“Anthropic's Claude Code team has shifted its code review process from line-by-line human review to using another instance of Claude for review, with humans performing higher-level acceptance testing.”