This paper develops a model of education production and uses it to study optimal school system and curriculum design. Curriculum design is modeled as a time-allocation problem. A school teaches students many skills and allocates time to different skills based on student characteristics. Our framework provides a novel interpretation of studies that find zero achievement effects at selective school admission cutoffs. We show that such findings may be consistent with highly effective schools implementing optimal curricula, rather than necessarily indicating ineffective schools. The interpretation depends on the alignment between measured outcome skills and skills emphasized in the curriculum. We test several model predictions using data from a prominent exam school and find supporting evidence that would be difficult to rationalize if selective schools were ineffective.