Hi everyone. 2nd submission attempt here of a very-indie, very-passion project after a no-op first try.

Backstory: ~10 years ago I started tinkering with a web based platform for experimenting around the fringes of SRS. The project had three major goals: to expand the sorts of content that SRS could address, for example doing musical exercises via midi instrument hookup; to hybridize SRS scheduling with other pedagogical protocols like prerequisite graphs, Elo ratings, procedurally tuning difficulty of specific exercises, and similar; and to permit usage-driven pedagogy discovery among competing protocols, in terms of things like learned pre-req edges, and surfacing aggregate learning bottlenecks to curriculum designers.

This project stayed at hobby level. It got a fair bit of personal in-house use, but I've always itched to put a more direct, sustained focus on it.

~4 years ago I found myself with a toddler.

So: I started throwing phonetic exercises into our in-house learning platform. We had great success with very modest daily-ish use. As her skills grew, I filled content ahead her. She was reading independently in earnest by around her fourth birthday, and hasn't stopped since.

Some time later, I learned that the established premium early literacy app sells at a $500/Month subscription. This irked me - not thrilled with the prospect of such an extreme income barrier on such an important early life skill, especially where the skill so clearly jump starts a student's self-efficacy and capacities for future learning. Here at home we had existence proof of a system that works well, and that, because of its mechanisms, is positioned to work even better at scale.

So I took the plunge. I dressed it up a bit, and built a phonetic analytics pipeline to run against English language data sets of phonetic decomposition and word usage. This helped to prioritize curriculum sequencing according to incremental utility, and also to do pretty strong procedural processing of / responding to children's errors.

I am pretty happy with the result! I think it's suitably inviting and friendly for the target audience (verbally strong, pre-literate kids) without leaning on attentional engagement hacks. To minimize screen-time, it is narrowly scoped to building and maintaining a phonetic skills inventory - the piece of early literacy where computer bookkeeping systems have the most comparative advantage over normal parent-child literacy activities (eg, reading a book together on the couch is going to beat any literacy-app narrative-slop). It is <<< $500/month.

There are still some (many?) rough edges here and there, but the app is in production now and is serving a handful of families who seem to be enjoying it.

Some links:

https://letterspractice.com/showHN - a technical demo of the app at runtime, showing some of the engine's workings in real time. A bit noisy I'm afraid.

https://github.com/patched-network/vue-skuilder - the framework itself, with docs at https://patched.network/vue-skuilder/

https://letterspractice.com/blog/a-is-for-access - my hello-world blog post for the app. Mostly an expansion of the origin story above.

I am taking all suggestions and commentary here - I am very new to the B2C space. Very happy also to be grilled on the design space around early literacy acquisition, the case for two- or three-year-olds to use software like this, or anything else.

Thank you!

Neural Word Embedding as Implicit Matrix Factorization (2014)