Live data from Hacker News

Show HN: Anki AI Utils

github.com

41–48 of 48 posts

Re: Show HN: Anki AI Utils

#42
post #30
post #25

I don't personally use Anki, not for any major reasons or ideological disagreements but mostly out of a "I do something else to learn" methodology. I've always liked the concept of Anki, flashcard learning, and needing some method of "deliberate practice" to reinforce the knowledge. My profile and even my dissertation explored the idea of practice through Chi and Wylie's ICAP framework [1] (interactive, constructive,…

Thanks for linking me to the ICAP framework. ICAP and your “I do something else to learn” methodology generally jives with the same thing I happened upon by chance after spending a lot of time trying to “learn how to learn” the best way; landing upon SRS and Anki specifically, as a tool; and then finding a much better process+system. Given how deeply involved you are in the space, I assume you’ve heard of, and possib…

I'm not familiar with Sung's videos, but after a quick perusal of his thumbnails I have some knowledge in the various cognitive science concepts he covers (Cognitive Load Theory, Flow, Mindmaps). I didn't really dig into educational theory until I started teaching Computer Science. I wanted to figure out how to better instruct my students after learning about the high drop/fail rates in intro courses. Once I started to experiment in my classes, I decided I should get the PhD for a pay bump.

Once I was in the program, I focused most of my research in reading where cog sci was being used for stem, but also for general practice research. I've been training martial arts for almost 20 years now, so some of the research was me double dipping in how to improve teaching CS and punching people.

Honestly, I still argue that martial arts' spaced repetition was a bigger influence on how I view learning. I need to allocate 2-4 hours 1-4 times a week for practice (4 when I was younger and could get away with it; correct due to immediate feedback from my partners; and have a giant support network of people through the US that make me vested in not only the art but their lives as well. I acknowledge the benefits of meta-cognitive methods like planning and self-reflection, but they feel more theory than application.

Planning is great until you're a novice that doesn't know what to train next. Then you are just a struggling student receiving negative reinforcement, which only amplifies any imposter syndrome you already have. Sadly, there isn't much research exploring how physical athletes learn beyond simple spaced repetition. There's some work in interleaved practice [1] but since physical training is more or less "solved", progress is slow.

Instead, I focus on the various lower-level practice activities so students can acquire subskills without needing to program. Then, I heavily encourage building a 'sense of community' [2], not through group projects (which have their own faults) but rather in simply "giving a damn" about your classmates' progress.

At the end of the day, I think learning is heavily a "time on task" [3] problem and determining how to structure lower-level practice and toy examples that encourage you to keep with it and break Carol Dweck's "fixed mindset" [4].

I'd like to dig deeper into how to properly structure practice across ICAP modalities, but the sheer number of variables and even determining how many activities should be in a practice is too complex of a problem without a very large sample size.

[1] https://effectiviology.com/interleaving/

[2] https://en.wikipedia.org/wiki/Sense_of_community

[3] https://www.thisiscalmer.com/blog/time-on-task-learning-stra...

[4] https://learning-theories.com/mindset-theory-fixed-vs-growth...

Re: Show HN: Anki AI Utils

#43

Earlier quoted context omitted.

For me it grew from a horror of the state I find myself when I just *can't* recall something I have known or have to lookup things several times. Then the satisfaction of never having that feeling again!

I can tell you haven't used Anki that long. No method is 100% perfect

I'll let you decide if using anki for 6+ years and with a 1000+ days streak counts at "long". In any case it doesn't have to 100% solve my memory issues to satisfy the goal I described above, it just has to solve it enough to make me feel okay with some forgetting because I literaly couldn't try harder.

Re: Show HN: Anki AI Utils

#44

Earlier quoted context omitted.

In that thread I first heard about anki smart notes : https://www.smart-notes.xyz/

That appears to be a subscription service. Surely there must be something using piper or another offline TTS program that works for free?

Well maybe not that less complicated than awesometts addon

Re: Show HN: Anki AI Utils

#45
post #26

Earlier quoted context omitted.

I am tempted to say yes, but in truthfulness it really depends on a lot of details. If we are just talking about the card creation phase, then I recommend other tools I made, especially voice 2 Anki (posted on hn at the same time but has not taken off for some reason :) ) If we are talking about the actual learning and memorizing phase, then yes, I think those tools are valuable. But that being said, I'm also sure th…

That's fair advice! Creating cards is also learning, so if you generate content using LLMs instead then you'll be missing out.

To expand my previous comment: writing a lesson to memorize always helps, but that does not mean that having some printouts from university won't be a good tradeoff or manually copying a whole book instead of renting it etc

Re: Show HN: Anki AI Utils

#48
post #10

Wow, so feature rich, congrats on the release. Love that the mnemonic generation takes into account your existing memory anchors (didn’t know that term). I’ve actually been working on a similar-ish Anki Plugin for about 6 months - it can autogenerate any field via LLM in bulk, as well as images and TTS. I’m not explicitly targeting the med school use case as much yours (I use it for language learning), and it’s more…

Thank you very much for the kind word but I have to return the complement. The app you are making is very feature-rich too, seems easy to use and affordable too! By the way, the new LLM by deepseek called deepseek-chat is very good and actually on par with Claude Sonnet 3.5, so you might want to give it a try. Openrouter and litellm help a lot. Also, my projects are not actually aimed at medical learners, I just happ…

Interesting - how do you use LiteLLM + OpenRouter here?

I would assume they're alternatives

Post reply on HN