How Duolingo Built a Chess Learning App with AI—From Prototype to 55 Hours of Content

Welcome to the latest issue of the AI Community Learning Series – a space to explore how leaders and innovators are reshaping industries with AI 🚀. In this session, Edwin Bodge, Group Product Manager at Duolingo, shared how the team took on their most unexpected challenge yet: teaching chess. If you missed it, catch the replay here.

Here are the five key takeaways from our discussion. 👇

1. Choose a Problem That’s Deep and Sticky

At first glance, launching a chess course might seem like an odd pivot for Duolingo. But as Edwin explained, chess shares something vital with Duolingo’s core learning philosophy: it’s a skill learned through structured repetition, long-term practice, and play. The team asked themselves, “What’s something the world wants to learn, but has too few on-ramps for?” Chess, despite its global popularity, is often presented in intimidating ways. Duolingo saw an opportunity to change that by applying their gamified, friendly UX to a space that felt exclusionary—and to teach chess as a language of logic and pattern.   

2. The Power of Scrappy Prototyping  

The original team was just Edwin and designer Tyler Murphy—no engineering resources, no backend team, no formal greenlight. Instead of trying to convince them with traditional methods, they hacked together a working prototype using spreadsheets, Duolingo’s internal tools, and a lot of persistence. They storyboarded lessons by hand, wrote content manually, and simulated what the course might feel like in the real app. They didn’t pitch with slides—they showed leadership.

This DIY, zero-permission approach echoed startup hustle and enabled them to validate the idea quickly and visually—earning support to staff their project with much-needed engineering resources not by alignment meetings, but by momentum.

3. AI Was a Multiplier—Especially When Used Thoughtfully

Rather than offloading whole workflows to AI, Edwin and team treated AI as a junior teammate. They used Cursor and Claude to help scaffold lesson logic and lesson formatting, but always reviewed and revised the outputs. They used ChatGPT to automate internal processes like updating infrastructure documentation or setting up experiment arms. And perhaps most interestingly—they used AI to act as a sounding board: a fast first-pass to iterate ideas before bothering a colleague.

The key wasn’t trusting AI blindly—it was treating it as an idea-stretcher, a thought partner, and a way to remove friction in the build process.

4. Design for Beginners and Test Relentlessly

Throughout the process, Edwin remained obsessed with making the experience friendly to absolute beginners. That meant running hands-on user tests—sometimes with colleagues, sometimes with his own family. One winter break, he sat down with his mom and had her test the entire flow, noting where she got stuck, where the copy failed, and how intuitive the mechanics really were.

The team used every piece of feedback to revise lesson sequencing, tweak illustrations, and fine-tune onboarding. They weren’t just teaching chess—they were unlearning decades of gatekeeping design patterns.

5. Getting Buy-In Through Empathy, Not Slide Decks 

When it came time to get executive support, Edwin didn’t walk in with a pitch. Instead, he asked the CEO and others to actually try learning chess through competitor apps. The result? Instant empathy. They saw firsthand how most chess learning tools dumped users into high-context strategy without helping them understand the basics.

This strategy flipped the conversation. It wasn’t about why Duolingo should teach chess—it was about how much better they could do it.

💡 Final Thought

“Chess has had product-market fit for 2,000 years,” Edwin joked. But what Duolingo did was something new: making it feel delightful, intuitive, and wildly accessible. The real lesson? With the right mix of AI tools, creative hustle, and human-centered design, you can break into any domain—and win.

What was your favorite insight from Edwin’s session? Reply and share what you’re excited to try in your own AI prototyping journey!