Peter Naur said that understanding programs is key to maintaining them. In today's AI-driven world, can coding agents really grasp the theory behind our programs? What happens if we don’t create a solid understanding of the programs we’re working on, because we rely too much on these coding agents?