DylanFox

LubaDylan, thank you so much for coming on the show.Link

DylanYeah, thanks for having me here. It's fun.Link

LubaI'm so excited to talk to you about, obviously, AssemblyAI, your start as a self-taught engineer, and so many more things about what it's been like to build a company. But I want to start big, with big flashy numbers. You guys have a million developers, 100 million API calls a day, over 3 million hours of audio a day, which is insane.Link

DylanYeah.Link

LubaIt's been 9 years in for you. How does it feel?Link

DylanYou know, yeah, sometimes when I think about how long I've been working on the company and building the company, it's like, "Wow, I can't believe it's been that long." But there's really been these, like, different chapters of the company. And so I would say, like, for us, we really started to accelerate our growth and make a lot of progress over the last, like, 2 and 1/2, 3 years. So I started the company in 2017. We were in Y Combinator's, like, very first AI batch. There were like 10 of us.Link

LubaAnd I think it was the only AI-only batch.Link

DylanYeah, they had, like, one. I think they maybe did it again after that, but it was, like, super early days. But to give you a sense of, like, the first chapter, I mean, I started the company, started working on the company, in 2017. We didn't raise our Series A until January of 2022.Link

LubaOh, wow. So, 5 years until Series A.Link

DylanUntil we did our Series A, because...Link

LubaPeople would have said, not going to make it, right?Link

DylanYeah, I mean, you know, for me it was like I really just had conviction that, like, a company like this could do really well. And it was also just a really fun thing to be working on. But our whole product vision from the beginning was, like, building voice AI infrastructure for other companies and builders to build on top of. And the market for that was, like, very small back then.Link

LubaIt didn't really exist.Link

DylanNo, it didn't really exist. And when we look at, you know, the, like, TAM numbers or whatever, you know, over the last 10 years, the, like, market for voice AI has increased by 10x and then it's going to increase another 10x over the next 10 years. And so we're in this really big expansionary market. But in the early days, in the first chapter, I mean, there were no LLMs. Remote work wasn't a thing, so there were no, like, online meetings.Link

LubaYeah. No Granola.Link

DylanThere was no, like, 5G mobile internet. So the amount of audio, the amount of, like, voice data that was flowing over the internet, the amount of AI applications that existed, you know, were in their infancy. And so it was, like, after COVID where that started to change.Link

And so the first chapter was actually, like, in many ways the hardest, because we weren't making a lot of progress. And it was me and, like, the first, like, four people, you know, that worked at the company. And it was hard because there wasn't really a market. But then over the last, like, two and a half, three years is when things have started to explode. So it's been a lot more fun. And the scale of what we're building has just really, like, kind of continued to reach new heights.Link

And I would say, like, broader, that's what I see. I think, like, right after ChatGPT, you had a ton of AI applications start to take off and get really popular. But it's really only been in the last 18 months that AI infrastructure companies like Assembly have started to take off, because the amount of AI applications that are scaling now, that, you know, are building on platforms like ours, are growing exponentially. Whereas, like, the first movers back in 2023 or '24, you know, got a lot of the kind of early adoption from end users. So there's been these distinct chapters of the company. And so while it's been a long time, I feel like the last two, three years have been, like, a totally different journey for us.Link

LubaAnd I want to spend more time talking about how you reached conviction in that beginning journey of starting AssemblyAI. But let's start with the big picture. Paint the picture to me now. What are you building?Link

DylanSo, you know, our product vision is to build voice AI infrastructure that other companies and really any builder can build on top of. So today, you know, companies like Granola, consumer apps like Tolan...Link

LubaMhm.Link

Dylan...and huge enterprises are building voice agents and apps on Assembly. So agents that you can talk to, or ambient agents that will help doctors take notes. So we have a customer, a company building on the platform called Athelas, Commure. They build AI scribes and other AI products for healthcare. And they have a scribe product that will help doctors take notes while they're visiting patients, update electronic health records after. So really agents and apps that are voice-focused, we help companies build those on our platform.Link

And so we do everything from building amazing models for voice. We have the best inference stack for voice models at scale. You know, over 100 million API calls a day. Almost 3 million hours a day of audio, and that's growing at, like, a really fast rate. You know, we basically put all that together on a single infrastructure platform just to make it really easy for builders to build voice agents and apps on top of.Link

LubaAnd how is what you guys are working on and doing different than what OpenAI or Google is doing?Link

DylanYeah, so we're really focused on voice and all of the use cases for voice. So, you know, it's this interesting thing that I've realized building an AI company, that, like, it's kind of easy to get 90% of the way there...Link

LubaMhm.Link

Dylan...with an AI product. But the last 10% is what creates product-market fit.Link

LubaRight. And what makes the product magical.Link

DylanYeah, exactly. And this is true if you're building an agent or AI feature or a company like Assembly. And so what I mean by that is, like, I'll give you an example: transcription models for voice agents. One thing that most of those models do a really poor job on is accurately transcribing, like, a really short utterance that a user will say. So if they'll say, like, "yes" or "mhm" or something. And if you don't get that correct, the whole agent experience, that you're talking to an agent about, will fail. And we have, like, tons of emails around, like, that specific thing. And that's one of, like, hundreds of things that we look at. It's similar on, you know, generating speech, on building an actual voice agent. And then having all that work at scale with super low latency. And so we really go deep.Link

And it's why customers pick Assembly. So, you know, a big part of that is our models are really good. But then they can run at huge scale with low latency. So they can be deployed at huge scale. We have a ton of different APIs and features to help customers really build voice agents and apps effortlessly. And we're constantly innovating there. And so, you know, ultimately, like, when I see companies pick to build on Assembly, it's, you know, they realize our models are the best. They realize we have the best infrastructure. We have the best platform to build on that will scale. And that ultimately helps them build a successful product.Link

And I think the best stat... you know, I can point to, like, benchmarks and scale and stuff. But the best stat, I think, that demonstrates this is, like, our net dollar retention is around 180%.Link

LubaOh, wow.Link

DylanAnd so...Link

LubaThat's crazy.Link

DylanAnd that's all usage.Link

LubaCompared to, let's say, competitors?Link

DylanYou know, I'm not sure. I...Link

LubaYou don't watch them.Link

DylanI would imagine... Well, so the unique thing about us is, like, it's all pay-as-you-go, usage-based. So when you use Assembly, you only pay for what you use. We don't have committed contracts, because we built this inference stack that gives customers the ability to scale up and down without needing to commit to compute or usage or turn servers online. And so when we look at our net dollar retention, that's 100%, like, live ARR that is real usage. It's not, "Hey, we upsold this customer on a bigger contract and they've got to grow into that and we're going to put, like, 10 CSMs on it." It's all real usage.Link

LubaThat's very impressive.Link

DylanAnd I think that's the best evidence of our infrastructure helping our customers and builders to build successful agents and apps with voice. And it's why, you know, of the million developers on the platform, over a million now, I think something like 40% of those signed up last year. I mean, it's all accelerating a lot, and it's, you know, we just really go deep on...Link

LubaOn the quality of the products and all.Link

DylanYeah, and the core things that make voice agents and apps successful. And builders realize that, and, you know, ultimately, I think that's why they pick us.Link

LubaI'm obsessed with voice. I feel like voice has unlocked so much for me personally, because as many times as I've tried to, you know, be a writer and, like, put my thoughts down in a very concise way on a piece of paper, or digital piece of paper... like, yeah, it works, but I'm more of a talk-to-the-computer type of person, or talk to a person, than write to a person. So you just made me so much more productive. Even, like, my recent thing is I voice-to-text kind of, like, a journal entry about my day to my Claude, or whatever AI assistant I'm using that day, to then analyze patterns. How do you use voice personally?Link

DylanWell, I mean, it's so interesting, right? Like, so, you know, we were talking about earlier, like, I have a Tesla and they put the Grok voice engine in there.Link

LubaI haven't tried it yet in the Tesla.Link

DylanAnd I think that actually is, like, a really good example of, like, voice pre-AI and voice post-AI. So you can hit the mic button in the Tesla and you can make a voice command. You know, navigate to here or whatever. And it doesn't understand you that well. It's kind of slow. You have to be very specific on what you're saying. Like, it's not smart. But if you hold the button, you can talk to Grok, which is an AI voice agent. And you can have a long conversation. You know, it understands you. You don't have to think about what you're saying. And it's a totally different experience. And that experience is so much better that you all of a sudden, like, build a habit around, oh, I'm actually going to rely on this thing and use it. Whereas the voice command feature, you press the button and, again, it, like, doesn't work that well.Link

LubaIt's like using Siri.Link

DylanYeah, you don't adopt it and it doesn't become sticky. And so voice has been a technology that's existed for decades. Like, you can call into, like, pretty much every customer service line you call into, you're going to have some crappy voice AI that you're talking to that was, you know, some legacy IVR thing. But there's been this huge shift in the last 18 months where the AI systems you can build around voice, if you're using the latest tech, are way better. And that creates these new user experiences that are possible for the first time.Link

I think of a customer of ours, like, they're called Siro. They're in New York. I don't know if you've heard of them.Link

LubaNo.Link

DylanBut they have an app where if you're an electrician or a field sales tech, a roofer, and you're going to a customer's house to give them a quote, the AI can take notes for you, give you feedback after the visit, and it's really helping these field sales techs be more productive, get coaching. They're earning... Jake, the founder, mentioned, I think, something like 20K more commission per year or something as a result of the help they're getting.Link

And that type of app wouldn't have been possible a couple years ago, because you think about the AI: it has to understand, like, who's the customer, who's the tech, and while you're walking around the house in this noisy, complex environment, the phone's in your pocket or being jumbled around or whatever. And so there really are these new use cases and new experiences that are possible for the first time ever with modern voice AI capabilities, and that's what we're constantly pushing to relentlessly unlock.Link

LubaWhat about you personally? How do you use your voice?Link

DylanFor me, like, I mean, we obviously, like, we have note-takers in all of our company meetings. When we're meeting with customers, all that information is context that other AI agents at the company can use, but also it creates more transparency across the company. I use voice input into my computer to send messages, to work with different applications. I use it in the car.Link

LubaWhat do you talk to Grok about when you drive the car?Link

DylanWell, it's mostly questions about the car. Like, "Hey, how does this work?" But I, you know, I've started using, like, GPT Live. OpenAI's new real-time bidirectional voice model's amazing, and when you're talking to ChatGPT now, it's like... it's really good.Link

LubaLike talking to a friend, yeah.Link

DylanYeah, it's really good. And there's still some rough edges around, like, how many people are speaking, and it can go off and...Link

LubaIt'll want to interrupt you too much.Link

DylanYeah, but it's amazing. And so I'm really adopting these across, like, personal life, how I work. And, you know, we were talking about this earlier. I have two kids, two small kids. And there's now these toys that are coming out that are voice-controlled that are cool. So, like, there's a toy called Stickerbox. You can press a button and describe the sticker you want and it will make the sticker and print it out. There's, like, a stuffed animal one of our customers built, and you can squeeze it and talk to it. We have one in our office. I'll show you.Link

And so I think for consumer, especially for kids, there's going to be some really cool new hardware that is going to come to market soon that I think will be really good for screen time for... yeah, I said it's adults, but also kids. So instead of playing with an iPad, you can actually interact with a toy and talk to it, and be in the world and, you know, present with other people. So those types of consumer experiences are really cool.Link

LubaUnless you're just interacting with your AI and you're not actually present. But maybe you're learning something from the AI.Link

DylanYeah, yeah. Well, you know, I think that's the thing about, like, voice: it's more of, like, a communal experience, because you're talking to it. So it's out loud.Link

LubaYeah, right. So other people can interact and participate as well.Link

DylanYeah, exactly. It's less of an isolated experience.Link

LubaSo in a couple of years, I don't even want to say 5 years or 10 years, cuz who knows where we're going to be in 5 or 10 years. What is the world that you imagine with voice?Link

DylanHardware and software: voice is going to be, like, a primary interface to all hardware and software. So I don't think it will, like, necessarily replace the keyboard or mouse or touch screen, but it will be another... you know, it's like you have the keyboard and mouse and you also have the touch screen, and I think you'll also have the ability to have voice input. So sometimes if you want to sit and type, you're still going to want to do that. But other times you want to talk to a computer and command it.Link

And I think, especially for, like... there's so many times where I need to order something. You know, I'm fiddling with these apps on my phone. It takes, like, 20 minutes to click through these buttons on my phone or whatever. And I want to be able to just talk to my phone and say, like, "Hey, I need to order this thing. Can you just, like, go get it done?" And those types of experiences will happen, and I'm excited for that, because I really think it will kind of free us from screen time.Link

LubaDo you think we'll live in a world where we have, like, the equivalent of Amazon Echoes everywhere in our rooms, and then the voice will be listening and then you can just, you know, order DoorDash or whatever?Link

DylanYeah, and I think also, I mean, you know, I think about, like, TVs, right? Like, I think you'll be able to just say to your TV, like, "Hey, you know, turn on the latest episode of whatever." And you won't have to wake it up. It will just be smart enough to understand, like, you're talking to it, and your voice versus someone else's. So I think you'll have, for sure, like, these new voice-first devices, but I think voice will also be integrated into hardware and software. Even things like a coffee machine.Link

LubaYeah.Link

DylanYou know, we have this complicated coffee machine in our office and...Link

LubaHave to read the instruction manual.Link

DylanYeah, no one knows how to use it, right? And so soon you'll be able to just go up and talk to it and say, "Hey, you know, make me this type of coffee."Link

LubaIt's so much more human to use voice than, you know, press the buttons or use a keyboard. Obviously, yeah, keyboard might still stick around.Link

DylanI think of that analogy because, you know, if you think about a coffee machine, first it was, like, you know, physical buttons you'd press, and now they have touch screens. But next they'll also have the ability to just press a button and you can talk. So I think it will, you know, it will be a primary interface into all software and hardware. And that's what we're building for, to unlock. And there's still a lot to do, and, especially for consumers, stuff has to be, you know, super low cost and it has to be running on the edge for a lot of use cases. So there's still a lot to do and build.Link

LubaYeah, that sounds like a really exciting future. I'm super excited for voice. I want to talk about your background and really understand where your conviction on Assembly and voice came from. Tell me more about what you were like as a kid. Like, if I were to meet Dylan who's 10 years old, what were his hopes and dreams?Link

DylanWell, 10-year-old Dylan was just living in my basement playing, like, World of Warcraft. So, you know, I was always, like, into computers, video games, and...Link

LubaAnd was someone in your family...Link

DylanMy brother was too. So, like, my older brother, he would, like, build the computers that we had. He would, like, order all the parts and, like, build them. And I think I, like, watched him do that, but then I also...Link

LubaAge difference?Link

DylanHe's 5 years older. Yeah. And so I also just, like, got really into video games, and then through that, like, building websites for, like, my teams for, like, Counter-Strike or whatever. Being on IRC. And so I was just kind of always on computers and really into, like, computers and technology. And then that's really always kind of stuck with me, and just this kind of interest in, like, new technology.Link

LubaSo why did a computer-interested guy go study economics? You studied economics at George Washington University?Link

DylanYeah, you know, it's one of those things where, like, you're 17, they ask you, like, what do you want to major in, and you don't know. So you just do...Link

LubaEconomics?Link

DylanYeah, you're like, whatever. I guess I'll just do this. And...Link

LubaWhat were your parents influencing you to become?Link

DylanOh, nothing. They didn't... Yeah, no.Link

LubaHow come they didn't care?Link

DylanI don't know. I mean, it wasn't, like, a "we don't care"...Link

LubaThey were just like, oh, as long as he's happy.Link

DylanOh yeah, he's going to, like, a school that seems good and this seems safe. So fine. So yeah, they never, like, pushed me one way or another.Link

LubaAnd you grew up in...Link

DylanIn New Jersey. Yeah, on the East Coast. Yeah.Link

LubaYou seem like an East Coast guy.Link

DylanYeah. And then, you know, in... yeah, a couple startups, and through that, like, got really into programming and learning how to code.Link

LubaSo when you got to college, were you like, okay, economics is, like, whatever. I'm not that interested. I'm just going to do side projects.Link

DylanYeah, pretty much. I mean, after, like... I think the last year of college I, like, barely went to classes, because I was just trying to work on these startups and...Link

LubaWere you considering switching majors?Link

DylanNo, I just wanted to, like, be done with college.Link

LubaWhat was so bad about college?Link

DylanWell, I mean, I just knew the classroom setting wasn't...Link

LubaWasn't for you.Link

DylanYeah, I always learned best by being, like, hands-on and doing stuff. So, you know, like, for me, like, I learned how to code and program by, like, building apps and stuff.Link

LubaHow old were you when you started learning how to code?Link

DylanWell, I started learning, like, HTML and stuff when I was, like, probably, like, 12 or whatever.Link

LubaOh, wow. Okay.Link

DylanAnd then I...Link

LubaDidn't cross your mind to go study CS?Link

DylanNo idea. That's a good question. I mean, when you're that young, I just feel like, at least for me, it's just like your surroundings influence you so much, and I don't think I knew anyone that went to college for, like, computer science...Link

LubaYeah, yeah.Link

Dylan...from my high school or whatever. So...Link

LubaYeah, that's fair.Link

DylanIt's just, I didn't even, like, think about it as an option. But then I got really into programming and stuff, and that's how I got into machine learning. I just bought a couple books about machine learning and taught myself, and then moved out here to San Francisco.Link

LubaSo when you taught yourself how to code, you were still in college?Link

DylanYeah. Yeah, I bought a book on PHP and Python and then just kind of...Link

LubaBut take me back to that decision. Like, why did you guys... why did you decide to buy some books about machine learning?Link

DylanWe were trying to launch this startup. We needed an app and a website. There were two. One was, like, a precursor to DoorDash. So it was, like, basically a way for college kids, if you needed, like, toiletries, to, like...Link

LubaUh-huh.Link

Dylan...you know, order them on demand. And then the other was, like, a...Link

LubaWait, what kind of toiletries would students need to order on demand?Link

DylanYou needed, like, shampoo or toothpaste or toilet paper or whatever.Link

LubaOkay. Okay.Link

DylanJust, like, you know, order them, and then they'd be there, like, within 24 hours, cuz we had, like, a facility, you know, on the campus, like a garage that we would deliver stuff from. But we needed, like, a website and app to, like, pay and have people order whatever. And so it was kind of, like, organically. We just needed that, and so I built it, and then was like, "Wow, this is really fun. It feels like a, you know, like a game," you know? And then...Link

LubaYeah, I remember my...Link

Dylan...deeper.Link

LubaYeah, I remember... I was also a self-taught engineer, or, like, computer scientist, whatever, programmer. But I started way later, when I was 21. I was studying chemical engineering, and then I realized that the ability to see anything produced in the real world by being a chemical engineer is very, very slow.Link

DylanYeah.Link

LubaAnd then all these computer science kids, they can just build stuff out and the feedback loop is much faster. So I remember I started with HTML, CSS, and, like, seeing things come up on the screen after, like, minutes. It was really cool.Link

DylanTotally. And now with AI, everyone has that experience, cuz you can just see AI build things for you, like, right in front of you, and it's so addicting. And that, like, process of, like, building stuff, like, getting feedback, iterating, was really just, like, fun.Link

LubaSo what happened with that startup?Link

DylanThey didn't go anywhere. Yeah, yeah. I had no idea what we were doing, and I think it was more, like, college, like, energy. It wasn't that serious. We just were trying to, you know, start something or whatever.Link

LubaTrying to be enterprising.Link

DylanYeah, exactly. But then after college I just kept kind of programming and doing, like, kind of contracting jobs, and got really into machine learning. And that's when I came out here and got a job as a machine learning engineer at Cisco.Link

LubaWhat about machine learning specifically made you so interested in it?Link

DylanYeah, it just seemed like a harder kind of programming problem, and more of a kind of technically challenging problem. And then also, you know, I think it felt closer to kind of where, like, the future was going. And, you know, it was, like, back in, like, probably 2015 or around that time. So, you know, neural networks started to become more popular, and it was, like, early days of, like, TensorFlow and deep learning frameworks and papers. So it just felt like this new area of computer science and of technology that was really, really interesting.Link

LubaSo you mentioned that it seemed harder. How come harder is the word that you picked to describe machine learning specifically? Why was that particularly a big motivator for you?Link

DylanYou know, I think I've always, and still am, just kind of motivated by, like, challenges. And even a startup, it feels that way. It's like you're building something. So it's always this, like, kind of building things. And that's even why I was really into these, I think, like, video games as a kid, because it was like you're building this world and you're getting to new levels and...Link

LubaThe quest.Link

DylanYeah, it's all kind of wrapped up in there, where it's like, you know, there's a lot to... it just feels like there's a lot to explore, and I like that feeling of you're kind of figuring things out.Link

LubaSo you graduated college, you were kind of entrepreneurial in college with a couple of startups, as you mentioned, but you decided to get a job at Cisco. Why not go and start a company after college?Link

DylanI considered it, but I just didn't know what I wanted to do. And I do think it's true, like, if you're going to start a company, you at least need to commit to it for, like, 5 years.Link

LubaMhm.Link

DylanAnd be ready to work on this thing for the next 5 years. And so I did experiment and try things, like a bunch of random ideas, but after a couple months I would kind of get bored or lose interest. And so I didn't want to force it. And then I also felt like I want to be out here, and I want to learn a lot more about programming and products and building things. And so the opportunity came up with... well, actually, it was first a friend of mine was trying to recruit me to be a developer evangelist at Twilio.Link

LubaOh, cool. Which is kind of similar to...Link

DylanYeah, yeah. And they didn't end up giving me the job. They rejected my application. But then I think through that experience I was like, okay, I want to actually be out here and, like, work at a startup, or just work at a technology company, and not be, like, kind of just on my own.Link

LubaI feel like Cisco's an unobvious choice in comparison to, say, Twilio at the...Link

DylanYeah, yeah, totally. There was this cool team, though, trying to innovate and build new products. And, you know, it was kind of this, like, incubator within Cisco that felt exciting, cuz you're building things.Link

LubaWhat products were they building?Link

DylanIt was trying to add, like, natural language interfaces to some of the Cisco, like, video conferencing hardware. And then that's when, you know, I kind of got into voice and natural language understanding and processing, and kind of fell into this whole space. And it was around that time that Amazon Echo came out.Link

LubaYeah. How long were you at Cisco for?Link

DylanI think it was, like, 2 years. Yeah.Link

LubaAnd what were the highlights of your experience there? So you got into natural language processing and voice there?Link

DylanYeah, I think it was this interesting thing where, like, I mean, I had started some, you know, those startups in college, and I think it was, like, seeing that even at, like, a big company, things are dysfunctional too.Link

LubaWere you surprised?Link

DylanI kind of, yeah, yeah, cuz, you know, it was my first, like, real job. And I think that was, like, calming in a way, cuz it was like, no one has it figured out. And that, like, a huge public company with, like, tons of revenue, everyone's still just figuring it out and...Link

LubaYeah, I feel like everyone has that realization at some point in their 20s, that, oh, there's just some confident person...Link

DylanNo one knows anything.Link

LubaNo one knows.Link

DylanYeah, exactly. And so that gave me the confidence to go and try to start my own thing. But I was probably, like, too early, right? Cuz, like we talked about earlier, it was, like, 5 years before we raised our Series A. It probably would have been much better to start Assembly in, like, 2020.Link

LubaDo you really think so, though? Because I bet it gave you a really great foundation to be the kind of person...Link

DylanYeah, like, I don't... like, it was super fun. But yeah, I think that the market back... I mean, no one wanted to invest in the company, cuz they were like, oh, what's the market for, like, you know, this stuff? And now it's so obvious, but...Link

LubaYeah.Link

Dylan...back then it was, you know, a really small market. So yeah, we were just a few years too early.Link

LubaSo how did you get into voice at Cisco? What happened exactly?Link

DylanWell, I bought an Amazon Echo. It was really the Amazon Echo, because, you know, I saw the commercial for it. I was like, wow, and then bought one and was blown away how well it worked. And, you know, we talked about the difference between, like, voice dictation in the car versus, like, talking to an AI in the car, and how different that is. And that's what the experience with an Echo felt like for me, where it's like, wow, this thing works, and it works so well, such that this is, like, a brand new user experience for me, and I'm building habits around it now, and I'm able to use it and rely on it, and I actually, like, reach for it and, like, want to use it. And it was the first type of voice product I ever used that had that experience for me.Link

And so then I wanted to start building my own stuff with voice and see what I could do, like, at Cisco with voice. And that's when I realized, like, there was really no platform for developers at the time to build with voice. It was, like, this enterprise technology, and it wasn't available to you to build with.Link

LubaAnd how long did it take you, from having that realization to actually quitting your job to be like, I'm going to go all in on this?Link

DylanIt was probably, like, 6 months or something. You know, cuz for me it was this combination of, like, a lot's happening in deep learning, and over the next 10 years it is going to continue to, like, advance and develop, and I wanted to be in that space.Link

LubaThat's fair, but I feel like in, whatever, 2015, 2016, probably when you started thinking about it, a lot was happening in deep learning, but it was still very, very far away.Link

DylanYeah, yeah, totally.Link

LubaSo why were you like, a lot is happening and...Link

DylanI felt like I could see that over the next 10 years, deep learning, AI, was going to get better and better, was going to, like, kind of revolutionize, like, so many products, and natural language, like, interfaces as well. And so I wanted to be kind of building in that direction. And then I felt like, you know, building developer tools that other people can innovate with, I always was passionate about.Link

And so it was this, like, combination of things for me, where it's like... I still love what we're building, and I love seeing people build really cool things with it that they genuinely, like, couldn't without our platform. And so it's like a tool that is helping people kind of unlock their creativity and their innovations.Link

LubaAnd you were quitting your job to work on this. Did you think that it would all materialize itself very quickly?Link

DylanI didn't have any expectations, to be honest with you.Link

LubaBut you didn't have, like, oh, you know, I'm going to work on this for 2 years, see how it goes, and...Link

DylanNo, I mean, I kind of just, you know, just started working on it, cuz at the time, you know, I was young and just, like, kind of jumped into it. And it was my first, like, real startup. I mean, I had of course, like, started some small things before, but it was my first real, like, startup.Link

LubaSo you were like, I'll figure it out.Link

DylanYeah, yeah. Whereas I think, like, you know, if I were to start a new company now, or even now when we're, like, launching a new product or something, you know, we know what we're doing. So it's like, all right, like, here's what we should expect and here's our goals and whatever. But also at the time, I mean, the market for what we're building, like, didn't really exist.Link

LubaDid it feel risky?Link

DylanNo, not really, cuz I mean, I knew I could always go get a job as an engineer.Link

LubaWhat gave you that confidence? Cuz I think, especially, like, now there are so many people starting companies and there's so much capital around that it really doesn't feel that risky. But back in 2016, 2017, it was really different.Link

DylanYou know, this is, like, pre-kids. You know, I got, like, Target furniture in my apartment. So it's like, you know, my personal, whatever you can call it, like, needs were very low. So yeah, I just didn't really think about it in that way at all. Yeah.Link

LubaInteresting. So I always wanted to start a startup as soon as I found out that that was a thing that one can do. And for a few years I was working as an engineer, then in product, and then it took me probably, like, 8 months to get comfortable with the idea of taking that leap of faith and risk. And what you said, of, oh, I can always get a job: I knew that rationally, but it definitely took some time to get emotionally comfortable with the idea that I'm not going to end up, you know, under the bridge trying to feed myself. So I'm very impressed with your, like, oh, yeah, it was just very easy.Link

DylanYeah, well, I had, you know... when, after college, I basically lived, like, on credit card debt, and then got my job at Cisco and paid it all off. So I'd already kind of done the rack-up-credit-card-debt, find-a-job, pay-it-off thing. So I was like, worst case, I'll just do that again.Link

LubaI see. Okay, fair. I did not do that.Link

DylanYeah, so I guess my risk tolerance was always, like, kind of high. Yeah.Link

LubaHigh. Does it run in the family at all?Link

DylanI don't know. I don't think so. I don't know where it's from.Link

LubaInteresting.Link

DylanYeah.Link

LubaOkay, so sounds like you just didn't really care about the risk. You had high risk tolerance. And you also had very low expenses. Were there any other considerations? Like, were you... I don't know, did you have to have a chat with your girlfriend?Link

DylanNo, no. I think, you know, a big part of it was, like, I felt like I could see where the broader AI industry was going back then. You know, I went to some of the first TensorFlow meetups, and I was reading a lot of blogs and books and papers at the time on some of the early deep learning systems for voice, for computer vision. And that also, I think, provided some mental safety, that it didn't feel like, okay, if I spend 5 years working on this and it doesn't work out, I'm...Link

LubaYou're still going to learn valuable skills.Link

DylanYou know, so if anything, it felt like a way to learn new technology really deeply and rapidly, in a way that I had conviction around was going to be really important.Link

LubaYeah, it's not like you were quitting to start a laundromat. Although you'd probably learn some skills there, too.Link

DylanYeah, or, like, you know, a, like, marketplace or something, you know. This was like... the core of the product has always been, yeah, this, like, AI, you know, R&D and, like, research engineering kind of work. And that to me just was like, I want to work on that. So I think that provided also some, like, mental safety, even if it didn't work out.Link

LubaYeah. I love that. Yeah, that's a very mature way to look at it. That you're learning new skills, and it's a technology you were interested in, and it's going to go somewhere, and you'll probably get a job if it doesn't work out.Link

DylanYeah, exactly.Link

LubaWhat about the co-founder piece, cuz you're running solo? Were you considering getting a co-founder?Link

DylanYeah, for sure. So when I applied to Y Combinator, it wasn't with the intention I was going to get in. It was just to apply and see...Link

LubaWait, what do you mean, without it? That's getting in.Link

DylanI...Link

LubaAnd actually I want to talk about it, cuz you have kind of a funny story of getting into YC. You applied much later than the deadline.Link

DylanYeah, so I applied a month late. And my plan was to, like... I'd raised a little bit of seed funding after I quit my job.Link

LubaMhm. Who did you raise from?Link

Dylan50 or 100K. Just, like, angels.Link

LubaFamily and friends, or, like, actually...Link

DylanYeah, like, angels in my network from out here. And, well, I mean, it wasn't that, like... I think we raised, like, you know, somewhere between, like, 50 and 100K. Just, like, a small amount of capital. And so my plan was, like, let me get to a V1 of the product. There were some people that I was friendly with that I was going to recruit as, like, co-founders, and then apply to YC later. But I don't know what made me apply when I did, but I applied, like, a month late to the summer batch.Link

LubaOne thing that comes to mind is maybe you just wanted to apply to clarify your own thinking, because the application kind of forces you to do that.Link

DylanNo downside to do it. But I remember thinking, oh, should I apply now or should I apply in the winter? And I was like, well, I'll just apply now and probably won't get in, and then I'll apply in the winter. Because, you know, it was just to increase the probability of, like, something happening that would be good. So I didn't put much thought into it. And then I went to Europe with my wife and got a text, it was like, "Hey, can you come interview?" And so flew back and interviewed. And, you know, Y Combinator, it's only 12 weeks. It's really short. And so because I applied so late, when I got in, it was starting the next week.Link

LubaOh, wow. And you guys were on vacation in Europe at the time.Link

DylanSo we were just, like, kind of, you know, working from Europe.Link

LubaOkay.Link

DylanBoth of us were founders. So we were just there for, like, a month.Link

LubaOh, your wife is a founder, too?Link

DylanYeah, exactly. Yeah.Link

LubaCool. Okay.Link

DylanAnd she is from Denmark. So we were just, like, kind of in Denmark and...Link

LubaIs she a tech founder? Or...Link

DylanYeah. Yeah.Link

LubaWhat is she building?Link

DylanWell, she started, like, an AI fintech company a couple years ago that got acquired. And then now she's, you know, doing, like, a bunch of different...Link

LubaVery cool. Yeah.Link

DylanYeah.Link

LubaWow, so two founders.Link

DylanYeah. Yeah. So...Link

LubaI want to talk about that. But back to the YC story. So you fly from Europe for your interview.Link

DylanYeah. And then got in so quickly that, you know, I had to decide, okay, am I going to spend, you know, like a third of YC, like, recruiting a co-founder and kind of doing that, or just get through the 12 weeks and then do it after?Link

LubaMhm.Link

DylanAnd then over those 12 weeks, like, we made some progress and raised... I think I raised, like, a million dollars after YC. And so then it was like, "Okay, I can, you know, just hire some people." Like early...Link

LubaFounding members.Link

DylanExactly. Like founding employees, founding team members.Link

LubaHow did it feel getting in? Like, you had zero expectations of getting in.Link

DylanI remember having, like, you know, a tremendous amount of anxiety about it, because you go and, you know, I think it was, like, 300 companies.Link

LubaMhm.Link

DylanAnd, you know, most people have teams, most people are further along. And...Link

LubaAnd you're like, I have nothing.Link

DylanJust starting.Link

LubaWhy do you think they let you in?Link

DylanWell, so the partner that we had was this guy Daniel Gross, who, him and Nat Friedman ended up, you know, doing a lot of AI investing.Link

LubaYeah, yeah.Link

DylanThey now are at Meta. And they both invested a lot of capital into the company after YC. But Daniel, he worked at Apple. He saw the state of voice technology there and was a big believer in the idea. So I think he probably, like, vouched for us, and, you know, it's why we got into YC. And then after YC, he also invested a bunch in the company. And...Link

LubaEarly believer.Link

DylanYeah, early, like, really early believer in the, like, product vision that we had.Link

LubaSo then take me to that 12 weeks at YC. You have a bunch of anxiety, everyone's much further along, you don't have a co-founder, you're just getting started. What was that like?Link

DylanYeah, it was a whirlwind. I think it was really hard to figure out what to prioritize. You know, cuz, like, you just want to make a lot of progress for the demo day.Link

LubaMhm.Link

DylanCuz that's this kind of end of YC, and then you raise capital. It's, like, the assembly line of YC. And so it was really hard to figure out what to prioritize, because you could, you know... for me it was like, you could train a new model, make it better, but that would take, like, 30 days. And then, you know, you can only do that, like, three times throughout YC, cuz it's 90 days. It's, like, three months. So it's like, do you do that, or do you just try to figure out who could use the thing that you have now? But again, the market was really small. So it's hard to know what to focus on: getting some traction or just making the product better.Link

LubaYeah. And I bet training the models took a long time also back then.Link

DylanYeah. So it was really hard to know what to focus on back then.Link

LubaSo what did you focus on?Link

DylanChanged week to week, is the reality, yeah. Yeah, really just changed. But we did find some other YC companies that became some of our first customers. And then that...Link

LubaAre they still around?Link

DylanYeah, one of them actually is still a customer. So, like, yeah, yeah. Yeah, it's a company called VoiceOps. Yeah, so they're still a customer. You know, VoiceOps, they have an AI for sales teams. And yeah, they've been around a while.Link

LubaYeah, they've been early.Link

DylanYeah, exactly. But yeah, they're based in New York. They're still a customer. If not the longest, one of the longest-standing customers that we've had. And they churned once and then they came back. So it wasn't a continuous story with them. But yeah, we found some early customers and we just kept building and...Link

LubaAnd who was "we" at that point?Link

DylanSo there were a couple of folks that I brought on board. I mean, it was really just, like, three of us. It was a guy, Guru, and Michael. And they're awesome. They started a company together...Link

LubaMhm.Link

Dylan...and just went through YC, actually.Link

LubaThat's sweet.Link

DylanAnd so it was a really small team. Yeah.Link

LubaOkay, so YC, you were like, what do I focus on? But in the end, it all worked out fine, because you've raised a million dollars, you had some customers. Let's get back to that story of conviction, cuz at the very beginning you mentioned quite a few times, like, okay, you were early and nothing was happening for 5 years. Should have started the company in 2021. What were those few first years like?Link

DylanIt was a lot of that feeling of, like, do you just go make the technology better, to cross these new thresholds, or do you take what you have and go find the market for that? And oscillating between that, because you want to make progress, and there always is a market for something. But the more impactful work has, like, the delayed gratification, which is, like, make the core thing better, cross these new thresholds, unlock these new kind of tranches of the market that you can then go sell to and acquire. So it was a lot of that back and forth.Link

But we needed some other macro trends to come together. So, you know, like, COVID was a big one, where all of a sudden remote work turned a lot of collaboration online. And so the amount of audio data going over the internet... we had, like, AI note takers and stuff that started to come out around that time. Podcasts got really big around that time. There's a lot of, like, podcast search engines, cataloging of, like, online content. Those use cases started to happen around COVID.Link

But then once you had these general-purpose LLMs become available to developers, and they could use, you know... and then you have, like, vector databases and this whole ecosystem of AI infrastructure come out, that's when this explosion of AI applications starts to really happen. Where now if you want to build an AI application, you have LLMs, you have... like, if you think about the toy example that I talked about earlier, where there's a toy that you can talk to and it will create a sticker for you and then print it.Link

For that to be possible, you need the voice, for sure. But you also need the image generation model to make the sticker. You need some other natural language processing technology to do, like, content moderation.Link

LubaMhm.Link

DylanYou need good Wi-Fi. You need the hardware to be able to ship that toy. So, you know, you need all that to come together. And now I think the reason why AI infrastructure companies like Assembly are growing so much is because there is this really broad ecosystem of infrastructure for people to build on now. Probably the biggest thing is these, yeah, coding agents now. So now it's not just these engineers, it's any creator, builder, can build stuff. And so we see this huge surge of builders that aren't, you know, product teams in a traditional sense, using our APIs to build workflows for their personal use or for, like, their small business. And that's where we see a lot of, like, new demand coming from.Link

LubaSo in 2020, there were all these inflection points that luckily happened for Assembly. For the first 3 years, not much was happening, as you pointed out. How much were you grappling with that question of, am I even working on the right thing?Link

DylanI never had that doubt of, like, is this the right thing to work on? It was more just, we could have over-rotated on, hey, the market for voice back then is, like, court reporting. So let's just go really deep there. But I think we always had this, like, kind of vision for: there's going to be all this new, like, cool stuff that people can build, and we want to just kind of get there. But it took a while to get there, where the use cases pre, you know, 2022 were more disrupting, like, legacy use cases versus the greenfield net new stuff that we wanted to be building for.Link

LubaAnd one thing is being convinced as a founder, where it's your thing, you were convinced that this is the technology that's going somewhere. I assume you still had at least those two engineers and maybe other team members over those first few years. How did you inspire and convince them to stay?Link

DylanYeah, well, I mean, I really view the first couple people as, likely as, you know... and, you know, we were all kind of really building it and kind of in it together, you know, and we all work together, and, you know, we're all here in San Francisco. And so we were just trying to, like, make it happen together. It was a lot of fun.Link

LubaSometimes, you know, if you're working on something and there isn't anything happening, especially if you're married, there might be a conversation at home of, like, okay, you know, is this going anywhere? But your wife is also a founder.Link

DylanYeah, yeah.Link

LubaWhat was the conversation at home? Was it, like, 100% support: I understand what you're working on, you believe in this, so keep going?Link

DylanYeah, totally. Yeah, I don't think we ever even talked about it, you know, in that way, you know what I mean? It's just like...Link

LubaYou do you.Link

DylanYeah, it was just, you know, fun to, like, try to build something.Link

LubaThat's amazing.Link

DylanYeah.Link

LubaWhat was that, you know, being two founders at home? I guess she sold her company now, so maybe her life is a little more balanced than running an AI fintech company. How did you guys manage to stay and thrive through the years of the grind?Link

DylanI think, luckily, like, her startup was taking off while mine wasn't. And then, you know, when mine was and is, like, you know, she's exited her company. So the timing worked out, luckily, where that was the case. But if it had overlapped, I don't know. It would have been much harder.Link

LubaAny advice for people who are in relationships where both people are founders?Link

DylanIt's fun. I mean, you know, we, you know, like, talk about the product a bunch and stuff, and, like, we've used each other's products. We can give each other, like, product feedback. So it's fun.Link

LubaThat's awesome. Lots of user research.Link

DylanYeah, exactly. Exactly.Link

LubaAnd how has your life changed since you've had kids? And, like, time management of it all.Link

DylanYeah, totally. No time for, like, exercise or anything.Link

LubaZero.Link

DylanNot really. I mean...Link

LubaYeah, I read somewhere that you used to be a CrossFit guy.Link

DylanThat, yeah, like, years and years ago. Yeah, yeah.Link

LubaI got it. I got it. No more CrossFit.Link

DylanNo, I mean, yeah, we have a 4-year-old and a 3-year-old. So between, like, working and then them, there's, you know, that's kind of all there is time for. So it's fun in the sense that it's, like, super full all the time. But I think just trying to... like, one thing I really try to do is, like, you know, always be home for, like, putting them to bed and, like, you know, having dinner and stuff. Even if it's just, like, you know, like 2 hours at the end of the day, and just, like, be really as present as I can for that. But it's hard to balance it all.Link

I think it's probably easier now than it ever has been, though, fortunately, cuz if I think about, like, you know, pre-Slack, or the ability to even, like, work from home when you need to, that would be, like, really hard. And we have an office here, we have an office in New York, but we also have a ton of people that work remote. So we're very, like, hybrid, kind of flexible company, which I think is why we actually have a ton of people that have kids at the company. And we try to be, like, as supportive to, like, parents as possible, cuz it just requires, like, flexibility.Link

LubaSo work, kids, no workouts, no friends? Or do you still have time for it?Link

DylanI mean, there's really just, you know, on the weekends, for sure. But it's really just, you know, the weeks are, you know, just the startup and family. Yeah.Link

LubaSo there's two acts to the company. Act one, as you said, lots of building, not much happening. Act two, now a lot of things are going on and there's a lot of customers, and you guys have very impressive growth. Tell me more about the second chapter. How has it been for you personally?Link

DylanIt's been a lot of fun. Because I think the, like, vision for the product and market that we had is really starting to come to life. And so, you know, every month there's tens of thousands of people that sign up for the API. We're launching new products and features all the time. Our customers are taking off. I mean, we have a lot of customers now that are billion-dollar customers... sorry, billion-dollar companies. You know, they've raised, like, tons of funding. When I think of, like, Granola, you know, I think of... you know, we have customers like Retell, Fireflies. So where voice is, like, the core of what they're doing, and they're taking off and they're super successful. So it's really cool to see, like, their success.Link

And then, you know, every couple months, it seems like there's these new... you know, we cross these thresholds and then there's these new use cases that come to life. So now, you know, we're seeing a lot of stuff with consumers, like these toys. I'll show you the sign on the desk over there. But seeing those start to come to market is really cool. So the pace of developments is just, like, much faster now, and it's really fun to just, like, see how much all that is developing.Link

LubaWhat have you had to learn as a leader from one chapter of the company to another?Link

DylanI think in the early days of scaling, you know, doing it for the first time, you tend to really, like, lean on, like, best practices, or, like, the kind of stereotypes, if you will, for, like, you know, how to build and scale a company. But in today's world, especially with AI, like, so many of the rules of, like, how you operate a company and how you, like, scale a company, I think, are changing.Link

LubaAre there any examples of you following the stereotypes that didn't work out?Link

DylanOne example is probably, like... it's kind of a boring example, but, like, you know, how we do, like, pricing and packaging for the company. So we started out by doing, you know, traditional, like, contracts, right? 12-month agreements. But we felt like, for our customers, that just wasn't the best, like, experience for them, a customer journey. It didn't kind of fit with what they were trying to do. And so we decided to go 100% usage-based and pay-as-you-go. And that's been amazing for our customers, but, you know, it's nontraditional in the sense of, like, how, you know, every company before us has done it.Link

So there's been a lot of small examples like that. But we are constantly applying that to decisions. So, like, where do we need to invest in the company? Like, how should we think about, like, the role of product at a company like ours? How do we think about, like, the hierarchy of the company? So we're only, like, 85 or 90 people or something at the company.Link

LubaYeah, that's not that many people for the amount of...Link

DylanNo, yeah, and...Link

Luba...customers you guys are supporting.Link

DylanYeah, exactly. And we've tried to be really intentional about that.Link

LubaThere was a period where you had to scale down.Link

DylanYeah, yeah. So, you know, we raised a bunch of money and then we scaled up kind of before we had figured out...Link

LubaWas it, like, the regular trope of raising a bunch of money and being like, well, we must hire a lot of headcount?Link

DylanYeah, and just kind of trying to scale before you're ready to scale, and before you really know, like, what's working and what are the right things to scale up, versus, like, where you're still learning, before you need to scale up. So we had to, you know, reset our go-to-market, really, was the big thing. It's, like, reset our go-to-market and kind of rebuild a new, like, customer journey for how our customers want to buy and...Link

LubaWhat was it like realizing that you needed to scale down, and actually going through that?Link

DylanWell, I think at some point it was just so obvious that, like, the way we're trying to go to market was, like, not a match with our customers. And so...Link

LubaAnd how were you going?Link

DylanWell, we were just trying to apply more of, like, a traditional...Link

LubaSales motion.Link

DylanYeah, sales motion to, like, selling AI infrastructure. And I think that was one part of it. But I think the other part is the market for AI infrastructure, still, though, in, like, 2023, was pretty early, because the kind of phases of AI in the market, I think, have been... I think, when did ChatGPT come out? It was, like, 2023. Yeah, so 2023, ChatGPT came out.Link

And then 2023, 2024, there were a lot of these, like, first movers in AI applications. Really smart teams bringing AI-native applications to market first. So I think about, like, Perplexity, a company like Jasper, Lovable, right? But there weren't, like, a ton of them. And we were always focused on infrastructure. And so it's really, from our point of view, not been until, like, last year where there's just been this explosion of applications. And that's accelerating even faster now with coding agents, cuz people are just building so much software now, that AI infrastructure companies are really scaling. And, you know, we see it with our peers and companies that we're friends with that are also in AI infrastructure. But that's really started in, like, 2025. And our metrics are representative of that as well.Link

So I think even scaling up the go-to-market in, like, 2023, it was still, actually, also too early. So it was a mix of, like, being too early, but also not having the right motion ready to scale. Whereas now it's like, you know, the go-to-market motions we have are, like, you know, they're working. They're, like, really, really easy to, like, understand.Link

LubaYou have playbooks.Link

DylanThe demand is, like, really strong. And so it's easy to say, "Okay, yeah, let's go invest here. Let's go invest here. Let's go invest here." And have the conviction that, you know, for the most part, it's going to work.Link

LubaIt sounds like now you're really drinking out of a firehose and everything's coming together. What have been the most challenging moments for you?Link

DylanI think the more challenging moments, honestly, were, like, the earlier days. But also I've been doing this so long now where I feel like I'm maybe... maybe desensitized to it, you know. Cuz there's... I mean, week to week there's ups and downs, you know. But I think, like, definitely now it feels like we have a good grasp on, like, the product vision, like, how we want to go to market, like, the markets that we're going after. The markets are more mature too that we're going after. So, you know, it's not, like, early anymore. It's, like, happening right now. So I think all those things help, and now it's really about, like, being kind of heads down and...Link

LubaI think you're the first founder I'm talking to who's, like, stress-free, figured it all out.Link

DylanWell, no. I definitely haven't figured it all out. Stress-free, but I think, yeah, maybe my coping mechanisms have just...Link

LubaEvolved.Link

Dylan...been well developed. The earlier years were, for sure, a lot more stressful, for me at least. Like, I think probably the most stressful time ever was, like, you know, going through Y Combinator, if I think about it. But we have a great team now, like, truly, and we have, like, great customers, and, like, the product's come a long way. So now it just feels like we really got to keep innovating on our product, like, keep helping customers, keep getting our, you know, name out there with companies that we can help. And so we're building out our, you know, sales team a lot right now, actually, go-to-market team. So things have really, like, come together over the last, like, 2 years.Link

LubaAnd what are some tools, if you have any in your toolbox, of managing stress, or managing, you know, anxiety about an important decision that you need to make, or just keeping moving forward?Link

DylanYeah. Probably the thing that's helped me the most is just having, like, a good network of, like, support network. So, like, our investors are awesome. They're super supportive, and, you know, I talk to them, like, all the time.Link

LubaMhm. Who are the people that you talk to all the time?Link

DylanWell, we have, so, Accel, Insight, and Smith Point are pretty big investors in the company, and pretty much weekly I'm talking to them. And they, you know, know exactly where the company's at. They're able to, you know, kind of, like, have a seat around the table and help. And also have a bunch of friends that are founders that are in a similar stage. And so...Link

LubaWho come to mind?Link

DylanLike Victor from Synthesia, like Edo from Pinecone. Just a bunch of folks I've met over the years where, you know, I can text them and...Link

LubaHave you guys ever tried any, like, standing dinners or Zoom calls, like accountability groups? Everyone's too busy, probably, to coordinate schedules.Link

DylanJust, like, you know, the WhatsApp text here and there.Link

LubaA recent thing that you've kind of asked for support or advice on?Link

DylanIt's less formal than that. I think it's more just, like, you know, you hang out. So you hear what they're dealing with. They hear what you're dealing with. You realize that everyone's dealing with the same stuff, and so therefore it's, like, you know, again, what we're talking about earlier. It's like, no one has all the answers. There's just ups and downs and you have to just kind of...Link

LubaBut it's reassuring to hear someone else's ups and downs.Link

DylanYeah, yeah, yeah. It's like group therapy. Yeah.Link

LubaDo you have, like, any executive coaches?Link

DylanNo, no.Link

LubaNever had?Link

DylanNo.Link

LubaWow. You don't believe in them?Link

DylanI think, for me, like, I like to get advice from people that...Link

LubaHave been there, done that. Your wife and friends and investors.Link

DylanYeah, so for me, that's, like, what I found is the most helpful thing, versus, like... Like, I'm sure they're helpful, and maybe they're a better fit for, like, other people, but for me, I have found better advice from, like, former, like, operators, basically.Link

LubaMhm.Link

DylanYeah. Or people that are currently, like, operating companies.Link

LubaYeah. What are you struggling with right now, if anything?Link

DylanReally trying to scale the company, because we have, like, amazing product-market fit and we, like, have a ton of demand. And so, honestly, the biggest thing right now, we're just trying to...Link

LubaManage the growth.Link

DylanYeah. Manage the growth of, like, existing customers, building the team to help with new customers, manage, like, the scale of customers we have on the platform, too. So we're really, I think, in a, like, fortunate position, where we're in a market where there's so much tailwind behind what we're doing. So we're just trying to, like, catch as much of that as possible.Link

LubaYeah. Speaking of building up the name, where does the name AssemblyAI come from?Link

DylanWhen I was starting the company, you know, it was this kind of thought of, like, you know, this developer tool for, like, building AI applications, and so this, like, assembly line of, like, AI primitives. And so I needed a name to kind of just, like, formalize what I was working on and make it more real. So I came up with that, and it kind of stuck, and then, you know, didn't put, like, too much thought into it in the early days.Link

LubaSeems to be working.Link

DylanYeah, yeah, yeah. Yeah. I think the only downside is, like, you know, sometimes people pronounce it as, like, "Assembly Al," cuz it's, like, AI lowercase on, like, mail or something. But yeah, the name wasn't, like, super important for me when I was starting the company, other than needing to have it be, like, an actual thing. So I needed to put a name on it.Link

LubaMakes sense.Link

DylanYeah, to make it official.Link

LubaSo I know people can really see that we are in Assembly's office in San Francisco. You recently moved to San Francisco from New York. Tell me a bit more about the reasoning behind the move.Link

DylanYeah, well, we're really scaling up our go-to-market team here. We have an amazing team and office in New York, and we're growing that too. But we're doing a bunch of hiring here. We have a ton of customers here, and so for me personally, I wanted to be, like, closer to that. But then I also lived out here for, like, almost 10 years, and then during COVID and post-COVID is when, you know, we moved. So my wife and I also, you know, we're excited to kind of be back here in the city and be in San Francisco. But yeah, we're building out our go-to-market team here. So, you know, I wanted to be here for that.Link

LubaMakes sense. Yeah. And just to wrap it up with a couple of questions. What are you still trying to prove to yourself?Link

DylanStill have this, like, product vision for what we're trying to build here, and there's so much still to do and build that I'm, like, really excited about. So for me, really, like, the reason I get excited every day to, you know, come to work and, like, build a company is because there's this really clear-in-my-head product vision that we have yet to achieve, and that we still have to do a lot of work to get to. And I'm really, like, motivated and excited about getting there. And getting closer to that, you know, vision that we have for the company, and making that more real.Link

LubaAnd what are you still becoming?Link

DylanI don't know. I think ultimately, like, I'm just, like, a builder, and want to, like, build the company into, like, what I think it can be. And so that's really, like, what I'm just, like, super, super focused on. You know, for me, it's never been about, like, the status or whatever of, like, me or the company, versus, like, I really think we can build something awesome here that can help other people innovate and be creative and build, like, really cool things for the world. You know, we've talked about that toy a bunch, but I'll show you. I think that's, like, a great example.Link

LubaSo, like, we need to show the toy on the camera, yeah.Link

DylanWe will, we will. But that's really, like, what I'm passionate about, is, like, just trying to make the company into what, you know, we think it can be.Link

LubaTotally forgot to ask you: why are you off X?Link

DylanI was checking it too much. And so I had to delete all the apps from my phone and block it on my computer, because it was distracting me too much. And then I unlearned the habit of going there. But I think I need to...Link

LubaDo you think you'll come back?Link

DylanFor the company, I think I need to. Yeah.Link

LubaThe company has a big X account.Link

DylanYeah, yeah, exactly.Link

LubaOkay, so since 2023, you've been completely cold.Link

DylanI would, like, visit it sometimes, but then I think after Elon bought X, you had to log in to browse it, and then I just didn't want to log back in, because it was too distracting for me. And then I just unlearned the habit. Yeah, yeah.Link

LubaAwesome. Dylan, thank you so much for your time.Link

DylanYeah, yeah, thanks for doing this. It was fun.Link

LubaThank you.Link

DylanCool.Link