Hacker News new | comments | ask | show | jobs | submit login
Incremental Learning – A continuous learning approach (2018) (kishorepv.github.io)
158 points by pps 8 days ago | hide | past | web | favorite | 21 comments

I have been personally experimenting with what I call "drip learning" - essentially, I am trying to learn a large amount of non-urgent information very slowly, piece by piece, over a very long timeline. For example, learning all the letters of the Greek (or Hebrew, or Russian, or Armenian) alphabet. If I sit down and attempt to learn them all it once, it quickly becomes overwhelming and retention is low. But, if you simply learn one new letter per day, it's very manageable - and you'll learn the entire alphabet in a month.

Combined with spaced repetition (Anki) to reinforce the memory, I think it's extremely powerful. Anecdotally, it works well - I can recall all of these items on demand, months later.

The next step (which I haven't quite figured out) is to multiply this across multiple domains. I.e. every day, learn 1 French phrase, 1 Spanish phrase, 1 German phrase, and so on. In my (nonscientific) experience, this is more effective than focusing on a single topic.

I like the idea but I think with languages it's better to stick with one till you a certain level (maybe b1/b2) then move on.

If you are going to Italy soon you can memorize "non lo so" to say "I dont know" very effectively. But if you are actually trying to learn and retain "some" Italian, you probably need to know more like "lo" is a direct objective pronoun for him/her/it and "so" is the verb essere conjugated in the present tense for io.

Little correction: "so" is the conjugated verb "sapere" (to know)

yep - thanks!

You may not be an Android user as I am, but if you are --- im excited for the V2 scheduler in Ankidroid, which will allow you to pull in cards from multiple subdecks when reviewing a parent deck. The desktop client already has it.

If that is added, or you're already just using the desktop client, then IIUC the feature you seek (1 new concept, per category, per day) could be acheived by setting the "new card" number on each deck to "1", provided you've separated each of the concepts into subdecks of a main deck.

I could have posted this exact post. I use Anki to slowly study subjects that I'm interested in, including foreign alphabets. Myself and my oldest daughter are doing Greek right now.

You can do this in Supermemo quite easily. It does have interleaving (mixing multiple "decks") by default, although you can do reviews from one deck if you want to.

You'd be losing FOSS, Linux, and IMHO my favorite feature, the mobile support by switching to Supermemo.

When I was using Anki in school I just used multiple decks. Usually one per class.

The most effective way to learn something is actually to sit down and learn it.

There is a significant time / resource difference between one word a day and constant study time where you do learn way more than 1 new vocabular per day in avg.

Right, but I believe his method was trying to optimize retention.

Depending on well you retain information through a block of constant study time vs. a much smaller, long-view approach to the same amount of material, could be the determining factor on which strategy you choose.

This is great. I'm actually working on an app that's literally designed as the perfect incremental reading/learning platform:


Here's how it works. You basically add everything you're reading to Polar.

Right now it supports PDF and captured web pages. It works offline and your data is yours. It's a desktop app with a webapp+mobile coming soon (like 2 weeks).

Polar supports suspend and resume of reading with "pagemarks" which are basically "boxes" covering multiple pages with a start and end.


This way it's very very clear what you've read and what you have not.

While you're reading you can annotate and take notes directly in Polar.

This includes comments, highlights, but also flashcards.

The flashcards can be sync'd with Anki which means you get spaced repetition built in so you NEVER forget anything.

Here's a screenshot of my current reading:


The progress bar on the right is how far I've read within that document.

All your annotations and highlights are yours. We support markdown export and the on disk format is simple JSON.

It's also Open Source and we'll have a web and mobile version soon. Works on Linux, MacOS and Windows.

It also supports cloud sync so if you have multiple computers you can keep all your data in sync.

If you just want to avoid the cloud you can not use the cloud sync.

However, if you do, you can also use the web version which is coming out soon and which also means you can use it on a tablet device.

I also have larger long term plans for Polar.

The idea is to build out a collaborative learning platform.

However, first I have to attract and audience together that is large enough that could benefit from collaboration.

So one step at a time.

Great idea. Is there any way to pull in kindle data?

Currently I read all my books and articles in kindle. I think I'll use polarized to read and organize the articles I read and use kindle to read the novels.

If I could see the total usage statistics in polarized that would be awesome.

The problem is the DRM .. you would have to use some thing like Calibre to port them to Polar.

It's legal to do this for compatibility purposes but IANAL.

Interesting, how about sharing annotations? I imagine that using a sorting algorithm like reddits comments could work very interesting on popular texts.

yeah.. it's on the TODO.. porting it to the web now specifically for this use case.

How would you compare and contrast using Polarized vs Evernote for the same tasks?

Evernote is better for notes.

Polar is better for creating notes from what you're reading.

I'm working on making Polar better for just plain notes. Getting there.

looking forward to epub support as well.

Working on it... need to build out a new reader though.

Especially given how much the icon reminds me of Calibre :D.

This sounds like the "Compounding Knowledge" post from FarnamStreet last week:


Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact