When AI Writes the Code: What Should Programmers Learn?

Generative AI challenges programming education because systems can now generate code, tests, documentation, and substantial parts of applications from natural-language instructions. This commentary argues that the shift does not make programming less important; it exposes curricula that equate programming competence with manual code production. It distinguishes coding from broader programming and software engineering, and places natural-language vibe coding within a longer history of abstraction from machine code to assembly and high-level languages. It further argues that conventional human-readable source code may become less central as AI systems increasingly translate human intent into machine-oriented representations. The educational response should not be to reduce technical depth. Programming education should combine human-centered problem definition, requirements, design, and evaluation with durable knowledge of algorithms, systems, security, complexity, and software failure. When machines can write the code, the value of programming education lies increasingly in deciding what the code should do and judging whether it does it well.