I have the habit of sending quickly message and only noticed the tone a minute later when it was already read. already proved that TypeSafe AI's Jev model is fast enough to sit in a save-triggered loop for code review, so the next question was: could the same model catch a bad message before it goe…