GitHub has enhanced its Copilot CLI with a new multi-model reflection loop feature aimed at improving code reviews. Released on April 11, this tool helps developers detect potential defects early, preventing issue accumulation. The GitHub Research team introduced an agent named "Rubber Duck," which automatically collects code review feedback from various AI models. GitHub reports that this feature is proving effective, delivering significant improvements in code quality.