Every Duolingo lesson is shaped by an AI system called Birdbrain, which constantly estimates how well a learner knows each skill and selects the 14 or so exercises most useful for that specific person at that moment. It works like a coach who watches you practice and adjusts every drill based on what you're struggling with — updating within minutes of each completed exercise.

Launched as V1 in March 2020 using a logistic regression model, Birdbrain estimates both learner proficiency and exercise difficulty simultaneously. Birdbrain V2 (May 2022) upgraded the underlying approach to a recurrent neural network — specifically a Long Short-Term Memory (LSTM) model — that compresses a learner's full interaction history into a 40-dimensional vector, updated in near real-time. The system can distinguish granular knowledge gaps, such as a learner who is strong in past tense but weak in future tense. At Duolingo's scale of 50+ million daily active users, the Session Generator selects roughly 14 optimal exercises from approximately 200 candidates per lesson, processing over one billion exercises daily.