Github pages:
https://emanueljankvist.github.io
Immanuel Jankvist
Karma: 52
I think this is really common, even if you are not using the “Everything on Back” approach. If I had to give it a name, I would go with “interference”, which seems to be the name from 20 Rules. This is not an endorsement of the 20 rules.
I previously used Obsidian with an Anki plugin for syncing to Anki one way (this also allowed card updates from Obsidian to Anki, but not the other way–hence one way). The upkeep became too heavy. I have since been trying to keep upkeep to a minimum.
I will update a bit however, since more people have now mentioned the same thing. Perhaps I am just doing something wrong. Thanks for sharing:)
After reading this comment I choose to make my font random (within a class of six readable ones). I normalised font height, but my line sizes are now all jumbled (I use max 50ch because I use my phone for Anki). Will probably keep this change indefinitely.
I really like this idea. Thank you so much for sharing!
Thank you for writing this! It’s great to see someone tackling Anki methodology in depth. I’d like to add my currently held beliefs about what makes for well-designed Anki cards and good practices. These are stated in definitive terms to make the critique tangible, though I acknowledge this means sacrificing some nuance.
Against the “Levels” System
The “level 1, 2, 3” hierarchy presupposes meta-knowledge about your knowledge. I think this risks wasting cognitive resources tracking what you know instead of learning more. I prefer closed-ended questions with single correct answers.
More Cards is Better
A couple of people have mentioned it here, but 2 new cards per day is very aggresive. I probably agree with Nielsen/Matuschak that you should make “more cards than you think you should.” While there are diminishing returns, there are also diminishing costs. A tightly interconnected web of facts has better retention than isolated ones.
Redundancy is a Feature
Hard agree with this point you make. Cards reinforce the pattern “circumstances → solution.” Multiple cards with slight prompt variations (epsilon changes) help you recognize solution spaces more generally. This is underappreciated.
Against “Everything on the Back”
This creates open-ended questions with multiple valid answers. You end up harshly reinforcing one answer while “forgetting” others that are technically valid (if this is really the preference, then it is hard to say anything against). Better alternatives:
Reverse cards (both-sides definitions)
Separate cards for object names vs. properties
Cloze Deletion: Format vs. Application
I disagree that clozes are bad. Every card type is technically cloze deletion (if there is any point in getting pedantic). Your real issue therefore (I assume) isn’t the format but excessive context, which leads to remembering “visual shape → answer” instead of “semantic meaning → answer.” Is this accurate do you think?
Practical Deviations
I break the 18-word limit regularly—mathematical theorems need their assumptions. I make one card for results, separate cards for assumptions. I also skip parent-child memorization by embedding more context directly in cards. This might be the fundamental tradeoff.
If you have run into any of these issues in your own practice or believe them to be non-essential, then I would be happy to know
The following seems a bit unclear to me, and might warrant an update–if I am not alone in the assessment:
Section 3 finds that even without a software feedback loop (i.e. “recursive self-improvement”), [...], then we should still expect very rapid technological development [...] once AI meaningfully substitutes for human researchers.
I might just be taking issue with the word “without” and taking it in a very literal sense, but to me “AI meaningfully substituting for human researchers” implies at least a weak form of recursive self-improvement.
That is, I would be quite surprised if the world allowed for AI to become as smart as human researchers but no smarter afterwards.
Basically agree: I have tried considering what is most likely to get me killed (in everyday life). This was a while ago. I am now pretty scared(/respectful?) of cars. Reflection seems to work okay at calibrating emotions. I hope LW gets a couple more posts considering emotions and virtue–perhaps a small shift from the utilitarian consensus.