What’s wrong with receiving code/security/MR review comments from AI?