Optimal. At this altitude, I can run flat out for a half mile before my hands start shaking. Can I answer your personal question? Now, what is the perfect time? I'm a cybernetic organism living over metal endoskeleton.
This episode is brought to you by AG1, the daily foundational nutritional supplement that supports whole body health. And I do get asked a lot what I would take if I could only take one supplement. And the true answer is invariably AG1. It simply covers a ton of bases. I usually drink it in the mornings and frequently take their travel packs with me on the road. So what is AG1? AG1 is a science-driven formulation of vitamins, probiotics, and whole food sourced nutrients. In a single scoop, AG1 gives you support for the brain, gut, and immune system. So, take ownership of your health and try AG1 today. You will get a free 1-year supply of vitamin D and five free AG1 travel packs with your first subscription purchase. So, learn more, check it out. Go to drinkag1.com/tim. That's drinkag1, the number one. drinkag1.com/tim. Last time, drinkag1.com/tim. Check it out.
Hello ladies and gents. This is Tim Ferriss and welcome to another episode of the Tim Ferriss Show where my dog is yawning and it is my job to deconstruct world-class performers of various types from military to entertainment to academia to sports, chess, everything in between and tease out the routines, habits, favorite books etc that you can use. In this episode we have Luis Von Ahn who is an entrepreneur and computer science professor at Carnegie Mellon University. He is known for inventing CAPTCHAs. We'll get into that. Being a MacArthur Fellow which is often nicknamed the genius grant and selling two companies to Google in his 20s. Luis has been named one of the 10 most brilliant scientists by Popular Science magazine, one of the 50 best brains in science by Discover, one of the top young innovators under 35 by the MIT Technology Review and one of the 100 most innovative people in business by Fast Company magazine. He has many things, many interests that keep him busy, but the primary is Duolingo. Luis is currently the co-founder and CEO of Duolingo, a language learning platform created to bring free language learning to the world. They have more than 100 million users, including yours truly. And that makes it the most popular way to learn languages in the world, and it is the most downloaded app in the education category on both iTunes, where it has a five-star average and more than 3,300 reviews, and Google Play. I first met Luis as an early investor in Duolingo. I tracked them down, which was very hard. I tracked Luis and his co-founder Severin down after you guys, my fans, actually told me I had to see it. It was in beta at the time and via Twitter and Facebook you guys informed me. I reached out to them. That's how that came together and every time I meet Luis, I learn something new. And certainly, I hope you will learn a lot in this episode. We talk about his favorite books and resources he'd recommend to entrepreneurs, language learning tips, the very clever way he caught cheating students at Carnegie Mellon, which is a great story. Early mentors and key lessons learned, the story of building and selling companies like reCAPTCHA, how to recruit and vet technical talent, the most surprising sources of users for Duolingo, and it goes on and on. So, we talk about quite a lot. Please say hello to Luis on Twitter, Luis Von Ahn, L-U-I-S-V-O-N-A-H-N. Say hi on the interwebs and please enjoy my conversation with Luis.
Luis, welcome to the show.
Thanks for having me, Tim.
And I thought we could start where a lot of people start in the morning, which is breakfast. You mentioned this briefly when we were doing a soundcheck. What did you have for breakfast this morning?
And you added some context. Is that the — you've had the same thing for breakfast for the last year.
Yeah, possibly two or three years. It's the identical Greek yogurt. It's the Fage Greek yogurt, the one that you have to combine the fruit but the spoon never fits in there.
And do you have meals that you repeat for lunch and dinner? I mean, have you standardized your daily routine in other ways besides breakfast?
Not that well. Breakfast is pretty standardized. My lunch is pretty different every day. We get free lunch at Duolingo and I just eat that.
Got it. Now, standardize — the word makes me think of testing. And when I think of testing, I actually got pinged by a former student of yours. And the student said, 'Ask him to explain how he catched — ' boy, here we go. I need to study English. Was it him or was it you? Catched.
Yeah, catched. No, I said this unfortunately.
Uh, caught cheaters in one of your classes at Carnegie Mellon, and I was hoping maybe you could explain how that went.
Yeah, I did. I did quite a bit. So I used to teach this class — I'm on leave, I'm a professor but I'm on leave — I used to teach this class called Great Theoretical Ideas in Computer Science, which is kind of the hardest class most people have ever taken in their lives. And it's basically discrete math and it's pretty hard. Now, a lot of people used to cheat. It's a large class. It's about 200 kids. A lot of people used to cheat in different ways. And I started becoming really obsessed with catching people cheating. I didn't even do anything bad to them after I caught them cheating. It was more like a fun game for me.
So for example, one of the things they couldn't do is — some of the assignments I said it is against the class policy to Google. Now, you can think what you want about that policy. I don't think it's that great of a policy, but that's a policy that I inherited. That class has always basically said you cannot Google for assignment solutions. So one of the things I did was I would sometimes assign homework and I would actually seed the answers to some of the things in websites that were crawled by Google but that I owned, and I could tell people's IP addresses and everything. I would catch people cheating that way. I basically set up honeypots for people to cheat and you know, you could say that's entrapment but that's what I did.
Do you recall any examples offhand?
Yeah, I mean, of the exact websites I don't know where they are or the terms. Yeah, one of them was — there was one that was called Girato's Puzzle. It's made up but I made that word up beforehand. I made sure there was no such thing on Google and I made a website that had the right solution, but it recorded everybody's IP address and at CMU you can figure out their dorm from their IP address. So I could figure out which person was actually checking. Turns out that time out of the 200 students about 40 Googled for the answer. And that was fun. I used to do all kinds of things like that and then the students were pretty — they were usually scared of almost everything being a trick. So I would do that in the first one or two assignments and then afterwards they would learn not to cheat.
So the story, the progression I've heard is that you would bring up slides silently. Did you do this? Where it was like a screenshot of whatever it is, the problem or whatever, circled, then a screenshot of the WHOIS for the domain site or the domain that people would go to with your name circled since you own it.
Yep. Yep. That's when you could see the whole class. You could see their faces change. It was pretty fun.
Now, but you didn't have to actually hunt down each identity, right? I mean, did you offer them a choice? I mean, tell me if this is true. So the story, the lore is that you offered them the option of confessing and take a zero for the homework, otherwise they would get reported. Or was that — or is it not — is it...
Who have you been talking to? That was the case. That was exactly the case because it took a little bit of effort to go and figure out who was who. So I would just say if you just confess, you get a zero on the assignment. And all kinds of people confessed. I even had some people confess that didn't even actually do anything. They just said, 'Well, you know, I don't know what the hell I was doing, so I'm just going to throw my name in there. I'd rather take a zero on the assignment.'
So if we rewind the clock, so that's as a professor. When did you become — at what age did you become a professor at Carnegie Mellon?
And was it clear that you were going to be involved with computer science from a very young age? Could you describe for people where you grew up? A little bit about your childhood?
Yeah, sure. I grew up in Guatemala. Which is, by the way, that's not where they keep the prisoners. That's Guantanamo.
I get that a lot. It's like, is that where they keep the prisoners? No, it's not. It's also not where they invented the avocado sauce. That's guacamole. Different. So yeah, I grew up in Guatemala. I started liking computers at about age eight. My mom — I really wanted a Nintendo because everybody wanted a Nintendo, or my friends all had Nintendos, and I told my mom, but my mom instead got me a Commodore 64.
A lot of Spy Hunter on Commodore 64.
Yeah. So, I played a lot of games there. But at first when I first got it, I was pretty pissed off because I thought, man, I wanted a Nintendo. But instead she got me that and she said, 'Well, you figure that out.' And so I had to figure out how to use the computer to play games. And then I started trying to figure out, you know, because she only would buy me a couple of games a year and I would get bored of them. I started trying to figure out how to get more and more games. And, you know, the best way to do that was basically to copy them from other people. So I started doing that at a young age. I think the statute of limitations has passed by now but I was basically pirating games on a large scale in Guatemala City is basically what I was doing.
And what did you learn in those early days of copying games that helped you later or gave you an indication that you had a proclivity for using computers in computer science?
I was pretty good at — I mean, most of these things had some sort of copy protection. Usually it was like hidden files or something that if you just knew how to list a directory by showing the hidden files you could see them. So it was not that hard but basically I started getting good at using computers and I was usually the better kid in my class at anything having to do with computers and I think that's kind of what got me going.
When did you leave Guatemala?
I was 17. I had just turned 17. When I came here for college, I wanted to become a math major and it turns out in this little country there was no way to become a math major. And so I left for college. I thought I would return four years later, but then I started liking it here. The whole not being scared of kidnappings, etc. That was pretty nice.
The luxuries of life like, you know, electricity reliably, all that stuff was nice. So I kind of stuck around.
And what did your parents do and what was your upbringing like? What were your parents?
Yeah, my parents were both doctors, medical doctors, but my mother — she was a doctor, but she only practiced for like six months. After that, she inherited a candy factory from her mother. So, my mother owned a candy factory. And that's what she did. And it was pretty awesome. It was like I grew up basically in a candy factory.
So it's like — there were I assume no Oompa Loompas or any such Willy Wonka elevators, which would be nice, but like you said, electricity not reliable, so no, there was none of that.
But there were huge machines that made humongous pieces of candy that had to be cut off and it was fun. I didn't really like it because I ate — I guess I ate way too much of it at first and then I stopped liking it. But it was really fun to take the machines apart and I guess people didn't like that but it was fun to take the machines apart.
So, you worked on taking the machines apart?
I did not work on putting them back together because I had no idea how to do that.
Must have been a real crowd-pleaser with the employees.
And did your parents do anything in particular in terms of parenting style or interacting with you that you could see doing with your own kids at some point?
My mom was great. I mean, she really — one of the things she did was she always told me — I've read this after the fact, but I noticed she did it. She always told me, you know, when you congratulate your kids, you always tell them, 'Oh, you're so smart.' She would say that I was smart every now and then, but most of the time she would say something like, 'Oh, you worked hard on that.' And she did that a lot. And I guess that's probably why I'm now a workaholic. But yeah, she did that a lot.
So I think I want to say Carol Dweck — I might be getting this researcher's name wrong — wrote a book called Mindset about this, how if kids believe that their intelligence is a fixed trait and they're told 'you're so smart, you're so smart' that when they fail they assume it is an unfixable problem, right? And then when you're told you worked really hard, congratulations, then you become more resilient and adaptable when facing setbacks and whatnot. What did your path look like through college and then immediately after graduation? Where did you go undergraduate?
I went to Duke. I had never been to North Carolina before. I showed up. I did not really realize that North Carolina was the South.
It does have North in the name.
It does have North in the name. But yeah, the first day I showed up I went to McDonald's because I had heard a lot — well, I mean we have McDonald's in Guatemala but it was a restaurant I knew. So I went to McDonald's and I could not — I spoke English but I could not understand the person asking me what I wanted to order. The Southern accent was way too thick for me. But I was at Duke. I was there. I studied computer science. I loved it. I loved basketball while I was there for exactly four years and then the day after I graduated I stopped liking it. But I loved it while I was there.
Yeah. Then I went to get a PhD in computer science at Carnegie Mellon. I went to Carnegie Mellon because of my PhD adviser, who's a guy named Manuel Blum, who many people consider the father of cryptography — you know, like encryption and all that stuff. So he's amazing. So I came to Carnegie Mellon to study computer science.
Where is Manuel Blum originally from? The US or is he from outside of the US?
He was born in Venezuela, but he lived there for like a year and then he grew up in New York City.
What did you learn from him? Are there any lessons in particular or principles or takeaways that you've carried with you or that you can remember from working with him?
Yeah, I learned a lot from him. I mean, he's very fun. He's an older man now. He's in his late 70s. And he always — by when I met him, which was like 15 years ago, I guess he was in his 60s, but he always acted way older than he actually was. He just acted as if he forgot everything. He really acted like he forgot a lot of things.
Meaning like an absent-minded doc from Back to the Future type of person or...
Yeah, but I think it was a bit of an act. Because at first, the way he treats his PhD students is he meets with them once a week and they have to explain to him what they're doing. So I had to explain to him what I was working on, which at the time, by the way, what I was working on was this thing that became very popular — CAPTCHAs, these distorted characters that you have to type all over the internet to prove you're not a robot of some type. That was the thing I was working on. And I had to explain it to him, but it was very funny because usually I would start explaining something and in the first sentence he would say, 'I don't understand what you're saying.'
And then I would try to find another way of saying it and the whole hour would pass. And I could not get past the first sentence. And he would say, 'Well, the hour is over. Let's meet next week.' And this happened for — this must have happened for months. And at some point I started thinking, I don't understand why people think this guy's so smart. I mean, he's won — for example, there's no such thing as the Nobel Prize in computer science because I guess there were no computers when Nobel was around. So there's a thing called the Turing Award and he won the Turing Award. So he's basically won the Nobel Prize for computer science. And I just could not understand how people thought this guy was smart.
But it turns out later, now I understand, this is basically just an act. I mean, essentially I was being unclear about what I was saying and I did not fully understand what I was trying to explain to him and he was just trying to drill deeper and deeper and deeper until I realized every time that there was something that I didn't have clear in my mind.
And that's something that — he really taught me how to think deeply about things and I think that's something that I have not forgotten. And when you understand something really, really deeply, you can explain it in a really short way usually. And that's great.
That's part of the reason, and particularly since I'm incompetent when it comes to physics, but I've always had a lot of respect for teachers who can develop a style like that of the late Richard Feynman, right? Being able to take very — perceived as complex concepts in theoretical or applied physics and explaining them using say an apple and a pencil to a third grader and being able to have them get it. Not just get it but remember it.
Yeah. I think that's true understanding. I think when you're able to explain something so simply that almost anybody can understand it, that means you really understand it.
And for anyone who hasn't seen it, it's a little hard to find. I have found it on YouTube before. Or I'm sure you can find it on some Chinese version of YouTube somewhere. But The Joy of Finding Things Out, which was a NOVA program, it's not that long, which is an interview of Richard Feynman about his experience with his father among other things and how his father taught him to explore the world. Really fascinating. So you mentioned CAPTCHA. Could you walk us through — or not walk us through — but what were the milestones for CAPTCHA and what ended up happening to it?
Sure. So the way this whole thing got started, it was about the year 2000. I had just started my PhD at Carnegie Mellon. I was looking for a PhD thesis project. I mean I had just started and I was looking for something. I really didn't have a research project. And this guy — a pretty awesome guy, his name is Udi Manber. He was the chief scientist at Yahoo, you know, works at Google. But at the time he was the chief scientist at Yahoo and Yahoo at the time in the year 2000 was really the biggest thing in the world.
He came to Carnegie Mellon to give a talk about 10 open problems that Yahoo did not really know how to solve.
And my guess is this was a recruiting talk. Now that I understand how the world operates, my guess is he was coming to recruit engineers of some sort. But, you know, he gave a talk about 10 problems that they didn't really know how to solve.
And there were a bunch of different problems there. I went home and I tried to — I said, 'Well, I'll try to figure out how to solve all of these.' I had no idea how to solve any of them except for one. And I didn't really know how to solve it, but I thought I could probably come up with something for that. And the problem that they didn't know how to solve is that at the time they had a free email service. I guess they still do. The problem they were having is that people were writing bots to sign up for millions of email accounts.
That's what they were doing. And then the people at Yahoo did not know how to stop this. They were trying all kinds of things. And whatever they did, the people who wrote the bots would kind of one-up them. And so I kind of thought about it. I thought about it with my PhD adviser, Manuel, and together we came up with this solution, which is why don't we test whether the thing that's trying to sign up is a human or a computer that's trying to sign up for millions of accounts. And we came up with this idea of a test. And the test is basically just giving some distorted characters and having the entity read it. And it turns out humans can read these pretty well whereas computers can't read it as well. And that was it.
It took a few months to come up with this idea. Then we developed it and then we showed it to him, to Udi Manber, and he had it running on Yahoo like two weeks later, which I also now in retrospect realize how big of a problem this must have been for them.
Yeah. That's a miracle for that to be implemented because it's a miracle for any large company to do anything in two weeks.
But he had it running pretty quickly after that. And from then on every other website started using these. So that happened. Then by — this was still the year about 2000. By the year about 2005 I had gone on to work on other things. But by then essentially every single website was using it. And at some point I did a little back-of-the-envelope calculation about how many of these were typed by people around the world. And it turns out the number I came up with was about 200 million. So about 200 million times a day somebody would type one of these CAPTCHAs.
And that's when I started thinking, I wonder if we can do something with this time because the thing is, each time you type one of these, not only are they annoying, but also they waste about 10 seconds of your time. And if you multiply 10 seconds by 200 million, you get that humanity as a whole is wasting like 500,000 hours every day typing these annoying CAPTCHAs. So, I started thinking, can we do something? I started feeling bad, first of all, because it's like 500,000 hours and it's kind of my fault. So I started feeling bad and I started thinking, can we do something good with this or useful? And that's when I came up with the idea. It's kind of the second round of CAPTCHAs. It's a system called reCAPTCHA, which the idea is that not only are people authenticating themselves as a human as they're typing these, but they're also helping to digitize books.
And the way that works is — okay, so first of all, digitizing books — Google for example is digitizing all the world's books or was digitizing all the world's books where the idea is you start with a physical book and you scan it. Now scanning a book literally what it consists of is sending the book to India and somebody in India taking a digital photograph of every page of the book. That's literally what scanning a book is. Now the problem is that the next step of the process, the computer needs to be able to decipher all of the words in these pictures of the pages. But computers can't do that very well or not as accurately as humans. So what we do is we take all of the words that the computer cannot recognize in these scans and we get people to read them for us while they're typing CAPTCHAs on the internet. So that's the idea of reCAPTCHA, that as people are typing these CAPTCHAs, they're also helping to digitize books.
And from the conception and development of reCAPTCHA to its — the technology was acquired or licensed — what was the ultimate...
reCAPTCHA turned into a company. So it was a company that — what we were doing. So at first this was at Carnegie Mellon. I was a professor at Carnegie Mellon and this was a research project at Carnegie Mellon but then at some point we realized that we could digitize a lot of things in a large scale because what happened is we launched this reCAPTCHA service pretty quickly. Facebook started using it. So basically anybody who signed up for Facebook was helping us digitize one word out of something.
And who provided the books? Was that...
Right. So at first nobody was providing the books really. We just had kind of some small amount of scans that we had gotten from somewhere. But I was giving a talk somewhere and in the audience was the CTO of the New York Times. He said, 'Hey, we have archives of the New York Times. We have 130 years of New York Times archives that we have scanned, but we just cannot finish the digitization process because the scans are not very good. So maybe you can do this for us.' And I said, 'Sure.' So pretty soon we started digitizing the New York Times and started charging them for doing that.
And at that point we had to leave the university because we were doing — apparently you cannot do work for hire inside a nonprofit university. Who knew?
And so we formed a company mainly because the New York Times wanted to pay us to digitize their archive. And so they started paying us — they were paying us quite a bit. They were paying us about $50,000 per year of content.
130 years. Yeah. So that was — we didn't need any...
Right? We didn't need any venture funding or anything. That's how we spun it out into a company and we were in the middle of that contract and we were going at it. But at some point, Google was having the exact same problem with their book digitization process — their computers were not as accurate as humans. So they kind of approached us and we ended up deciding to sell the company to them.
Question on Carnegie Mellon and technology transfer. So for those people who are not familiar, if you were to go to say a place like MIT or many universities, if you develop — and I'm sure this varies in the fine print — but if you are a faculty member or maybe even a student and you develop technology at the university, am I correct in saying that in some cases that is the intellectual property of the university? How did you...
So how did you navigate that with Carnegie Mellon?
So most universities — what happens is if you are an employee of the university or if you're a student and you are funded somehow through the university, so for example PhD students that are funded through grants or something, then the university can take ownership of your intellectual property. And usually also universities have some sort of licensing agreement in case you want to form a company out of a university. So this is actually what happened with Google. Google was developed inside Stanford and there's many other companies like that and usually there's some sort of licensing agreement where if you start a company the university now owns a certain fraction of your company and that varies per university and exactly the terms of this vary per university.
At Carnegie Mellon it's pretty nice. What's nice about it is that it's really simple. Carnegie Mellon owns 5% of your company, period. Now if you raise funding, their ownership dilutes, but at the time of company formation, if you're taking technology out of the university, they own 5% of the company.
And that's it. And what's nice about that is you don't have to fuss with them. Many others have some sort of negotiation or variable amount depending on milestones or whatever and you end up having to negotiate all kinds of things but here it's just a standard thing.
Yep. So Carnegie Mellon owned 5% of reCAPTCHA. That's what happened.
And looking back at Udi's presentation and the 10 open problems and how you recognized it later in retrospect as a recruiting trip most likely for engineers. Have you developed a particular approach or what is your approach for recruiting engineers for your startups? Like for instance in the current day or when you were just putting together the initial team for Duolingo? How do you recruit and vet engineers?
Yeah, so the initial team was actually relatively easy at Duolingo. I basically took — because I was a professor and essentially I was teaching a class that every single student had to take in computer science. So I pretty much knew who the best students were.
You got to see all the fresh meat coming in.
I knew who the best students were in the whole university. So it was relatively simple at the beginning. I essentially just emailed my top 10 that I knew and that was it. And some of them already had jobs. And some of them were very interested and I hired some of them. But those I didn't even have to interview because I knew them. I had taught them and I knew how good they were. Things changed after that. Now that we have to hire more. Now with Duolingo, we have about 60 people. And what we do, we go to universities. We recruit from usually kind of East Coast universities. So MIT, Harvard, Princeton, CMU, etc. We usually go there, we give a talk and then we get...
Usually, you know, technical but entertaining. But I think what's most useful is showing that you have really difficult problems. That is what works the best.
Because difficult problems attract good people.
Yeah, difficult problems — they're like, 'Oh, I'm intelligent, I'm going to go solve difficult problems.' So, you basically pick the hardest problems that your company's working on and give a talk about how they're so hard and your solution and how your solution is so clever, but still could be improved and it could be improved by you if you come work for us. So that's usually the type of talk that you give. And then we get a bunch of resumes.
Then we interview. We're very strict on the interview. This is something that really matters. We're pretty strict on our interview in terms of making it a high bar for hiring. I think that's something that I've learned over the years that hiring mistakes are very expensive.
So, what would be an example of that strictness? What is it?
Well, we do — this is, you know, a lot of technology startups follow a very similar routine. And Google follows a very similar routine. It's basically — so we do two phone screens. So the first thing that we do is we call them. So first of all we do a resume screen. So we look at what university they went to and did they have a good GPA, have they had good internships etc. And after that, the better ones we call them on the phone and we do an interview and it's usually — because it's on the phone it's not that difficult of a problem but it's some computer science problem that they have to solve, like I give you some numbers in some way, can you sort them in some other way by only doing this type of operation or something like that.
And you give a problem. Usually we do it through Google Docs. So they have to write their solution so that we can watch it while they're writing their solution.
So we do that, we do two of these phone screens. So if they pass the first one, we do a second one. And then after that we bring them on site and then we do usually four different interviews on site where some of them they have to solve some problems on the board and these are problems of the form — I don't know — this kind of logic type...
problems that are related to computer science. And then afterwards they have to do a pair programming with one of our engineers. So they sit there and actually try to fix a bug that an actual bug that you know it's usually the same bug but an actual bug that has occurred at Duolingo. They have to fix it. Now, so pair programming, for those who aren't familiar, it's often talked about as a great way to learn how to code or to get employees up to speed, right? So, you have Pivotal Labs and other companies like this that can bring on say coders who know one language and get them up to speed in the language you need them to be proficient with in a startup and do so very quickly and help them to build prototypes and so on. But in this case, you're using it to vet someone. So you have one of your engineers sitting at a computer and then you have someone else sitting right next to them and what are they doing?
Well, usually we let the interviewee drive. I mean they're the ones using the computer and so our engineer is usually helping them, you know, of the formula. Oh, you know because they're navigating our codebase. So our engineer is usually telling them oh you know go to this directory to open that file. That's where that would be. And then they essentially have to track down a bug and solve it. You know, fix it.
That's usually what they do. And it is amazing how much you find out from doing that for just 45 minutes.
And, do you then take people for any additional test drive? For instance, Matt Mullen's been on the podcast, one of the lead developers of WordPress and now CEO of Automattic. And when they hire people for Automattic, which is fully distributed, they do text or IM. I guess it would probably be IM interview as opposed to voice. And then they'll audition them for two weeks. And I think the standard rate is something like $25 an hour whether it's a CFO or just an entry-level customer service person or a coder. And they just they do it in off hours, you know, at night or on the weekends. But do you audition people after that or is that
We don't, that's it. After we do two phone screens and four on-site interviews, we are done. We've tried the auditioning at first we were kind of doing that but we found that a lot of people would tell us well you know I have an offer from Google.
Right, yeah, I'm not going to put up with this circus.
I'm not going to put up with two weeks of crap from you. I'm also in school, you know, especially for fresh grads, right? I'm also in school and I have like finals coming and so we found that they just were pretty, they were not that open to that.
Got it. I'm saying got it a lot today. I guess that's my catchphrase. It's like my Buzz Lightyear to infinity and beyond for this conversation. Well,
I also never realize that Tim Allen did that voice for those people who are wondering who Buzz Lightyear's voice is actually attributed to in real life. But I digress. I wanted to ask you if you were advising say a very bright young would-be entrepreneur who is going to be starting their first tech company. Could be any type of company. If you could just give them two or three books or resources to help increase the likelihood of them succeeding, what would you point them to?
Let's see. So I like, okay, I like Peter Thiel's book Zero to One.
I love that. You know, one that I've recently started really liking, I guess I like it because it's kind of, it makes me feel good because they have very similar problems that I've had. It's the Startup Podcast.
Yeah. Yeah. Yeah. The Gimlet Media guys.
Yeah. I love it. Because they, you know, it's a lot of times it's just like, 'Yep, we had that exact same problem.' And I love the fact that you guys completely fumbled it and so did we. And now I feel better about myself. And so I think it at the very least makes you feel good about yourself.
So Zero to One, Startup, the Startup Podcast which everybody can find on iTunes or in a podcast player. I like Overcast quite a lot for listening. What else might you
I like this one but it's not for very new entrepreneurs. I found it very useful. I read, I actually, well I didn't really read it. I don't read books. I just listen to them.
Mhm. Reading. I don't know, for some reason I just stopped reading like 10 years ago. But I quote unquote read twice. The Hard Thing About Hard Things.
Right. That's Ben Horowitz.
The first time I found it not very useful because it just was stuff that was just not, it was not applicable to me because you know I was not at the time running a company really and I had never really run a company, you know, reCAPTCHA only reached you know like nine people so that was it and whereas he talks about problems of you know companies with hundreds of people. But now I recently read it again, you know, in the last like month or so and it is amazing how much good advice there is in there. For somebody that is running a company of around, you know, in our case 60 employees, it is amazing. I loved it.
So what would the profile be of the entrepreneur for whom that book would be appropriate or most useful do you think?
I think somebody that is running a company that has already, you know, more than like 30 people is my guess. That's my sense.
Yeah. The Hard Thing About Hard Things has been recommended to me maybe not within this podcast but certainly offline by at least four or five of the best tech entrepreneurs who have been on this podcast. It's come up repeatedly. I mean but these are people who are also running as by definition larger organizations. The Zero to One book by Peter Thiel is fantastic and I would also encourage people who are interested in digging even deeper to try to track down the Blake Masters transcriptions or notes from the original class that was taught at Stanford which formed the basis for that book because Peter goes very very deep into specific hiring decisions and equity and negotiating term sheets and things like this. He gets very very granular on things that were not put into the book because they wouldn't be applicable to a very broad audience necessarily. It would have seemed too inside baseball, but the original class notes are just fantastic. And he brings in guest lecturers and so on who are a lot of the icons in Silicon Valley that the both of us would recognize. Silicon Valley. So Silicon Valley has this vaunted mythological place in the minds of a lot of people. It's, you know, the Mount Olympus of tech, the epicenter of all good things related to innovation, etc.
Why have you chosen to remain in Pittsburgh and what have the benefits, the pros and cons been of staying there as opposed to going to SF or New York City?
Listen, Pittsburgh is absolutely the best place on earth.
No, I actually I like Pittsburgh quite a bit. And I'll tell you, it is really unclear. One of our board members, Bing Gordon from Kleiner Perkins, he visited us, and he said to me, would you start Duolingo in Pittsburgh again? And I said to him, honestly, I don't know. Well, you know, his answer at least made me feel good. He said, 'Well, you're, you know, you're like in the 99th percentile of startups, so things seem to have worked out for you, so you know, you're fine.' But it's at first, you know, we started here because I was a professor at Carnegie Mellon and Carnegie Mellon's, you know, a top school for computer science. And we were around here and we kind of just never left. And what we tell ourselves is and I think it's true. It has been a pretty awesome recruiting tool for engineers. We've been able to recruit really amazing engineers. You know as a company we when we were only 20 people we had like eight people with PhDs in computer science. Which is a pretty hard thing to do anywhere else including Silicon Valley. So I think we've had a great time recruiting engineers. I think for that it's been very helpful. I think it's also been very helpful in terms of being able to, you know, we're not affected so much. You know, one of the things that I noticed about Silicon Valley is there's definitely trends that happen that come and go. And we're not affected too much by that. And I like that we're, you know, we're kind of, we know what we're doing with Duolingo. We know what kind of company we're creating and we're sticking to it and, you know, it's not the case that next time there's the next whatever Yik Yak or whatever app is the next awesomest thing, you're changing your whole product to look more like theirs.
Right? Yeah. You're outside of the echo chamber.
Yeah. And that I think that has been helpful for us. And now can I tell you for a fact that we would have been less successful in Silicon Valley? I have no idea. And another thing is we're now really starting, I think up until now I don't think it has been a problem that we've been in Pittsburgh but we're now starting to have a couple of issues. One of them is in terms of the infrastructure of things. So, for example, finding office space here was a pain in the ass because everybody here wants a 10-year lease because that's who they rent to, you know, 10-year leases. And we're like, yeah, we cannot take a 10-year lease. I assure you that in 10 years, we will either have gone away or, you know, massively grown, but we're not going to be the same size. So, you know, the infrastructure here is just not the same. And the other thing that we're having a harder time is with people with previous startup experience. There's not very many here. Google has a pretty big office here. It's about 600 engineers. And so we can hire good engineers from Google, for example, even seasoned engineers from Google. But people with previous startup experience, especially outside of engineering, there's not very many here. I was having a chat with the founders of Shopify not too long ago because I was the first adviser to Shopify in 2008 or 2009 and you know they IPOed 6 or 12 months ago something, six or nine months ago I guess and they've done very well and they're based in Ottawa, Canada.
And they had, they said it's actually been a huge gift in a way because they are the only game in town. So, from a recruiting standpoint, not only are they the go-to tech company in a lot of respects, but on top of that, they don't have a lot of attrition because people are settled in Ottawa and they're not getting poached by Facebook, Google, Uber. They're not going into that bloodbath of bidding wars.
So, we see exactly that for us. That's great. I mean, so, you know, it's funny. We sometimes go to events, you know, a lot of times these tech events or sometimes our investors have events for, for example, one of our investors is Union Square Ventures. They have these amazing
I'll try to throw a picnic with some paper plates for you guys.
Yeah, you should. You're also one of our investors. You have not had us in any of your events. Okay.
So one of, you know, Union Square Ventures, they have these great events where they bring in everybody, you know, for example they have a marketing event and it's all the people that are in the portfolio companies that do marketing or they have a design events and it's all the people that do design in the portfolio companies. And so we send people to all of these events and it's pretty amazing how when you're talking to people it really does seem like the average tenure of a person in one of these companies and a lot of them is like a year and a half. It's like after a year and a half it's like going to the next one. Whereas for us it's, I mean people really don't leave. I mean there just that because we really are the, in terms of startups we're not exactly the only game in town. I mean, that's unfair to say, but there's not very many games in town.
Do you, what do you think the likelihood is that you will stay, that your HQ, the majority of your employees will be in Pittsburgh in say two years time?
I think that's pretty high. We just started an office in, well, if you could call it an office, we just started an office with two people in San Francisco. You met one of them, Gina.
And they're there. They're working off of the Google Capital Office for now, but they're getting their own office pretty soon. And I think we're going to have, I think we're probably going to have about in the next year, we're probably going to staff it with about 15 people. So, you know, it's not going to be super tiny, but it's still not going to be the biggest one.
And for those people who are not familiar with Union Square Ventures, Fred Wilson, who's one of the, I think he's still a general partner there, isn't he? Yeah. He has a fantastic blog for people interested in tech and venture capital entrepreneurship, which is AVC, and I think it's just avc.com if I'm not mistaken, but you could certainly Google Fred Wilson and AVC and it'll pop right up. How many employees did you have at reCAPTCHA when the company was absorbed by Google?
We were between seven and eight and then inside Google we started, you know, the team grew but that was it.
And between that point I assume you kind of rested in peace for a while as they say. I mean you had, did you remain, were you at Google for a period of time?
I was at Google for about a year and a half.
Year and a half. And when did and how did Duolingo get started? I mean, what were the, when did that idea start to germinate?
That started right after the acquisition with Google. I mean, I was in a pretty good place in my life where I didn't really have to work anymore. And I wanted to do something that was related to education. And I started, you know, at the same time I had a PhD student, Severin, who is my co-founder, he was my PhD student, we went, we wanted a PhD project for him and I thought well you know I really am passionate about education let's try to do something related to education. Now my views on education have always been very influenced by where I'm from because I'm from Guatemala and it's a very poor country. Now the thing about Guatemala is and poor countries, a lot of people say that education is the thing that can bring equality to the different social classes. But I always saw it as quite the opposite. I always thought that education actually was something that brought inequality because what happens is the people who have money can buy themselves the best education in the world. And because of that they remain having a lot of money. Whereas the people who don't have very much money barely learn how to read and write and therefore remain not having very much money. So I wanted to do something with education that would give equal access to everybody. That's what I wanted. Now education is very general. So we decided to concentrate on just one type of education which is learning a foreign language which you know if you're in the US it doesn't sound like that big of a deal but outside of the US or generally outside of English-speaking countries this is a massive deal, generally learning English. There's 1.2 billion people in the world learning a foreign language. And so you know we decided that we wanted to do something to teach foreign languages. But this is a funny market, the language learning market, the majority of the people learning a foreign language, 800 million of them, satisfied three properties. First of all they're learning English. Second the reason they're learning English is to get a job or a better job. And third they're of low socioeconomic conditions. Okay. So most people are trying to learn English in order to get out of poverty. That's the people who are learning a language. But at the same time, most of the ways there are to learn a language are usually very expensive. Like well, especially before Duolingo came around, like Rosetta Stone, for example, think about that. It's like between $500 and $1,000. So, you know, this was kind of the irony. It seemed like you needed a thousand bucks in order to get out of poverty. And so we thought we could do better. And that was the idea with Duolingo was to give a 100% free way to learn languages. That's what we wanted to do. And
How did you develop the idea or the model and what did those early conversations look like?
Yeah, at first we started thinking, well, if it's going to be free, we got to figure out a way to pay for it. A good idea.
We hadn't discovered what we now know, which is you can always raise more VC funding. I'm kidding. That's not a good way to pay for things.
Yeah, eventually mom and dad, i.e. VCs on the board stop giving you allowance.
Well, you know, somebody on our board said a good thing. He said eventually we can find a bigger fool.
I think that's about right. So yeah, that's not our business model. But at first we had a business model which we've actually moved away from. It was a, I thought a pretty clever business model but there are reasons for which we moved away from which I'll tell you about. But our first business model that we thought was okay, here's what we're going to do. Much like the idea with reCAPTCHA where as people are typing CAPTCHAs we're actually capturing some value from them. Could we do the same for learning a language? Could we get people to give us something of value while they're learning a language? And what we thought the answer to that was yes and we thought we could get people to help us translate stuff while they were learning a language. So the idea and this still works and we're still doing it. So the idea is the following. After somebody would learn a lesson on Duolingo, they would learn a lesson about food like they would learn all the food words. At the end we would say, 'Hey, if you want to practice the things you just learned, here's this document that comes from the web that is related to what you just learned.' So, in this case, it may be like a food blog or something. And we would say, it's in the language you're learning. Do you want to help us translate it to your native language to practice? And what would happen is some people said yes to that. And a number of people would get the same document and they would kind of all translate it together, kind of edit each other's translations etc. And they were translating from the language they were learning to their native language and at the end we would come up with one translation and that was a translation and in some of the cases those documents came from for example CNN. CNN is one of our clients. They wrote all their news in English. They sent it to us and we had our people who were learning English translate it into their native languages and then we would send the translations back to CNN and CNN would pay us for having translated their documents. This is still going on. CNN is still one of our clients. So it's been about 2 years and they're still at it. So that means the quality of the translations is high enough that they haven't fired us. So that was the business model and it was a pretty good business model but we moved away from it because we realized the more we worked on it the more we realized we started turning into a translations company as opposed to an education company. So because that's kind of where the money came from. So we started just more and more developing translation and translation and not working on the education side. And at the same time, we realized that also translation is a crappy business to be in because it's kind of a race to the bottom how much you charge for translations. You know, if you can do it for one cent a word, you can always find somebody else that can do it for half a cent a word. So we decided to move away from that.
Was willing to try for half a cent a word.
Yeah. So we decided to move away from that. And now we have different business models, but that was our original one.
Can you elaborate on the other business models?
Sure. We have two now. The main one, the one that we've been at it for longer. It's English language certification. Well, generally language certification, but what matters here is English. So the idea is this. There are
Just to, sorry not to interrupt but I will briefly because I think that to put in perspective for native English speakers who might not get the magnitude and maybe you can call BS on this or confirm but I've heard for instance that in China alone I mean we could talk about a place like Brazil also but there are more people learning English in China than all the native speakers of English combined.
Yes. There are 400 million people learning English in China.
Yeah. Still for those people listening, big market.
Yeah. No, I mean learning English is humongous, right? I mean this is for teaching languages that is what you want to teach. I mean there are, you know, it's funny there are companies in China that are venture funded that only teach English and that only teach English to hotel employees. That is a big enough market for a startup.
So I mean that's how big the market is in China for learning English.
But I threw you off course that you were talking about the certification.
Yeah. So the thing is this with certification. So you know exactly with English, people need to learn English but after they've learned English they also need to prove that they know English in particular to get a job or to go to school etc. So about 2 years ago we started getting a lot of emails like that of people saying thank you for teaching me English. I did not have the resources to pay for English software for classes or something and now I can learn it because of Duolingo. But now I have a problem. I need a certificate that says that I know English because I need to apply for a university or I need to get a job etc. And they're requiring some sort of certificate that says that I know English. And so we thought well let's look into this whole certifying that people know English this whole market. It turns out this is a humongous market. It's about $15 billion a year are spent on people certifying that they know English. You use that for example if you want to, if you're from outside of the US and you want to apply to enroll in a US university, you have to take a test that proves that you speak English. That test is called the TOEFL. If you want to get a job at a multinational corporation in a non-English-speaking country, you usually have to take a test to prove that you know English. If you want to get a work visa in the UK, you have to take a test that proves that you know English. So, there's these standardized tests. In total, $15 billion are spent a year on these tests. They're all very similar to each other. There's a few of them. There's like three or four that are the big ones. The way this process works is you have to, it's kind of similar to taking the SAT. You have to take this test. You pay about 250 bucks for it. You usually have to take it, well you always have to take it at a testing center. So you have to go to a place to take the test. Because of that you usually have to make an appointment several weeks in advance and so several weeks in advance you make an appointment you pay 250 bucks you go somewhere you take a test and you get your results another maybe four weeks later. So at the end you spent eight weeks, you spent 250 bucks and you have a certificate that says you know this is your test score for English and that's this measures your English proficiency. Now so that's the process. And this sounds kind of annoying but it's actually way worse because most of the people that are taking this are in developing countries. Now there the thing is 250 bucks in the developing countries a month salary. Also the testing centers are not in every city. They're just in certain cities. So, you have to travel to take a test to prove that you know English.
Right? And just to also add some color to that, I mean, traveling might not be, it's not hopping on a bus and going 20 minutes necessarily, right? If you're in a village in Peru and you're trying to get to Lima, you know, it's several hours.
And it also cost you money.
So this actually happened to me when I was applying to come to school here in the US. I had to take this test to prove that I knew English. What happened is Guatemala ran out of tests. They there were no more seats for that year in the testing centers. So I had to fly to the neighboring country of El Salvador to take the test. So for me, this cost a thousand bucks and it involved a trip outside of the country. So we thought you know when we saw that we thought man we can do a lot better than that. That's really ridiculous. It seems like technology from a hundred years ago. So we launched this thing called the Duolingo Test Center which is basically an app. It's also a website that helps you certify your knowledge of English. And the idea is that it's only $20. It's not $250 and also you take it from a device. You don't have to go to somewhere to take it. So you don't have to go to a testing center. And now the tricky thing here was that when you know the reason people have to go to a testing center to take the test is to make sure that they're not cheating. Because you know you could for example send your cousin to take the test instead of you or you could show up with a bunch of books. So the way we had to prevent cheating with our test. And the way we've decided to prevent cheating is, or were able to do it is, when you're taking the test from your phone, we actually turn on the front-facing camera and the microphone, and we actually record you taking the test. So, we record your face and the ambient noise and everything, and then a real human proctor watches you take the test and make sure that you were not cheating and you were actually looking at the screen, etc. And it turns out this works.
That's very cool. And how hard has it been or how, let me ask a better question. How have you been able to get buy-in from companies to accept this?
Yeah, that's been the hardest part probably, you know, the technology we figured that out relatively easily. The hardest part has been getting accepted by universities and companies etc. But we're making good traction. We're, let's see, so in terms of universities so some departments of Harvard University take it some departments of Carnegie Mellon University are starting to take it some departments of the Max Planck Institute in Germany are starting to take it and this year we're going to run a study with 12 very well-known name brand universities here in the US like you know Yale and that type of name. So 12 of them are going to be running a study where all of their applicants are going to be taking both the standard test that they're taking and our test and the idea is they're going to look at the correlation between that and if it's high enough then all these universities are also going to start accepting our test in addition to the other one. So we're doing that. We've been also doing a lot of work with different companies. So Uber for example is a good one. They have in many countries they're starting to roll out a new Uber service called Uber English which is you know your driver speaks English. And we have a partnership with them where the idea is that in order to certify that the drivers speak English they have to take the Duolingo test. So we're doing a few of these companies. And the way we're getting it done is basically by going at first to kind of the blue chip places, you know, the Harvards of the world. And you know, after that, the sale is much easier because you say, look, it's good enough for Harvard. Probably should be good enough for you.
Yeah, the H name works in a lot of places. MIT too, very powerful in Asia.
The, so we've been talking about English. What other languages are currently available on Duolingo for people to learn, for English speakers to learn?
Uh there's a bunch. You know, I used to know them all. Now I don't. So, you know, we have the big ones like Spanish, French, German, Italian, Portuguese. We have some lesser-known ones. We have like, Irish, Esperanto, Esperanto is a funny one. Danish, Swedish, Russian, Ukrainian, Turkish. We, you'll notice we have no Asian languages right now.
It's because it's harder to teach them. We're working on it, but it's just we're, it has not been our highest priority because it's kind of low ROI in the following respect. The investment is high because they're actually harder to teach them and the return is not very high because if you actually look at from for English speakers, for example for English speakers of the people who are learning a language that are English speakers 52% of them are learning Spanish 26% of them are learning French then the next big one is German at 6% and Chinese is somewhere at 4% or so so you know ROI is not super high and this is why we haven't done it. But, you know, we want to do it.
But a lot of my fans want Japanese. That's been a very common.
I know. That's the most requested one.
It's the most requested one. And it really is just, we're on it. We're on it.
Now, if you when you develop the ability or the curriculum, you code the program for say English speakers to learn Japanese. Does that, how much additional work is required like percentage-wise to have it move in the opposite direction to teach Japanese people English for instance? And does that dictate which Asian language you might choose? For instance, I know that Korea, well, China would be a huge market, I would imagine, for teaching English. But Korea as well, right, where they have super teachers who fill up entire stadiums for particular subjects. That's a lot of, that's a mouthful of a question, but does if you develop it in one direction, does it make it easier to then teach in the opposite direction?
Yeah, it does. And we've done that for a lot of languages. So we do already teach English to Chinese speakers. We teach English to Japanese speakers and to Korean speakers and Vietnamese speakers and Indonesian speakers. So we do teach English to most Asian large Asian languages. So that makes it easier to go the other way around. I mean truthfully the main problem that we have not yet really worked on solving, it's not that it's not impossible to solve it in terms of ideas. It's just we just haven't gotten to it. It's teaching you the writing system.
Yeah, the orthography is tricky.
Yeah. And it's pretty tricky particularly for the ones that don't have an alphabet. I mean, like Chinese, that is it's pretty tough. And so that's just, we just, you know, we have a long list of priorities. This is somewhere in there, but it is not the highest one.
I bet a lot of folks out there who would want to learn Japanese. I mean, speaking as someone who's studied a lot of Japanese and lived there, I would imagine a lot of folks who are hoping to learn Japanese are doing it for out of cultural interest. They're doing it, speaking of English speakers or conversational interest, they may not be. I would imagine if you gave them the choice and maybe people can tell you, it's just Luis Von Ahn on Twitter, right?
So that's Luis Von Ahn. Maybe folks who are listening can let Luis know. But
Yeah. If you gave them the option, say of conversational Japanese without any writing system, no hiragana, no katakana, no kanji in a year or 6 months or whatever it might be versus say the entire ball of wax, the whole nine yards with writing systems in hypothetically two or three years. I bet a lot of people would take the former.
Yeah. Yeah. And we've seen that. We've actually started, so you know what we have started doing, so we did this with Russian. So Russian you also have to teach the alphabet. It's much easier because there
Yeah. Much easier though. I mean there's only, it's only about 15 letters that are different. So you basically just have to learn 15 letters. But we've started doing this. We basically have started teaching romanizations. For people who don't want to learn the alphabet, we just romanize it. And that seems to be working out relatively well. So we may end up doing
That something like that for Chinese like pinyin. We may end up doing that.
Pinyin can work here. I'll throw you know what if I would be so happy if this happened and I think people would probably come after me with pitchforks but there is a separate romanization approach to Chinese that I think is actually better for Mandarin. This is how I learned it but it's less standard which is why most people default to pinyin. So with the pinyin you'd have to deal with the diacritical marks meaning the slanting lines to indicate tone right. So in Chinese you have four tones for people who are familiar. So you'd have like ma ma ma ma which are all different things.
Those sounded like four things of the same thing for me.
Exactly. And so Chinese tongue twisters are a... They are crazy. But there's a system called GR or romad. So it's like I think it was developed in Taiwan. And what's cool about that is they build the tone into the spelling. And so if you remember the romanization, and I'm highly visual, so I remember the spelling of words, then you automatically remember the tones. So students who are taught using GR have a in my experience, and this is what was used at Princeton, which has a very good East Asian studies department, one of the best in the world. The retention of tones and consistency of tones of Princeton students was outstanding compared to many other schools that used pinyin. Of course, that wasn't the only difference in the pedagogy. But for instance, if you had like gu which is country or like it's like past tense kind of I have gone, you know, something like that. Then you would have and I haven't done this in so long, but it would be like guo would be the second tone gu and then first tone gu would be guo and then fourth tone would be like guo h and that h at the end indicates it's a falling tone. And I really want to resurrect and popularize this system because I feel like it's so much more effective. It's just not as common. So it's not often used. Anyway, that's my I'll throw that one.
That's good to know. You know, sometimes when we do things we've done some things that are a little you know, kind of unexpected and we end up getting really good results. I mean for example we started teaching Irish which I think a lot of people were very surprised by the fact that we launched Irish before like you know I don't know Turkish or Russian or Japanese.
Right so how do you justify that investment?
Yeah, well it was very easy was the point. You know that was actually done by volunteers not by us and we didn't have to make a single change to the app that was just our system just allowed it to just go through with the volunteers. So it was just done by volunteers so we didn't have to be involved almost at all. So that was the reason was just because it was relatively easy. But you know one of the funny things is when we launched it there's 94,000 native Irish speakers but we have about a million people learning Irish on Duolingo like actively.
So we're you know if we put something there a lot of times we can actually have a big impact on the language itself just by putting it there.
Huh. So Duolingo extensively at some point could have a role not just in preserving but even resurrecting languages depending.
Yeah we've thought about that. We thought about doing and we have a few well you know one of the funny ones we're working on is well not us but volunteers at Klingon. That one's an interesting one.
Do you have any plan Do volunteers have any plans for Dothraki or anything like that?
Well, you know, a few of those they're kind of I think they're like copyrighted or something.
Oh, yeah. You're right. I remember I tried to do something in I spoke to the gentleman who developed Na'vi or the language for the Na'vi people in Avatar, which is largely based on Japanese people might find interesting to know. But yeah, so I see. So you might not have the ability. That's so stupid. Wow.
I not we're we kind of asked for permission for Dothraki and we were told no. And then you know for that massive Dothraki textbook market that they want to capitalize on.
I know. We and we thought, you know, we probably thought we could like anyways do it, but we didn't really want to get into trouble. So whatever, right? Got it. Yeah. Yeah. Don't really want to fight HBO or whoever owns the copyright over that.
I still want to watch their shows is the thing.
Uh, now this is a bit of a left turn, but actually before we move on, I wanted to flash back to Google for a second, but how do you know or do you know how Duolingo compares to say college instruction for any language?
Yeah, we spend time we actually have a we spend quite a bit of time measuring how well Duolingo works. We have a team that's whose job it is to do that. The we have a lot of internal studies. The one that we have that's external so done by somebody else not Duolingo people. What they did is they measured how much people learn on Duolingo and the way you measure how people learn is basically you do this thing called pre and post test. So you have people take a test before they start doing Duolingo. Then they do Duolingo for a while, then they take the test again. And then you see how much they improved. And the measurement they got was that if you use Duolingo for 34 hours, you learn the equivalent of one semester of college, the first semester of college of instruction in that language. So that's the semester and usually a semester takes a lot longer than 34 hours of active time. So that's you know we think that that's pretty efficient.
Do you have any and feel free to punt this if you don't want to get into it, but I've had enough people ask me that I figured I would relay it. Do you have any plans for allowing people to practice conversation or oral skills?
Yes, we have plans. It is not what you would think. Everybody, this is everybody would like us to do this one thing, which is basically pair up our users with each other or pair up our users with teachers of some sort and do kind of a video, you know, like a Skype type thing. A lot of people have asked this question, you know, have asked us for this. A lot of people inside the company have thought that this would be a good idea, but that is not what we're going to do. But we are going to do conversation. I think within we've been working on this for a while and we're pretty happy with where we're at and I think we're going to you know we've been working on it mainly in terms of you know the hardest part has been just coming up with the right way to do this. And we think we have a pretty good way of doing it that we're probably going to launch in maybe I don't know 3 4 months.
That's exciting. All right. So, and it's going to be it I'm very excited about it because it's really I think that's the main thing that Duolingo is lacking right now, like the ability to actually practice real conversations.
Um, and I think we're going to be able to do it pretty well.
Very cool. So, this is we're recording this mid January 2016, so hopefully I guess Q2 Q3.
Yeah, we're hoping I'm hoping. So, this is going to be one of those things that's going to get better over time. I think that we're going to have something in Q2. So let's say you know May-ish. But I think every month it will just get better over time.
So the next question that I've been asked surprisingly often which may not be surprising to you is does Duolingo have any plans or could you foresee Duolingo moving from teaching just languages to teaching anything and for whatever reason a lot of people have asked about philosophy which I would be curious to see actually implemented but how do you answer that question?
Yes, we do have plans. We definitely have plans. From the beginning our goal was to do all of education. We've stuck around with only languages for a while, but we do have plans. So we're on it. We are I can tell you we are going to launch another app this year. I cannot tell you what the app is about. But we are definitely going to launch another one. It is not philosophy. We are not the second subject that we are going to teach.
Irish philosophy. That's a good one.
Yes. Yes. It's Irish philosophy of the 17th century.
No, it's not that. But we are going to launch another app this year. That's for sure. So we're definitely going to get into other areas of education and you know the way we see this for one we're you know if you look at the education category of you know either the iTunes or Google Play we're consistently in the vast majority of countries we're the number one app in all of education and we only teach languages whereas I think there are many of these education apps that teach like everything basically and you know with just languages we've been able to do a pretty good job. So we think that this approach of not a video lectures but a really kind of gamified mobile native app to teach you something. We think that that approach is a pretty good approach and we're going to try to replicate it with other things. It's not going to look just like Duolingo. It's going to you know they're going to look like you know each subject requires different things. I have no idea how you would teach philosophy, but I can give you pretty good ideas of how you would teach you know certain areas of math, certain areas of physics or reading and writing. A lot of these things can be taught pretty well with an interactive app.
Can't wait to see it. I'm going to hit the rewind button to go back because I wanted to pick up on this and the story may or may not be interesting but you mentioned you were at Google for one and a half years.
Does that mean you left before you finished vesting?
Okay. So for people who can you explain how that usually works to people who aren't familiar with when companies get acquired?
Yeah. Usually when companies get acquired, so when you hear of an acquisition price, and this is, by the way, I'm going to generalize a lot here because every deal is a little different, but when you hear of an acquisition price, let's say your company was acquired for 100 million bucks by a larger company, you know, like a Google or a Facebook or something, usually that acquisition has usually two components. Kind of an upfront payment. That is some fraction of it. And then the other component is what they call a stay bonus which is basically you get it if you stick around for a while. And it makes a lot of sense for companies to do this because when they acquire another company, they don't want everybody to leave.
Yeah. And especially for you know multi-million dollar acquisitions where you know there's several tens of millions of dollars and especially the founders usually after you know the next day they see themselves sitting there you know with several tens of millions of bucks in the bank. You wouldn't want them to leave the next day. So that's why you know you hold out for another large chunk that says you know if you stick around for one year you get this much. If you stick around for two years, you get this much. If you stick around for three years, you get this much, etc.
What would be a typical percentage of the total payout?
Uh, you know, I don't know what a I have sample size of Well, of the ones that I actually have knowledge of. I have sample size of three. So, and there is variance. So, I don't know. I mean,
Would it be fair to say that it's often more than 50%?
No, the stay bonus is often less than 50%. It's from my experience. But not a lot less. Let's say, you know, this I'm just going to throw out, but let's say it's 40%.
