"Human's need to review the code" vs "Fable does the planning and 20 agents write the code". Both of these things cannot be true?

The user is seeking to understand the apparent contradiction between the necessity for human programmers to review AI-generated code and claims of 100x productivity boosts where AI agents write code autonomously. They note that even AI models like Opus emphasize the importance of safe code review, yet there are also assertions about developers leveraging tools like Claude and 20 agents to write code even while they are sleeping, suggesting minimal human oversight.