Glyph Lefkowitz: What Is Code Review For?

Humans are not good at finding bugs in code due to limitations like fatigue and inattentional blindness. Automated tools like tests, formatters, and linters should be used to catch errors, freeing humans from repetitive checks. Code review's primary purpose is not bug detection, as the author emphasizes. Code review is instead a process failures and a tool for team acculturation, sharing knowledge and best practices. Code review also fosters a culture of learning and avoids stagnation by incorporating new perspectives. LLMs, unlike humans, should be treated adversarially when it comes to code review. LLMs cannot learn from feedback in the same way, thus requiring stringent automated verification tools. Code review for LLM-generated code necessitates robust automated testing to ensure quality. The author concludes by reiterating that code review's focus should be social, building knowledge and providing encouragement. The author encourages leveraging code review to promote growth and team alignment.