> I guess difficult part would be to detect the fact which is worth to be added as a card.

The time to memorize something is supposed to outweight the time you'd need to look it up. Say the target is five minutes - in that case, it's important to follow the Pareto rule as well, and to discard any cards that take you more than that to memorize in total, including the time you recall them. In practice, it means discarding over 20% of cards if possible, to learn 80% that are easy to remember.