Conversation

This feels like such an obvious build on SRS, I feel like it must either exist or have a flaw I’m not seeing. One weakness is that the association graph will scale as n^2 in number of cards, so random learning of associations will be slow unless you prime the deck’s associations
Quote Tweet
I’d like a spaced repetition system that supports Hebbian learning synapses between cards, where you can thumbs-up any pair of cards that accidentally show up in review in sequence as an interesting juxtaposition so they show up together more often in the future. Model as an MDP.