We initially aimed to teach math but shifted to teaching languages, particularly English, because we realized it would have a broader impact and reach more people.

We realized that instead of creating standalone apps for new subjects like math, we should integrate them into the main Duolingo app to enhance user experience.

The way we educate mathematics often fails to accommodate different learning styles, leading many to abandon the subject before finding their unique approach.

With AI, we can generate high-quality educational content quickly, allowing us to expand into new subjects like chess, math, and music, enhancing our learning offerings. We started working on the chess course eight months ago, initiated by two people who had no coding or chess experience, and they managed to prototype it successfully before we added engineers to productionize it.

In the early days of Duolingo, we realized that keeping users motivated was the biggest challenge in language learning, which led us to incorporate game mechanics into our lessons.

Students struggling with math should seek enrichment opportunities outside the classroom, such as math competitions and online resources, to foster their interest and skills.

We initially underestimated the complexity of teaching languages, but through A/B testing and hiring experts in second language acquisition, we significantly improved our teaching methods.

Our strategy involves expanding our English learning product to include various formats, such as one-on-one and one-to-four teaching models, catering to a wider age range of students from young children to teenagers.

The challenge is to make our models good enough and easy to use so that even non-expert musicians can find them genuinely musically engaging, focusing on generating interesting sounds rather than just melodies.