Nobody knows how the whole system works

Modern software systems have grown so complex that no single person can understand them completely, from hardware interrupts to application logic. This reality predates AI but is accelerating with LLM-assisted development. While frameworks that obscure underlying mechanisms ("magic") have always been controversial, the benefits of abstraction have historically outweighed the risks. The fundamental challenge isn't new—engineers have always specialized in specific layers while relying on abstractions for the rest—but AI is shifting where those boundaries lie and who needs to understand what.