Really awesome article. It’s exciting to hear that Quizlet is starting to move in the direction of long-term learning for users.
I had a quick question about the algorithm you guys developed:
It sounds like it was originally based off of the SM-2 algorithm and increasing delays by a little over twice the interval sounds like e-factors.
What challenges did you guys run into when implementing the algorithm without allowing the user to set scores (I remember supermemo uses 6, and Anki uses 4 different scores to set)?
And what prompted the shift from short to long-term studying solutions?