We're a little ways off from launching this multi-lingual approach, but it reflects some of the values and ideas behind Sunlo – you don't learn so you can translate; you learn to connect with people and groups, for a time and place and a range of contexts that's personal to you.

(Sunlo will use language processing tools like tokenising and vectorising to categorise HUMAN-MADE flashcards and create "smart playlist" style sets of cards to add to your deck, but every flashcard, phrase, and translation will be written by people, for people.)

Obviously with a pure DIY flashcards app you can do any language to any set of translations, but Sunlo will be the first app I've seen that is truly any-to-any (and any-to-many) and still offers a kind of structure for learners to get started using curated public resources. I hope it works!

Actually this is something I think is pretty neat about our flash card model: if you're learning Hindi and you speak French and English, Hindi phrases that have translations into both languages will show both translations, which can help with comprehension when the structures and grammars differ.

The app uses user-generated content to populate the flash card library so it can support any language to/from any other language as long as we have users. But from the start we'll be focusing on seed content for English speakers to learn Hindi, Tamil, Spanish and French.

But ultimately these are just details the user shouldn't even have to know about. We hope the app is easy, intuitive, fun to use, and most importantly an effective tool for learning languages 🧄🥦

Second: the scheduler is dynamic so the user (or the client app) can select cards with different difficulty scores for more dynamic, adaptive user experiences, such as giving the user an easier card if they've been struggling, or testing out different mixes for their effect on session completion.

First, we wrote the scheduler entirely in Postgres as an RPC to record reviews and a view to fetch cards with their FSRS metadata attached. We hope to release our PLv8 implementation as an open source package in the future, if there's interest.

- invite friends and send them cards

We also wrote our own version of the FSRS (Spaced Repetition) algorithm, directly in Postgres using PLv8! Our version of the algo builds on the amazing work of the FSRS community to do a couple things I don't think others have done before: