DUOL - NASDAQ

Good afternoon and welcome to Duolingo’s first quarter earnings webcast. My name is Debbie Belevan, Head of IR. Today after market close, we released our year-end -- our quarter-end shareholder letter with our Q1 results and commentary, which you can find on our IR website at investors.duolingo.com. On today's call, we'll have Luis von Ahn, our Co-Founder and CEO; and Matt Skaruppa, our CFO. They'll begin with some brief remarks before opening the call to questions. [Operator Instructions] And please note that this event is being recorded. Just a reminder that we'll make forward-looking statements regarding future events and financial performance, which are subject to material risks and uncertainties. Some of these risks have been set forth in the risk factors of our filings with the SEC. These forward-looking statements are based on assumptions that we believe to be reasonable as of today, and we have no obligation to update these statements as a result of new information or future events. Additionally, we'll present both GAAP and non-GAAP financial measures on today's call. These non-GAAP measures are not intended to be considered in isolation from, a substitute for or superior to our GAAP results. And we encourage you to consider all measures when analyzing our performance. And with that, I'll turn it over to Luis.

Hello, everyone. Thank you, Debbie, and welcome. I'm proud to report that we kicked off this year with another great quarter. We had strong user growth, top line results, profitability and free cash flow. In the first quarter, our user growth exceeded our expectations. DAUs increased 62% year-over-year to $20.3 million, with all major markets growing nicely. Our strong user growth, of course, helps us deliver on our mission to create the best education in the world and make it universally available. But it also helps us increase paying subscribers, which are up 63% year-over-year to $4.8 million or 8% of MAUs at quarter end. This user and subscriber growth led to bookings and revenue climbing 37% and 42% year-over-year, respectively. And thanks to our continued financial discipline, this quarter saw us post our highest profitability ever. As a result of this outperformance, we're raising our top line and profitability guidance for this year. Matt is going to walk you through our updated outlook shortly. I've said many times before that the hardest thing about learning a language is staying motivated. That's why I'm proud that our engagement numbers continue to improve, with our DAU to MAU ratio reaching an all-time high of 28% compared to about 25% a year ago. And the number of DAUs with a streak longer than 7 days grew to nearly $14 million. Our goal is to keep learners coming back to our products every day, and we mainly do that by continually innovating and improving them so that they are fun and effective. This quarter's shareholder letter focuses on how well the recent advances in generative AI complement our existing competitive advantages like our data moat, our beloved brand and our unique way of teaching. As a former Computer Science professor, I've always believed that humans and computers working together can accomplish incredible things, and applying this synergy to benefit the greater good has long been my passion. I also feel very fortunate to be among the companies with the best chances of taking advantage of the rapid advances in AI. As you'll recall, last quarter, we announced our new higher-tier subscription offering, Duolingo Max, which is powered by GPT-4. We are proud that we are one of the few companies that have launched a live consumer-facing product with this technology. That we did this so quickly speaks to the talent of our team. Also last quarter, my shareholder letter reminded everyone that our freemium business model enables us to grow organically, keeps competitors at bay and provides us with enormous amounts of data that we use to make our products better. Because of our large user base, we're able to test Max features on small fractions of users and iterate rapidly. This is a great example of how our model works in general and how we'll use generative AI in particular. As you can tell I'm very excited about all the possibilities I see with AI and Duolingo Max, but I should emphasize that we're still in the early stages of rolling Max out. We'll continue to update you about the progress we're making over time. And with that, I'll turn it over to Matt.

Thanks, Luis. To recap our impressive results, in the first quarter, we delivered 37% bookings growth year-over-year, which was about 42% on a constant currency basis. We had a net loss of $2.6 million compared to a net loss of $12.2 million in the year ago quarter. And we posted our highest quarterly adjusted EBITDA of $15.1 million, which was a 13.1% adjusted EBITDA margin. We also had our highest quarterly free cash flow margin of about 25%. Based on the strong start to the year, we feel very good about our Q2 and full year outlook. For Q2 2023, we are issuing guidance of $128 million to $131 million in total bookings, $122 million to $125 million in revenue and an adjusted EBITDA margin of 11% to 12%. And for the full year 2023, we are raising our guidance to $552 million to $561 million in total bookings, $500 million to $509 million in total revenue, and we are updating our adjusted EBITDA margin range from 11% to 12%, which reflects an incremental margin of about 32%. Because of the strong trends we saw in Q1, we are guiding to continued strong top line growth, with bookings growing at 30% year-over-year at the midpoint and revenue growing at 37% at the midpoint. The combination of strong top line growth and continued discipline on operating expenses is why we feel good about raising our adjusted EBITDA margin guidance at the midpoint. In Q2, we expect each non-GAAP expense line item to show operating leverage year-over-year with R&D showing about 1 point of improvement as a percentage of revenue, G&A showing about 3 points of improvement and S&M showing about 1.5 points of improvement. I know that the S&M line would show about 2.5 points of improvement, but for the roughly $1.5 million of S&M spend there was time shifted from Q1 of this year into Q2. As to the seasonality we expect for the rest of the year in adjusted EBITDA, we are guiding to an 11% to 12% adjusted EBITDA margin for Q2. Our Q3 margin will be lower than Q2, and that's the quarter in which the largest portion of our new hire start. And then in Q4, the margin will expand because that quarter is typically our strongest revenue quarter. We don't expect a material step-up in costs between Q3 and Q4. As Luis mentioned, we are excited about Duolingo Max, but we are still in the early days of rolling it out. We have not yet included any material amount of bookings or revenue in our guidance for this new higher tier. We will keep you posted on the progress in the coming months. Finally, we ended the year with approximately 48.3 million fully diluted shares outstanding using the quarter end close price. And as we mentioned on the last call, we expect to end the year with about 2% dilution from equity issued to employees. And with that, I'll turn it back to Luis.

Thank you, Matt. And I just want to take this opportunity to thank our amazing team whose collective passion and commitment to excellence helped us deliver another excellent quarter. And now we would be happy to take your questions as long as they're good. I'll turn it back to Debbie to manage the queue.

All right. Thanks, Luis. [Operator Instructions] So our first question comes from Mark Mahaney of Evercore.

Well, I hope this question is good. So we haven't talked to...

It better be, Matt.

We haven't talked -- you haven't mentioned yet one of my favorite topics, which is Math. So could you just talk about what you're seeing early on in terms of the interest in that? And then could you just talk a little bit about the China market, too? So that's kind of wax and wane, but I think it's been more waning -- or no waxing for you, becoming stronger. So how much of a contributor that's been to you in terms of your MAUs and, if at all, your paid subs?

Well, thank you for the questions, Mark. Of course, Math is also one of my favorite topics. As you know, we launched a Math -- an app to learn Math a few months ago. It's doing really well. It's growing entirely organically, and it's growing very nicely. I should mention, of course, this is still the very early days and it's still also a very small team. So we're just basically have a long list of features that we still need to add and we're working on doing that. I don't know if there's anything else to say other than I'm loving using it. And very soon, there's actually going to be more content in it, more advanced content. So we're very happy with that. In terms of China, we're also very happy with our progress in China. It's one of our fastest-growing countries. I should remind people, though, I mean, yes, China is one of our fastest-growing countries, and we seem to be doing well, but it is still a small market for us. It's probably 2% to 3% of our revenue, give or take, but it's growing very nicely, and we're very happy with it. Thank you, Mark.

Next question comes from Justin Patterson of KeyBanc.

Great. Two, if I can. First, just a big picture one. We've seen a lot of companies in gaming, in dating, [indiscernible] around monetization and engagement. Could you talk a bit, Luis, about just some of the guardrails you have in building product in a way where monetization initiatives don't necessarily impede the monetization or impede the consumer experience? So that would be the first question. And then since you alluded to it in your prepared remarks, would love how you're thinking about just leveraging AI more over the future? Whether that's something that investors should think of as just broadening the set of educational apps that Duolingo can participate, so even moving beyond math over time. So even just building deeper, more immersive experiences and existing apps. So moving up from what might be a casual learning experience to really gaining mastery of a competency.

Great questions, Justin. In terms of the monetization guardrails, let me just describe to you how we develop product. The way we make our product better is we run a lot of A/B tests. I mean we're running hundreds of A/B tests every quarter. And for each A/B test, usually it's trying to improve something -- so it's trying to improve either how well we monetize or it's trying to make the app more engaging or it's trying to teach better, et cetera. It usually has a goal. Now in addition to the goal, for every A/B test we run, we also have these guardrail metrics. So for example, if the goal of an A/B test would be to improve monetization, we may do something where like we make the subscribe button bigger or something like that. That's just -- this is just a silly example. The goal of that would be to improve monetization. We also look at what it does to our engagement. So we look at are people spending more or less time on the app. We look at whether people are learning more or less from that A/B test, et cetera. And we only launch experiments that do not mess up our guardrail metrics. So in this case, if making the subscribe button bigger, would make it so that people are spending less time on the app, we would not launch that experiment. That's what we're doing. And it's actually worked out -- we've been doing this for years, and it's worked out really well to be able to grow our monetization, which -- without having an impact on our user engagement, which just to remind you, for us, internally, we believe user engagement and user growth is the single most important thing because from that, all kinds of good things come up. I mean, when we grow a user, even if it's a free user, we have the chance to convert them into paying subscribers over the next several years because they'll continue coming to Duolingo. So this is how we think about that. In terms of AI, we're extremely excited. I'm personally extremely excited about AI. Since we launched Duolingo, the goal has always been to make something that can teach you as well as a one-on-one human tutor, but without the one-on-one human tutor because one-on-one human tutors are expensive and not very scalable. So that's been the goal. And we've been using AI from the beginning to be able to do that. Now the main way in which historically we've used AI is we've looked at all the exercises that our users solve -- and at this point, we're getting about 1 billion exercises solved by our users every day. And we use that to improve how well we teach. And in particular, we use that to try to give users the right exercise at the right time. So we try to make sure that you get exactly the right exercise, and we use all the data from our users. We've used -- we've built a very sophisticated model called Birdbrain that can do that. Now over the last few months, we've had this amazing thing of generative AI. And ever since that came out, ever since we got early access to it, we started developing features for it. And what's amazing about it is it's like having a really good writer on staff. It allows us to just have a really good language that can come pretty quickly. So that has allowed us to develop new features. For example, we developed 2 new features as part of our new higher-tier subscription offering called Duolingo Max. One of them is called Explain My Answer, which basically explains whenever you have a mistake, it gives you a human language understandable explanation. And the other one is called Roleplay, where you can basically role play or you can pretend to be something like -- pretend to be ordering a croissant in France or something. And so that gives you conversational practice. So, so far, we've decided to use AI to try to teach closer to a human. That's also, of course, going to allow us to get into other areas. So we are going to invest -- be investing in making, for example, our other app, our Math app, for example, better; Duolingo ABC better. We're also going to be using AI to create content faster and cheaper. So there's the part that is online, that is live, where people are interacting with AI, but there's also the part that is off-line where we just generate a lot of the content faster and cheaper. So in our case, we just think there's a massive opportunity to make our apps teach better, be more engaging, which is really important and also to have lower costs.

Yes. And just to jump in there, Justin. I think the one additional point I would make that I think is our not-so-secret weapon is Luis on this. You can tell by his answer there how excited he is. I mean it's really been a dream of his to have this type of technology so you can teach in this way. I mean he really wrote his PhD thesis on how humans and computers could work better to learn better together. So it's just a moment in time. We're pretty lucky that he's leading us through this. That's great.

And your next question comes from Ralph Schackart of William Blair.

I'm just curious on DAU to MAU is strong yet again and saw further reacceleration in the quarter, which is really impressive. Just curious if there's anything you'd call out there driving that really strong growth once again. And then second question, I know you've been testing pricing in different international markets. Just curious what sort of trends you're seeing there and how users are responding to different pricing plans?

In the case of the DAU and MAU, you're right, we keep getting the product, our main product, the Duolingo language learning app keeps getting more and more engaging. Our DAU to MAU ratio keeps getting better and better, we're at 28% now. And the reason for that is just because we keep running more and more A/B tests that make our product more engaging. That's really it. And so the types of things that we do, we make our street feature more prominent or just more intuitive. We make our social features better so that users get their friends to come back. So in general, we just have a very high-performing growth area. And for us, our growth area, what it does is it grows our DAUs. And of course, our DAU to MAU ratio is going to get better and better because nobody in this company is looking at MAUs. So MAUs, it's nice. We report them and everything, but we are trying to grow our DAUs because we believe that getting somebody to come every single day is the right way to learn a language. You can't learn a language by coming once a month. So that ratio keeps getting better and better. And I would expect that to continue getting better. Your next question is about regional pricing. So -- just to remind everyone, when we IPO-ed a couple of years ago, we had the same price everywhere in the world. We, of course, knew that, that was not the most optimal price. Over the last couple of years, we have tried prices in every country. And at this point, in most regions in the world, in most countries in the world, the price makes sense in that it's pretty correlated with the GDP for that country. So the prices make sense. There has been an increase in -- that has implied already an increase in our bookings. I should say the increase was nice. We took it, of course, but it wasn't life-changing because what happened with the prices was, in wealthy countries like the U.S. or Europe, the prices remained pretty similar. Sometimes they went up a little bit, but they remained pretty similar, whereas in poorer countries, the prices went down quite significantly. And what that did is you got a lot more people to buy. But you got to remember, these are poor countries where digital subscriptions are just not as mature or sometimes people don't even have payment methods. So we saw the decrease of prices in these poor countries as an important step in monetizing them. But we also understand that over the years or over the next few years, more things have to happen in order for us to monetize this as well as we're monetizing the wealthy countries.

Next question comes from Andrew Boone of JMP Securities.

One on Max. I would assume that Max is attracting a more advanced learner to the platform. Is that true? What are you seeing in terms of the type of learners that are actually adopting Max? And then secondly, just a broader question on LOMs more broadly. How do you think about this change in the competition for Duolingo? Are you at all concerned there? What comes top of mind as you think through that?

Thank you, Andrew. Okay. In terms of Max, I should say, so we're -- we're super happy that we launched Max exactly the day that GPT-4 was announced. So we're super happy that we put that out there very fast. . Now of course, when we put that out there, we didn't give it to all our users. We gave it to only a small fraction of our users. And the reason for that is because that's how we do product development here at Duolingo. We start with a small fraction of our users and then particularly for complex features like Max or like our home screen redesign or like the family plan or anything, it usually takes us about a year to optimize it on small groups of users and then we start giving it to more and more users. That is just how we do product development. That is what's happening with Max. So at this point, we can't really give, for example, metrics for Max or anything like that because we just don't know where this will end up. But what I can say is there is a lot of demand for this higher tier, which is good, we're happy with that. There's also demand for the actual features. And now to your question about whether this is attracting more advanced learners, the honest answer is we don't know. It is -- a number of people are buying it. It's hard to know whether they're more advanced learners or not, but they're very interested. They may be more committed or wealthier is probably the things that I would say rather than more advanced, but it's just hard to know what it's doing. I should also say because when we were preparing. Matt would kill me if I did not say that you should not include Max in your models because we ourselves are not including Max in our model. It is too early to tell. So that's the thing with Max. With the language models and competition, this is just not something we're particularly worried about. There's enough moats for us. For example, large language models are this great thing that allow you -- they train on the whole worldwide web. And they have this very generic information. And so they're -- whenever you ask it to write something, it's basically what the worldwide web would say, it's kind of like the average of what the worldwide web would say. What they don't have or they're not trained with, for example, data about how people learn a language or they're not trained with all the data we have. We have a lot of data. And so we -- the way we use these large language models is kind of on top of how we do our AI for our own stuff. So we have this huge data moat of all the people that are learning a language with us, and it is, I don't know, 10x, 20x larger than any of our competitors. And so that's one thing. We also have the distribution to -- even if we just apply the large language models without any tweaks, which we're not doing, we're actually tweaking them. But even if we did that without any tweaks, we still have much larger distribution than anybody else. So that's a big thing. We also, of course, have other moats like the fact that our product is so engaging and our brand and also all our engaging characters like our owl being passive aggressive and stuff like that. So this is just not something that we're particularly worried about in terms of competition. Hopefully, that answers your question.

Next question comes from Ryan MacDonald of Needham.

Congrats on the awesome quarter. Luis, maybe just piggybacking off of that topic. As you think about sort of Max and sort of the evolution of it as you sort of mature and scaling it out, I think one of the benefits of why sort of ChatGPT sort of gained so much adoption so quickly is not only was it powerful, but it was also free. And learners got to experiment with that -- with it. So as you think about rolling out the features more broadly with Max, how do you balance sort of functionality with pricing, given obviously Max is a sort of fairly steep increase relative to the core Super Duolingo?

Yes, that's an excellent question. It's something that, over time, we're going to see what features belong somewhere -- or what features belong where. At this point, we're happy that we have -- we have the standard free tier. We've always had Super Duolingo over -- for the last 5 years, we've had kind of the pay tier. We knew that adding an extra higher tier was going to be good for the business. So GPT-4 was a good kind of excuse, a good reason to add a higher tier for the business. And we have that. And so now we have these 3 tiers. And over the next literally years, it's going to take us years, we're going to be figuring out where to put each feature. And we're going to do what we think is best for our business, and it's important to mention, of course, we care about revenue and everything, but the single most important for us is our free user growth because that leads to good stuff everywhere. So yes, if we see that certain things make more sense in the free tier, we'll put them there. Now the one thing that I'll mention, which is important to mention is providing particularly live access to large language models is not super cheap. For example, in our case, we use OpenAI, we have to pay them for GPT-4, et cetera. So for now, we keep these features in the highest paid tiers so that we can actually pay for that cost. Over time, though, we expect that the cost of querying large language models is going to go down quite significantly, and that may allow us to do certain other things, for example, for the free tier. So it's just something that is going to evolve over time, and it's kind of a little early to tell where these features are going to end up in kind of on an ongoing basis.

Makes sense. I appreciate the color. That was my good question. That was my lame and sort of bad line. I apologize if I missed this. But what really struck me on the metrics is the nice jump quarter-over-quarter in paid subs. And if I missed it, I apologize, but was there anything that you would -- in terms of a specific region or sort of specific feature that you'd attribute that large jump that you saw sort of fourth quarter to first quarter on the paid subs side?

Well, there's a number of things. So our paid subscriptions just have been consistently growing ever since we launched our subscription. That's just basically -- they just keep growing and growing and growing. And there's no single reason for that. I mean we just get better and better at converting our users. And we do that by a number of things. I mean, one of them is just making the subscription more interesting by adding more features to it or improving the features for it. But also by merchandising, we get a lot better at knowing when to advertise the subscription, what to say to you to get you to subscribe. So all of that just -- we run enough A/B tests that gets us to more and more subscriptions. I think that's the main answer. That's just the standard thing that keeps happening every quarter. So there was a jump -- there's been a jump in, I would say, every quarter since we've been public.

Yes. Since we've been public, we've seen a nice trend up in conversion basically every quarter. So I think this recent quarter was just kind of continued strength in our conversion trends based on what Luis said.

Next question comes from Mario Lu at Barclays.

The first one is why is Love Language not a real TV show?

Listen, a lot of people have asked us for it, but we're concentrated right now on our main business, I think, is the answer. But it is -- internally, I would say if we had a vote, half the employees would vote to make it a real show or more than that.

Got it. You had my vote as well. The first one is on DAUs. You said it's a very important metric that you guys track. It accelerated in the first quarter. I guess any main drivers for this acceleration? I know you guys talked about compounding. But is this social media like TikTok, like newest promotion, anything to call out? And how should we think about DAUs for 2Q and rest of the year?

Yes. So I mean, the main reason our DAUs keep growing is because our product keeps getting better. That's it. I mean we can track that. We know that, and it's this compounding effect of just it becoming stickier and stickier. That's the main reason. Now there are other reasons, for example, you mentioned marketing. We keep getting more efficient and better at marketing. We have really found -- our marketing team has really found its stride in terms of what are the things that -- the levers that work and don't. For example, we have found that organic social media is really good for us. TikTok is an example, but it's not just TikTok. I mean we do really well in Twitter, we do really well in Instagram. So we just found that we have a brand that is very good for social media. And it's organic, it's not paid stuff. So that works really well. We have found that influencers work -- and this is paid, paid influencers work really well in certain countries. Particularly in Asia and Latin America, paid influencers work really well. And we have also found that some small amount of performance marketing in -- usually in cheaper markets works also really well for us. So these are the things that we have found work well. And combined, all of this has just -- keep accelerating our DAUs. In terms of what to expect for Q2, I'll let Matt speak to that. But I think -- the one thing that I'll say is it's very nice that our users keep -- the growth of our users has been accelerating for, I don't know, how many quarters in a row by now. Obviously, this can't go on forever. That's the one thing that I'll say, I don't think we can have 300 quarters of accelerating user growth. At some point, we run out of people.

That was going to be my point is we feel really good about those 7 quarters since we've been public. It can't go on forever, and we can still have really strong user growth, even if it doesn't accelerate. So it's not a requirement that things accelerate, just we want that to stay strong, and we like the trends so far in this year.

Great. And then just one more on gross margins. If we look at COGS, I assume a big portion of it is app store fees. There's this new ruling from the Apple vs Android court case where apps can now go direct to consumer, if you will. I guess what is Duolingo's strategy in terms of potentially going to direct-to-consumer and kind of expanding gross margins out of it?

Yes. So I mean, historically, we've been pretty clear that we view the app stores as good partners. Luis mentioned part of our advantage earlier is being -- our incredibly widespread distribution, and that's in part because of the app stores. Our margins on our subscription products have gone up because more and more users are staying on our platform for longer than a year, and that reduces our app store fees. And that's our primary way to increase subscription gross margins over time. As far as the other taking different [techs]. That's not really been something -- we've experimented with it in the past, but it's not really been a focus, and I don't expect it to be a focus for us in the near term.

And the next question comes from Arvind Ramnani of Piper Sandler.

Thanks, Debbie. All the good questions have been taken. So you get my stinky questions. So I wanted to ask about Duolingo English test, I know it's not a big part of the story, but it's kind of -- I think it drives a decent amount of optionality if some of the dominoes would fall in place. Where does that stand in terms of like the broader acceptance from universities? Or I think you said the British or the Canadian government are they sort of like big dominoes that will get it rolling?

Yes. Thank you for the great question. I mean the Duolingo English test is something that we're all very excited about. Me, personally, I'm very excited about the Duolingo English test. As you said, it provides quite a bit of optionality. It's quite complementary to the Duolingo language learning app because if we are able to really win at that, that would help us own the score that people use to talk about their language. What we really would love to get to is to a point where as opposed to people saying, when they're asked how much French do you know as opposed to people saying, "Oh, I took 4 years of French in high school," we want people to say, "Oh, I'm a Duolingo 65." . That's -- of course, that's going to take a while. But that's kind of where we want to get to. And the Duolingo English test is one of the ways in which that type of score would get a lot of legitimacy. So we're very excited by it. It keeps growing. So it's growing nicely. Now of course, I should mention that it's important to say it's about 10% of our revenue. The growth of the Duolingo English test is not going to be as smooth as the growth of our language learning app, which is just the kind of this very predictable, very smooth growth because the Duolingo English test has just these externalities, like, for example, governments accepting it, universities accepting it. So the growth is going to be just less smooth. But we're very happy with the progress. We've made a lot of progress with among U.S. institutions. We're making now good progress with Canada, Australia and the U.K., and we keep growing the number of accepting institutions there. Now you mentioned governments, yes, we are working on acceptance by the U.K. government and Canadian and Australian government. Of course, the U.K. being the bigger one. That's just going to take some time. They have a call for proposals, for tests, which hasn't even gone out. We don't know when that's going to go out, and it's -- they've been talking about it for the last year. So at some point, it will go out. It's completely outside of our control. And so that will -- that's the thing about dealing with governments.

Yes. Terrific. And then just in terms of kind of your Math, like if you can kind of share some of -- I don't know if I missed it, but any kind of metrics on sort of usage, anything you're sort of willing to kind of share in terms of kind of progress you made on Math?

Yes. We're not sharing exact metrics yet, but we are very happy with the growth is what I can tell you. I mean it's growing very nicely. You can probably look at there's ways to track metrics external, and you can see that it's growing very nicely.

Next question comes from Nat Schindler of BofA.

I think I'm last here, but -- Luis, you're a computer scientist and a language learning expert. You're a perfect person to ask this question because I've really been asked this a lot recently. I've been told this is going to happen. I really want to know. I'm trying to figure out how is the search box going to teach me French?

You mean how is a search box like a ChatGPT going to teach you French?

Yes, it's really amazing. I keep getting told again and again that ChatGPT is going to teach me languages. And I'm truly a little hard -- hard pressed to figure out how that would work. Do you have any ideas?

Well, I'll tell you this. I am sure there exists some savant out there who can probably learn French by just querying something like ChatGPT. Similarly, for hundreds of years you've been able to learn French by reading books. Books have been out there. You can actually -- in fact, some people, a small number of people can actually learn a language by reading a bunch of books. But your average person does not have enough motivation or enough aptitude to be able to really learn a language by reading books. Similarly, I don't think that they can just sit there with just a search box and be like, "Hi." And then it just -- usually, your average student just does not even know what to ask. And also the other big problem is keeping yourself motivated. We've always thought, and this is something that I think sets us apart from most educational companies, we've always thought the hardest thing about learning something is staying motivated. And the reason we believe that is because the technology to learn anything has been there for like thousands of years, it's called books. You could actually learn almost anything. You want to learn quantum mechanics. You actually can. You can go and read a book. Just turns out the vast majority of people choose not to because it's really boring. So the hardest thing about learning anything is staying motivated. And so that's the thing, and that's what really sets us apart with Duolingo, that we really are -- we are going to use things like GPT-4, but we're going to combine it with things to keep you more motivated and also guide you through it. I mean if you have no guide, it's very hard. I mean this is like asking you to just learn Math by yourself. You can, but it's -- the vast majority of people just won't do it. I think it's...

And a more serious question. How do I size your market? I know you brought up the 1 billion people around the world who are actively learning a foreign language. And my guess is, of your current subscriber base, half of which probably are in the U.S., all of which -- the bulk of which are adults. I'd say 100% of that group at least was not actively learning a language before Duolingo. So how do I use what's the -- what could possibly be the way to know where you've come in your penetration cycle? Because if I'm looking at this accelerating DAUs and MAUs, -- and I'm just wondering, everybody is trying to figure out when and where this goes.

It's a really good question to which -- the most honest one is we, ourselves, don't really know. I'll tell you the things we know . We know a few things. There's about 2 billion people in the world that are learning a foreign language. They're spending about $60 billion a year. That is -- we know that. We also know -- which is very similar to what you said -- if you look at a country like the U.S., we're massively growing the market. 80% of our users in the U.S. were not learning a language before Duolingo. They were not in the market. So we also know we're growing the market. . And it's because we make it so easy and we make it so -- we make it so that people can use Duolingo with no friction, they essentially get very engaged with it, and then they feel good that they're -- well, they're learning a language, they feel good about that. So it's hard to say exactly where this will stop. I mean, I wish I knew. If I knew, I would tell you. But we do know that there's a lot more room. I mean we have -- I think, the latest numbers is 73 million MAUs. There's a couple of billion people out there actively learning a language. So even if you just -- even if we weren't growing the market, there is still more room. But again, we're growing the market. So it's -- my answer to you is, I wish I knew.

And we have no more questions. So I'll turn it back to Luis.