Reviews have become expensive, rewrites have become cheap (ishmeetbindra.com)

Recent observations in software development reveal a significant shift in the economics of handling AI-generated code, particularly with large language models (LLMs). While LLMs tend to produce technically correct code, they often do so in an overly complex manner. This has made code reviews more time-consuming, as developers are left to assess and manage the intricacies introduced by the AI, which defaults to building rather than leveraging existing libraries. Consequently, the reviewing process has become more expensive due to the repeated discussions around unnecessary complexity. In contrast, rewriting code has become a more efficient solution. Developers can now quickly request AI assistance to simplify over-engineered code or remove superfluous features. This shift enables a more agile approach to refining code, allowing for expedited iterations that were once considered laborious. As a result, project management is evolving; developers are investing more time upfront on planning and scope definition to prevent complexity early on, leading to a more streamlined implementation process. This transformation not only alters workload dynamics but also empowers developers to be more assertive in refining their code, as the cost of iteration is significantly reduced.