🔗 What are code reviews even for?

AI didn’t break code review; it highlighted existing issues. Brian Houck discusses how the surge of AI in code production has outpaced human review. At Meta, lines of code per diff increased by 106%, with diffs up 51% due to AI, while the timeliness of reviews has declined. Developers idealize spending only 7% of their time reviewing code, indicating we must rethink the purpose of code review itself. It's traditionally about more than just defect detection; it's a space for knowledge transfer and architectural understanding. Rethinking habits around code reviews before AI can be effective is crucial. Using AI wisely can enhance the process, but organizations risk eroded collective understanding by automating too much. Effective code review is about preserving knowledge and fostering collaboration, not just speeding up the process.