Spaced repetition is a way of memorizing stuff effectively, typically via flash cards. It's essentially the opposite of cramming, but with more structure than simply "study all semester rather than just the night before the test". There's a fair bit of research supporting using spaced repetition, but seems to to be no solid consensus on any exact formulation.
Suppose I want to learn the US state capitals, I could start with fifty flash cards all mixed together, but that's a little overwhelming
I could instead start with five, and once they're memorized moved on to the next. But by the time I reach the end, I might have forgotten the first five.
Alternatively, I could mix each subsequent five into the ever-growing deck, so I'm still revisiting old ones while gradually adding in new.
If you want to be really advanced, though, you can revisit each card based on how well you know it. Got it wrong? Revisit in five cards. Knew it right away? Revisit it after reviewing all the others. Got it, but it was a half-guess? Somewhere in between.
It's worth pointing out that under any of these systems, I can remove the card for Austin, Texas, because I live there, and should leave in Kentucky because Frankfurt? Isn't that in Germany?! But somewhere in the middle there are edge cases. Spaced repetition is about sorting the flash cards from known to unknown, including properly sorting the "maybe known, but iffy" cases in the middle.
There are various software programs for implementing this, Anki being the most popular, but people have developed systems for doing it with physical cards as well. If you want physical cards, I'd recommend starting with the Leitner system: you have some number of buckets of cards (probably 3-5), arranged from "review this most frequently" down to "review this least frequently". If you get a card right, you move it to the less frequent bucket, and if it's wrong, to the more frequent. I'll review bucket 1 every session, bucket 2 every other session, bucket 3 every third sessions, and so on. (At some point, often bucket 5 or 6, you consider a card fully memorized and don't review those at all anymore.)
You'll note I haven't specified a starting bucket. I'm not actually sure if there's an optimal starting state. I could see arguments for starting with them all in the "least well known" bucket or the middle bucket.
Personally, I'd suggest (and I think most programs do this) starting with a subset in the "least well known" bucket, and add in new ones as that bucket gets emptier. You'll probably need to tweak if based on what you're learning and how quickly you pick things up.
If you want to read a lot about spaced repetition by someone who spends time reading research and rigorously experimenting with self-improvement, I'd recommend Gwern's spaced repetition page.