A conversation with a 747 pilot sparked reflections on how AI coding agents are changing software development. The pilot, who had flown the iconic aircraft for two decades, noted that unlike his static role, programming once offered continuous learning and improvement. Now, with coding agents handling increasingly complex tasks, developers face a similar plateau.

Large language models have evolved from helpful assistants to autonomous feature implementers. Where developers once learned through debugging and research, they now often hand off entire problems to AI agents. The author, who works at an AI lab, acknowledges the productivity gains but worries about the loss of skill development. Reading code produced by AI differs fundamentally from writing it yourself.

This shift mirrors the pilot's experience - performing the same tasks without growth. The solution isn't rejecting AI tools but maintaining deliberate learning practices. Writing code manually for educational purposes, attempting solutions before consulting AI, and understanding your domain remain essential. As coding agents become more capable, developers must consciously preserve the learning that once came naturally through the development process.