Repository Memory Makes AI More Useful — and More Dangerous

A repo-aware agent produces changes that look team-native. The problem is that “looks team-native” is not the same thing as “is correct.