uh we're ramping up manufacturing. We're going to do our best so everybody can get one. So, is the 2026 strategy really get solid on the supply chain so you can scale a ton and then are you focused on like ruthless ramping of digital ads or do are you more likely to go let's get Walmart, let's get Target because I can imagine it's selling really well on both but you probably don't want to do both as an early stage startup at the same time. You want to be really have your ducks in the row in a row, right?

Yeah. It's sort of a classic trap that startups get into where Target is like, "Yeah, we'll buy a million units or we'll buy 200,000 units." And then they sit on the shelves because you don't have the marketing to support them. So, you kind of have to ramp up appropriately. Um, so for us right now really, it's like we've got this core image to sticker generation loop down cold. Um, where do we go from there? You know, kids are making characters. Like some kids have this like stickman and Cappy Barrett character and they go on adventures over and over again. So, like let's name that character and then reuse it. Let's let them send that character to their friends box. Let's let them upload their own likeness into the box and say, "Hey, let me do me and my mom on a picnic in the moon or like me riding a dinosaur or what have you." And like basically what these kids are doing is what we feel like is we're basically enabling the next generation of storytellers, right? the next Spielberg, the next Ryan Cougler, the next Miyazaki. They're growing up with the cigar box as their like imagination accelerant and it's letting them tell these like little graphic novels or little stories where they're reusing these characters and they're basically developing their own IP. Um, so leaning more into that and creating more features for them. And honestly, I think the other thing that we are just so happy about is how it brings families together cuz it's for kids obviously, but like it's really fun to use, you know? And like the nice thing is is like when it's like I'm like at the thing at the table, you know, and like the dad is like in the background and like hovering and then he's like, you know, let me do one. And then like, you know, the uncle or the mom or the cousin are like all coming together and they're like, oh, you did that. Let me try to remix your dragon and make it like a metal dragon. Let me make your metal dragon breathe fire. Let me make it breathe fire in a volcano. Um, so it really brings everybody together. So, I think focusing more on these sort of like whole family experiences as well as the single player sort of like experience with the individual kid as well is something that we're uh my co-founder is really focused on. Um it's something he really believes passionately in is uh using to bring people together and as a really positive activity for parents and kids uh to do.

What have you guys done on the on the p like on on the actual material side? I know you know people are very afraid of receipts. I don't I personally not

put out a post about seats. I saw you guys have avoided like BPA and BPS, but maybe talk about some of the decision-m there.

Yeah,

this was another thing that was really difficult. Um, so BPA and BPS are uh obviously a hot button topic, but if you produce paper, thermal paper uh and you test it, often there are trace amounts of BPA and BPS in the paper. Uh, and for us that wasn't good enough. You know, I was like, just because it's below the California Proposition 65 limit doesn't mean that it's, you know, BPA or BPA. Completely free.

Are not safety.

Yeah. Yeah.

Yeah. Exactly. And for us, it's like it's a kids product. It's safety from the ground up. Everything from the images that are generated to everything that happens with the device. So, you can touch all these stickers. There is no BPA, no BPS. We tested in batches from the factory. It was really difficult to do. It actually delayed our launch by a full four or five weeks. We could have launched four or five weeks earlier if we were willing to accept sort of like this trace element of BPA uh BPS in the paper, but we were like, "No, we like refused to release this."

That's going to be a good that's going to be a good angle because I imagine you guys are already getting copied by a bunch of different manufacturers.

Knockoffs.

And it's like, yeah, you can use the knockoffs, but are you there's you know the kids. Uh talk about safety with regard to image generation. Obviously, some of these engineers go crazy.

Get just fair warning.

Put that dragon in a bikini. Uh, let's not have that happen. Uh, we want the dragon breathing fire. I want to know about what models you like, how model agnostic you are, because I imagine that if you even if you just change the model to something better, even if the style looks better, you're like, well, that doesn't look like the dragon I'm used to generating. So, there's some sort of like lock in there. At the same time, you might be optimizing for cost. How are you thinking about the software side? Yeah, I think you're hitting the nail on the head there. Um, and for us it's like it's really a product for kids, right? And I think Chat GBT when you do image generation, it wants to make everything look sexy. Um, that's kind of the bias, right? You even upload your own photo. It's like it wants to make you look super buff like you're a superhero.

Well, that's because I put that in my in my prompt by default. It has to or else it's it's a But yes, it looks very like polished, refined, smooth skin. It looks it's a glow up filter. There's something going on there. Uh, No, for us the systems matter.

Safety is number one, you know. So, it's a it's basically a multiffactor system. So, we have a patent uh for the sticker box device itself, but we're actually also working on a patent right now for safety in the sticker box system. And it goes all the way through from the text all the way through to after the image is generated. So, when you say the text, we analyze the text for safety. We remove bad words. Uh we analyze the prompt for, you know, how likely it is to generate something safe or unsafe. Then we go through an enhancement process where the sticker is steered towards being something geared for a child. So it's actually pre-written in some different ways. Um, in order to do that, then we generate the image. And I'm oversimplifying a little bit here. And then even after we generate the image, then there's post image generation checks. And those post image generation checks are on

I don't even know how many variables, right? There's uh, you know, all the sorts of bad things like hate speech and nudity and violence and, you know, all sorts of different things. And we basically have all these confidence intervals and scores that go into it. But it's this huge system like you think about airplanes, right? And airplanes are like this complicated system of like many many many checks overlapping with each other. And that's what creates like the robust safety system. Uh it's a similar system here, right? Where we have many many many checks at basically every area of the process because the last thing we want um is for a kid to get something that's unsafe or for a parent to question it. Like for us, our big mission is what if AI were built for kids? And I think that the boogeyman in our heads is that social media was not built for kids. Social media was built for adults.

Was not built for kids.

Grock was not built for kids. I mean, AI is not built for kids. It's built for businesses, right? It's built for adults and it's built for businesses. But for us, we're like, what if we could just reimagine AI completely and just say, okay, what if this was built for kids? And that's why we didn't put a chatbot in a teddy bear. That's why we made a sticker image generator, right? Which completely different direction.

So, talk about uh latency with all those checks. Uh I imagine none of this is happening on the device. You're going to the Wi-Fi to the cloud to your servers. Uh hydrating the text with an LLM, that's not super fast. Then generating an image isn't super fast. Then sending that back down to the box and printing it, that's not super fast. Uh how fast do you think you can get this? Do you want to do it on device at some point? Um, do you how and and then what user experience sort of tricks are you using to make the latency and the lag tolerable?

Yeah. Um, so great questions. Um, it's actually really fast. Um, mostly because of the work that my co-founder Bob and his great engineering team have done. So I think the top

Yeah, Bob is

Bob Whitney. Yeah,

Bob the Builder. Let's go. Incredible. Incredible nominative determinism. [laughter]

I wonder who you're going.

I mean, there's not a better moniker, honestly.

I think the sticker I might be I might be off by a couple a second here or two, but I think time to sticker is like six to eight seconds. So, you push the get the sticker in like six to eight seconds and that's with all of the content moderation, everything you do there. But one of the nice tricks that you mentioned, right?

Oh, yeah. And because you're not doing like chatgbt level HD 4K, it's probably faster to actually generate the image.

Exactly.

But I think the other thing that you said that's really interesting is once you say it, right? You say it and then it hears it and then it displays your text on the screen and that's a nice moment that's an intermediary moment that makes it feel not as slow, right? Or it makes it feel a lot faster because you're like, I just said this thing. Whoa, there it is on the screen. I'm reading it back. the words and then you're probably

and then there's a pregnant pause basically and then the image comes in. But that moment between saying what you have seeing it on the screen and that like like just slight delay to when the image comes out cuz it appears on the screen and then prints basically at the same time. It's like a lot of suspense and it's basically like relief that happens kind of in your head, right? Like it's like when a joke lands, you know? It's like you have something happening happening happening and then there's the punch line. So there's this sort of tension that's built up like in the process and then it lands and it's just it feels really good because you're like, "Oh my god, I said that and now it's real."

That's very cool.

Uh you have a background at Grailed. Can we expect any sticker box fashion [laughter] partnerships? I want the Bayga weave version of the sticker box. I want the chrome hearts. I want, you know, [laughter] full

full.

I want something I can, you know, leave on on the table. Yeah. Well, I mean I I am interested like do you do you see yourself more uh expanding vertically or horizontally? Do you think you might be like a multi-product toy company or more of like a camera and printer company that maybe uh winds up making a device that's delightful for an analog photographer who goes on photography trips and wants to print things quickly? Uh do you see yourself focusing on a particular demographic over time or expanding one way or the other? Yeah, I think right now we're really focused on AI for kids. So, like what does that look? And in terms of like multi-devices, we obviously have tons of ideas, you know, um it's a really fun space to be in, but I think that there's just so much gold inside of this creation machine. Yeah. Right. So, it's like kids maybe not might not be the best illustrators. They may not be able to be animators, right? They might be able to use Photoshop or or whatever super well, but they can still bring their ideas to life with the AI. They can still print them. So, I think focusing more on creativity and investing more in the sticker box is definitely where we're going for the time being.

3D printer.

Do a do a do a two billion do a $2 billion training run

for your own model?

Yeahve cost that much. Give them two billion. Do you think you'll wind up fine-tuning and running a custom model or some open- source uh like fine-tuned on some optimized hardware or or the cost doesn't matter? Definitely this year my uh my co-founder Bob is very interested in creating our own model. So definitely from the ground up on our own model with our own training data because when you talk about safety that is the number one holy grail you know not to put a pun on it. Um like you want the training data to be completely safe and that way the model can't generate anything that's totally

and that's what really matters. And then in terms of in terms of like the business model, obviously it feels like razor and blade model because you're buying the box, but then you're also buying ink and paper and that's how you make money or do you have a subscription as well?

Yeah. So there's no ink, it's all thermal printing, so it's all just black and white. Um, so there are the paper refills. Um, but we want people to use the device, so we're kind of just pushing as much paper at people as possible. It's uh

very very cheap. Um, or like you know, we're not we're not trying to make a lot of money off of it right now. We want people to use the device. Um, so you get uh let's see, 200 stickers for $6. Um, which is pretty good. Um, and free shipping on all of those stickers for everybody who's watching. So if you buy stickers off our site, it ships for free and you can buy as many as you want.

Um, and then yeah, it we're trying to put more features in. So like let's say like you have a box and I have a box. I send you my tiger character, my tiger astronaut, and you use my tiger astronaut in your box and maybe we make a story together and my tiger astronaut and your wizard go into the castle and then we go underwater or we do these things. So like basically creating more features for people to create together for them to create more with each other and just do more sort of like more ideas and more um more more intellectual property and more content basically creation sort of

Ken in the X chat says what about licensing deals would would you ever do

we're joking about Supreme and Chrome parts but Disney and stuff that's

they've got a year one email you okay

yeah one year exclusive

arnappo.com stickerbox.com.

Yeah. I mean I mean uh the the uh uh what's it called? The the Yodo player has a bunch of uh cards that you can buy that have specific stories, licensed uh music, licensed uh short stories that are read aloud. And I could see some sort of partnership there being super lucrative where you could buy this pack and then be able to generate this type of character. That sounds

that's exactly what we're thinking. Sticker packs of parent companion app. You unlock Teen Mut Ninja Turtles. You unlock Hello Kitty. you unlock

whatever you want. Um, and then honestly, let kids create their own stuff, right? If kids create their own dinosaur superhero, let them sell that sticker pack, right? And then other kids can download it as well.

That's exciting.

But yeah, licensing it would be a huge play because it's super fun.

I want to see Sticker Box Millionaires, eight-year-olds [laughter] that just

I mean, Gra had this ecosystem, too. Uh, fundraising news. I want to hit the gong. Give us the fundraising news.

Oh, yeah. Yeah. So, we raised $7 million. Did that go

for the kids?

For the kids.

I'm so glad.

Always for the kids. We do it for the kids. But yeah, $7 million. Uh Maveron, who is behind Love Every which is another Monasuri kids product. They were investor. Um and then Beth Ferrer, sorry, we got we had Jerry Lou from Maver. And then we got Beth Ferrer from Serena Ventures. Serena Williams fund also put in a significant amount of money into our round. Uh

do you have kids yet or is Bob carrying all the weight there? [laughter]

Uh yeah, Bob for now Bob is but give me six months.

Six months. Let's go. Let's go.

Breaking news here.

Let's go.

Yeah. Seriously, my wife I'm sure my wife is watching. [laughter]

It's amazing. Well, thank you so much for uh

Yeah. So great to meet you. I'm excited. I can't wait to get mine.

What a white pill and uh and good luck with this year. Sure. It sounds like it's going to be a massive one and uh please come back on the show when there's more news. We'd love to chat.

I really appreciate that. We're trying to be a positive force for good. Uh you know, there's a lot of negativity around AI and I think for good reasons in a lot of cases, but for us it's really um

use as much use as much water in the data science [laughter] as you want.

I'm happy with my electricity bill going up because

for this my electricity bill can go up.