The Learning Revolution Was Never Waiting for LLMs

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.

I’m excited about modern AI (LLMs) as a new tool in the toolbox for education.

But it’s not some magical unlock that suddenly makes a learning revolution possible. It’s an add-on.

The reality is that we’ve had the tools we need for decades, and we just haven’t been using them.

Much of the science of learning is stable yet barely leveraged. Key learning-enhancing practice strategies (like mastery learning, spaced review, retrieval practice, interleaving, etc.) have been known and well-replicated since the early to mid 1900s, yet they remain relatively unused.

And you don’t need LLMs to build software that leverages these techniques.

The AI behind Math Academy, for instance, is old-school. It’s an “expert system” that emulates the decisions of an expert tutor regarding what tasks a student should work on at any given point in time.

There’s a knowledge graph encoding structural relationships between thousands of math topics. And then there’s an algorithmic reasoning system that looks at a student’s answers, figures out what they know (and how well they know it), and decides what learning tasks are going to move the needle most given their personal knowledge profile.

None of this requires LLMs. It takes coding chops, mathematical modeling, learning expertise, teaching experience, and oil tankers of elbow grease… but not LLMs.

For decades, the real technology bottleneck was student access to computers. The missing tool was Chromebooks, which are now commonplace in schools.

Now all the pieces are in place for a learning revolution. That’s what Jason Roberts realized a while back, and that’s what we’ve been working on ever since: pulling these pieces together into a scalable learning experience and tuning the highest-leverage dials for each individual student to maximize their learning.

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.