This week, I made a video about Duolingo. Quite an interesting company, which is a market leader in its field. AI is a big bogeyman in this investment story.
On TikTok, they currently have fewer viewers due to a previous controversy.

This week, I made a video about Duolingo. Quite an interesting company, which is a market leader in its field. AI is a big bogeyman in this investment story.
On TikTok, they currently have fewer viewers due to a previous controversy.

I myself have been very satisfied with Duolingo’s free version, and I don’t mind at all that grammar isn’t covered, because I can always learn it through other means if needed. I’m not a big grammar fan; I learn languages more by just listening and reading. For a lazy person, it’s great that Duolingo diligently reminds you to study every day. I warmly recommend Duolingo as a tool for language learning. However, it’s also worth reading some easy-to-understand book alongside it.
Is AI more of a bogeyman or an accelerator? According to management comments, Duolingo has successfully accelerated its content production and significantly lowered the production costs of its own courses with the help of AI. Furthermore, their speed of collecting behavioral data and developing addictive features is at a level that language models cannot match.
I have a streak of almost 2000 days on Duolingo and I study languages in many different ways. Duolingo’s role is to additionally offer new words, repetitions, review, and daily language practice. It can be used more lightly or more heavily - or for example, in a game-like manner by collecting points. With a family subscription, you can get even more people included for the same fee.
Max AI usage is still being explored, it clearly requires concentration, courage to immerse oneself in spontaneous foreign language use in unexpected situations.
So, I personally see value in this in daily routine and repetitions.
Just out of curiosity, you’ve completed “two languages” during a 450-day streak. That’s quite an accomplishment unless you have a strong foundation in both. I have an estimated 700-day streak with my main language (a couple of others at the beginning) and I am at level 48. It’s considered completed when the level is 160. However, I started the language from scratch. I don’t see scores for Swedish and German as I’ve done so little of them, but they probably won’t run out right at the beginning. Or do the courses in the free version end so early that they are immediately completed?
It’s hard to say about grammar and such quality as people learn differently. For me, even Duolingo’s grammars have worked perfectly well (cf. school and mandatory rote learning). Lots of repetition and sometimes several of those grammar lessons etc. types. Or then errors are reviewed.
I haven’t thought about ranking, but you never know. But a new language has been learned with Duolingo, and I can manage basic things in the country in the local language, I don’t complain.
I played through German first; I indicated my starting level as a beginner and did the course from start to finish. I reached the final unit, and in my opinion, the course ended at a rather basic proficiency level. I would have liked to continue much further, but there were no more new units available. Next, I wanted to review Swedish. For that, I have a background of standard East Finnish school Swedish from decades ago, and I stated my starting level as something like intermediate knowledge—I don’t remember exactly what the term was. I think the app initially gauged my skill level and skipped a large portion of the early units. The course was over very quickly.
Different languages have different course lengths on Duolingo, and as I understand it, the number of units doesn’t depend on whether you have a free or paid subscription.
According to AI, the courses are as follows:
“Duolingo’s courses are constantly evolving, but according to the latest information, they look roughly like this:
German: about 130–150 units (in the old version 114–156, a new ~137–292 unit course is being rolled out)
Swedish: typically 55–85 units depending on which course version and device you are using (older 55 units, newer CEFR version 85 units)
Italian: estimates vary; previously about 51 units, newer updates have reported 66–79+ units”
I have now reached 474 days, and most of them were spent on German. Swedish was finished very quickly, and now I’m at “score 17” (Section 2, Unit 23) in Italian. Just don’t ask me about the logic behind those scores, sections, or units—no idea.
I bet Duolingo increases course lengths fairly easily by utilizing AI, so courses can probably lengthen quickly in today’s world. What that means for the quality of the courses, I don’t know. These courses for different languages have been slightly different in structure, but for example, the stories included in the courses are the same across different languages. So now I know in German, Swedish, and Italian that the guy who lost his passport at the airport runs to ask if the passport fell into the taxi, even though it was in his own hand. AI can presumably build more similar stories nowadays in a nanosecond in any language.
Have you used the Duolingo Max AI speaking partner service? What do you think of it?
I often do my exercises late at night, so spontaneous language practice for speaking situations doesn’t really suit me then. When I’m more alert, sure.
Duolingo was a very volatile stock even before that “boogeyman”.
The app has specifically been utilizing AI for some time now to select and refine teaching materials and to enhance users’ learning experience.
It’s very difficult to imagine that people would start creating similar tools for themselves on a large scale using ChatGPT or similar, when a high-quality ready-made application is available at a reasonable price.
Duolingo’s stock fell today due to a CFO transition and fourth-quarter preliminary data. However, user numbers grew by 30 percent and bookings are estimated to exceed targets, but investors were spooked by the leadership change and the company’s announcement to emphasize product development at the expense of short-term results. Gillian Munson will start as the new CFO in February.
“We’re prioritizing teaching better and user growth, and we executed on that strategy in the fourth quarter,” said Luis von Ahn, Duolingo’s Co-Founder and CEO. “We plan to continue to invest meaningfully in the product, even when it involves near-term tradeoffs.”
There are definitely the ingredients for a really interesting story here.
I do understand the argument that AI could eat almost all software. But still, all key metrics at Duolingo are moving in the right direction.
Duolingo is an interesting case. I see them as being in a strong position because they already have brand awareness and a customer base. Arguments that AI can be asked to teach any language are, in my view, secondary. Sure it can, but the experience is unlikely to be as good as Duolingo’s, and why would using AI be free forever anyway?
The real risk, in my opinion, is that due to AI-assisted live translation, people won’t need to learn languages to communicate. Meaning, when every pair of AirPods translates a conversation live, will anyone be bothered to keep up their Duolingo streak? (Use Live Translation with your AirPods - Apple Support)
I still believe in people’s desire to learn, and perhaps the superficial language learning offered by Duolingo is exactly what is wanted when more demanding conversations are handled by AI.
Does anyone have experience with the AI conversation partner feature available in Duolingo’s paid version? I consider it a decisive feature—if Duolingo develops a functional live-conversation-based AI “teacher,” it will dominate the entire language learning field for a long time due to its high brand recognition, scale, and existing customer base.
Duolingo’s value isn’t in teaching a language, but in gamifying it. You’ve got streaks, random rewards in diamonds, a social dimension, course difficulty levels precisely honed to fit, etc. Every psychological trick is in use. However, the price for this is that unpleasant things like grammar and comprehensive vocabulary grinding are avoided. Even if generating a language learning app with an AI agent were easy, getting the balance right wouldn’t happen overnight.
Additionally, Duolingo has the option to expand into new verticals—starting to teach new things. For instance, various musical instruments, visual arts, poker… even for languages, not every language pair is covered yet. Duolingo’s own pace of execution is also accelerating with AI, and they are leaving others in the dust.
On the other hand, could someone build an app so good for free that paying customers vanish? For example, with music, as the supply on Spotify has exploded, many people end up listening to fewer bands than before. Oversupply causes confusion, and a familiar brand becomes even more important. Will the same happen with software? I’ve bought a position in Duolingo stock and I’m watching it slide toward zero. Usually, a situation where the stock price and the business performance move in opposite directions is a buying opportunity. What about now?