When Code Gets Cheap, Understanding Gets Expensive

Markus Poppastring, reflecting on his experience at medical transcription startup Heartland Information Services, warns that the current drop in code production cost due to AI mirrors the offshore outsourcing wave of the early 2000s. Back then, the expensive part wasn't writing code—it was understanding it well enough to change it safely, debug under pressure, and explain decisions to the next developer. The code produced offshore was often good, but knowledge lived in one time zone and responsibility in another.

With AI-generated code, the problem is worse: "the knowledge may not exist anywhere. There is no human on the other end who once held the full picture. The code has been committed, syntactically correct but devoid of intent." This echoes the thesis of Prediction Machines: when a fundamental input gets cheap, value shifts to its complements. In software, the complement of production is understanding.

The author argues that the scarce resource isn't producing code—it's reading it, navigating it, knowing which parts matter. He quotes Joel Spolsky's 25-year-old observation: "it's harder to read code than to write it." The solution, learned from outsourcing, is to invest deliberately in shared context, documentation, code review, and treating comprehension as a first-class engineering concern. Developer tools should focus on helping us understand existing code, not just write new code faster.