But I'm not too crazy about spaced repetition. I used it to learn all the capitals in the world, and after about a month I managed to do all of them without error. But after two years of barely using the knowledge, I only remember half.
Of course, I could keep reviewing it every few months to keep it fresh, but that's just not an efficient system, when you think of how many things I'm supposed to (and do) remember.
For me, at least, the best way to remember things is by tying them in as many associations and metaphors as I can, and that gives me a pretty reliable recall. It has some downsides: It's a bit more work than spaced repetition (requires creativity, for example), and it's not that good for unconnected data (but then, are random facts that useful anyway?). But I think that as a long-term method, it's much more solid.
> I don't see how cards would help me with mathematical ideas. You have to get them on a fundamental level, and then they're hard to forget.
I find it too easy to trick myself into believing I understand a concept on a fundamental level. But often that "understanding" slips away, and six months later when faced with an example problem out of context I struggle to solve it.
That's basically the spaced repetition model. You need to forget a little bit, so that working memory isn't saving you, and the recall process takes effort. That's what builds long term memory. Spaced repetition systems just extend that process beyond the length of the book, in a time shorter than rereading a book.
> Quite frankly the idea of knowing /by heart/ the name of the greek goddess blah blah blah he uses as an example would bore me solid.
I agree, the names of greek gods is a bad example, unless you're a student of mythology. Spaced repetition is hard work, so probably only deploy it on things that matter. A better example might be Bayes Theorem. Incredibly powerful, but unintuitive and easily forgotten. Cards for Bayes' theorem might include the purpose: "Bayes' Theorem calculates how to adjust our prior beliefs given new evidence", as well as cards about the formula, or an intuitive visualization.
Or, maybe you want to learn more about the linux internals and spend a portion of your time memorizing the meanings of signals and errnos as a small part of a larger program of study. Sure, you can look these up in a man page, but part of the value in knowledge is knowing things exist. How many developers do you think know of SIGUSR2? Or... SIGBUS ;)