Connor Zwick: Making Language Immersion Possible Through AI
Estimated read time: 1:20
Summary
In this insightful discussion, Connor Zwick, co-founder of Speak, shares his journey from a tech-savvy youngster in Wisconsin to a pioneering figure in AI-powered language learning. He reflects on his experiences with successful ventures like Flashcards Plus and his passion for AI's potential to transform language education. Zwick delves into the challenges and revelations faced while establishing Speak in South Korea, emphasizing a focus on providing immersive, conversational learning experiences over traditional grammar-based methods. The conversation highlights the lessons learned and the future of AI in education.
Highlights
Connor Zwick started building tech companies in middle school, highlighting early talent 🌟.
Speak was founded on the belief that AI can replace human tutors for language learning 🔄.
Connor's insight to focus on non-English speaking markets has been pivotal for Speak's success 🌏.
The company's targeted approach in South Korea highlights the value of market-specific strategies 🏢.
Speak provides immersive language learning by encouraging users to converse naturally 🤝.
The flexibility and persistence in Speak’s strategy show how Founders can adapt and thrive 💪.
Connor Zwick and his team emphasize the importance of scalable software models in tech innovation 📈.
Key Takeaways
Connor Zwick's journey highlights the power of curiosity and persistence in innovation 🚀.
Starting with small tech projects can lead to significant breakthroughs and global impact 🌍.
The importance of product-market fit is crucial; it's all about iteration and feedback loops 🔄.
AI's potential in education is vast, with language learning being a prime example 🧠.
Targeting non-English speaking markets can open up massive opportunities 👏.
Understanding cultural nuances is key to successful international expansion 🌐.
The future of AI in education looks promising with potential for personalized learning experiences 📚.
Overview
Connor Zwick's entrepreneurial spirit was evident from a young age, as he began building small tech projects during middle school. These humble beginnings laid the groundwork for his future success in the tech industry, eventually leading to the founding of Speak, a revolutionary AI language learning platform.
Speak distinguishes itself by focusing on immersive, conversation-based language learning, primarily targeting non-English speaking audiences. This approach stemmed from Connor’s realization that traditional grammar-focused methods were less effective for real-world communication. The choice to concentrate on South Korea, a market with a strong demand for English learning, proved to be a strategic masterstroke.
Connor's journey is a testament to the balance of technical innovation and cultural understanding. By emphasizing scalable, adaptable software solutions, and through persistent iterations, Speak has not only catered to a specific market need but also paved the way for the future of AI-driven education, aiming to make high-quality learning accessible to all.
Chapters
00:00 - 00:30: Introduction and Host Welcome The chapter titled 'Introduction and Host Welcome' begins with an introduction accompanied by music, followed by a warm welcome to the show 'generative now.' The show focuses on conversations with innovators in the AI space, exploring the impact of AI on current and future societal dynamics. Michael McNano, a partner at Lightspeed, a global venture capital firm with early investments in successful companies like Snap, Affirm, Nest, GrubHub, and Giphy, hosts the show.
00:30 - 01:00: Introducing Connor Zwick This chapter introduces Connor Zwick, the co-founder of Speak, a company innovating language learning through an advanced AI-powered tutor. The transcript highlights Connor's entrepreneurial journey starting in middle school, including his ventures in teaching coding online and founding Flashcards Plus.
01:00 - 02:00: Connor's Early Tech Background The chapter discusses Connor Zwick's early involvement in the tech industry, highlighting his acceptance as a Thiel Fellow immediately after high school. It also mentions his innovative spirit, as demonstrated by sneaking into Berkeley classes with his co-founder to learn about AI and machine learning, showcasing his passion and dedication to the field. The discussion sets the stage for an insightful conversation with Connor, the co-founder of Speak.
02:00 - 03:00: Flashcards Plus Success The chapter titled 'Flashcards Plus Success' features a conversation with Connor, who invites listeners to understand his background. Connor reflects on his formative years, growing up in suburban Wisconsin with limited travel experiences. An essential aspect of his youth was having access to a computer from a young age because his mother owned one. This early exposure to technology plays a significant part in the narrative of his development and future success.
03:00 - 04:00: Lessons from Flashcards The chapter 'Lessons from Flashcards' begins with a personal narrative highlighting the early and profound impact of technology. The narrator recalls growing up with hand-me-down computers from a family member, which sparked a deep interest in technology from a young age. Despite parental concerns about excessive computer use, this early exposure led to a self-taught journey into programming, specifically JavaScript. This technological enthusiasm also inspired the narrator to start several small businesses, showcasing an entrepreneurial spirit.
04:00 - 05:00: Education and Early Ventures This chapter titled 'Education and Early Ventures' details the speaker's involvements during middle school and high school. They were actively teaching coding and building programming projects for others during that time, which became a continuing pattern in their life. The speaker reflects on using old Windows computers, wishing they had a Mac instead. They mention trying older versions of Windows, such as before Windows 95, possibly 3.1 or 3.31.
05:00 - 06:00: Teal Fellowship Experience The chapter discusses distant memories of older computers, including printers with side ribbons. The experience of building early coding platforms, similar to Code Academy, is highlighted. The narrative portrays a nostalgic look back at the early days of online coding education and how it was set up on some of the initial computing machines.
06:00 - 07:00: From Teal to Entrepreneurial Pursuits In this chapter, the narrator reflects on their Middle School experiences with a website called TTS plus tsu.com, where they learned programming by reading articles. Discovering that anyone could submit articles and get paid, they contributed content on learning Ruby, JavaScript, and CodeIgniter. To their surprise, the website accepted their submissions without questions, leading to a steady income, marking the commencement of their entrepreneurial journey.
07:00 - 08:00: Coco Controller and Hardware Challenges The chapter discusses the unexpected rise to popularity of a young author, who was a major contributor to a well-known coding website. Despite being only 11 years old, the author was offered a senior position as an editor, which required moving to Australia. The chapter captures the moment when the author had to reveal their age and the challenges that arose from this situation.
08:00 - 09:00: Transition to AI Interest During the transition to AI interest, the narrator talks about their early fascination with technology, specifically mentioning their use of a Windows 311 computer and writing articles as a child. During high school, the narrator identified a common problem among classmates using physical flashcards for studying, and developed an application solution to address this issue, drawing on their technological interests and skills. This period marks the beginning of their deepened engagement with AI and technology.
09:00 - 10:00: Immersing in AI Learning The chapter titled 'Immersing in AI Learning' begins with the author discussing their strategy for digitizing flashcards, something akin to the size of index cards, for use on mobile devices. This initiative was part of a side project as they had not yet finished another app. Despite the low technological standards of the time, the flashcard app managed to resonate with users. On its first day of release, it secured 12 downloads, matching the number of people the author had informed about the app. It reflects modest success and the author's early foray into app development.
10:00 - 11:00: Reinforcement Learning and Berkeley Experience The chapter "Reinforcement Learning and Berkeley Experience" discusses how the author's app initially received 25 downloads and saw a rapid increase to 300 downloads the following day. This unexpected success led the app to become one of the top education apps in the App Store for several years. This success story illustrates the author's pivotal journey into technology, emphasizing how impactful and life-changing the experience with the app was. The narrative reflects early App Store dynamics and development trends.
11:00 - 12:00: Pioneering Speech Models In the chapter titled 'Pioneering Speech Models', the transcript discusses the early era of app development where applications were mapped to real-world objects. The focus is on flashcard apps, which were popular with editorial teams due to their clever replication of physical index cards. The description includes the nostalgia and humor of recreating these physical formats in digital form during the early days of app development. The transcript ends with a query about the outcome of the flashcard project, whether it ended or was sold.
12:00 - 13:00: Vision for AI's Future in Speak The chapter delves into the journey of the speaker regarding their AI project named Speak, which significantly impacted their personal and professional identity. Initially, it started as a passion project during high school and continued through college, which the speaker eventually left to focus more on their entrepreneurial journey. Post-high school graduation, the speaker describes a pivotal moment of moving to San Francisco, spurred by interest from a venture capital firm. The project, eventually sold several years later, served as a foundational experience, opening numerous other opportunities for the speaker.
13:00 - 14:00: Founding Speak and Early Challenges The chapter, titled "Founding Speak and Early Challenges," discusses the early stages of founding a tech company. It begins with the founders being approached by venture capitalists who encouraged them to leave college and work full-time on their project. Despite being inexperienced and originating from Wisconsin, they were urged to relocate to Silicon Valley to develop their app further. This chapter highlights the transition from a personal project to being thrust into the tech industry, driven by significant investment interest.
14:00 - 15:00: Refining and Testing Speak's Model This chapter discusses the journey of learning necessary to build a successful technology company, despite initially lacking knowledge about the startup ecosystem. The narrator reflects on their experience, being an avid follower of tech news without understanding the practical aspects of creating a venture-backed business. They reminisce about early perceptions of knowing everything contrasted by the complex realities encountered in tech entrepreneurship.
15:00 - 16:00: Focus on Korean Market The chapter titled 'Focus on Korean Market' appears to centralize on the theme of missed opportunities and hindsight in business development. The speaker reflects on their initial lack of confidence in a project involving flash cards, despite positive feedback from users. They express a sense of regret for not having pursued the idea further, admitting they underestimated its potential at the time due to their inexperience. The narrative suggests that the speaker has learned from their past mistakes and acknowledges the evolving dynamics of the world, implying a shift in their perspective towards more ambitious undertakings.
16:00 - 17:00: Speak's Market Strategy and Expansion The chapter titled 'Speak's Market Strategy and Expansion' covers lessons learned from previous experiences with flash cards that have been applied to the development and expansion of Speak. A crucial lesson highlighted is understanding what product-market fit looks like, which is instrumental in shaping their current strategies. The narrative hints at a deeper exploration of Speak's journey, although some details are reserved for a more comprehensive discussion later.
17:00 - 18:00: User Experience and Language Teaching Approach The chapter explores the concept of user experience in the context of language teaching approaches. The speaker reflects on their early experiences with understanding effective practices, emphasizing the difficulty in defining what makes an experience successful. They share insights from their work with a language learning platform, highlighting the years of effort in refining the product before achieving a satisfactory product-market fit. The narrative underscores the importance of continuous iteration and the need to respond to user demands effectively.
18:00 - 19:00: Current Success and Future Plans This chapter discusses a personal retrospective on the key learnings and experiences gained from building a company. It highlights lessons in the mechanics of company building, collaboration with education companies, and partnership dynamics. A significant part of the chapter focuses on the acquisition process by 'ch' before they went public, detailing the negotiation and company valuation procedures.
19:00 - 20:00: The Impact of AI on Education The chapter discusses the importance of discerning between valuable and nonvaluable components in technology and education. It highlights a case where a code base was entirely scrapped and rebuilt using better technology, emphasizing the lesson of focusing on what truly matters in development and education.
20:00 - 21:00: Potential and Challenges of AI Expansion The chapter titled 'Potential and Challenges of AI Expansion' emphasizes the significance of achieving product-market fit. It draws an analogy between the effort required to push a boulder up a mountain and the ease with which it rolls down once it reaches the peak, highlighting the transformative moment when a product becomes self-sustaining. The chapter asserts the critical importance of attaining product-market fit, suggesting that other factors, such as press coverage, are meaningless unless they contribute to this goal. This lesson is described as incredibly valuable, underscoring the necessity of reaching a point where a product can naturally progress and succeed in the market.
21:00 - 22:00: Conclusion and Invitation to Join The chapter titled 'Conclusion and Invitation to Join' emphasizes the importance of focusing on core activities that genuinely matter for progress. It highlights that attending events or conferences should have clear intentions such as networking for hiring or partnerships, and fundraising should directly contribute to enhancing user value. The speaker shares a personal lesson on maintaining a ruthless focus on critical elements throughout their career. The chapter serves as a call to action to remain dedicated to fundamental priorities and extends an invitation to join this approach for meaningful advancement.
Connor Zwick: Making Language Immersion Possible Through AI Transcription
00:00 - 00:30 [Music] hello everyone and welcome to generative now this is the show where we talk to the builders who are creating the world's most exciting AI products and companies we get their perspectives on how AI will impact the world we all live in today right now and in the future I am your host Michael mcnano I am a partner at light speed a global Venture Capital firm that has been one of the earliest investors in companies like Snap affirm nest GrubHub gify and many
00:30 - 01:00 many others including a bunch of AI companies in fact we've invested over a billion dollars across more than 50 AI native companies and today we've got an awesome guest for you we've got Connor zwick the co-founder of speak the company that is changing the way that people learn languages through the world's most advanced AI powered tutor before speak Connor got his start building tech companies in middle school first by teaching coding online and then founding flashcards Plus which was
01:00 - 01:30 acquired by CH in 2013 this was such an awesome conversation we talked about Connor being accepted as a teal fellow right out of high school and then even sneaking into Berkeley classes with his speak co-founder to learn all about Ai and machine Learning Without further Ado let's get into the conversation with Connor zwick co-founder of speak Connor great to see you thanks for doing this yeah of course of course excited to get into it what I like to do with these is I like to go all the way
01:30 - 02:00 back you know to to really understand speak I feel like we need to understand Connor so can you take us back to the beginning give give us some information on your background growing up the formative years of Conor's Wick wow okay yeah all the way back um okay let's see originally uh you know I'm from Wisconsin you know typical kind of suburban American uh upbringing hadn't really traveled that much um but one thing I did have was a computer from a pretty young age my mom um had her own
02:00 - 02:30 little business and she had computers that she would pass down to me and so I was always just like you know from as as early as I can remember I was always like on the on the computer uh you know definitely my parents thought I probably spent too much time on the computer uh but yeah I just uh uh you know growing up um was always super interested in technology um you know kind of taught myself out of program when I discovered what JavaScript was um started a bunch of like little businesses you know
02:30 - 03:00 middle school high school around like uh teaching people how to code and uh building like you know programming for people and stuff like that and that's kind of like yeah that that was kind of like where where I came from um and I've been kind of just doing that ever since what kind of computer were you using uh when you when you were doing all this man I wish I was using a Mac but I was using old windows like man I think what was the first Windows like what was the one before Windows 95 and then Windows 95 3.1 3 31 right like something like
03:00 - 03:30 that I Distant Memories of the older computers too they were like uh the old like printers with the little uh side ribbons um and yeah it was fun stuff back then um everything was so cool you you built like the original code academy you were you were teaching coding online um from some of those early machines yeah it's a funny it's a funny little Story I mean uh I I don't know I think code academy may have been kind of parallel to this but when I was um you
03:30 - 04:00 know when I think I was in Middle School there was this website called TTS plus tsu.com and Nets and I like learned how to program when or I was like reading the Articles from the site um and I realized oh anyone you know can submit articles to this to this website you can get paid and so I started submitting a few of them um for like learning Ruby learning JavaScript um code igniter all this all this stuff and uh they didn't ask me any questions they accepted the Articles and they started sending me like you know it was like good money um
04:00 - 04:30 and I just started writing more and more of these articles I became like pretty popular uh one of the like primary authors of this website that had a lot of views was it was definitely one of the primary websites back then uh for learning how to code and uh eventually they offered me the uh position of editor and I would have to move to Australia and be like a salaried employee um and then I had to like level with them and be like look guys I'm only 11 years old I uh think this going to work out yeah I didn't realize the the
04:30 - 05:00 age we were talking about here oh my God that's amazing maybe 12 yeah so okay so you're so you're writing these articles uh you're playing with the the windows 311 uh computer what what what what happens in high school um before you go away to college yeah so I was a student at the time so I built an app to solve one of my own problems which was like everyone in my you know classes they were using flash cards like the index cards to like study things and uh I was like oh this is you know an iPhone is
05:00 - 05:30 you know approximately the size of one of these index cards I'll just make an app that like digitizes those flash cards and people can do it on their phones um and so I build that as a little side thing release it I still haven't finished the other app yet at this point um and it just like really hit a nerve I mean the the bar back then was pretty low but like I remember the first day I I like logged into iTunes connect back then it was iTunes connect I think and it had like 12 downloads and I was like okay cool like I told 12 people makes sense like that's that's
05:30 - 06:00 cool that everyone downloaded the next day I had like 25 downloads and I was like where did these people come from and then the next day I had like 300 downloads and then it literally just like cascaded from there um and it literally it went on to be like one of the like the number one education app in the app store for like a bunch of years and that's like really how I got into technology I think um was like the experience with that app was very like life-changing that's fascinating yeah that that that reminds me that sort of in the early days of the App Store development there was a lot of this sort of of it was it was like that um it was
06:00 - 06:30 like the skoric era right where everyone were was mapping apps to like real world things right in the case of this it was flash cards and the Apple team the editorial team loved that stuff right they loved like promoting all these apps that like mapped these real world things it's like like the red line on the top and the blue and all the little like lines of what like an actual index card looked like it was so funny we were all figuring it out together the in the early days you know that's awesome what like give us the story of flash cards did you did it wind down did you sell it
06:30 - 07:00 like what happened to it from there the short version of the story is that I I eventually did sell it um but it was like many years later actually so it was always this thing that I was working on it felt like a huge part of my identity I I went to college for a bit before I dropping out um and I probably sold it you know a year or two after that but yeah I mean it was definitely the thing that opened the doors for me like the the day after I graduated from high school um I was on a plane to San Francisco and I moved to San Francisco uh like there's a VC firm that like
07:00 - 07:30 tried to get me uh to come out for the summer and not even go to college and just work on this fulltime because they wanted to invest um and so it it definitely like was the the thing that really kind of got me into the world that I'm in today it was always just like my own project and and as these VCS are trying to draw you out to Silicon Valley and and all like did you know anything about starting a tech company or I mean were you just a kid in Wisconsin that happened to build an app like you know what I mean like there's
07:30 - 08:00 all there's this whole other world to building a tech company did you know anything about that absolutely not absolutely not I mean I I would I was like an Avid Reader of tech fronch right like I I followed the like I followed the news and I like uh you know I remember when Twitter came out and all that but I didn't know what it was I didn't know the nuts and bolts the Dynamics of what it takes to like build like an actual Venture style business at all um you know I I I like to joke that like I feel like when I was that age I I thought I knew everything and I like AB
08:00 - 08:30 absolutely knew like AB nothing at all um like one of the reasons I didn't lean into flash cards even though I had all this poll from users that were actually like using it and I had actually built something people loved um was I just didn't think it was like a big idea or ambitious enough um and you know in retrospect I could have totally leaned into it and I you know probably made it even better and bigger than it was um but I just had no idea what I was doing back then yeah I think the world has changed though at this point right yeah
08:30 - 09:00 for sure um do you feel like okay so so so maybe a lesson was you could have leaned into it but again we'll get to the full speak story so we're skipping ahead a little bit but were you able to draw any lessons from flash cards that you that you still to this day have now mapped to speak yeah I think like um I mean I think one of the biggest lessons was was just realizing what product Market fit Pole looks like
09:00 - 09:30 um I think I was fortunate that I had that early and I saw what that felt like looked like it's very hard to Define right it's like this famously difficult thing to articulate but um I think that allowed me to realize at least for in the case of speak and I'm sure we'll get into this but like it was many years of grinding and choosing not to scale something that was subpar product Market fit um and like continuing to like iterate until we actually felt like theet was pulling us and telling us
09:30 - 10:00 something um so that was like definitely in retrospect one thing I learned I do think like just the nuts and bolts of building it I learned a lot of the like mechanics of building a company and what that looks like um how to partner with other people like uh we partnered um with a few other like education companies I guess like one really big thing was like it just learned like how when eventually when it was acquired by uh ch um right before they went public uh like the negotiation process there how to like Val value a company what
10:00 - 10:30 what that means what what what are like the actual things of value that you're building versus not like for instance they scrap the accelerator code base immediately when they took over and rebuilt the entire thing from ground to look exactly the same but like in a better technology the code was not worth a single you know scent uh but uh but yeah like I I think that was like a big lesson um it was just like what is like actually valuable versus what's not valuable and really like ruthlessly focusing on on the stuff that really matters um yeah those are some of the early lessons I think I think the
10:30 - 11:00 product Market fit lesson was probably very very valuable right like there's a there's a big difference between pushing the boulder up the mountain and it rolling down the other side of the mountain on its own and like I think if you've never felt that before it's it's pretty easy to talk yourself into oh it's going down naturally right and so that that was probably like hugely valuable and also like the fact that like that's the only thing that matters like I feel like I learned at a relatively young age like press doesn't matter unless it gets you product Market
11:00 - 11:30 fit and like actual durable traction conferences don't matter like going to like events don't matter unless you're like going to actually meet people that you can like hire or work with fundraising doesn't matter unless it can advance you know the actual value you're providing for your users and capturing right um and so I do think like that was probably like a very fortuitous thing to learn young in my career uh is just like the focus just like the ruthless focus on like the core things that matter and being able to kind of like see past all
11:30 - 12:00 the all the things that I think distract a lot of people yeah for sure that's an awesome story um okay let's get into so okay wait so you graduate and then you said you got on a plane right yes you got on got on a plan so what happens after that plane ride I I hated San Francisco I literally just stayed in my house the entire time so it it's not like the you know fun post High School graduation summer that I'm sure some people had um and then at the end of it um I decided to go to university um I was lucky to get into Harvard
12:00 - 12:30 honestly mostly probably because of the app like it was like the differentiator yeah I never really thought about that that um but it makes total sense right like that building an app or a company could be a part of your college application but I guess it makes total sense it's just R how do you differentiate if you're if you're you know like a Harvard or Sanford or something how do you like you get so many people they all look the same on paper um anything unique I think is probably really valuable I don't know I never paid attention to that stuff up and I never really put much stock in it
12:30 - 13:00 um so I I I went to school for one year um absolutely loved it I mean i' I've always like it was filled with amazing people it felt so like selfish to me to just be able to focus on like learning things and following my curiosity and all that I was still working on flashcards but I was like coming out to San Francisco all the time because of flash cards to meet with people or to do things um and towards the end of it um I think I just started thinking about the fact that there was was an opportunity costed me being at school um I also
13:00 - 13:30 simultaneously met met a really good friend at Harvard that I felt like would be really fun to work on something with and we got really excited about a different idea and so long story short we actually ended up uh dropping out of school um I applied for and got like the teal Fellowship which is the second year of the teal fellowship and we got into the YC summer 12 um cohort uh and we dropped out of school and joined YC that was kind of like the The Next Step tell
13:30 - 14:00 tell us a little bit about the Mystique of the the teal Fellowship it it feels like it's this uh super exclusive cabal of very very successful founder I mean I feel like if you meet a really really successful founder there's like a good chance they were a teal fellow um tell tell us about that what was that like and how does that all go down does somebody just do you apply does somebody reach out to you yeah I can't remember if someone eventually reached out but I definitely did do an application um I you know I feel like the first few
14:00 - 14:30 batches in particular it got so much hype and press um that like when I was doing it there's I think it was like CNBC or NBC or some someone like literally filmed the documentary during our entire like finalist weekend and there was like this like five-part TV series that was like airing it was like totally overhyped and they made us feel like we were like what's the like school called the academy and X-Men you know where there's like the for the like the Misfits and mutants it felt like that um
14:30 - 15:00 it was absurd uh but it does seem to correlate with like really successful Founders well I feel like especially back then like you're just looking at outliers like who who the hell that like at that age is thinking about this stuff and like actually wants to drop out of school or like didn't even go to school um because they were working on something ambitious and impressive and like if you look at the original lineup of people like half the ideas were absolutely crazy um like there was like asteroid mining there was an asteroid Mining Company everyone was like doing
15:00 - 15:30 something like you know Laura Deming wanted to stop aging she probably actually made a remarkable amount of progress in the last decade um but anyway like it was it was like a ridiculous cohort of people and very like extreme outcomes I think in both ways right like you look at the top half and it's it was like one thing and I you know honestly like I feel like half at least at the time I'd be curious to see where they are now like I felt like the other half were going to like end up like you know homeless or something like it was just like there was such a level of like kind of almost Insanity to the
15:30 - 16:00 group um or just like extreme um but yeah like it was amazing like it opened so many doors like especially back then like anyone was willing to meet with you um it was definitely a huge opportunity to like build a network early um which I've never been good at like doing intentionally but um it was like super fortuitous I think probably the biggest thing though is like the early Community aspect of that and like um all of us being in this thing together it was like move out to San Francisco take two uh
16:00 - 16:30 first time I hated it thought I was never going to move back and second time um you know I was now still very young but like you know still couldn't go to bars but like we ended up moving into a house with a bunch of other teal fellows and like uh it was just like an it was like a very cool Community they did really good like um quarterly kind of like off like trips with everyone so we would like really get to know each other and bond and so a lot of my best friends did like a bunch of people in my I got
16:30 - 17:00 married in may like a bunch of my uh uh you know Grooms men were teal Fells from that you know from that cohort and my co-founder for speak was a teal fellow and um a lot of the investors in speak are teal fellows I don't know it's it was a very very cool Community um and yeah really interesting so the co-founder that you met at Harvard that that was your speak co-founder that that's Andrew and you both got into teal and you both got and you applied to YC for speak at the same time so uh not not
17:00 - 17:30 quite um we applied we did a different company oh okay and my co-founder for the first company the first company was called Coco controller it was basically like a physical device that you would put on your phone and it would turn it into like an entire gaming console um and we can we can definitely go into that but that was the first company and um was kind of doing that in parallel with flash cards and then um Andrew my co-founder for speak was someone I met through the de fellowship and we were just friends um the funny story is my co-funder for the first company after
17:30 - 18:00 that was wound down he went back to Harvard finished um and now he's our coo so we kind of got the gang back together like a year and a half ago um and it's been really awesome uh in a parallel universe he would have just started speak with us instead of going back but you know certain things are I guess just inevitable so yes uh give give us like the quick on Coco controller like quick maybe quick lessons and maybe how long how long did that last that that entire cycle that was what we did YC with the first first time around I remember halfway through the batch we had office
18:00 - 18:30 hours with Paul Graham my idol and he basically predicted the next year and you one or two lines he was like you guys are in deep you're going to have a horrible time fundraising why because it was Hardware or yeah we were like 19 working on a hardware project we like like I taught myself electrical engineering to build the first circuit boards I had no place you know doing this we Mar we made it remarkably far like we were build like we were building units in China in a factory we we came close to getting a a distribution deal
18:30 - 19:00 with Apple but at the end of the day like it was super hard to pull off um we weren't the right team to build something like that at that time like we would have needed to fundraise like a lot and we weren't able to tell the story effectively so we didn't know enough about fundraising I think like the honestly probably the main thing I learned was like do software um yeah uh but but also like scalable yeah like there's just there was there's a lot of stuff there but uh but you know was it
19:00 - 19:30 was definitely that was the first time we raised money I learned a lot more about how to fund raise um it was very valuable like I definitely like am just a huge believer in like second time Founders versus firsttime Founders because you just like if you're an investor you get like someone they learned on someone else's dime you know and now you can like now they've learned a lot of like the initial mistakes gone through those pitfalls and then they can just learn to at least avoid that first set there'll still be more but um you know I think that's like that all that stuff was what I learned yeah with was there any element of I I I know you've
19:30 - 20:00 been way ahead of the AI curve and I'm sure we're going to get into that in a few minutes but was there any element of of AI and Coco controller there wasn't it was it was in the very early days like I think like image that had just come out like a year or two into the controller um and so by the time we were winding that down like the first few papers were kind of coming out and um all that was starting to ramp up um and so that was actually probably part of the reason why we decided to kind of
20:00 - 20:30 throw in the towel with it partially just because we we were I've always been from a young age intellectually very interested in AI it some it almost feels like you're like playing God where you're inventing intelligence like it's just like the in my view like the most incredible thing humans can do or achieve is create other sentient intelligence um or anything even close to it um and so I've always been interested in it but I think part of it was so for us like once we started seeing these papers come out it was
20:30 - 21:00 unavoidable it was like a gravity well in terms of just like we were of course going to start thinking about it working on it so so this was almost kind of an inspiration for for shutting down coko controller yeah like I don't really remember the exact sequence I think I think we' been like we'd kind of already given up on it Apple had given us Apple was making our lives very difficult we were working through the audio jack like Square they didn't like that um they didn't like that we were trying to become a platform grams so they were like using all their levers to discourage us um but yeah uh you know it
21:00 - 21:30 was probably in the back of my head as well interesting okay so so so you wind down Coco controller then what basically the same time I'm winding down co coco controller I sell flash cards so I've been operating flashcards kind of in the you know on the sidelines while I've been working on this really like not doing anything literally just letting it kind of in maintenance mode it was still really popular getting you know millions of downloads and users um and and yeah I think
21:30 - 22:00 like it was probably within a matter of months where I I I had I had both basically shut or one shut down one and one was acquired so after flash cards was acquired I was like uh the one thing I knew for sure is I was never going to actually go work at ch um so I was like an adviser on paper helped the transition but that was pretty much it and then I like they try to get you to stay yeah they wanted me to stay um obviously you you'd rather have the the you know founder stay but who knows maybe they didn't I was like it's this Punk that like mostly got
22:00 - 22:30 lucky um like no actual software like real software engineering abilities um and uh you know at that point who who knew the scy morphism maybe wasn't as good um but uh anyway I think like I I remember like just kind of having a little bit of like an identity crisis like I was like I always had flashcards as part of my identity um I was like a Founder type what am I going to do next I thought for a hot second like maybe I
22:30 - 23:00 should go back to school threw that out pretty quickly um though um and then yeah like all this was kind of happening at the same time but I think I was already kind of becoming starting to become really obsessed with AI and I just realized all the papers I'd read so far every like I just couldn't stop thinking about it um I I wanted to spend all of my time reading papers and trying to like Implement things but I felt like I needed to learn a lot more before I could Implement a lot of stuff there were certain things that I just didn't
23:00 - 23:30 understand at a deep enough level to really grock a paper and so at that point um me and my future co-founder Andrew for speak um we were roommates at the time um and and Colton actually our coo now my old co-founder all three of us were basically like we want to study AI I think we want to kind of dedicate our livs to this um Colton decided to go back to school go that route we decided Andrew and I decided uh you know we're not going to do that we're going to we're going to we're like we are in the center of it
23:30 - 24:00 we're in the like all these papers are people at either the big tech companies or Berkeley and Stanford like pretty much universally there's not going to be good people on the East Coast even to like learn from um and so we just decided we're going to spend 18 months of our life doing nothing but just trying to become as like knowledgeable as possible about everything that's happening and that's basically what we did like the the biggest thing we did was we literally just um like looked at all looked up all the people from all the big papers um and we just tried to
24:00 - 24:30 like spend as much time with him as possible so like Andre karpathy I remember like he was someone he was super nice like we literally cold cold reach out to him we're like we'd love to buy you coffee can we just like interrogate you about you know the latest blog post he wrote um so he he would meet up with us which is amazing that he was willing to do that um we ended up uh I think shortly after that we ended up discovering there were some pretty good courses that were just coming online for all the latest stuff at Berkeley which is remarkable because
24:30 - 25:00 literally none of this this discipline didn't really exist a few years prior so like having a a university course on like deep learning in 2015 or whatever it was not not a thing nowhere like nowhere and Berkeley was I think one of the first like real I mean there was like some super early like machine learning stuff of course but not like the the interesting Cutting Edge stuff and um uh we found a course at Berkeley it was like a graduate level course um it was about reinforcement learning which was like the newest like kind of most interesting uh can you just explain
25:00 - 25:30 it for listeners reinforcement learning yeah so um you know if people are following gen they they you know may be they may have heard of this concept of like rhf right reinforcement learning human feedback um and basically the idea is that this was like um like when people were you know uh releasing models to like you know beat games alpha go um uh or like Starcraft or any of this stuff this is what reinforcement learning is it's essentially like um strategy based intelligence so it's like
25:30 - 26:00 uh making the correct move over time and kind of keeping a strategy in mind to execute over time and do and like basically self-learning that so like just like trying a bunch of times until you get like the rewards and so that's why it's called reinforcement learning anytime you do a good thing you say good robot and you give it a reward anytime it does a bad thing you don't give it the reinforcement and so very much like inspired by how the brain works and how humans learn um so this was showing a lot of really interesting promise it was very intuitive that it would become a
26:00 - 26:30 big like maybe the path to AGI and so a lot of people were focused on this um it ended up kind of mattering with like large linguage models because rhf was a big unlock and so open AI even though they're well anyway it's tangent but open AI like was super interested in RL in the beginning everyone was and then pivoted to llms but like the RL stuff actually ended up being very useful still um but anyway yeah we we found this course in 2015 back to the back to the story and uh um we were like what if we just show up
26:30 - 27:00 um so it like had just started I think it was like September of that year and I remember going to Berkeley and we like take the BART there me and Andrew we have our backpacks on we have like our notebooks and we're literally walking I think it's like was like the first week of school because it was like the first lecture and we remember getting stopped by like an undergrad and he was like hey my name's you know whatever and he introduces himself and he's trying to make friends because he thinks we're also just like these undergrads going to school and we felt like such imposters cuz we literally were uh and we end up
27:00 - 27:30 like getting into the building and finding the classroom and we were expecting a giant lecture hall and it's literally like 12 people and it's like a graduate level course about this like pretty specific topic and so there's no one in the course it's like it's like no one had ever even really heard of reinforcement learning they didn't even know what it was and so we just like walk in and sit at the back of the classroom everyone's staring at us it's very clear everyone kind of knows each other from other courses and they're like who are these people um but
27:30 - 28:00 everyone's a little too awkward to actually like confront us and including the I think the lecturer definitely knew for sure and uh yeah it was it was uh so we just literally would go every single time um the so so the the lecturer ended up being John Schulman who ended up being one of the co-founders of open AI um he like he was the one teaching it even though Peter Abel was like the actual Professor but like I'm sure AIA like had less less knowledge of reinforcement
28:00 - 28:30 learning because it was such a new topic than John Schulman did who was like I think grad a grad student at that point maybe maybe um but anyway uh it like yeah it was like literally I'm sure the people that were in that room all if I knew who they were like they probably have all gone on to like join research labs and such did he ever confront you did he ever did he ever stop you hey guys I I know you're not in this course what what are you doing here yeah never did you ever have to submit any work we so everything was on a public um everything was on a public website and
28:30 - 29:00 so it was amazing he would just give us the website and everything was there he also was I think there was a period where it was literally like three or four weeks in a row where he was literally making the courses we went and he like was too busy and so he didn't create homework and so there literally was no homework for like three or four weeks and then he finally reveals this thing and he's like okay this is going to be your new environment where you're going to like build your models and then it'll test on this and I'm pretty sure this was like the first version of the open playground or gym which is like their big reinforcement learning thing
29:00 - 29:30 that he ended up doing at open a and he literally like spent four weeks on it to do our like to basically give us like a homework environment to build models in um like you know casually producing this for a student so it was pretty cool did did you ever tell him this story I don't I actually don't I'm actually pretty good friends with his uh his wife now um I I I don't think I have it'll be hilarious at some point I will um that'll be a great yeah that'll be an amazing i' be very curious if you rememb two randos in the back of the classroom uh that's
29:30 - 30:00 incredible okay so so you learn everything you need to know about AI by sitting in a Berkeley class that you had no business being in then what yeah everything was published on archive we like it was such an you know open time in terms of all these papers being published basically like what we did was our our entire strategy was we want to just like Implement as many papers as possible build and Learn by doing and um we eventually started building speech models models um and so uh like one
30:00 - 30:30 thing that I think uh was very very very important for speak was we were building a um a speech recognition model and while we were building it um we ended up deciding why don't we try to like actually understand more than just the words that people are saying also people are like speech recognition models today are super bad for people speaking with accents or in Lou environments like it's they basically were not like they not accurate whatsoever like basically not useful and we were like what if we try
30:30 - 31:00 to intentionally find accented data on YouTube um we'll find like the BBC for example all that's in in British English we'll try to find stuff in like uh Indian dialect English Etc um and maybe we can make the model robust by including all this data um and we ended up building a model that could detect what accent people were speaking with and was incredibly robust to Accents as well and it was like basically state-of-the-art and and it was like a weekend project we did um and I think for that that was like a
31:00 - 31:30 very kind of you know holy moment for us because it the results were like mind-blowingly good um and it really laid laid bare to us it was like probably the first thing that we did that was somewhat novel and it laid bear to us that these things are really really powerful and we're clearly like people are only we're only mining the surface of what these things can do and in our minds I think it became very clear that it was if you could extrapolate over time it was merely a
31:30 - 32:00 function of the amount of data that you have and like the size of the model in the compute and invariably there'd be like like ways to make the models be even like more efficient for training and everything but as you do that the accuracy of the model goes up and up and up and up like this and we have like the entire internet to train on and so it is just a matter of time until these models get to the level where at some point they're like better than a human right um and at the time at that at that point
32:00 - 32:30 that's like an inflection because for all these models they're doing tasks that right now you know fundamentally the only thing like machines can't do and so as you approach human you start to become really useful as like a a helper tool right like um it's like an assistant tool in the way that like co-pilot helps people you know write code but it's not going to really write a program all by yourself but once you surp and that's super valuable by itself but once you surpass humans that's where it's really a true unlock um and we became convinced that like
32:30 - 33:00 this is just a matter of time and um the kind of like Mantra that we've had from from day one is basically that like in the next 5 to 10 years we'll be able to replace humans for language learning but that's kind of how we think about most things it's it was basically just like it's only a matter of time there's an inevitability here these models clearly have a lot more juice to squeeze and um I think that's like where everything started from was just like technological vision and really just like complete conviction like it was it wasn't just a
33:00 - 33:30 vision we knew this was going to be the case um it was very very clear um like that the models would just continue to get better I mean um maybe not like AGI on a certain time scale but like but like the models would at least in narrow cases surpass humans and that was really like the the underpinning for speak once we knew that once we knew that we thought at least speech models could get to the point where they were eventually superhuman we could see that like all the pieces would unlock to to to do a lot of things and the thing that we kind of immediately got to was language
33:30 - 34:00 learning we envisioned like we want to build we were super obsessed with the idea of like okay if in five years we can we have a model that's sufficiently good what is like the thing that we want to be building what feels like the most magical and we wanted to make sure that it was something that humans could form a relationship with uh something that felt really magical and that could really help them become like better versions of themselves for a variety of reasons we got really focused on language learning and and that was kind of like the way we started to speak we knew absolutely nothing about language
34:00 - 34:30 learning but we were convinced that over time there'd be this thing that would be unlocked and we'd be able to help literally millions of people improve themselves in this in this way that really like changes changes people's lives so I want to drill down into an Insight you had before we dive into the speak story which is you mentioned that back then and and by the way was this 2015 2016 when you're having this realization about the data so it feels like post chaty BT you know there's this there's this big realization that if you feed these
34:30 - 35:00 models enough data they do exactly what you just said right and and and that seems to be a bit of a an Insight or a breakthrough moment that I feel like a lot of people have had over the past year it sounds like you guys could could tell this was going to happen in 2015 am I understanding that correctly and do you feel like there were a lot of people like you that that knew that this would happen with enough data and enough time it it wasn't like a secret no no no no this was not a secret this is not just us like like uh it was just a smaller group of people right um that
35:00 - 35:30 that knew I think people that like less people were paying attention but I mean when we did YC we so we did YC for speak again second time and um that was Winter 2017 by that point so we started in 2015 through 2016 started speak at some point in late 2016 and then actually did YC um by that time there was already a mini AI hype cycle that was happening like there were there were a fair number of AI companies that existed in that batch and there were VCS that were really
35:30 - 36:00 interested in Ai and it was very hyped like I remember a lot the most one of the most common questions we got was like well you're not actually trying to like teach people languages though right this is a trojan horse to collect data so that you can like build models and then offer an API to other companies and do a B2B angle right and we were like no like we uh we actually just like think that when this happens we can unlock like a consumer use case that like billions of people want and we think that'll be a really big company and they kind of looked at us like we were crazy um but like there were a lot of people
36:00 - 36:30 that were interested in like the data plays to unlock big AI models and that was that was not like a super I mean I think we had the unique thing about us is we had a lot of conviction to persevere even though because we really did understand it wasn't going to happen tomorrow it was going to happen in 5 to 10 years um and we would we needed to like climb the steps to get there over time so we'd be in a really good position to continue to build towards the long-term vision and so I think a lot of these companies like an interesting question is like of this earlier a AI like hype cycle like what
36:30 - 37:00 com big companies ended up emerging and I think almost all the value if you think about it went to the incumbents like Google like you know they were able like the YouTube algorithm is so much better now because of deep learning as just like one example ad targeting huge unlock like literally um you know billion tens of billions of dollars of value probably um so so there are lots of unlocks but it all went to incumbents and maybe there were a few like um you know pick and shovel companies like scale um that that emerged but there weren't many but it seems like to your point about
37:00 - 37:30 the YouTube algorithm and you know I can think of other companies that have had similar breakthroughs it's probably more on like you said the sort of deep learning side Less on more of the generative side which has really emerged over the over the past year or so is that fair to say I mean I guess what's really the difference is I mean I I consider generative AI the creation word yeah but I'm thinking about speak how it's talking back to you right or I'm thinking about mid Journey or these things that are actually like creating
37:30 - 38:00 content and and experiences I I don't know maybe maybe I'm mixing the two I mean I think it's it's a little bit nuanced there's there's clearly I think something different now but I I think really what's happening is that all these models are like the transformer was created the the attentional all you need paper I think came out in like 2017 something like that um and so a lot of these models like uh you know they've like all this is existed I I don't really like think of the like for example whisper is superum level speech
38:00 - 38:30 recognition um that's Transformer architecture and it's it's just generating a transcript given an audio input right at any any model is fundamentally generating something it might just be like a classification or a recommendation but it is like still generating some sort of output from an input um and so I I I always like am a little bit amused by the like the fact that we use we're using this term generative AI I've I've like C I've started using it as well just because it's the term it's like the par L now at
38:30 - 39:00 this point um but like I don't really think I think the big unlock was more just like that these models got really good like GPT 3 came out in I 2020 or 2021 um it just wasn't as good because it was like Da Vinci 1 or like Ada before that and um the thing is like these models are just continuing to get better the curve is is starting to become even sharper over time right and I I think like the only thing that's happened is like B basically it's continued to get better to a certain point that unlocked value and really
39:00 - 39:30 more than anything um chat PT proves that the UI mattered right like chpt was just the da Vinci 2 model which was available for I think nine months before chat PT came out anyone could have built chaty PT basically um if they had just like released that no one did and then they did and everyone paid attention um and so I yeah I think really the function for me is like it's the same thing it's like everyone was working on llms have been around for at
39:30 - 40:00 you know at least three four years at this point and it's just a matter of like getting above a certain like utility level I think so you decide you want to do something in speech but you don't really know what yet give us the story of of speak from sort of the beginning to product Market fit and and and when did that happen you know we talked about product Market fit earlier you now have the perspective that you know when when it's real and when it's not when did you know it was real and and give us the story to get there yeah well it definitely took a few years of
40:00 - 40:30 grinding um we knew nothing about language learning I think even to back up a little bit further we had this technology story in our heads but I think one another thing that was super helpful that's a very like uh a very specific memory for me is that Andrew and I spent an afternoon and we just posed the question in five years from now what kind of company do we want to be running at that point we decided we think the best way to do this is not to go into research but to start a company had you started the company yet I don't think so yeah this was probably pre
40:30 - 41:00 starting the company yeah got it okay still just an idea okay got it yeah when we were doing research we didn't even know if we wanted to start a company necessarily we were intentional about like the best like for our given motivations we we weren't necessarily sure that we wanted to do a company but um yeah basically like we were like in five years what would be like the most kind of incredible kind of company or position to be in and we became really captivated by the idea of building something magical um something that really um felt like novel and magical
41:00 - 41:30 and and a big part of us for that was like something that you could actually interact with and it would it it would be like another human in the sense that it would feel like it understood you and it was and you were going back and forth in some way um and we felt like it was clear that like chat bots in general were a ways off and it was less clear to us said that would be 5 years from now but we felt like uh and also when that happened it it would be something that like a giant Tech platform company would
41:30 - 42:00 probably do and not us um we didn't we didn't think that we would be the ones to be able to do that yeah and also it it's just like such a general problem space that it felt like they would be in a much better position to do it um and we always knew it would be like the hardest problem to do like the bar would be that much higher to be able to do that however like one thing we realized was like language learning is actually the perfect thing here because it's actually okay if the model isn't perfect it's okay if it makes mistakes um and people are willing to talk with it anyway because it's way better than the
42:00 - 42:30 alternative which is not having anyone to talk with um except for like a human which is judges you and is inconvenient and is expensive um and we realize it's like one of the like we thought about like what the categories could be of like what people want to interact with we always we knew we wanted to do something more like straight consumer we knew we didn't want to build for Enterprise for this um and um it's actually pretty limited like we we pretty clearly identified early on like language learning is one of the few things that adults do um where they're
42:30 - 43:00 willing to interact with a system like this and have it not be perfect and it's one of the only areas of self-improvement and like uh especially learning that that adults do in Mass uh in general um and it was perfectly aligned for uh what we saw as like this the speech recognition would get to human level faster than the language side would if that makes sense what what do you mean not perfect like like it's teaching you the wrong things or just the language that it speaks to I mean
43:00 - 43:30 let's start with the speech recognition let's start with the speech recognition our first our first wedge here was the speech recognition we knew we could do that well um and better than better than what Google could provide or anyone else because the whole idea was we can get all this accented data from our users and then use it to train a model that's specifically good at like understanding them and so let's just start there but like even if we don't pick up every single word the user says perfectly all the time they'll still use the product that was a hypothesis it because it'll still be useful but like if if a chat
43:30 - 44:00 bot like says kind of crazy things every once in a while or doesn't really like understand your words all the time like you're just not going to use that the bar is too high it's not like a finance application or something like where the stakes are really high we're just having we're having a conversation it's it's fine if it's not perfect okay got exactly yeah yeah yeah um and so it's actually one of the few cases even today where you have back to what I was saying earlier like there's a difference between like an assistant to tool where the human is in the loop right co-pilot or even for like legal right like it's
44:00 - 44:30 going to be a very far time from now where you know we'll we'll trust a legal review that's done by Ai and there's no human in the loop right the stakes are way too high and mistakes are costly um and so there's actually very few use cases even today where you can have the AI fully replace the human with no intervention from a human whatsoever um because for most things the stakes are just too high um and so Chad is one of those I think language learning is another um I don't think there are that many right now um when that happens for
44:30 - 45:00 other Industries it'll be a really big deal for those Industries too but I think that the hallucination rate and like a variety of other technical challenges are still too great that it hasn't happened yet um and I think that's like one of the most significant kind of like Trends got it got it that makes a lot of sense actually and it's really good context and framing for what can be a viable AI application in 2023 or 2024 okay so so you know you want to build a speech product because you feel like there's a tolerance for mistakes
45:00 - 45:30 then then what did you did you know what you wanted it to be at that point we had um you know very little of a vision around the specific way that we were going to teach people um like knew very little about language acquisition um and so no we started from scratch on on all that and that was like a huge challenge for us um I think there were a few points where we narrowed narrowed it down like something we fundamentally knew from very early on was we wanted to focus on speaking first because it was
45:30 - 46:00 the area that was most important to people and the least supported there are already apps like dualingo that helped you with vocabulary building basic grammar all that that was covered that's been covered for a long time it's like less way less valuable um but speaking wasn't and it was also the area where we could build the best models to actually like replace that that piece and actually get people speaking the other thing we realized is that we needed some sort of like long-term Vision where we could build steps along the way and continue to improve a product Market fit and it was still valuable in the intervening years we did not want to
46:00 - 46:30 research project for 5 years until the models were good and then we're just going to go releas something we knew that like getting actual product Market fit was tough it was going to require a lot of iterations and we just wanted to like launch things and rapidly move um and like maybe the product Market fit in the beginning wouldn't be as strong but we knew that over time it would just we would you know be with the rising tide um that's probably one of the things that you learned from flash cards that you need to be shipping product and getting feedback otherwise like I mean so many first time Founders just build stuff for years before they ship it but that feedback is so important it's
46:30 - 47:00 something I learned with the controller as well and it was a hard a limitation with Hardware even though I knew I needed to be doing that I couldn't do it it's because it was Hardware um so yeah 100% um and then yeah I think like um from there we we quickly honed in on the idea that like okay it's a vitamin for like people like us that are native English speakers that want to learn Spanish to go to Mexico City or something but people that are trying to learn English wow you don't realize that living in um you know a place where most people are speaking English but if you don't live in an English speaking country it is like such a common need
47:00 - 47:30 and it is an absolute painkiller and man there are way more of those people in the world than the other way around so we quickly honed in on teaching English as well at least that is that is a brilliant Insight it it maybe it's such a subtle thing but um especially given that you know so much of tech silic and Valley obviously based in the US you you naturally think of products for People based in the US right like that said are going to Mexico City and and clearly that's where products and companies like dingo have been very very successful
47:30 - 48:00 right maybe it seems obvious but like how did you have that Insight um it's a simple one but it's so powerful I think like very early on we realized like even for me and Andrew like we realized well we're not that motivated to learn a language like okay do we know people that are and so the motivation piece came up very early and then I think we just naturally as we were thinking about the market we realized like wow but what about the people that are in the US that need to learn English okay well what about like oh the fact that English is the lingual franka it's the way that a
48:00 - 48:30 Chinese person speaks to a Korean person as they speak in English um and then we start to really uncover it it's a total blind spot for a lot of people in the Western World though especially the English speaking countries um and so yeah I mean uh I think that was a a really big insight and you know also at the same time like a very hard pill to swallow because all of a sudden we were operating this meant operating in markets where we didn't speak the language we weren't familiar with the culture we weren't we didn't even like understand how to grow an app in these
48:30 - 49:00 other markets and it was like a cerebral Insight but it was very like uh difficult to operationalize it if that makes sense um and it was definitely like something that was required like a lot of like commitment again to to the idea that like longterm this is the right strategy yeah and then from there we realized okay even if we're doing that now we're talking about teaching to a Chinese speaker a Korean speaker a Spanish speaker a French speaker a um you know a Hindi speaker right like all these different language groups that
49:00 - 49:30 were so vastly different how do we start um and so we ultimately uh we could go into it if it's interesting but we ultimately decided we're going to even go more specific than this we're going to just pick one country and one language and we're going to get really really zoomed in to that place uh at this point we've been struggling to find product Market fit for a while I should have like probably how long like how long is this journey to be like hey we need to be star in other countries other languages like I think probably it took us like 6 months to like even understand
49:30 - 50:00 what like a early iteration could look like that made sense and like testing a bunch of bad ideas first and then we realized you know pretty shortly after that okay we need to teach English and then it took us probably another three or six months to realize okay we need to go small we need to go specific we need to go pick a country um and then it took another year um to it um so we were grinding for w for years um but yeah we ended up picking um we ended up picking South Korea um how do you do that like I
50:00 - 50:30 mean yeah I wouldn't even know where to start yeah yeah it was uh well uh yeah I think like it was a combination of many things uh this is probably one of the most common questions I get from like everyone from prospective employees from um investors from random people um it was a combination of the fact that number one um we had a product out in many markets and we saw that Korea is one of the places where people were using it in really interesting ways and
50:30 - 51:00 so there was natural like this population of users clearly there's something going on here let's take a closer look I also ended up literally just traveling to a bunch of different countries and talking with our users for this early product and Korea was one of those countries and the thing that immediately stood out when I was in the country and I had done the research right we we saw that like a crazy statistic that South Korea spent 1% of their entire GDP on learning English um um and all this like stuff around like the national obsession with it and the
51:00 - 51:30 idea that no one but it's still like one of the lowest like English speaking ability uh countries um but really I think one of the things that cemented it for me was the fact that when I was driving down a street in gdam with all the big skyscrapers um my my friend who was kind of my tour guide was pointing out different skyscrapers and was like that is uh English learning academy that's an English learning academy every single one of those windows you see in this giant skyscraper those are all Eng classrooms there are people here that are going after work or after school and they're studying for 3 hours late into
51:30 - 52:00 the night every single day English um and I was like okay this is crazy this Market is really interesting um and we realized like okay hyper competitive market lots of noise out there lots of different English Solutions and companies and products super noisy but if we can build something that's truly technologically differentiated true product differentiation right that like that Pro proverbial 10x um then we'll get really good signal here cuz we also know people are really eager to to
52:00 - 52:30 actually get success here and um they're willing to try anything and so we'll get that market pull more here than anywhere else we'll get the signal we care about for this long-term product Market fit which is the only thing that matters um here more than anywhere else um and so that's basically how we did it I I I can't imagine like a more appropriate or more fitting metaphor for potential disruption than you driving down the street and seeing English teaching skyscrapers and saying to yourself we're
52:30 - 53:00 going to go we're going to go disrupt that with people that are teaching that don't speak English by the way right that is crazy um so okay you realize South Korea is your target market um by the way you said you had a I want to back up a little bit you said you had a friend as a tour guide like what was this you did you have like a person in each of these cities like giving you a sales pitch for why their country should be speaks target market like explain that yeah some of them I did some of
53:00 - 53:30 them in in some way or another I just like you know it's like um Andrew's parents are from Taiwan so like when I was in Taiwan his parents were showing us around in Korea it's actually a really funny story as well this is where this is the single biggest Roi I got I I guess I met my first co-founder as well but the second or equally positive Roi um for going to Harvard for one year which is that one of my roommates freshman year happen to be someone that uh was originally from South Korea lived there until he was like 11 or 12 and then moved to the United States and so
53:30 - 54:00 he was like one of these rare individuals very rare individuals who's like perfectly native in both places and still like has a foot in both um and so I actually um he's like a very good friend of mine um and I texted him when I decided I was going to go to Korea and I was like Hey how do you feel about going to Korea and visiting your family all expenses paay trip from um you know on my company and he was like all right I I'll I'll take time off right now this sounds awesome um and uh so he flew flew
54:00 - 54:30 there with me um and we had an amazing time because he was we were like really good friends but he also is this he was working in New York at the time at Rent the Runway I like a a product or data analyst or product analyst and uh he was asking like when we were doing interviews with users he was like he would just talk with them for like 10 minutes and he'd be like I'd be like what was that entire conversation he was like well okay I knew you wanted to ask this but ultimately like this is the more interesting thing so we just like kind of had this entire conversation
54:30 - 55:00 don't worry I have notes here but basically this is the takeaway and it was like much higher quality interviews um because he was just like so good and so smart um so maybe that was also a reason we went to South Korea you know uh unknowingly um because the interviews were so much higher quality um but yeah I was in the back of my head that was like in November in the back of my head I was like okay um that was that was awesome doing that with him by December we decided a few weeks later okay we're going to go to South Korea this is it um
55:00 - 55:30 and so I texted him again and I was like hey I have another crazy idea what if you quit your job move to San Francisco in a week and join speak as basically our first employee and help us uh figure out South Korea uh and he's like okay let me think about it and then I think literally later that day he was like all right I'm in um let's do this and he's still at the company to this day absolutely like pivotal role what's his role literally he's done every role Under the Sun up until now he he like
55:30 - 56:00 figured out our content pedagogy strategy from scratch became an expert there did that like really pushing the boundaries there I think it's one of our strengths is like how much better we are in terms of language acquisition than anyone else we' basically like invented our own Playbook um now he's like now he's doing product work um but he's been yeah he's done so many different things and was like he's just like classic kind of like Barrel versus ammunition type um you you know you could just give him something but like uh he is like the secret weapon that allowed us to figure
56:00 - 56:30 out South Korea because the last thing I ever imagined in my life before starting speak was that I would be for like a good three or four years exclusively operating a South Korean consumer app company yeah it's such a crazy concept right I mean especially you're somebody that's built a product in the US before um you know as you and I have talked about I'm I'm a former founder I've built products here in the US and it's so I I think it's so important for the found ERS to be able to see that sort of real time feedback between what you know
56:30 - 57:00 what you're shipping and what the users are saying like in real time you can't you can't see that I mean I I don't think you speak the Lang or maybe maybe I missed that part maybe you do now so like what is that like what is it like building a product for another Market especially a startup where like that that signal is so important I think especially even consumer right like getting consumers to drop everything in their busy lives and go spend their not even their money their time on an experience especially not like an addictive entertainment experience or a
57:00 - 57:30 game but like something that like takes real motivational willpower um this is why it took so long to figure out because it's like it was like I I like to joke that like every company has like their one unique challenge they have like the one thing that's like really weird about that company for us it was the fact that the founders had never been to South Korea before we didn't speak the language and we were operating in you know San Francisco and not in Soul um that was that was our Challenge and like it very much felt like in many
57:30 - 58:00 ways it was like one hand tied behind your back when it came to like product iteration and product work um and really force you to like do the like do the best practices not not take the shortcuts um which I think really strengthened us and the other really big benefit was that when it came to International expansion which we have been heavily focused on this year um it's not like most companies when they start International expansion they they're like oh okay we're in live in the US how do we go operate in like the
58:00 - 58:30 UK or something for us uh and it's like a very new muscle to build um for us we've our first Market was International expansion like we've been an international expansion expert company from the very beginning so we had four years of doing that before we went to another market and so um you know we knew how to hire we knew the Playbook we knew like how to localize we knew how to like design in one language but then ship something in a different language we knew how to like do the interviews everything like was something we had like down to a
58:30 - 59:00 science um in terms of like operating and expanding to New Markets um from the very like from you know many years of doing it in South Korea um which is another funny kind of consequence so once you decide to go all in on South Korea at this point you have the product available globally now you're just deciding to focus is that right and then and then how long to till things really start working in South Korea yeah it probably took after being exclusively in South Korea it probably took like a few months until we started getting like
59:00 - 59:30 some real poll and then another like year before we were like okay this is something that we definitely want to like the foundation is there and we want to continue to iterate on top of this core like vision and and model um and so there were a lot of like little things that all incrementally added up to like a cohesively good experience but um I yeah like I I think it probably took you know at least four more quarters until we were like at that point and then we started like really thinking about growth um and trying to actually like scale that up got it talk talk us
59:30 - 60:00 through the actual speak experience I mean right now we've been talking mostly about the insight to be in another Market to teach English as a as a language but for people that haven't tried the product um and I imagine there there there will be some that haven't given that it's you know it's it's targeting teaching English um what is it like compared to say dingo for learning a non-english language yeah well the name speaks kind of says it all right like the entire idea is we
60:00 - 60:30 want to build an experience that feels as close to possible to just kind of having this you know across the table experience where you're chatting with this like friendly tutor type and they're just teaching you from the ground up how to become conversationally fluent by getting you to speak out loud a lot like a ton like you're speaking more than 50% of the time and we're not teaching you grammar we're not teaching you like weird vocabulary um we're literally just going to focus on teaching you the like core
60:30 - 61:00 building blocks of actual fluency as if you were in the country or you were like a kid learning for the first time like we're just exposing you to the language and instead of for example teaching you you know uh like a classic thing in America or in any language uh school or uh you know classroom is like the idea of like okay we're going to teach you how to conjugate this verb in every single form and then we're going to teach you the the present tense and then the past tense and then the future tense and the like XYZ grammar and you're going to learn every single one of those
61:00 - 61:30 forms and memorize it and then we're going to teach you some vocabulary and then you're going to be able to say sentences like um I eat at the library I ate at the library um you know like sentences that you will never use in real life at all right and for us we're like we're not going to teach you any of that we're going to teach you like the the three most common ways to use the word eat in actual phrases you're going to use like you know I um I love to eat out or I like to do this thing or um I um what did you have to you know for
61:30 - 62:00 breakfast yesterday oh I had a blank we're not even teaching you eight we're actually teaching you I had because that's the thing that you're actually going to learn and use um The High Frequency stuff and we're not teaching you the underlying logic necessarily of the grammar but we're going to teach you like the phrases and the patterns that you need so you can be like I had a or um I had the blank with the blank um and then we teach you a bunch of relevant vocabulary words that fit in into those sentences and then we marry it all together by having you learn a bunch of different phrases and patterns that you
62:00 - 62:30 repeat so many times that it becomes automatic those words are now collocated together so they all roll off the tongue automatically and you're not translating in your head you're never translating and then we married all together by having now all of a sudden you know five or six different patterns and phrases and a little bit of vocabulary that fits with those things and we put you in a situation where those are actually all the building blocks you need to functionally go do something that you need that you care about like I I want to talk about like going out to eat or I
62:30 - 63:00 want to be at a restaurant and be able to order food and all of a sudden you realize oh wait I can do that now I've just unlocked this like little part of the universe of fluency of this language like for this specific use case I can go do that now at a fluent level it it makes so much sense um you know anytime I've ever learned a language uh and I'm sure you've had similar experiences it's about the immersion right it's about talking to somebody it's not about going through those hypothetical exercises like you mentioned so is is is the Insight here that enabled this uh basically that you could take large
63:00 - 63:30 language model you know style application like a Chachi BT and you could take you know really really great text to speech and you could effectively enable a human style voice voice conversation uh is that really the Insight that led to this product breakthrough does that make sense yeah I think that's part of it different than like a rules-based system like I'm sure dingo has right right yeah I think that's part of it I mean I think the reality is that we we knew that these we
63:30 - 64:00 didn't have like the complete answer on day one but what we knew was that all these models were going to get better and better and better and surpass humans and we'd be able to use them in a myriad of different ways um but probably much more important than that and to our approach and I think this is super relevant for anyone doing anything in the generative AI you know world and you know a lot of people know this intellectually but they still I think make mistake is like you need to actually solve real problems and it's not just about a technology in search of
64:00 - 64:30 a solution um but like for in the case of speak a huge insight for us and we've seen this like this we've seen this mistake play out with all the other companies that are now trying to do the same thing is it really matters that we've been able to build an expertise and actually cross functionally fuse the machine learning and the technology piece and the the vision part and building that technology with the content approach that I was just talking about which was hard hard-learned lessons by doing lots and lots of interviews with users measuring progress over time and just
64:30 - 65:00 years of iteration and building that expertise and and figuring out how to marry it with the machine learning for today and tomorrow but then also like building novel product experiences that can that can combine the content in the pedagogy with the machine learning capabilities of today and tomorrow and doing all that together to actually solve people's problems and not become just you know obsessed with what we can do with the latest AI demo and and just kind of glue it all together into a product that people want to use once cuz it's flashy on Twitter but then they never come back and they don't retain yeah
65:00 - 65:30 there's so much AI tourism happening with products right now people trying things just because they're doing new things but to your point you need to solve real problems for somebody to keep coming back and and for you to have that product Market fit so what does product Market fit for speak look like you know you you you you know what it feels like when did you feel it um maybe maybe a way to answer that would be give us a sense of at least what you're comfortable sharing the scale of this thing in South Korea right now yeah yeah
65:30 - 66:00 I mean I think for me like at a high level way I think about it is like it's like Topline growth and retention those are like the things that matter like if you're growing naturally you don't it's not not because you're like an absolute paid marketing machine or something like that but it's just kind of happening without without you doing anything and people are sticking around and retaining those are the signs those are like pretty much the the only scorecard that matter in terms of like yeah where we are today in Korea like I you know we're the number one most popular education
66:00 - 66:30 out there uh I think we've had roughly 6% of the entire population have tried and used our app um now they're not all paying subscribers yet but uh we think we can get there um but yeah it's it's it's quickly becoming like a if it's not already a household name um it's it's getting there um and we've helped a lot of people um I think which is the most important part like a lot of people have substantially improved it's really cool it's so cool to see that like talk to users um and
66:30 - 67:00 like see how much progress they've made uh and uh yeah that's awesome so now like you said this year you've been very very focused on expansion of new markets any specific markets that uh worth mentioning yeah like uh there's I mean there's it is it I think it's a remarkably transferable product which has been awesome to see um and we yeah like uh Japan has been a really really interesting Market to us it's in some ways you know a lot of similar learnings just more um for Japan it's bigger um
67:00 - 67:30 even more intention in some ways which has been surprising um and then yeah generally like uh the we're seeing this as universal like we're launching we're live in many markets now we're live in over 20 countries we want to like double that by the end of the year um and basically the goal is to be globally available as soon as possible um and uh it I do think like the really cool thing about software and like the entire Vision here is it's not like we're scaling a tutor Marketplace that is you
67:30 - 68:00 know something that is tricky and like just squishy to actually be able to scale the quality across the new markets and figure out those Dynamics like a like a door Dash or something but it's just software it's just a it's just a way to get people talking and like this everyone learns languages and fundamentally kind of the same way you need to just like practice it and use use an experience like speak um so overall like what's been really encouraging is just like it does feel like anywhere where there's fundamental motivation people are willing to like go give the product a shot and then if it works for them they're they'll stick
68:00 - 68:30 around so if you want to be universally available then I I I'm guessing at some point you switch to teaching English speakers other language is that is that part of the plan absolutely yeah the United States is even though people here it's it's very us-centric Anglo Centric you know it's just an incredible Market there are lots and lots of people here duelingo has built a fabulous business here um um and so we think there's a huge opportunity here and uh yeah there'll be more to announce for that
68:30 - 69:00 very very soon I think cool do you feel like there are opport opportunities for speak to help teach people things you know other areas of Education through AI or do you feel like speak will always be focused on the namesake which is speaking and language I think like obviously there's a lot more to be done in the language space even just like other parts of it B2B for children in classrooms for really Advanced users Etc lots and lots of things left to do there
69:00 - 69:30 uh but I think that at the same time one of the one of the like biggest and most exciting opportunities when it comes to like Ai and changing the world and like creating real value for the world is education education I think traditionally has been like you know an underperforming not as exciting area of like venture capital investment and companies in general the companies just aren't that exciting there's not that many success stories there's a whole graveyard the actually like ironically
69:30 - 70:00 the reason why it's so exciting now is also the reason why there's been nothing exciting so far right which is the the reason that like technology hasn't disrupted education yet people are still learning in the same way that they learned basically 100 years ago basically since the Industrial Revolution when we invented the modern classroom um and there's just so much more room for improvement like fundamentally you go back 2,000 years the way that like PL learned or uh you know any anyone back in ancient Greece was the best way the way that the the you know most privileged people learned
70:00 - 70:30 was like a one-on-one tutor um and so that hasn't changed since then but the fact remains that like virtually everyone don't like you they don't have access to that so um that's the holy grail for Education it's like it's like one of the only things in education that's been like scientifically actually proven out the you know blooms 2 Sigma effect um in terms of outcomes for personaliz education um and so I think like there's just a tremendous opportunity here I think we're well positioned for it but I'm also not
70:30 - 71:00 arrogant enough to think that like we have the right to win here um there's so many there's G to be so many interesting things I'm really fascinated to see what Khan Academy does I think they're that's a really cool um opportunity for them as well um but but yes we're certainly going to be thinking about all sorts of things outside of language eventually or that at least that's the ambition if you weren't building speak where else do you feel like there's huge opportunity within AI at the moment yeah I think like there's a a few different Frameworks you can think about here but
71:00 - 71:30 like I think one of the biggest ones here is like thinking about the stakes of a of a use case and there's a certain amount of value that's unlocked when you become an assistant and that's like that's a total category of use case co-pilot for X that will be a thing for sure that's the majority of use cases I feel like right now especially on the Enterprise side I think that the fundamentally the most interesting long enduring companies though are playing in the spaces where the stakes are low enough or the models will get
71:30 - 72:00 sufficiently good enough to be above the the stakes if that makes sense um such that you can have something that truly replaces a human altogether the economic like you know gp4 isn't that much cheaper right now than a human um and so like when you can really replace the human altogether that's where you get the true kind of explosive margins of software and that is where you like really unlock the like real value um and so you know what that means I think
72:00 - 72:30 people can think about that for themselves but I would say like the fundamentally like the most interesting things I think are like where that is an opportunity uh not to say that the like co-pilot fra isn't also very interesting um but that that's like one way I would think about it it it's really easy to see like the upside case for that and and the way that it'll give so much time and productivity back back to people I mean how do you feel about the the downside cases of that like replacing humans net across the board um what what
72:30 - 73:00 will happen in that scenario I mean yeah I like I don't think I'm an expert in that I would just caveat it with that like really take it with a grain of salt but if I were to speculate like I'm fundamentally an optimistic person I think I believe that technology historically has almost always been mostly used for good um there's obviously always the flip side of that I do think there's a danger if we like basically get to AGI too quickly and we don't we're not um able to you know
73:00 - 73:30 properly handle that that's that's like the risk that I'm more concerned about candidly I think in terms of like the economics of it like everything is accelerating this will happen faster than the last technology wave but I do believe that fundamentally we're we're just in a better world with this technology we're able to continue to like remove the Yoke from humans in terms of doing the work that isn't fulfilling for the people is just purely a paycheck and fuel economic growth and be able to just unlock better better
73:30 - 74:00 jobs um so I fundamentally think that's the case but I I also do know there's there's totally a flip side here um for the people that are you know maybe later in their careers and less willing to like make it a transition and that's really tricky um I don't have the answers there but at the same time this is inevitable like I I I do think like fighting fighting against it is is kind of like trying to you know fight quick or swim up a river that's just flowing the opposite way like it's just it's it's hard to do um and so I think it's
74:00 - 74:30 it's one of those things where we need to just figure out how to embrace it you you mentioned early on earlier in the conversation that um when you were thinking about speak it was pretty clear like like when when it was going to get good enough by by feeding it enough data and you you also mentioned earli on that um you know one of the one of the coolest things you could do with one could do with AI is is build effect effectively sentient AI like I'm sure this question gets asked to you often but you've been ahead of the curve for
74:30 - 75:00 so long like when is that going to happen when is sentient like when is like basically AGI going to happen yeah exactly yeah um well the I feel like you know the the classic thing here is like it's going to be like 5 to 10 years away or or Beyond um and um I like I think it's very tricky I think it's like it's very unclear what that even means means um can gp4 pass the Turing test yeah absolutely that's what we thought that's what we thought AGI
75:00 - 75:30 was up until recently are we going to continue to move the goalposts probably human level intelligence is a little bit arbitrary um it's important in the narrow case I'm not sure it's important in like the macro case I think the the bigger like concern is like if there's like a runoff where it just continues to accelerate really quickly and exponentially and that's like the you know the super intelligence scenario I think it's pretty clear that's probably not that likely at least it's not going to be like oh we're going to blink our eyes and it's going to happen I think we're fundamentally very clearly
75:30 - 76:00 constrained by uh Nvidia basically um and gpus and it five years ago when everyone was thinking about reinforcement learning I was more concerned about that but like if it's going to happen with this current for and also by the way I think there's like some hard limitations like the transformer we're just getting we're squeezing a lot of juice out right now but we're going to need to make some breakthroughs to get to the next level I pretty convinced of that so that's why it's in determinate how long it will take because there will need to be some fundamental new breakthroughs I believe um but yeah I think I think
76:00 - 76:30 fundamentally like five years ago when we you know when everyone was focused on reinforcement learning it was less clear what that like what the kind of like um limiting factors would be for accelerations and I think now it's like pretty clear compute and energy are are those so like it's only going to get so much like Nvidia will only get so much better at making really really good A1 100s and stuff um over the over the next time and I don't think we're close yet in terms of having the compute to be
76:30 - 77:00 able to easily do like to be able to get to the point where that's the case when compute doesn't become the limiting factor that's when I would be more nervous about this happening and having the runaway scenario but as long as that's the case look you can monitor how much electricity uh certain you know super uh super computer cluster is using um and like that's a pretty good proxy for kind of like the level of of complexity and reasoning that the system can use so um so that's why I think like
77:00 - 77:30 I do think like open AI approach here is very reasonable and I I do think they genuinely care about AI safety and all that and I I think like I'm pretty sure they view it not that differently than that as well Connor uh anyone listening to this interview I'm sure finds speak to be fascinating so I I have to ask is speak hiring absolutely thank you for asking that uh um yeah I mean basically if there's anyone that's super interested in this uh especially if you're technical we'd love to talk to you um definitely uh
77:30 - 78:00 anyone on the kind of engineering product design machine learning stack uh we're looking to hire across the board there um and like you know definitely the rights kind of offs higher or growth higher as well um so yeah definitely um please reach out you can just go to speak.com careers uh to to find out more and get in touch with us and and for people who are in the markets where you're currently focused how can they try the product yeah just go to our website speak.com and and you
78:00 - 78:30 can download the app from there um and always love feedback um especially if it's in our new markets sure there's things we can do to improve H how can they give the feedback um there's a there's an easy way to give feedback with directly in the app um or you can just shoot us a line at uh contact at speak.com awesome or feedback atp.com yeah awesome Connor Thank you so much this was an incredible convers appreciate you giving us all this time and uh hope to have you on again sometime yeah thanks for having me this is awesome great
78:30 - 79:00 questions thank you so much for listening to my conversation with Connor follow generative now wherever you get your podcast on Apple podcast and Spotify and YouTube and wherever else you listen also if you enjoyed the episode please do us a favor and rate and review it it really really helps if you want to hear more from me or light speed follow me at Mig Nano on X or Twitter or LinkedIn Instagram everywhere and if you want to follow us at Lightspeed you can find us at Lightspeed
79:00 - 79:30 VP on all those same platforms thanks again to Connor for the awesome interview and generative now is produced by lighted in partnership with pod people special thanks to everyone our production team for making this podcast possible we will be back next week with another awesome conversation so don't miss it see you next time