The potential for publicly controlled education on the decentralized internet

This week, after stewing on design details and prototyping for months, I released a version of my dream spaced repetition system (SRS) project that I'm finally happy with, https://memsora.com/. If you're not familiar with the concept, the short version is that it's flash cards with reviews scheduled by an algorithm that always tries to keep you just barely on the edge of remembering, exploiting how we reinforce memory through recall and making information stick easier, for longer.

If you are familiar, though, you might instead be thinking: "isn't this just [insert favorite SRS tool]?" Yes, this idea isn't new, and what I'm building isn't revolutionary, groundbreaking work. However, I think it's still important, for a couple reasons:

First, to get this issue out of the way before getting to the real topic of this post, with memsora I'm attempting to address issues I have with the existing options for SRS study. The most popular choice for this, Anki, does the job well and has a large body of material already to use, but has some critical UX issues in my eyes. I won't dwell on these too much, but one of the big ones is the lack of write-in prompt reviews that force physical action that has been shown to aid memory. Some other options exist, although many of the benefits seen by other contenders are locked behind closed-source paywalled services that may only cater to creating study material for a specific topic rather than having open contributions.

And support for open contributions is where I want to focus here. One of the main uses that has popularized SRS software is for language study. I've been a habitual language learner for almost 20 years now, and while the landscape has grown greatly in that time, it has also heavily been exploited. A handful of names have emerged vying for the growing market in language learning, rushing to secure their share through means dependent on exploiting passionate volunteers and underpaid professionals, prioritizing addictive user retention over evidence-based learning or even basic accuracy, and using their sway to decide what information is and isn't worth learning to the world. And what is the result? I know real people, proud to have checked in daily with a green owl for several years, who can't hold a conversation in their target language.

I believe the emerging decentralized internet enabled by developments like atproto stand a chance to flip this unfortunate direction on its head. Not just for language either, although I think it's easy to see the benefits there, but for anything we learn and teach online. I don't see it happening right now, but I see the potential and hope to play my own small part. Today, it is easier than ever for someone to independently publish educational material online in a way that properly belongs to them, and as developers it is easier than ever for us to create the tools to enable doing so more effectively. There is a future, if we can steer ourselves towards it, where education online is owned by educators, and is focused on truly being effective.

What does this look like for atproto and memsora? Here's something that could happen today: Someone creates a course online in the form of a bunch of static pages, one page for each lesson. There's a lot of ways to host these, but let's say we want to host from the PDS (since the creator owns everything that goes there), served via something like https://wisp.place/. Now, at the end of each lesson, they could link to a memsora deck containing all the items that the student just learned, which the student can import into a course deck containing everything they've learned so far. Now, as they keep studying, they'll have periodic reviews of what they've learned.

This is just an example. As the atmosphere evolves and the options we have for building out better decentralized services grows, the model of how we take advantage of this is likely to change. But the central theme should remain that creators of educational material control the information they are distributing, and decide how it is made available. And in that regard, one of the greatest motivations for memsora is that, to serve this end, we need tools that are designed with this sort of decentralized model in mind first. I am offering the (admittedly very light right now) computational resources needed to serve the appview for this, but the data involved always belongs to the people making decks and the people studying. Not by my own design choices, not bundled up in caveats, but because in this ecosystem there is no other way. And that just might help us make something with people with the right priorities in control in the end.

Let's become smarter together.