Understanding the impact of the COVID-19 pandemic on education requires a solid grasp of the impact of student autonomy on learning. In this paper, we estimate the effect of an increased autonomy policy for higher-performing students on short- and longer-term school outcomes. We exploit an institutional setting with high demand for autonomy in randomly formed classrooms. Identification comes from a natural experiment that allowed higher-achieving students to miss 30 percent more classes without penalty. Using a difference-in-difference-in-differences approach, we find that allowing higher-achieving students to skip class more often improves their performance in high-stakes subjects and increases their university admission outcomes. Higher-achieving students in more academically diverse classrooms exerted more autonomy when allowed to.