- say we calculate some kind of hash either based on the learning item or based on datetime and then based on that choose color, frame, font and so on based on that
- if we make it consistent per item, do we strengthen initial ◊ memory but later pay when transfer fails (even just transfer to differently colored cards)?
- if we randomly vary the looks all the time, does that fight cue interference of the from "if I see a white card with black Times New Roman it's 'me gusta'" that you would otherwise have with every single card?