Rethinking Homework in the Age of AI

Traditional homework has long served as the cornerstone of mathematics education, but it is increasingly showing its cracks. The feedback loop is too slow. Students may wait a week or more to discover fundamental mistakes. The rise of AI has blurred the line between authentic effort and outsourced solutions. In this talk, I present an experimental AI teaching assistant, developed over the summer, that reimagines homework not as an assessment tool but as a space for learning. The assistant provides immediate, guided feedback, transforming assignments into opportunities for active engagement rather than delayed evaluation. To align incentives with this new model, homework is graded by completion, rewarding time invested in learning under AI guidance rather than penalizing errors. I will share both the motivation for this shift and my early experiences with its classroom impact, and invite discussion on how such tools might reshape our pedagogical practices.