The author argues that reviewing AI-generated code doesn't equate to understanding it. They emphasize that while it's common advice to treat code reviews like mentoring junior developers, it's misleading, as reviewing doesn't lead to the same knowledge as writing code, which can lead to misunderstandings when bugs need to be reproduced.