Jev’s context should have signals/features to operate on. Same way LLMs can use CV & code to analyze an image.