After the algorithm-design work, this is where I’ve leaned on Anki the hardest, and where it most obviously pays off. A language is almost pure memory work at the start: thousands of small, arbitrary mappings (this squiggle means that sound, this sound means that thing) that no amount of “understanding” can shortcut. You simply have to have seen them enough times. That is exactly the shape of problem spaced repetition was built for. I think about the work as happening along four axes, reading, listening, speaking, and writing, and Anki touches the first two heavily.
I’m learning Japanese, and the writing system came first. I worked through James Heisig’s Remembering the Kanji, which teaches the ~2200 standard kanji by hanging an imaginative story on each character’s components. Every kanji became one Anki note: a keyword on one side (Autumn, Interval, Renunciation), the character to be recalled and written on the other. I keep them as a flat, alphabetised index so I can see the whole set at once, and so other notes can point back into it.
Then comprehension. I use a beginner podcast, Nihongo con Teppei, as raw listening material, and I have a small pipeline around it. I take an episode, produce an interleaved transcript (the Japanese sentence, then romaji, then English), and explode it into Anki, one card per sentence. The front is the spoken Japanese sentence; the back is the English, a short audio clip of just that line (cut out of the episode so I can replay the exact sentence), and a link to the kanji note for every character that appears. So a single listening card reaches back into the kanji deck: the network of notes references itself. I do the same trick with whole films, clipping the video of each line onto the back of its card, so the sentence is reviewed with the scene it came from.
A nice accident: one of the episodes I ankified, number 1498, is Teppei explaining osarai / fukushū, the practice of reviewing the same vocabulary several times a day so you don’t forget it. The content of the card and the reason I made the card were the same thing. Spaced repetition was teaching me spaced repetition, in Japanese.
What works is recognition: reading a kanji, parsing a heard sentence. What Anki cannot hand me is production, actually speaking and holding a conversation. You can drill the inputs endlessly but the outputs need a mouth and another person. This matches the through-line of the whole project: Anki plants and waters, it does not make you fluent on its own.