Cem Kansu on How Duolingo Built a $500M Business Being "Unhinged"

When "Unhinged" Becomes Your Product Philosophy

At most companies, calling a product change "unhinged" would be an insult. At Duolingo, it's the highest praise.

Cem Kansu, Chief Product Officer at Duolingo, has helped grow the language app to 100 million monthly active users and $500 million in annual revenue over 8 years. His secret? Breaking every sanitized product norm in the book.

When someone says, "Oh, that change is unhinged," at Duolingo, that means you've done a good job. Like, you've really unhinged this change. Our app icon is one. When Duo does something really weird in our app, we're like, yeah, cool, that's really unhinged.

— Cem Kansu

The philosophy shows up everywhere. They've changed their app icon multiple times to show Duo looking sick, wrinkly, or melting—with boogers dripping from his nose. Users who haven't opened the app think it's personalized punishment. It's not. But the guilt brings them back.

Three Growth Levers That Compound Over Time

Duolingo didn't explode from one viral feature. The growth came from layering retention mechanics that compound. Kansu points to three main levers built over 8+ years.

First: Streaks. The consecutive-day counter existed since launch, but Duolingo kept improving it. They made it more animated, more visible on the homepage and profile. The higher your streak, the more painful it is to lose. Some users have 1,000+ day streaks. That's not retention—that's religion.

Second: Leaderboards. Weekly XP rankings tap into pure human nature. You don't want to be at the bottom. You want to climb. More lessons = more XP = higher rank. Simple, effective, relentless.

Third: Push notifications. Not generic reminders. Passive-aggressive, character-driven messages written in Duo's voice. If you ignore them for days, the final notification reads: "Hi, it's Duo. These reminders don't seem to be working. We'll stop sending them for now." It feels like a text from a pissed-off partner. And it works.

If we send you a notification for a few days and you don't come back to the app, we don't keep spamming you forever. We stop at some point. And the last notification that you receive is meant to be like a text you receive from a significant other that is kind of pissed off.

— Cem Kansu

Freemium After Promising "Never Ads, Never Subscriptions"

When Kansu joined in 2016, Duolingo had zero revenue. CEO Luis von Ahn had publicly said the company would never have ads, never have in-app purchases, and never have subscriptions.

Kansu built all three.

The shift wasn't about locking content behind paywalls. Education companies monetize courses. Duolingo refused. They wanted to charge for bells and whistles, not learning. So they tested ads, then optional in-app purchases like streak repairs, then subscriptions starting with ad removal.

Today, subscriptions are over 80% of revenue. Free users and paying users learn the same amount of French. The paywall is for features like unlimited hearts, personalized practice, and now AI-powered conversation tools.

It took endless A/B tests—ad load, session timing, Plus vs. Super branding, purchase flow colors. Not one magic bullet. Just relentless experimentation.

AI Isn't Killing Duolingo—It's Powering It

Every time OpenAI ships real-time translation, someone tweets "RIP Duolingo." Kansu laughs it off. Google Translate has done real-time translation for years. Language learning demand only went up.

People don't just learn languages to be transactional. They want to connect with heritage, talk to relatives, order cappuccinos in Italian. Real-time translation has lag. You can't work a knowledge job with a 15-second delay in every conversation.

Real-time translation tools have been around for a long time. Google Translate can do real-time translation quite well, and it's been able to do that for a long time. That has not reduced the need to learn languages. The demand for learning languages has only gone up over time.

— Cem Kansu

But AI is transformational inside Duolingo. The team uses it for two big things: content creation and interactive features.

Generative AI slashed the time to build and iterate courses. What used to take human teams weeks now happens in days. They can A/B test new teaching methods faster, launch new languages faster, improve courses faster.

The second use case is Video Call—a feature where you call Lily, Duolingo's emo teenage character, and practice conversation in real time. The AI tailors vocabulary and speed to your proficiency. She remembers past conversations. If you told her you're going to Mexico, she'll ask about your trip next week.

It's the closest thing to practicing with a real human. No typing crutch. No scaffolding. Just speaking. Duolingo tested it internally for a year, iterating every few days, before rolling it out to learners.

International Growth Beyond Localization

Most companies think international expansion means translating UI strings. Duolingo goes deeper.

First, they build courses that market needs. If you want to grow in Japan, you need a Japanese-to-English course. Second, they solve product blockers—WeChat login in China, performance issues, proper localization.

Then comes marketing. Duolingo doesn't spend heavily on paid acquisition. Instead, they build local TikTok accounts that post unhinged content in Japanese, Spanish, French. They work with influencers. In Japan, they run TV ads because TV still works there. They don't apply the same playbook everywhere—they learn what works per market.

The goal isn't just virality. It's brand love. Duo's personality makes Duolingo a brand people care about, not just another education app.