
A Good LLM Doesn’t Make a Good Teacher
When people learn that our speaking apps run on advanced AI, they ask which model we use. It is the least interesting part of the story. Choosing a capable model is the easy part; any developer can connect to one. What decides whether a learner actually improves is whether the system understands how people learn to speak a second language. That knowledge does not come from the model. It comes from us, from years in bilingual classrooms, instructional design, and second language acquisition research. This post is about how we build AI tutors as educators, not just engineers.




