Duolingo, Inc.

Good evening, everyone, and welcome to Duolingo's first quarter 2025 earnings webcast. Today, after market close, we released this quarter shareholder letter, a copy of which you can find on our IR website at investors.duolingo.com. On today's call, we have Luis Van On, our co-founder and CEO, and Matt Scarupa, our CFO. They'll begin with some brief remarks before opening the call to questions. Analysts will be able to ask a question by using the raise hand feature. And please note this event's being recorded and all attendees are in listen-only mode. Just a reminder, 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 in our filings with the SEC. These forward-looking statements are based on assumptions 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 now I will hand it over to Luis.

Thank you, Debbie. Normally, we start this call with prepared remarks, which in this case probably would have included the term AI like 17 times. But as I'm sure you've read in our shareholder letter, things are going great. So we're going to get right into questions. Shoot.

All right. Thanks, Luis. Now I'll pass it over to the moderator who will begin the Q&A portion of the call.

We will now move to our question and answer session. At this time, if you would like to ask a question, please click on the raise hand button which can be found at the bottom of the black bar at the bottom of your screen. You may remove yourself from the queue at any time by lowering your hand. When it is your turn, you will receive a message on your screen asking to be promoted to panelist. Please accept, wait a moment, and once you have been promoted, you will hear your name called and you may unmute your video and audio and ask your question. Analysts are only allowed one relevant follow-up from their main question. We will now pause a moment to gather the team and allow the queue to assemble. Just a moment as our first question is connecting. That will come from Curtis Nagel with Bank of America. Please unmute your audio and ask your question.

Taking the question, maybe one first for Matt and then one for Luis. Matt, could we just talk through the cadence of the gross margin for the year? Really nice outperformance on some of the max efficiencies. I'm still expecting, you know, I think, or expecting a 50 bit beep in 2Q and then ramping up in 2H. I guess not what, you know, why not more of a benefit in the second quarter? And how should we think about kind of run rates going into next year in terms of gross margin?

Thanks, Curtis. Appreciate the question. So we lay out in the shareholder letter, you know, the detailed gross margin guidance. So just to kind of reiterate what we put in there, you're absolutely right. We continued to expect for the full year about a 150 basis point decline year over year. We did see Q1 come in better than we expected. We expected about a 300 basis point decline, and we saw about a 200 basis point decline. For the most part in Q1, max was on plan in terms of its impact on gross margin, and we saw a slight benefit on ads, which helped. But as we lay out in the shareholder letter, we do expect Q2 to sequentially, compared to Q1, decline 50 basis points. And then we'll see gross margin ramp up in Q3 and Q4, such that by the end of the year, we end roughly back to where we started the year at the end of Q4 of last year.

Okay, awesome. That's helpful. And then, Luis, for you, just maybe expectations in terms of the long run. language modules, chess, music, math, right? Do we still sort of do these largely as tools for, you know, engagement, for streaks, adding value to existing users or, you know, maybe increasingly so as potential monetization tools? What are your thoughts there?

Yeah, thanks for the question, Curtis. So we're very happy, which is just added chess. It's not quite yet live in the app, but it's going to be live very soon. I should mention something amazing about the chess is that it really started with a team of two people, neither of whom knew how to program. So they were not programmers and they basically made prototypes and did the whole curriculum of chess by just using A.I. Also, neither of them knew how to play chess. And we started that for several months. And eventually, when they had a really good prototype, we had a whole team to professionalize it and put it in the app. So we're very happy with it. This is, at the moment, mainly going to be to increase users who we think there's a huge demand for chess. There's hundreds of millions of people who want to play chess. And the same is true for math and music. I mean, they're growing really well. We are monetizing all of these subjects, by the way. The way you monetize them is the same way we monetize language learning, which is if you want to turn off ads, you can pay to subscribe. So we're monetizing them already, but compared to language learning, they're a lot smaller. But yeah, we're very excited about it.

Okay, thank you. Our next question will come from Ralph Shackert with William Blair. Please unmute your audio video and ask your question.

Great. Thanks for taking the question. Matt, maybe I'll start with you. Maybe kind of give us a sense of any macro softness you may or may not be seeing. I think, you know, during the 2022 timeframe after going public, you didn't see much there, but just give us a sense there and on the DAOs kind of, you know, how broad based was that? Then Luis, just on Gen AI, you know, not focused on Macs, but Obviously, you've been working with AI for a long time. Just give us a sense, when you're testing now and developing products, what does GenAI allow you to do in terms of testing at scale and to develop products at scale? And then how does that sort of inform your roadmap to really sort of leverage that to create, I guess, more products over time and potentially be able to roll them out faster? Thank you. Yeah.

Thanks, Ralph. So when it comes to the broader environment, you can see from our Q1 results, which were great, and the strength of our guide that we feel confident about the roadmap for 2025. We have a lot of exciting opportunities ahead of us. And so that's what you see in the guide. So far in Q2, we've been monitoring the metrics we always monitor, and we haven't seen any real change in the trends based on any macro environment. So we feel good about that. I just remind you also that we're a global business. A majority of our bookings and revenue come from outside the US. And we offer a really attractive, in my view, value to price relationship. Our average Revenue producers in the $6 range. So we feel good about it. And that's what you see in the guide, Ralph.

And in terms of GenAI, I mean, in general, GenAI is allowing us three places where it's helping us. I mean, one is data creation. Teaching requires quite a bit of content. It used to be the case that, you know, a lot of this content was made by hand with some automation, but a lot of it by hand. By now, we're able to make the content for all of our subjects mostly close to 100% automatically. So that allows us to just have a lot more content that is made a lot faster and much cheaper. So that's one place where it's helping us. Another place where it's helping us is just creating new features that were just not possible before, like video call with Lily. That was just impossible to practice conversation with a computer two years ago. And then the last place is it's just helping us with efficiencies everywhere in the company. And the fact that, for example, chess is a really good example. This is why I mentioned in the last answer, the fact that two people who were not engineers really were able to get to a pretty far in chess. I mean, they didn't make the production version of chess, but what they made is really, really good prototypes. And so this is really, you know, this is showing what the future is going to look like. We're going to have a lot of people in the company that are playing around with new subjects or with new ideas. And I think that's going to speed up the development of our features. Great. Thanks, Luis.

Thanks, Matt.

Our next question will come from Chris Contarich with UBS. Please unmute your video, audio, and ask your question.

Predicting the question. Maybe the first would just be on the Mac side of things. Any update on retention trends for Macs and percentage of overall subs? And how should we be thinking about kind of where Macs retention is versus Super at this point? and i guess my second question would really be on the 148u language courses kind of pushing on that thread a little bit more here um you said it took 12 years to add the the first hundred i guess is it more are we at the point now where we kind of reach saturation here as far as the available courses that you're looking to offer here and it's about pushing deeper into certain markets or Is there another 100 courses potentially that we should be looking here that'll be driving those, really driving the funnel here? Thanks.

Yeah, so on the... The max retention rates, we feel good about how max is progressing. It's about 7% of subscribers at quarter end, which is an impressive set of growth. And it helped deliver a really strong Q1 for us. And it's going well. I think on the retention rates, it's kind of as expected. It's early days, but we feel confident that it's going well on the overall retention rates for maxes at this point.

Yeah, and in terms of the language courses, we added 148 new language courses. All of this content was done with AI, and it really took us about a year to add these 148 language courses. And like you mentioned, the previous 100-ish courses took us about 12 years to make. So this is an incredible speed up. Now, to clarify what these courses are, these are not new languages. So these are the same languages that we were teaching, but for other base languages. So, for example, we used to if you were a Portuguese speaker, you couldn't learn Korean, but now you can. So that's the type of thing we're basically filling in all the gaps that we had at this point. You know, we're teaching all the major languages to the majority of the countries that we serve. So we're very happy with that. We're still going to add other languages, likely. And there's demand for that. But there really is diminishing returns. I mean, in total, different languages, we only teach about 40 of them. And these 148 courses mainly pertain to about eight of our top languages that are now we're teaching in the different geographies. So there's not that much in terms of new courses to add. But we are going to continue improving the courses that we have and also getting them to teach to more advanced levels.

Our next question will come from Brian Smilick with JP Morgan. Please unmute your audio, video, and ask your question.

Great. Thanks for taking the questions. You know, on daily active users, I mean, continue to see such strong and impressive growth rates. I believe last quarter you had talked about mature regions, you know, growing still very strong. So just curious, you know, what are you seeing on the product optimization cycle that's driving better resurrected user growth? Anything that stands out to keep in mind there? And then secondarily and kind of related as well, more on the English learning side, you know, still under indexing towards English language learning. You know, as you continue to scale course units published, add new content and efficacy over time. You know, how should we think about user growth across, you know, English language learning and the opportunity longer term? Thank you.

Yeah, thank you, Brian. In terms of DAUs, it's a similar story to last quarter. I mean, we're getting very strong growth. I mean, we just posted 49% year-over-year growth on DAUs, which, to remind you, that is lapping the same quarter, but the year before was doing 60% year-over-year growth. And before that, another 60% year-over-year growth. So it's really growing quite fast. The growth is from all regions. Of course, some regions are growing faster than others, but really all of them are growing. And the same thing is true as last time. Some of our more mature markets are actually growing the fastest. So, you know, we're very happy with that. And what that tells us is that we really are not reaching the point of saturation in really any market. And it's because there really are, you know, the total number of people learning a language in the world is about 2 billion. You know, we just posted 130 million active users. So there's a lot of room for that. In terms of what drives the growth, it's the same story as has been. Two things drive our growth. One is technology. product improvements, and the product improvements basically get improved retention of the product and also get people to tell their friends, so word of mouth. And then the other thing that drives the growth is our excellent marketing campaigns. I mean, this last quarter, you saw that we had this campaign where DuoDAO faked his own death. Um, we, you know, we couldn't, we didn't know he was going to do that. Um, this is a kind of quirky guy that decides to do this type of stuff. Um, and but that campaign, for example, got 1.7Billion, 1.7Billion impressions, not dollars, 1.7Billion impressions. Uh, and, you know, the cost of that campaign really was essentially nothing. So that's what's generating the growth. In terms of English learners, we're very excited by it. We're still under indexed, as you rightly said. If you look at language learning as a whole, 80% of the people that are learning a language are learning English, but on Duolingo, It's a little around 50%. That percentage is increasing, which is good, because English learners in general are growing faster than the rest of our growth. So we see a lot of potential there. And just to remind everyone, the reason that English learners are now growing faster is because we've added a lot of content for them, in particular, intermediate and advanced content. And that's really helping us grow. Now, one of the things that we've said last time and we're going to continue saying here is it's going to take some time for this adoption to happen for English learners because the word got around. I mean, the main way Duolingo grows is through word of mouth. And the word got around that we were mainly good for beginner English learners. And now that we have content for more intermediate and advanced learners, the word has to get around. We're trying to speed that up with some marketing initiatives, but generally we expect that most of this is going to be word of mouth. Um, and so, you know, it'll in, in, in some time, and I don't know what that timeframe is, but it's not months. It's probably years. We do expect that English learners are going to, you know, hopefully get to the 80%. Um, but you know, it's going to take some time.

Awesome. Thank you. And I guess, um, just one quick follow-up on English learning. Um, you know, could you share any insights on just overall conversion rates? And I mean, right, like how that would vary versus a Western world user, anything would be super helpful. Thank you.

Yeah, I mean, English learners, a lot of them are in regions that don't convert very well. Some of them are in regions that convert very well. For example, Japan. Japan converts very well. But if they are in low GDP per capita countries, they don't convert very well. However, we are seeing something really awesome with Duolingo Macs. And it is that they really have, because Max really helps you try to practice conversation a lot more, they're a lot more interested in it. So we're seeing more adoption of a video call with Lily by English learners than non-English learners. And we're also seeing that, like we said, 7% of our subscribers are paying for Max. But actually, if you restrict to English learners, that number is actually higher than 7%, and non-English learners, that's lower than 7%. So the average is 7%, but English learners, the fraction of subscribers that are English learners that are max is higher than for non-English learners. So we like that.

Awesome. Thank you.

Our next question comes from Justin Patterson with KeyBank. Please unmute your audio video and ask your question.

Great. Thank you very much. It looks like cost optimizations are occurring faster than expected. You've got better line of sight on AI costs. How does that change just your desire to make Duo Max more affordable or even just provide more features to some of those English learners? And then related to that, the 3D Lily update for video call looks pretty interesting. Talk about how you think that's going to influence both engagement with Lily and conversion rates. Thank you.

I can take the first part, and you can jump in on the next one. Sounds good. Justin, on the optimizations, it's actually, in our opinion, trending as we expected. So there were some smaller optimizations that we were able to kind of do in Q1 that were faster than we expected, but they were pretty small. In general, the optimization of video call and max is on trend with what we expected on the last call. And that's good news for us because it does lead to margin expansion in the back half of the year, like we thought. And we also saw evidence in the first quarter that, you know, rates were coming down for those type of API calls. So, you know, just in general, we feel confident with our, the current course and speed of those optimizations.

Yeah, and related to max cost, I should say, it is still the case that max in certain countries, like India, India is a good example, is too expensive. We have not lowered the price. Max in India is about 70 US dollars per year. We feel confident that over, you know, probably in a few months, we'll be able to lower that price because, you know, prices are coming down. And at that point, we're going to see more adoption. But it's going to take a little bit of time. So that's Max. In terms of 3D video call, we're super excited about it. I mean, if you actually played with it, for some users already have it, not everybody has it. It's a lot more engaging. Now you can start asking Lily about her dog, which you can see it right there and everything. So I think that's going to increase engagement. And the good news about engagement is it increases a number of things. It increases word of mouth. of Max itself. So basically people are like, you should get Max. I know you're a Duolingo user, but you should get Max because you can talk to Lily. It also, of course, increases retention. So it's just generally going to be good. I don't know exactly what the results are going to be of the A-B test. I mean, this is an A-B test at the moment, but we're pretty excited about it.

Our next question will come from Nathan Feather with Morgan Stanley. Please unmute your audio video and ask your question.

Hey everyone. Thanks so much for the question. Really encouraging results. It's exciting to see the expansion and additional language courses, but given that there is a diminishing return as you add additional languages, do you ever think your relative prioritization list for new content production now between further language courses, more advanced content, new subjects,

Yeah, in terms of relative prioritization, I mean, we're at the moment, of course, language learning is our largest business and it is growing very fast. So the majority of our efforts still go into that. Now, in terms of creating content there, we may not be adding new languages, but we're always improving the content for the languages that we have and making it more and more advanced. And what's nice about this is because we now have this pipeline on AI, we can actually regenerate the content super fast, but in an improved manner. So you'll just see us improve things a lot, you know, and it pays to do that. So, for example, you'll see us improve our Spanish course, which is one of our largest courses. And, you know, people all over the world will be able to learn Spanish faster. We are, of course, also investing in math and music and chess. We're spending less effort on that because they're much smaller businesses than language learning. But you'll see us, for example, for math, you'll see us vastly increase the content because of AI. Within the next few months, the content is probably going to quadruple that we have in math because of AI in, call it, I don't know, a quarter or so.

Very helpful. Thank you.

Our next question will come from Arvind Ramnani with Piper Sandler. Please unmute your audio video and ask your question.

Hi. Hey, thanks. Thanks, Louise. Thanks, man. Good background. I love it too. Well done. I use it for all my calls, all my Zoom calls.

Yes, I believe you.

Look, I mean, you know, you launched math and music back in 2023. Now you're launching chess. And, you know, I'm really trying to better understand your business strategy. I mean, I fully appreciate that, you know, the investment costs are minimal. Even with math, I believe it was just a couple of engineers. I appreciate that part of the equation. But also, with the past couple of years, you've insisted that don't look at math and music and chess is going to be free. Don't look at it as monetization. But there is clearly a business strategy for expanding product. I'm just trying to get a better understanding of that.

Yeah. To clarify one thing, All of our new subjects, math, music, and chess, when it launches, which is going to happen in a few days, they all have monetization. They do make us money. And it's because the same subscription works for everything. And if you're not subscribed and you're doing math, you have to see ads at the end of a math lesson. And all the standard things that get you to subscribe will get you to subscribe. So we do get subscriptions from people who are using math or people who are using music. And soon, it'll be people who are using chess. But, you know, they're proportional to the number of daily active uses that we have in those subjects, which is significantly smaller than language learning. Now, they're growing. The good news is these new subjects are growing a lot faster than language learning. And now they have millions of daily active uses. So there is some monetization there. It's just we're not breaking it out yet. But in general, the strategy is really to become not just a language learning app, but an app that teaches you subjects that take a long time to learn. And these are all going to be subjects that hundreds of millions of people want to learn and that take a long time to learn and that are also good for the world. We believe that. getting better at math, getting better at music, getting better at chess, are all things that make people smarter and are good for the world. So that's the strategy at the moment, and we're very excited. But we've known, and you are right, we've been kind of cautioning people. We know that it takes a while to grow each of these subjects. I mean, it took us 10 years to grow language learning to a big scale. So it takes a while to grow these subjects, but we're very happy with the growth so far for them.

Perfect. And just a quick follow up. What's the sort of like common theme in learners? I mean, I listened to an interview a couple of weeks back where you talked about you're looking for large markets and all of that. But is there a common thread where like the person who learns math is also learning music or learning? What's the common thread?

There's definitely overlap in all of our subjects between them. You know, the general common thread is people who want to improve themselves in one way or another and who used to probably a lot of these people used to spend all of their time on playing either mobile games or on, you know, doom scrolling on social media. And now they're spending some of that time improving themselves either by getting better at math or music or language or something. That's the common theme. And there's some overlap between them. It's not 100% overlap. There's many, many users only learn one thing. But quite a few users are learning, you know, multiple of our subjects, including multiple of our languages.

Perfect. Thank you. Excellent.

Our next question comes from Wyatt Swanson with DA Davidson. Please unmute your audio, video, and ask your question.

Hey, guys. Just kind of following up on chess here, could you provide details around timeline and plans for rolling out the course to more users' devices? What point do you think it'll be widely available? And what are your plans around driving engagement and making it well known that you also teach chess? And then just looking at a little bit further, how do you expect it to contribute to engagement revenue, you know, 12 to 18 months or even further down the line?

Yeah, thanks for the question. OK, so Chess, I literally a few minutes ago promised the product manager of the Chess team that I would not give a specific date. But it really is in the next, you know, next few weeks. It's like very soon. It'll be there. It'll first be on iPhones and then soon after it'll be on Android phones. You know, by the end of the year, this will be widely available to everybody online. uh in terms of you know how we're gonna get the word across it's kind of similar to math and music most of it is word of mouth but we're gonna speed it up with our marketing uh certainly you'll probably see a lot of uh you know kind of tick tock or uh you know instagram videos of our uh our mascot doing you know unhinged things related to chess um so that that's gonna get the the word out we're probably going to uh you know we're gonna test our different we're gonna we're gonna uh point our marketing engine towards this this and including also towards math and music um so i think we're gonna you know we're gonna see quite a bit of growth from this for for these new subjects in terms of the contribution of chess it's similar to math and music um there is a business there as in already people are are going to be as soon as we launch it people are going to be able to pay for a subscription for chess and that removes the ads and and also all the things that a subscription does so there's going to be a contribution there but it will depend on the daily active uses. And like I said in the previous answer, it takes time for daily active uses of a given subject to grow from zero. So I don't really know what the contribution is going to be in 12 months, but compared to language learning, it's probably going to be small.

Got it. Okay. Thank you, guys.

Our next question will come from Edison Kai with Citi. Please unmute your audio video and ask your question.

Hi, Louis. It's great to see such a strong performance given that it's really early in China right now, but a very good performance. And I do have one question. Marketing campaign and the market that you're diving into, because that's what you talked about, how your market campaign is shifting from growing the... user base to let the users know how your in-depth English content is growing. And now you are doing chest contents, which I think is another rocket, given that there are already mature provider, like service providers like chess.com. So how do you think that your market campaigning and user base growing is going to shift from current situation, given you're diving deeper into the market in both language learning and the new markets like chess playing, which is more competitive. So I would like to know more about that.

Yeah, thanks for the question. OK, so our marketing, you know, what's what's amazing about our marketing is that it's incredibly efficient. I mean, most of our marketing efforts are, you know, social media efforts, either on TikTok or Instagram or, you know, different social media, you know, sites, including Chinese sites. And we basically have a really good playbook for doing this type of marketing on social media where we don't spend very much money. The majority of our efforts are going to continue in the same way that they've been so far, which is making content that goes viral almost every single day. some content related to Duolingo, either made by us or made by our audience, goes viral. And by that, I mean it gets millions of views. That will continue, and we're not going to stop investing in that. Additionally, what I mentioned last quarter is if we are trying to get things like the word out that Duolingo is now good for more advanced English learners, we're probably going to do some things related to that. This includes things like influencers. It also includes things in certain geographies. For example, in Japan, We ran a TV campaign that actually had a video call with Lily. So you'll see us do stuff like that, but the majority of our efforts will continue being kind of the owl doing unhinged things. And when it comes to things like chess, probably the owl will do unhinged things related to chess. And I think that'll get the word out that we teach chess.

Got it. Thank you, Luis.

Our next question comes from Alex Sklar with Raymond James. Please unmute your audio video and ask your question.

Great. Thank you all. Luis, just starting off on packaging across your tiers, you've talked about testing, pricing, and also potentially moving some certain features across the tiers. What have you kind of learned so far this year from those tests? How are you thinking about any change in terms of new features still to come, including those in the premium version versus wanting to put them in one of the paid plans?

Yeah, a great question. I mean, you know, with packaging and pricing, it's the same story that it's always been, which is we are testing a lot of things and we continue testing. I mean, for example, we've continued testing pricing in terms of, you know, geographically in different countries, we're testing different prices. This last quarter, we actually tested a price increase for new users and, in fact, increased the price for Super Duolingo for new users because of that. So you'll see us continue doing that going forward. We're always testing things. We have not moved things between packages. So for example, between tiers at the moment, Max still contains role play and video call and explain my answer. And super still contains what super has been, you know, has been having this whole time. You'll see us change a few things. I mean, not change, test a few things. I don't know what we'll end up launching. So far, nothing, no changes have launched. But you'll see us doing that. It is likely that some of the Max features that are cheaper for us to give, we may bring them down to packages. For example, Explain My Answer is relatively cheap for us to do. That is not the core feature of Max. The core feature of Max is video call, but that's a feature that's in Max. We may try it on Super or even Free. So we'll try that, but at the moment, nothing has changed.

Okay, great. Matt, maybe one follow-up for you. You've given some stats on annual and family plan penetration in the past. How does that mix look for your max base versus the other super base? Any difference in kind of annual or family plan adoption between those tiers?

Yeah, sure. I think that, you know, the... The theory of the case for us is providing the right package to the right user at the right time and servicing that. That's primarily how we drive Max and Super and family plan adoption. And right now, the mix is different for Max than I would say annual or family plan adoption. that'll evolve over time, right? It's still very early days. It's 7% of subs is max. And so as we move features around and then decide how and when to surface family plan for max to users, that'll evolve. But it's too early to kind of home in on what that's gonna look like in the longer term. All right, great. Thank you both.

Our next question comes from Ryan McDonald with Needham. Please unmute your audio video and ask your question.

Thanks for taking my questions, and congrats on a great quarter. Two questions on AI. Luis, first for you, can you talk about the level of urgency internally in terms of new generative AI features, particularly as we're seeing more funding going into the space for gen AI and language, whether it's speak or Preply, but then also the announcement that Google has some lightweight gen AI language learning now? And then conversely, on the cost side, Matt, you know, there was this email, obviously, that you published about internally sort of this new approach in evaluation of use of AI and as we incrementally hire, you know, let's see if we can automate things first. Can you just talk about sort of how quickly that approach can translate to improve the unit economics? And does this change your view on the long-term sort of structural profitability of the business? Thanks.

Yeah, thanks for the question. I mean, we believe that AI is really transformative for our business, and I think it's probably transformative for a lot of businesses, but we believe AI is very transformative for our business. It helps us teach better. It helps us create content a lot faster. It helps us create content that was just infeasible to do before. I mean, a good example is we just released 148 language courses. We wouldn't have done that if it wasn't for AI. We are really going all in as a company on AI because it's a technology that particularly applies to us because the main improvement in AI the last couple of years is large language models and language is what we teach. So it particularly applies to us. So you'll see us, I mean, we really are, you know, a lot of features are being developed that are related to AI. And you'll see us develop the right features for the user. So, you know, we really want to stay ahead with it. And we believe that we're far ahead of everybody else on this. I should mention that in terms of my email to everyone, I did say that we're going to be AI first, we're going to try to automate everything. In terms of cost, we're not changing, and I'll let Matt talk a little bit about this, we're not changing our estimates because yes, this is going to save us some costs on things that humans used to do that now AI can do. However, We're going to apply all of that. I mean, many of those same people, but we're going to apply not just with people, but all of that to develop AI features. We think this is an amazing opportunity for us. So the cost savings are going to be offset by the fact that we're just investing a lot in making the right features. I don't know if Matt has more to say about that, but that's how we're seeing it.

No. Ryan, just to underline that last point, you know, we... have done things because of AI that we wouldn't have been able to do before. So that's not taking the savings to the bottom line. That's investing in the enormous opportunity ahead of us. And that's what we're going to continue to do. I think we first and foremost prioritize reinvesting to go tackle all of the opportunities. And that's how we're going to use this.

Excellent. Thanks again for the color.

Our next question comes from Andrew Boone with Citizens JMP. Please unmute your audio video and ask your question.

Hi, guys. Thanks so much for taking the question. I wanted to ask about subscriber conversion rates. It's been a steady progression up kind of over the last four quarters. You know, call it 10 basis points a quarter if you look at kind of trailing 12-month mouths. Can you speak to the incrementality of max, though, as it affects that? Should we expect kind of that step up as you guys do start to unlock maybe what is max in terms of a lower price or connect that metric to max and incrementality of that subscription product? And then I wanted to add on to Ryan's question, right? Like Zuck just talked about this, but he talked about the unlock of what is an AI engineer that's going to be a mid-level engineer that's coming in the next kind of year. Luis, he dropped out of college. You have a PhD, right? Help us understand your view of the AI software developer and how do we think about the efficiency of kind of your OpEx and workforce on a go-forward basis?

Matt, you want to take the first part? Do you want to you want to start with the second part first? Oh, sure. I'm happy to talk about the professor.

I mean, look, In particular for engineering, AI is getting a lot better. That is just the case. So we're going to see a lot more efficiency from engineers. We're already seeing it. In terms of what it's good at and what it's bad at, I think AI coding is really good at creating something from scratch. It's not as good at taking a large code base, which is what we have on Duolingo, fully understanding it and making modifications for it. It's not as good at that. And that's going to take some time to be really good at that. And so my sense is what's going to happen, I truly do not believe that over the next, at least the foreseeable future, it's going to be the case that we're not going to need engineers. In fact, if you look, all of these companies, even the companies that are developing you know, the AI are hiring engineers. And there's a reason for that. I think we're going to need them and we're going to need them in part because they're going to be so much more effective and they are inefficient than they are now. So I think that's what's going to happen. We're just going to see them be a lot more efficient. Another thing that is really exciting is that now also non-engineers can do stuff, maybe not production code, but just kind of play around with things. One of our goals internally at the company is that, you know, 100% of our employees, and I'm not just talking about engineers, really 100% of our employees will have coded something with AI. We want everybody here to know how to automate things. And I think just the whole workforce is going to get a lot more efficient because of that.

Andrew, in that context, what was your first part of your question?

So I want to focus on the incrementality of max, right?

So if I think about the conversion rate... Sorry, I was thinking about the... Matt was worried that he's going to have to vibe code something himself. Oh, I love vibe coding. I do it most days. So when we think about the incrementality, Andrew, when we run our A-B tests for max, again, we are not optimizing for... in the very specific sense, we're not optimizing for a Mac subscriber or an ARPU or this or that. We are optimizing for platform LTV. And so when we run our A-B tests, we have to take into account, you know, are these... Max subs incremental or not in figuring out the overall impact on platform LTV. And that's why we launched Max was because we ran enough of the experiment to figure out that they were incremental. There were enough incrementality to it that it was platform LTV positive. And we still think that's true. And in fact, we think it'll become more true over time. And so that's really my answer to your question is that, again, we are not trying to drive a max penetration rate. We're not trying to drive a subscriber penetration rate. We are trying to drive platform LTV higher by driving mix shift to higher LTV plans. That's max. And we see also, you know, there's a lot of... Part of the Q1 outperformance was driven by the fact that super subscribers were converting up to max subscribers. And that's what I said to the answer to another question, which was showing our current users, free and paid, the right plan at the right time to get to the right LTV. So we feel really good about the incrementality and the overall platform LTV increase we're seeing from max subscribers.

Thank you.

Our next question will come from Mark Mahaney with Evercore ISI. Please unmute your audio video and ask your question.

Okay. Thank you. I may have missed this, but I know you disclosed the max penetration, I guess, at 7%. Did you provide an update on math and music adoption? And then you talked a little bit, Luis, I think, about pricing for Super, and I think you're your logic is kind of, I don't think you've changed pricing in super in quite some time. And as you improve the product and you prove more, the functionality, make it better and better that you, you warrant, you, you, you create greater value so that you can, you know, potentially ask people to pay more for it. So I've noticed this long way of answering, where are you, you know, are you thinking about raising prices kind of across the board on, on super and, and then what would be the key factors that would make you decide to do that now versus, you know, never doing it or doing it next year. Thank you.

Thank you, Mark. In terms of adoption for math and music, there's no update from last time that we talked about it, which we said we had about 3 million daily active users learning either math or music. It's higher now. It's growing. And math and music are growing faster than language learning. But compared to language learning, they're still small. And now in terms of prices, we're just always testing different things. So at the moment, we're testing prices in different countries and everything. My sense is there will be increases, but I don't know what the A-B test will tell us. And generally, we're just going to be data-driven on this.

Thank you.

Our next question will come from John Coluantoni with Jefferies. Please unmute your audio video and ask your question.

Great, thanks for taking my questions. First, I wanted to come back to the topic of incrementality of Macs. Can you give us a sense for what portion of recent Mac subscribers are new to Duolingo and what portion are intermediate English learners? Just to give us a sense for that incrementality. And second, turning to gross margins for Macs, you'd originally characterized the subscription tier as dilutive to gross margin, but accretive to gross profit dollars. Has that changed as more cost-effective AI models have been released since you first made those comments? Thanks. Yeah.

Yeah, so we're seeing, like I mentioned to Andrew, we're seeing good incrementality. We like the increase in platform LTV for Macs. A lot of that comes from the fact that a lot of new to the platform or new to paid subscribers are choosing Macs for the first time. But we're also seeing a relatively high proportion of folks upgrading into Macs. So of total Mac subscribers, a good chunk of those are coming from folks who currently are paying us and then start paying us more. So in general, the bookings incrementality of max is attractive. And we like how that's trending. In terms of the gross margin for Macs, nothing that you mentioned has changed. It's still the same lower gross margin percent, higher gross profit dollars. That will change over time as two things happen. One, as the price of the generative AI used for video call, for example, comes down. And we saw in Q1 new models released with lower pricing. And then we're also going to run optimizations in the back half of the year. So we would expect the gross margin of Max to increase. But again, the goal of Max is right now to drive incremental subscriptions and bookings in LTV and still make a really attractive gross profit dollar return. We're on track, like we mentioned on the last call, to do that better and better throughout the course of the year. Thanks so much.

Our next question will come from Ross Sandler with Barclays. Please unmute your audio video and ask your question.

Hey, guys.

So I want to ask a question sort of for Matt Chalupa, but maybe for Luis, because we might need a PhD to answer this. But yesterday, Apple was finally dealt a fairly definitive blow on the injunction with Epic around, like, high-quality developers like Fortnite or Duolingo redirecting users to pay for subscriptions off the app store billing. And it seems like you won't be discriminated against if you do that now. So I know 60% of bookings comes from in-app payments, I think is what it says in your filing. So Luis, I guess, have you guys looked at testing the flow of directing people to the web? And Matt, what percent of COGS does this represent? and kind of where do you guys see this going? Is this an opportunity to drive a little bit of leverage?

Yeah, I mean, first of all, this ruling is brand new. Obviously, we saw it. The app stores, both Apple and Google, have been great partners to us. One of the things that they really help with is just ease of payment. I mean, people's credit card is already in there, and it's usually just kind of like a double click that you have to do when you pay. So that's great, and that has really helped us. My sense is that if we were to try to send people to the web, we wouldn't have to pay Apple or Google, but we would see slightly fewer people actually converting because there's more friction. We're obviously going to test something here, given the ruling. We're obviously going to test something and we'll see what happens. But at the moment, I can't give you results. It's hard to know exactly how much this will help. It may help, but it's hard to know by how much.

Yeah. And then on the gross margin of the COG side, Ross, I mean, it's the majority, well over a majority of our COGs are payments to payment providers. And so to the extent that you would take a chunk out of those, it would be definitely accretive to gross margins. But as Louis said, it's early days for us and we're going to run A-B tests and green machine any change we make. So it would happen over time and you know, we would just have to wait and see. It certainly seems like it's, you know, potential optionality, but again, it's not something that we run experiments on given the recency of the news. We would be foolish not to test this.

Thanks.

Our next question comes from Shweta Kajaria with Wolf Research. Please unmute your audio video and ask your question.

Hello. Thanks for taking my question. Luis, I have one for you. What is your vision for the video call feature as it evolves over the next one, two, three years? What are you working towards? What's a good, not end state, but close to an end state solution look like?

Yeah, I mean, for video call, the main thing we want is to be able to get you to practice conversation. That's a very important part of language learning. And so the more time we get people to spend on practicing conversation, the better they'll get at the language that they're learning. We know that. So what we want to do is just get people to spend more time on it. How do we do that? By making it a lot more engaging and by making it a lot snappier. So, for example, right now, there's still a little bit of latency between when you say something and when Lily responds. We need to decrease that. It is still the case that... You know, sometimes she cuts you off a little too soon. We need to get better at she needs to get better at knowing when the conversation has faded out. Humans are actually really good at basically saying, I now know you're not so interested in talking to me, so I'm just going to leave. Humans are very good at that. Lily is not very good at that. So, you know, there's there's that type of stuff we got to get better at. And then, you know, we're going to have a whole world for her. where she's going to be able to, you know, change her background immediately and start pretending that she's like, you know, your taxi driver or something. So we're really, the idea is to practice as much language as possible and to increase engagement. And the more we increase the engagement, the more people will tell their friends that not only to use Duolingo, but if they are a Duolingo user, they should get Duolingo Max to be able to practice with Lily.

Thanks, Luis. I'm showing no further questions. This concludes the question and answer session of the call. I would now like to turn the call back to the host for closing remarks.

Just want to say thank you. Thank you for the great questions. But more importantly, thank you to our excellent Duolingo employees for yet another great quarter. The team is firing on all cylinders. So thank you and thank you, everybody.