What's on the Box vs. What's in the Box (rjcorwin.github.io)

Boris Cherny, the original creator of Claude Code, recently articulated two contrasting perspectives on the role of AI in generating code. The first view asserts that AI can accelerate coding while allowing the output to remain reviewable, with a submitter qualified to explain and maintain it. The second perspective suggests treating AI-generated code as a “black box”—focusing solely on the output without understanding the underlying code. Cherny advocates for a balance, emphasizing that prototype code may function as black boxes, but production code necessitates thorough scrutiny, especially when dealing with sensitive areas like financial transactions. This discussion is significant for the AI/ML community as it addresses the critical balance between leveraging AI's efficiencies and maintaining code quality and safety. Cherny raises an important analogy: while the plan for the code represents “what's on the box,” the actual code represents “what's in the box.” Understanding the code is vital to ensuring that it behaves as expected and meets both operational standards and user needs. By prioritizing code comprehension alongside efficiency, engineers can better evaluate whether the code works, can be owned, and is truly desired, ultimately enhancing the reliability and trustworthiness of AI-generated solutions.