Lately, I’ve been thinking a lot about what really makes learning experiences stick—especially those that stay with us for years. As someone who spends every day designing learning experiences at an innovation-focused company- Desklight- I’m always searching for new ways to make learning more authentic and transformative. This search is what drew me to Andy Matuschak’s recent reflections on the future of learning. His ideas challenge some deeply ingrained assumptions and point toward a world where “learning” and “doing” are inseparable—and where technology can finally help us close that gap.
We’ve all had moments—maybe launching a startup, tackling a creative project, or diving deep into a research challenge—where we were fully absorbed, learning tons in the process, yet it never felt like “studying.” It felt natural and authentic and invigorating. Those periods stand out not just because we learned something new, but because the learning was baked into something we genuinely cared about. It wasn’t a chore, it was deeply connected to our values and goals.
The question is: Why don’t we feel like that more often? Why, most of the time, do we find ourselves either forcing authentic projects without enough understanding or slogging through formal training that never seems to stick?
For decades, there’s been a tug-of-war in education between purely “immersive” learning (just diving in) and more explicit, guided instruction. Both camps have valid points. Immersion sparks motivation and real-world relevance. Guided instruction respects cognitive limits and ensures people actually remember and master the material. But so often, our attempts to blend the two fail—ending up with uninspiring projects on one side and forgettable lectures on the other.
Matuschak’s insight is that we need a genuine synthesis. We want learners to jump into authentic projects, but we also need well-timed guidance and conceptual scaffolding. It’s not about compromising between two half-measures; it’s about building a new approach that does both well.
Right now, the dominant narrative about AI in education revolves around “chatbot tutors.” While helpful for quick questions, these chatbots often sit on the sidelines. They might clarify a concept, but they don’t join the learner in their actual workflow. They don’t know what the learner’s project is, can’t see the code they’re editing, the text they’re reading, or the communities they’re curious about. That distance can make the experience feel hollow—like calling tech support in another room, rather than having a partner right beside you in the workshop.
Matuschak imagines a more integrated role for AI, one that’s woven directly into a learner’s real-world context. Instead of a separate Q&A box, AI would understand your background, your current goals, the specific tasks at hand, and the tools you’re using. It could guide you through code implementations, visualize complex concepts right in your Jupyter notebook, connect you with a local meetup on brain-computer interfaces, and suggest just the right reading in a textbook. And crucially, it wouldn’t just hand you answers—it would help you truly understand the underlying ideas, strengthening both motivation and long-term retention.
One of the stickiest challenges in education is ensuring that new knowledge doesn’t evaporate after a week or two. Research on memory is clear: repeated retrieval and spaced practice make a world of difference. Matuschak’s work with something called a “mnemonic medium” (like in Quantum Country) shows how weaving short, well-timed reviews into the learning material helps readers remember complex ideas for months or years, not days.
Imagine if, instead of occasionally quizzing ourselves or relying on rote flashcards, we integrated spaced review directly into the meaningful project work we’re already doing—where the practice is always contextualized, always relevant to what we care about. The result: stronger, more flexible understanding that feels like part of who we are, not just something we “learned once.”
The standard chatbot tutor concept falls flat when compared to what real human mentors or colleagues offer. A great tutor or mentor understands your goals, helps you navigate complex tasks, adapts to your growing skillset, and models the culture and values of the field you’re entering. More than just answering questions, they help shape how you think, how you approach problems, how you see yourself as a contributor in that space.
Can AI help approximate some of that? Possibly. Not by dumbing down learning or locking it into a text box, but by joining us right where the action is—embedding itself in our real work, connecting us with communities, and personalizing guidance based on a deep, evolving understanding of who we are as learners and creators.
What resonates most deeply with me is the moral imperative in this reimagined approach. Instead of viewing learners as “deficient” and trying to “fix” them, we should be empowering them to pursue their genuine interests and grow within them. The role of technology in this future isn’t to micromanage people into correct answers. It’s to give them the tools and guidance they need to tackle meaningful challenges—and in doing so, help them internalize knowledge deeply and authentically.
In other words, the ultimate promise of AI in learning isn’t about standardized tests or efficiency metrics. It’s about helping people explore their frontiers of interest, connect with like-minded communities, and build knowledge and skills that genuinely matter in their own lives. It’s about turning “study time” into real-life growth—into doing, making, and becoming.
As a director of learning design, I find this vision both challenging and exciting. It’s not a simple matter of adding a chatbot or layering in some quizzes. It’s about designing entire ecosystems—tools, resources, social connections, and reflective practices—that let learning unfold naturally as part of the creative work we’re already doing. It’s about making “learning” inseparable from “living.”
This may be a long journey, but if we get it right, we’ll move beyond the limited notion of AI tutors as question-answering bots. We’ll instead create rich, integrated learning environments where guidance, mastery, and authentic engagement flow together, bringing our best learning moments into the everyday.