For language learning, I wish there was an audio-first flashcard app that changes up the example sentences every time. Right now, I'm using Anki to learn Japanese vocabulary from N5 to N3[1]. I know the words and the example sentences well enough to read N3-level text, provided I know the grammar. But when it comes to listening, I struggle to understand even N4-level spoken Japanese. Anki just doesn't offer enough variety for me to truly internalize what the sound means in different contexts. Plus, seeing the text before hearing the audio tricks my brain. I think I'm learning the sound, but it's an illusion because I already know the meaning from seeing the word first.
[1] I feel like Anki offers diminishing returns once you get past N3. Advanced words usually have subtle nuances that you can only really pick up through rich context, like in a full paragraph or a TV scene. Native-speaking kids can understand complex words in context because they have a deep grasp of a smaller, simpler vocabulary. That’s why I’m focusing on mastering high-frequency, simple words first to build a learning flywheel. I'm hoping this will eventually let me pick up new words naturally through reading and listening, just like a native kid does.
I use Anki to learn French, Chess openings/tactics/techniques, to unscramble letters for scrabble, for Pub Trivia... The options are kind of limitless.
As a mid-30s guy who has well passed the neuroplasticity of his teen years, it's a godsend for me.
To echo the author's thoughts though, I can't prove empirically that I learn more effectively using Anki (or spaced repetition) than other methods... Only anecdotally. I have a shockingly poor memory, but now I'm B2 certified in French and an ~1800 Elo on chess.com .
I'm glad the article acknowledges that flashcards are just one small part of learning. When I first got into spaced repetition, first with the Mnemosyne Project and later switching to Anki, I discovered that the efficiency I imagined was partially illusory.
* Memorizing things often takes much more time than learning things naturally as you use them because it takes extra time out of your day.
* I often lacked the associations that would normally help reinforce a concept or fact because I used brute force a-single-super-simple-concept-at-a-time memorization instead of more natural methods where context helped me gain a better understanding.
* Breaking things down into very narrow, simple, one concept cards is more difficult than I imagined.
* Creating mnemonics is really helpful but can be time consuming, and you don't know which cards where you will need them until you repeatedly forget those cards. Someone on HackerNews about a year or two ago recommended using AI and that did help a little, but it didn't take long before I realized that the AI created mnemonics feel so similar and less connected than time tested mnemonics that I find them less effective.
* Since brute force memory is slower I often learn slower, which is less efficient than learning a groups of related things together at a pace where the concepts together give you a better understanding than learning one at a time. (Sometimes you need to slow down because your not getting the concept, but going too slow is less efficient also.)
I still use spaced repetition but I realized it's not the amazing revolution that I first imagined that it would be.
Practicing your retrieval is actually one of the best ways to retain knowledge of something. Flashcard programs like Anki are really great because it identifies where you need more work and drills you on your weak points -- it feels awkward working constantly on your weak points, but you get quantifiably better results with the flashcard method it uses.
Some people criticize flashcards as optimizing for rote memorization and deemphasizing understanding, but you'll never achieve understanding or mastery in general without a solid platform of knowledge to work from.
My problem with Anki is that it's very, very inefficient. It will take much more time and effort to memorize the vocabulary words you learn, and losing those words is very quick. It's much better to use SRS with actual and varied sentences.
I think deliberate practice is what's really core to improving any skill, including memory.
Spaced repetition is an effective way to review things but its biggest benefit is a process that's easy to be consistent with.
Somebody else can have equal or better performance with other technique but just like dieting, it doesnt matter as much what method you use as long as you stick with it.
I feel like the actual core mechanic at improving is the actual act of "recall", it doesnt matter what you do if its a form of recall it is effective and very awkward in practice because you just sit there waiting for your brain to do a mysterious thing
When I was doing rote memorization and flashcards frequently (some years ago now) I observed that remembering things became a lot easier for me.
I also find my verbal fluency is directly affected by how much pure social time I have in my schedule. It makes me think its one of those 'use it or lose it' things and that I need to schedule more time with people.
Really curious exactly how you learn things like chess with flash cards. French makes sense as I would guess you just have a word or phrase in both languages.
... That I got by scraping the Lichess database, favoring common patterns that appear within +- 600 of my current Elo.
From there, I have Claude build me a script to convert each of those positions into a .png, then create me a deck with all the cards, et voilà. The front of the card is the position, the back is the best move in that position with a small explanation.
Every ~2-3 months when I see that most of the cards have matured (according the the Anki spaced repetition scheduler, I build a new deck around my new Elo.
I also play a lot. Prior to ~1000 rating I got away with spending 90% of my time in Anki and 10% playing online games, but lately it's been pretty 50/50. In higher ratings, playing real games tends to translate into wins more effectively for me.
For studying openings, it's almost the same thing, but the back of each card is the book move for my opening + the name of the opening the opponent chose.
> Some people also use LLMs to generate flashcards. And of course, the result will be those impersonal, mediocre cards.
> I won’t say LLMs are useless for this. But from my trials, I get about 1 card that’s useful to me out of 10, and even that 1 card still needs rewriting.
I don't know the specifics of how the author tried to do so, but from what I've seen the majority of attempts are, let me drop a chapter of a textbook and say "make flashcards." If that is what we are talking about, then yes, LLMs are useless.
In my mind, though, this is sort of like looking at the very first GitHub Copilot LLM autocomplete from a couple of years ago and concluding, yeah it's nice for one-liners, but it cannot write an app.
If you create a framework around your card-creation AI so that it can use tools, and verify its work to ensure common card-creation pitfalls don't happen, you can get pretty high-quality cards. In my experience, you go from a 10~20% hit rate to a ~90% hit rate, which in my mind is good enough. I got to ~75% quality just from a two extra LLM calls that would assess a potential card against a standard set of rules (adapted from [0]). There are huge Pareto gains to be had here.
I've generated thousdands of cards over the last few months this way. I let the AI add it directly to Anki via AnkiConnect. Then, if when I go to review I find a card that my AI created and I don't like it, I just delete it.
Removing the limitation of card creation is really quite compelling, and I think the area is still highly under-invested in. Would be cool to see a generic framework evolve that one could use. For now, I've been using a personal fork of clanki [1].
I once took a psychology course with my girlfriend at the time. She and her pals would be up all night studying with flashcards. I'd just walk into class, learn about it, never study, and get A's on the tests.
When they asked me how I was doing it I explained that I'm there to learn understand the topic and I don't give a damn about the test. So I just let my curiosity lead the way, it causes me to ask questions in and out of class, email the professor about them, do my own research and experiments. Despite not letting the test be my guide, this prepared me for the test anyhow.
I'm glad they work for some people, but flashcards to me seem like they provide a shallow kind of understanding. I don't want to remember the equation, I want to be able to derive it in a pinch, and flash cards don't give me that.
I think that really depends on how you use the cards. I know medicine students who clearly are passionate about their subject, but just need to get the reminders on individual terms because there’s such a wide range of topics they need to be prepared for, and lots of very similar words.
I also know that in math courses, there’s certain formulas with similar names I simply need to remember because deriving them on a test would be too time consuming. Or I could imagine in chemistry there’s various special cases that you’d want to be able to recite on a test.
I’ve never tried it for language learning but people talk highly of it, so there’s definitely some truth in their utility there.
I think the trick is flash cards are fantastic supplementally with other studying strategies.
It depends on the subject. For classes/subjects that are very terminology-heavy, with a lot of dependencies between topics, memorizing things is a pre-requisite to the "deeper" understanding, especially under time-constraint of a semester or whatever. I think most people would agree that it's ideal to "naturally" explore a subject in order to get a deeper understanding of it, but one can't always be expected to 1. Be able to deploy that "curiosity" for all subjects at any time and 2. Be able to adequately achieve competence in a subject within the given timeframe
What about classes that weren't interesting to you, but you had to pass anyway? Personally I was the same as you for classes that I found interesting, but for other required classes I went for the shallow understanding, memorize for the test approach.
The best advice I've ever read about flash cards is if you are dreading to review, because you are forgetting your cards or they are too complicated, you are writing cards wrong. Learning is supposed to be fun! Also, Common Core should ship Anki decks. I seriously think so many problems with education stem from students not realizing that memorizing is actually very easy with FSRS, and thus struggling and hating learning.
Voluntary retrieval (the memory "mode" of flashcards), even for simple pairs, is perceived as highly effortful by almost everyone, on a similar tier to doing mental math.
It's incredibly efficient for learning, and achieving your goals quickly can be very rewarding, but I don't think "learning flashcards should be fun!" is a reasonable expectation for the vast majority of people.
Most people don't use flash cards at all. So if their experience with them is terrible then why would they keep doing it? Also effort can be fun. Working out is effortful and most people don't do it. So, they should go to the gym and try to have a good time. If they aren't, maybe the advice to do some 5x5 program or Jocko whatever workout is not good for them. It's just a heuristic to stay motivated. But yeah, my friend just finished med school and memorized like 30k extremely complicated cards. He did not "have fun".
It's a balance. Maybe a helpful analogy would be a book -- yes, reading a book is effortful, and yes, "almost everyone" does not read. Still, I think most wouldn't consider it crazy to say that reading can be "fun."
My sense is the same as the authors that LLMs + an Anki-like -> mediocrity. But I feel like there ought to be such potential there! Even just things as simple as rewording the same question a bunch of equivalent ways to avoid recognizing the sentence structure...
Downside for AI potential as a whole, framed broadly: We don't seem to be good enough yet at identifying what friction is functional and what we should strive to automate/eliminate.
There's an arena/research report by Ozzie Kirkby and Andy Matuschak at https://memory-machines.com/ which you might find interesting.
I predict that things will be 'good' within the next 5 years, but your intuition is correct (and the SOTA models are often producing worse prompts than older models).
I do a really lightweight version of flash cards. Everytime I'm learning a new tool or tech, I grab oversized notecards (my favorite are 8x5" dot-grid cards). I put a label at the top, and create bullet points of each item i want to remember. I then review. No individual cards for each item or anything. Just all the things grouped on one card as bullet points.
For example, I'll have a `sqlite` card, and put all the commands and everything on it, as I learn them. I'll use it as a cheatsheet, but then also a few minutes of mindful review. This for the toolings that I want to know well enough to not get slowed down googling the commands. I do this for a lot of CLI tools, but also things I need to remember about the business of my company and working across group, etc....
Eventually the five or six working cards I have, get put on a pile and new ones come in.....
> How do I actually use flashcards? My software of choice is Anki. I am not completely satisfied with it. The UI looks dated, the WYSIWYG HTML editor is clunky, and the undocumented file format makes potential porting and interoperability tricky.
And we still love it. I'm on the same page. This phenomena feels oddly satisfying.
> From this perspective, fields that require deep understanding, like math, require memory just as fields with a breadth of shallow knowledge do, though in different ways.
I'm interested in understanding how others use Anki for conceptual subjects like pure math or physics. I believe many fundamental rules in Spaced Repetition (e.g. like keeping cards concise) are thrown out the window for conceptual subjects.
I took my first real analysis course last semester, and I made flashcards with pen and paper for every single non trivial definition, theorem, lemma, and corollary that we covered in lecture.
Analysis definitions and theorems get really complicated with intricate and difficult to follow logical chains, and there are a lot to remember.
These definitions and results don’t mean much on their own without exploring their neighbourhoods by proving relevant things, and I could have learned these definitions and results by just doing proofs. But being absolutely sure I could recite every theorem and definition definitely helped me on the final exam.
I think if you’re learning algorithms (like find the area under a curve) in a calculus course for example, flashcards might have more limited value, as in that case problems are relatively short and you’re better off just running through your set of algorithms a ton of times by doing problems.
I also took a group theory course last semester and I memorized every definition and result from lecture via flashcard, but didn’t practice using them enough by writing proofs. I ended up with like 2 or 3 out of 10 complete proofs and the rest half finished on the final exam because I had the right starting points, but not enough practice using what I knew in unexpected ways. Still passed somehow.
Yeah most of the advance assumes you have the data ready at hand and just need to phrase the cards right, get the number of words right. Whereas for conceptual domains the biggest problem is: how do I encode this as question-answer pairs at all? What I want to read more of is people sitting down and writing in the first-person perspective how they go about it, like Michael Nielsen does here: https://cognitivemedium.com/srs-mathematics
hey fernando, I read your article a lot and it's helped me a lot in my own spaced repetition so thanks from me!
a note on your request, have you seen this video before? Andy has some custom PDF reader he built with flashcards built-in, and it's two hours of tacit flashcard creation centered around quantum mechanics: https://www.youtube.com/watch?v=OFuu4pesKf0
Perhaps "Using spaced repetition systems to see through a piece of mathematics
" [1] might be of interest for you. I have read author's "Augmenting Long-term Memory" [2] and have incorporated a lot of his advice into my Anki practice.
For me, it's quick access recipes (breakfast pancakes for kids), what was the name of the glacier that we hiked to last year, behavioral prompts etc.
Same caveat as in the article: Spaced repetition is just one (minor) part of learning math/physics. It alone won't get you anywhere.
For math - particularly higher level math, the most obvious use case is definitions. There are so many!
You can put theorems in there, but it is a bit challenging on how to phrase it. A single theorem could result in several smaller flash cards.
I think what works better is taking a theorem, finding a representative problem that is solved via that theorem, and make the problem statement the question. The downside of this approach is each card takes longer to process as this is not just plain recall, but actively solving a problem. For this reason, I keep such cards in a separate deck and review them only when I have time I can dedicate (e.g. spending well over a minute per card).
Depends on how you use the flashcards. You can use them to memorize definitions and equalities, and you can also use them as quiz questions which excercise your reason and not simply your memory. For example, you make a flashcard for each excercise question in your textbook. Once you identify what you're struggling with, make more flashcards of that same problem type to avoid remembering the solutions. This will take you from a shaky understanding to much firmer ground pretty quickly.
Honestly just making the flashcards and elaborating on/modifying problems you're struggling with will take you a very long way.
I actually tend to keep my cards super concise. I treat Anki as a way to practice fundamentals, like memorizing certain formulas. Anytime I try to add conceptual stuff to cards I feel like I'm only memorizing one specialized version of the thing and it doesn't feel super useful.
> IMO you want to be actively trying to map the new concepts to things you already understand, and constantly working to update your mental model.
It's not an either-or.
Where SRS comes in handy is when you have to take long breaks between your study sessions (due to job + family). Have you ever tried learning an advanced math topic where you get to work on it for a few days, then may have to stop for a few weeks (or even months), then resume, and repeat over and over?
Chances are, no matter how intense you study during those few days, you'll likely forget important definitions/theorems in the periods you don't.
SRS takes care of those gaps.
Case in point - many years ago I put a lot of my intro to statistics course in flashcards and actively reviewed them. I hadn't done actual statistics for over a year, and then made a (false) claim here on HN. Someone gave me a counterexample using the chi-squared distribution. And it was amazing that I could recall the basic properties of the chi-squared distribution, and enough other theorems to verify what he said without consulting any book.
I've never used the chi-squared distribution for anything before or after.
(Sadly, I stopped using those cards years ago so I've forgotten the material!)
I think of it like drills in a sport. If your practice is 100% drills, you'll be pretty bad. But drills give you an awesome foundation to do the really complex stuff intuitively.
> prefer your own flashcards to other people’s flashcards, at least for fields that require deep understanding
For me, much of the value of flashcards comes in the making of them. Part of it is thinking about what each flashcard will say, and part of it is the action of writing it down in handwriting.
I tried to make an auto flashcard generator but ran into the issue that one word can map to many senses. But most word frequency datasets don't disambiguate the sense. So if you want to include all the senses for a word while ranking words by frequency they all get the same starting position.
This is a big part of why language learners have largely moved toward sentence mining as the preferred way to build an Anki deck.
Getting your words from real-world contexts, and keeping that context on the front of the card, largely eliminates the ambiguity problem. If a word has multiple senses, it gets multiple cards with different example sentences to illustrate each one.
It also helps a bunch with words that don’t really have a concise translation to your native language. For example the French words “mur” and “paroi” both mean “wall” in English, but the contexts where you use them are quite different. An example sentence helps with that, and getting that sentence from an even richer context such as a book or article you’ve read helps even more.
It’s also, frankly, just more enjoyable. I’ve come to view frequency lists as an antiquated tool. I needed them in the 1990s when good authentic-context study materials were hard to come by, but the modern Internet has made so-called immersion-based learning methods so easy and inexpensive I’m frankly mystified that people still cling to the joyless, almost mechanistic methods we were stuck with in the previous century.
Thank you, its good to hear some of what the state of the art is. My natural language processing studies at university are around the vintage you mention. I will have a go at this...
Yeah, NLP is a different beast from human language learning.
The most salient difference here is that NLP wants to automate as much as possible for reasons that are specific to NLP.
But for human language learning a lot of automation is actually harmful because manual effort tends to be good for Ebbinghaus’s arguably more important but less popularly appreciated discovery: memory encoding quality.
The fun is in making the cards truly yours, by writing them yourself based on your experience. After experimenting with generated cards, I throw them away. They were semantically correct, but not relatable/memorable.
Used to use Anki for foreign language learning. Guessing it would have been useful to memorise calc, chem and physics equations if it had existed when I was young.
Flashcards are brilliant. Anki is finally usable after they ditched the hot garbage algorithm they were using. Previously I've used the Leitner method and I stil think that's the best one for me.
[1] I feel like Anki offers diminishing returns once you get past N3. Advanced words usually have subtle nuances that you can only really pick up through rich context, like in a full paragraph or a TV scene. Native-speaking kids can understand complex words in context because they have a deep grasp of a smaller, simpler vocabulary. That’s why I’m focusing on mastering high-frequency, simple words first to build a learning flywheel. I'm hoping this will eventually let me pick up new words naturally through reading and listening, just like a native kid does.
As a mid-30s guy who has well passed the neuroplasticity of his teen years, it's a godsend for me.
To echo the author's thoughts though, I can't prove empirically that I learn more effectively using Anki (or spaced repetition) than other methods... Only anecdotally. I have a shockingly poor memory, but now I'm B2 certified in French and an ~1800 Elo on chess.com .
Do I still forget things all the time? Yes.
* Memorizing things often takes much more time than learning things naturally as you use them because it takes extra time out of your day. * I often lacked the associations that would normally help reinforce a concept or fact because I used brute force a-single-super-simple-concept-at-a-time memorization instead of more natural methods where context helped me gain a better understanding. * Breaking things down into very narrow, simple, one concept cards is more difficult than I imagined. * Creating mnemonics is really helpful but can be time consuming, and you don't know which cards where you will need them until you repeatedly forget those cards. Someone on HackerNews about a year or two ago recommended using AI and that did help a little, but it didn't take long before I realized that the AI created mnemonics feel so similar and less connected than time tested mnemonics that I find them less effective. * Since brute force memory is slower I often learn slower, which is less efficient than learning a groups of related things together at a pace where the concepts together give you a better understanding than learning one at a time. (Sometimes you need to slow down because your not getting the concept, but going too slow is less efficient also.)
I still use spaced repetition but I realized it's not the amazing revolution that I first imagined that it would be.
Some people criticize flashcards as optimizing for rote memorization and deemphasizing understanding, but you'll never achieve understanding or mastery in general without a solid platform of knowledge to work from.
Spaced repetition is an effective way to review things but its biggest benefit is a process that's easy to be consistent with.
Somebody else can have equal or better performance with other technique but just like dieting, it doesnt matter as much what method you use as long as you stick with it.
I also find my verbal fluency is directly affected by how much pure social time I have in my schedule. It makes me think its one of those 'use it or lose it' things and that I need to schedule more time with people.
What do you do for topics like chess?
- Checkmates-in-one - Checkmates-in-two - Defensive technique (avoid checkmate/material loss) - Winning material - Endgame patterns - etc (~5 more)
... That I got by scraping the Lichess database, favoring common patterns that appear within +- 600 of my current Elo.
From there, I have Claude build me a script to convert each of those positions into a .png, then create me a deck with all the cards, et voilà. The front of the card is the position, the back is the best move in that position with a small explanation.
Every ~2-3 months when I see that most of the cards have matured (according the the Anki spaced repetition scheduler, I build a new deck around my new Elo.
I also play a lot. Prior to ~1000 rating I got away with spending 90% of my time in Anki and 10% playing online games, but lately it's been pretty 50/50. In higher ratings, playing real games tends to translate into wins more effectively for me.
For studying openings, it's almost the same thing, but the back of each card is the book move for my opening + the name of the opening the opponent chose.
> I won’t say LLMs are useless for this. But from my trials, I get about 1 card that’s useful to me out of 10, and even that 1 card still needs rewriting.
I don't know the specifics of how the author tried to do so, but from what I've seen the majority of attempts are, let me drop a chapter of a textbook and say "make flashcards." If that is what we are talking about, then yes, LLMs are useless.
In my mind, though, this is sort of like looking at the very first GitHub Copilot LLM autocomplete from a couple of years ago and concluding, yeah it's nice for one-liners, but it cannot write an app.
If you create a framework around your card-creation AI so that it can use tools, and verify its work to ensure common card-creation pitfalls don't happen, you can get pretty high-quality cards. In my experience, you go from a 10~20% hit rate to a ~90% hit rate, which in my mind is good enough. I got to ~75% quality just from a two extra LLM calls that would assess a potential card against a standard set of rules (adapted from [0]). There are huge Pareto gains to be had here.
I've generated thousdands of cards over the last few months this way. I let the AI add it directly to Anki via AnkiConnect. Then, if when I go to review I find a card that my AI created and I don't like it, I just delete it.
Removing the limitation of card creation is really quite compelling, and I think the area is still highly under-invested in. Would be cool to see a generic framework evolve that one could use. For now, I've been using a personal fork of clanki [1].
[0] https://supermemo.guru/wiki/20_rules_of_knowledge_formulatio...
[1] https://github.com/jasperket/clanki
When they asked me how I was doing it I explained that I'm there to learn understand the topic and I don't give a damn about the test. So I just let my curiosity lead the way, it causes me to ask questions in and out of class, email the professor about them, do my own research and experiments. Despite not letting the test be my guide, this prepared me for the test anyhow.
I'm glad they work for some people, but flashcards to me seem like they provide a shallow kind of understanding. I don't want to remember the equation, I want to be able to derive it in a pinch, and flash cards don't give me that.
You did better than they did because the test didn’t test wrote memorization.
I also know that in math courses, there’s certain formulas with similar names I simply need to remember because deriving them on a test would be too time consuming. Or I could imagine in chemistry there’s various special cases that you’d want to be able to recite on a test.
I’ve never tried it for language learning but people talk highly of it, so there’s definitely some truth in their utility there.
I think the trick is flash cards are fantastic supplementally with other studying strategies.
Voluntary retrieval (the memory "mode" of flashcards), even for simple pairs, is perceived as highly effortful by almost everyone, on a similar tier to doing mental math.
It's incredibly efficient for learning, and achieving your goals quickly can be very rewarding, but I don't think "learning flashcards should be fun!" is a reasonable expectation for the vast majority of people.
Downside for AI potential as a whole, framed broadly: We don't seem to be good enough yet at identifying what friction is functional and what we should strive to automate/eliminate.
I predict that things will be 'good' within the next 5 years, but your intuition is correct (and the SOTA models are often producing worse prompts than older models).
For example, I'll have a `sqlite` card, and put all the commands and everything on it, as I learn them. I'll use it as a cheatsheet, but then also a few minutes of mindful review. This for the toolings that I want to know well enough to not get slowed down googling the commands. I do this for a lot of CLI tools, but also things I need to remember about the business of my company and working across group, etc....
Eventually the five or six working cards I have, get put on a pile and new ones come in.....
And we still love it. I'm on the same page. This phenomena feels oddly satisfying.
I'm interested in understanding how others use Anki for conceptual subjects like pure math or physics. I believe many fundamental rules in Spaced Repetition (e.g. like keeping cards concise) are thrown out the window for conceptual subjects.
Analysis definitions and theorems get really complicated with intricate and difficult to follow logical chains, and there are a lot to remember.
These definitions and results don’t mean much on their own without exploring their neighbourhoods by proving relevant things, and I could have learned these definitions and results by just doing proofs. But being absolutely sure I could recite every theorem and definition definitely helped me on the final exam.
I think if you’re learning algorithms (like find the area under a curve) in a calculus course for example, flashcards might have more limited value, as in that case problems are relatively short and you’re better off just running through your set of algorithms a ton of times by doing problems.
I also took a group theory course last semester and I memorized every definition and result from lecture via flashcard, but didn’t practice using them enough by writing proofs. I ended up with like 2 or 3 out of 10 complete proofs and the rest half finished on the final exam because I had the right starting points, but not enough practice using what I knew in unexpected ways. Still passed somehow.
> These definitions and results don’t mean much on their own without exploring their neighbourhoods
Were these epsilon-neighborhoods?
I wrote a bit more about this problem here: https://borretti.me/article/the-applicability-of-spaced-repe...
a note on your request, have you seen this video before? Andy has some custom PDF reader he built with flashcards built-in, and it's two hours of tacit flashcard creation centered around quantum mechanics: https://www.youtube.com/watch?v=OFuu4pesKf0
For me, it's quick access recipes (breakfast pancakes for kids), what was the name of the glacier that we hiked to last year, behavioral prompts etc.
1: https://cognitivemedium.com/srs-mathematics
2: https://augmentingcognition.com/ltm.html
Same caveat as in the article: Spaced repetition is just one (minor) part of learning math/physics. It alone won't get you anywhere.
For math - particularly higher level math, the most obvious use case is definitions. There are so many!
You can put theorems in there, but it is a bit challenging on how to phrase it. A single theorem could result in several smaller flash cards.
I think what works better is taking a theorem, finding a representative problem that is solved via that theorem, and make the problem statement the question. The downside of this approach is each card takes longer to process as this is not just plain recall, but actively solving a problem. For this reason, I keep such cards in a separate deck and review them only when I have time I can dedicate (e.g. spending well over a minute per card).
Honestly just making the flashcards and elaborating on/modifying problems you're struggling with will take you a very long way.
There are no "rules" for how flashcards should work.
IMO you want to be actively trying to map the new concepts to things you already understand, and constantly working to update your mental model.
It's not an either-or.
Where SRS comes in handy is when you have to take long breaks between your study sessions (due to job + family). Have you ever tried learning an advanced math topic where you get to work on it for a few days, then may have to stop for a few weeks (or even months), then resume, and repeat over and over?
Chances are, no matter how intense you study during those few days, you'll likely forget important definitions/theorems in the periods you don't.
SRS takes care of those gaps.
Case in point - many years ago I put a lot of my intro to statistics course in flashcards and actively reviewed them. I hadn't done actual statistics for over a year, and then made a (false) claim here on HN. Someone gave me a counterexample using the chi-squared distribution. And it was amazing that I could recall the basic properties of the chi-squared distribution, and enough other theorems to verify what he said without consulting any book.
I've never used the chi-squared distribution for anything before or after.
(Sadly, I stopped using those cards years ago so I've forgotten the material!)
For me, much of the value of flashcards comes in the making of them. Part of it is thinking about what each flashcard will say, and part of it is the action of writing it down in handwriting.
Getting your words from real-world contexts, and keeping that context on the front of the card, largely eliminates the ambiguity problem. If a word has multiple senses, it gets multiple cards with different example sentences to illustrate each one.
It also helps a bunch with words that don’t really have a concise translation to your native language. For example the French words “mur” and “paroi” both mean “wall” in English, but the contexts where you use them are quite different. An example sentence helps with that, and getting that sentence from an even richer context such as a book or article you’ve read helps even more.
It’s also, frankly, just more enjoyable. I’ve come to view frequency lists as an antiquated tool. I needed them in the 1990s when good authentic-context study materials were hard to come by, but the modern Internet has made so-called immersion-based learning methods so easy and inexpensive I’m frankly mystified that people still cling to the joyless, almost mechanistic methods we were stuck with in the previous century.
The most salient difference here is that NLP wants to automate as much as possible for reasons that are specific to NLP.
But for human language learning a lot of automation is actually harmful because manual effort tends to be good for Ebbinghaus’s arguably more important but less popularly appreciated discovery: memory encoding quality.
It is important for language acquisition too, but the language involves a lot more rote memorization than the above.
I mean, you can put whatever you want on a flashcard. e.g "Derive the fundamental theorem of whatever", "Prove this theorem" etc.
Also music has a extreme level of "stuff you just need to memorise".