July 2024. Part of “Letters from the Lab”, a series of essays on my research written for patrons. You can also listen to this essay (29 minutes).

If you want to make an educational technologist’s eyes sparkle, just mention “The Young Lady’s Illustrated Primer”. It’s a futuristic interactive schoolbook, described in Neal Stephenson’s The Diamond Age, where it lifts a young girl out of poverty and into sovereign power. It’s my field’s most canonical vision of a wildly powerful learning environment. If you ask a technologist interested in learning what they dream of achieving, most will answer: “building the Primer.”

Fifteen years ago, I’d have given the same answer. With my weak skills and understanding, the Primer was far enough away that I couldn’t see its details properly, but it was still enchanting enough to drive me emotionally. As my practice grew, and as I earnestly considered what it would mean to build the Primer, I started to notice the vision’s serious flaws. Gradually, I came to see it as fundamentally unworkable, even while it still deeply compelled me. Now I feel haunted by the Primer. I know it’s not what I want to build, but some part of my mind won’t let go of that vision until it has something else it can grab onto.

In fact, I think my whole field is haunted by the Primer. That’s not Stephenson’s fault: it’s ours. Our shared canonical vision remains a plot device from a science fiction novel because we haven’t managed to articulate something better ourselves.

I want to exorcise myself of the Primer. I want to clearly delineate what makes its vision so compelling—what I want to carry in my heart as a creative fuel. But I also want to sharply clarify the lessons we shouldn’t take from the Primer, and what it simply ignores. Then I want to reconstitute all that into something new, a vision I can use to drive my work forward.

(Audience note: I’m writing this essay for technologists entranced by the Primer’s vision, so I assume familiarity with The Diamond Age. If you haven’t read the book, don’t expect to follow this discussion in detail.)

What I want to take from the Primer

In Nell’s adventures with the Primer, I recognize something precious and scarce from my own life. Her interactions with the book remind me of my most rewarding, highest-growth learning experiences—except that the Primer produces those rare experiences basically all the time, for every topic. That’s my central connection to the book. The Diamond Age compels me because it depicts a girl who spends most of her life in those states I cherish.

With this framing in mind, I think the right approach to extracting ideas from the Primer is a selfish one. When I think about “building the Primer”, I’m not thinking about an educational intervention, a schoolbook to help children learn a curriculum; I’m not thinking about its morality tales. I’m thinking about what *I* want, as a capable, curious adult. I want to use its ideas to invent enabling environments for myself—to help me better think, learn, and create.

Immersion

My most cherished learning experiences have involved diving into a topic, trying things, getting my hands dirty, living and breathing it. In the real world, this kind of learning is somewhat rare. Many topics just seem too abstract. Or, when I try to “just dive in”, I’ll struggle to find an effective way to engage. But this is mostly how we see Nell learn: through immersive, hands-on action.

The main way the Primer makes this possible is through dynamic media. Its interactive representations and simulations expand the range of topics where immersion is possible. When it wants to introduce a conceptual topic, it begins with concrete hands-on projects: Turing machines, microeconomics, and mitosis are presented through binary-coding iron chains, the cipher’s market, and Nell’s carrot garden. Then the Primer introduces extra explanation just-in-time, as necessary.

That’s not how learning usually works in these domains. Abstract topics often demand that we start with some necessary theoretical background; only then can we deeply engage with examples and applications. With the Primer, though, Nell consistently begins each concept by exploring concrete instances with real meaning to her. Then, once she’s built a personal connection and some intuition, she moves into abstraction, developing a fuller theoretical grasp through the Primer’s embedded books.

We needn’t (and shouldn’t!) think of these dynamic media from a purely didactic perspective. The Primer’s nanomolecular microscope helps Nell dip her toes into cellular biology, but it’s also a tool which makes expert biologists much more capable. It’s not a toy representation, a cognitive dead end which Nell must discard as soon as she builds an intuition. It’s a tool which can grow with her into legitimate practice, a tool which in fact expands the frontiers of practice for the entire field. My collaborator Michael Nielsen has long argued that this is true of all our most powerful representations. If you make experts more capable, similar ideas will often also help novices; but if you focus on educational use, you’re unlikely to transform real work in the field. Mathematica is a great modern example of this: it was invented to support frontier research in cellular automata; happily, it also allows novices to more easily build intuition for complex mathematical ideas.

Responsiveness

Have you ever stumbled on a stairwell or lobby which felt acoustically spring-loaded—where when you hum or clap, you feel like you’re plucking an instrument, and the whole room vibrates back at you? That’s what my favorite intellectual and creative experiences have felt like. My every move is absorbed, reflected, amplified back at me through the work or the people around me. The whole situation is alive, responsive. It’s a rare and precious feeling, and the Primer offers Nell something like it all the time.

This responsiveness rewards curiosity. It’s not just that the Primer patiently answers Nell’s endless childlike “why?” questions. When she expresses interest in anything, the book finds tractable ways for her to start meaningfully engaging. Then it helps her go deeper over time, as in this passage:

In the plot, Nell planted some carrots, thinking about her friend Peter who had vanished so long ago, and in the window boxes she planted some geraniums. The Primer taught her how to do it and also reminded her to dig up a carrot sprout every few days and examine it so that she could learn how they grew. Nell learned that if she held the Primer above the carrot and stared at a certain page, it would turn into a magic illustration that would grow larger and larger until she could see the tiny little fibers that grew out of the roots, and the one-celled organisms clinging to the fibers, and the mitochondria inside them. The same trick worked on anything, and she spent many days examining flies’ eyes, bread mold, and blood cells that she got out of her own body by pricking her finger.

Responsiveness also manifests through rich feedback, which the Primer provides continuously and often immediately. When learning somersaults, for instance, the Primer shows Nell a recording of her movements and points out where her form could be improved. But it does the same trick with abstract topics, too: Nell learns Boolean logic by manipulating waterways, and punchcard programming by manipulating an organ. The Primer chooses (or invents) representations which offer live interaction and feedback, minimizing the distance between idea and action, between action and mental update. It’s true that immediate feedback can sometimes subdue reflective contemplation, but I generally find I want more of it than I can easily get in most topics.

Assurance

My highest-growth experiences weren’t easy or stress-free. They were often quite demanding. But critically, in these periods, I felt my hard work paying off. I had confidence that if I kept pushing, I could reach whatever I was chasing. Likewise, part of what makes memory systems so rewarding is a sense of total assurance: if I want to remember a fact, I can add it to my memory system and be quite sure that I’ll retain it indefinitely. That certainty drives a positive feedback loop: learning works, which feels good, so I’m eager to learn more, and so on.

The Primer creates this kind of assurance for Nell. It lets her struggle—she’s often depicted attempting a puzzle many times—but it never lets her fail. She grows to trust that she can handle anything the book throws at her, because she never encounters any topic she can’t eventually learn. My memory system makes me feel that way about remembering things; I’d love to feel that kind of certainty about learning any topic or skill at all, as Nell does.

One way the Primer achieves this feeling is through dynamic scaffolding. In my highest-growth experiences, I’ve managed to find a perfect route through the difficult terrain I’m traversing. At every moment, I have the support I need to make progress, but not so much that what I’m doing feels boring or fake. And as I become more skilled, that support smoothly fades away, so that my frontier of mastery keeps expanding. These routes are rare. Much of the time, when I “just dive in”, I end up diving into a wall, or stuck in trivialities.

But the Primer provides and adjusts this kind of scaffolding for everything Nell learns. Sometimes it does that over a span of years. For instance, it initially reads aloud to her, then supports her as she sounds out unfamiliar words, then leaves her to read fine print silently. For some topics, the Primer scaffolds over days and weeks, as in the Castle Turing sequence in which Nell progresses from unary alphabetic codes to Turing machines. The Primer even provides dynamic scaffolding within individual sessions, as the interventions in this passage illustrate (my comments in brackets):

Harv and Nell were trying to build a fire. There was a pile of wet logs Harv had chopped up. Harv also had a rock, which he was striking against the butt of a knife. Sparks flew out and were swallowed up by the wet logs.

“You start the fire, Nell,” Harv said, and left her alone.

Then the picture stopped moving, and Nell realized, after a few minutes, that it was fully ractive now. She picked up the rock and the knife and began to whack them together. … Sparks flew, but there was no fire.

She kept at it for a while, getting more and more frustrated, until tears came to her eyes. But then one of the sparks went awry and landed in some dry grass. A little curl of smoke rose up and died out. {This was not an accident, but a scaffolded hint provided in response to her frustration.}

She experimented a bit and learned that dry yellow grass worked better than green grass. Still, the fire never lasted for more than a few seconds.

A gust of wind came up and blew a few dry leaves in her direction. {Again: not an accident, but a response to repeated failure.} She learned that the fire could spread from dry grass to leaves. The stem of a leaf was basically a small dry twig, so that gave her the idea to explore a little grove of trees and look for some twigs. The grove was densely overgrown, but she found what she was looking for beneath an old dead bush.

“Good!” Harv said, when he came back and found her approaching with an armload of small dry sticks. {Note that Nell finding the twigs triggers his return.} … Soon they had built up a roaring bonfire.

Emphasis on the emotional

My highest-growth experiences are all driven by obsession: I’ll find an idea utterly beautiful, or I’ll fall in love with a community, or I’ll feel an almost righteous need to realize some captivating vision. It’s not a cold utilitarian calculus; it’s a hot emotional fuel. It can be hard to stoke these fires, though, and easy to accidentally lose the spark if I make the wrong moves. One reason the Primer’s vision sticks with me is that it emphasizes the emotional.

The book’s full name is The Young Lady’s Illustrated Primer. It recognizes that beautiful imagery can enchant. The Primer isn’t an encyclopedia; instead, it uses mythic stories, vivid characters, and immersive environments to create and maintain emotional connection. And the dialogue is performed by a professional who devotes her life to expressing care for Nell through her voice.

As much as emotions can draw me in, they can also hold me back. The biggest bottlenecks to intellectual and creative work are often emotional, and the Primer prepares Nell for that. It meters out setbacks and new challenges strategically. Its scenarios are designed to create resilience, self-efficacy, and independence, as much as to help Nell learn specific topic knowledge.

Another important emotional move: the Primer places Nell in a fundamentally active stance, framed as the author of her experiences. There’s a built-in narrative structure, but Nell’s choices and actions turn each page. It feels to her as though she’s causing the story, not passively consuming it. My own highest-growth experiences all have this authorial emotional texture.

The Primer also recognizes that learning is an act of identity construction. When I learn about a topic, I’m not just learning facts and abstractions. I’m becoming, at least partially, the kind of person who practices that discipline. I’m taking on some of that domain’s values and perspectives. I’m changing the way I view myself and the world.

Much of what the Primer does emotionally is too patronizing and manipulative—more on that soon. But the main lesson I want to take here is that it’s right to care enormously about these issues. Anyone looking to extend human cognition must engage just as deeply with emotion.

Wrong lessons to take from the Primer

The Diamond Age is a novel, not a research paper, so the Primer is presented as a single whole, rather than a list of system design properties. It’s too easy, then, to treat that vision as a single whole when saying “I want to build the Primer”—there are no other handles to grab onto. This is a mistake. The concept has serious and foundational flaws. So I want to be very clear about the lessons I don’t want to take from the Primer.

Authoritarianism

The Primer has an agenda. It is designed to instill a set of values and ideas, and while it’s supportive of Nell’s curiosities, those are “side quests” to its central structure. Each of the twelve “Lands Beyond” focuses on different topics, but they’re not specific to Nell, and Nell didn’t choose them. In fact, Nell doesn’t even know the Primer’s goals for her—she’s never told. Its goals are its own privileged secret. Nell is manipulated so completely by the Primer, for so much of her life, that it’s hard to determine whether she has meaningful goals or values, other than those the Primer’s creators have deemed “good for her”.

The Primer is built on a foundation of what Ivan Illich has called “our pedagogical hubris”: that is, “our belief that man can do what God cannot, namely, manipulate others for their own salvation.” Its design is not only patronizing and infantilizing, but (I believe) immoral. I wouldn’t want to be manipulated in this way, so why would I want to do this to others?

Ideology aside, there’s a central contradiction here. The Primer’s main goal isn’t to teach any specific domain knowledge. Finkle-McGraw commissioned it to develop creative, independent thinkers. The problem here is that Nell spends her entire intellectual life thinking about what the Primer tells her to think about. We never see her embark on a substantial creative project with her own initiative or responsibility. I simply don’t believe this can work. If you spend your whole life solving well-defined problems with well-defined solutions, you’re unlikely to suddenly pose interesting problems or find startling solutions.

These problems force us to reject much of the Primer’s structure. Its central spine, the story of Princess Nell, relies on Nell’s essential passivity. The same is true for the way that the Primer covertly weaves lessons into the adventure, so that Nell learns important skills without realizing she’s learning. Control must be inverted, so that while guidance and support are offered, learners have ultimate responsibility for their own agenda.

Isolation

When I ask people about their most rewarding high-growth experiences, they usually tell me about times when they were immersed in something that really mattered to them, like a startup, or an art project, or a competition. Lots of learning happened, but learning wasn’t the point: it was subsidiary to some other meaningful purpose, often pursued in community with others. By contrast, Nell is immersed in a fantasy world where the primary aim is learning. It’s isolated from any external meaningful purpose: Nell doesn’t have any, as we’ve discussed. She’s collecting trophies (keys), waiting for her real life to begin.

Worse, the Primer is also isolated from other real people. Its characters exist only to manipulate or to be manipulated by Nell. Miranda, her ractor, reads from a script and can’t even see Nell’s responses. Insofar as The Diamond Age comments on AI as a substitute for human connection, it suggests an enormous difference between Nell’s devoted experience with Miranda and the Mouse Army’s experiences with an AI-generated voice. But that misses the more important point: in both cases, the relationship is fully scripted by an AI designed for manipulation. Differences in voice acting hardly seem to matter. Outside the Primer’s world and at school, the girls do make real personal connections. But the book doesn’t involve itself in any of their authentic social interactions, other than some protective interventions in Nell’s youth.

The problems with both kinds of isolation become vivid when I think about the kind of enabling environment I want for myself, rather than what I think might be “good for” others. I’m suspicious of isolated learning—learning for learning’s sake. I most enjoy learning when it’s part of some larger meaningful activity, and when it helps connect me more deeply with other people I enjoy. That doesn’t mean I think learning needs to be utilitarian or instrumental. The passionate pursuit of a curiosity can be a larger meaningful activity, and it’s one I most enjoy pursuing with or for others. When I study philosophy, for instance, the material becomes meaningful when I debate it with others or when I bring the ideas into some vivid question in my life.

As with the Primer’s authoritarianism, we can’t change this without inverting many of the book’s central structures. Instead of isolating people within a virtual world, we must weave support into life as people participate ever more richly in the real world.

Gamification

Many of my most rewarding high-growth experiences were tremendously fun, even as they were stressful and challenging. But (particularly in school) learning is often an unpleasant grind. Many technologists have wondered: why can’t learning be fun all the time? Wonderful games like Portal and The Witness do reliably cause players to learn complex ideas without explicit instruction—so couldn’t we teach everything through games? For that matter, could we make a game so fun that people will play it without even realizing that they’re learning?

Apart from the issues of authoritarianism and isolation we’ve already discussed, I believe this aspiration cannot succeed for most topics. This perspective doesn’t take seriously how games work as a form. Advocates of this approach are satisfied with making mediocre games: games which can’t compete on their own merits as games, games which are doomed to fail without coercion.

Games are designed first and foremost to be fun—or beautiful, or engrossing, or exhilarating. Games are an aesthetic medium, and (generally speaking) they compel our participation insofar as they compel us aesthetically. It’s true that in some games, players end up developing certain skills or understandings along the way. But that doesn’t mean we can make a great game that teaches anything. You’re seeing the survivors. These games’ designers tried and discarded dozens of gameplay ideas in search of something aesthetically compelling. Then, only after they’d satisfied the primary constraint of making something fun, or beautiful, or whatever, the designers figured out how to ensure people would learn what they need as they play. Most mechanisms are not fun. Good games come from a demanding selection process which works the other way around: first, find the fun. There’s no reason at all to believe that for any arbitrary abstract topic, one can always “find the fun” which implicitly teaches it.

Even if you could make an effective educational game for any topic, you’d be competing along the wrong axis. Educational technologists often aspire to make games which are so fun that people will play them despite their educational content, or maybe even without awareness of their educational content. But if you remove coercion from the picture, and we’re imagining that people will choose to play whatever they find most fun, that means you’re in competition to out-fun Fortnite, or League of Legends, or Stardew Valley, or whatever—games exquisitely optimized to be as viscerally enjoyable as possible. This is the Primer’s strategy, and it’s only plausible because Nell’s life is otherwise so utterly impoverished of meaningful activity.

What if you don’t compete on pure fun, but on some other aesthetic value? After all, provocative artistic films and mindlessly entertaining films coexist in the market. Rather than “most viscerally fun”, you could create interactive learning environments which compete to be the most profound, or awe-inspiring, or joyfully perplexing. But even in that framing, you’re still saddled with a demanding extra constraint, one that your competitors don’t have. You need to produce an irresistible aesthetic experience while also ensuring that the player thoroughly learns some disciplinary topic. Sophisticated films and novels have carved out a successful niche, but how many of these works also manage to teach an abstract topic? It’s hard enough to make a profound aesthetic experience, and hard enough to convey an abstract idea thoroughly. Doing both simultaneously requires rare genius, no matter the aesthetic axis.

I’m absolutely not arguing that learning can’t be fun or profound. In fact, my most cherished learning experiences were more engaging than my favorite games. The difference is that those experiences weren’t structured first and foremost around an aesthetic goal—or around learning, for that matter. Those experiences were primarily about doing something that I really cared about: creating something, participating in a community, answering a question, helping a friend. Learning and aesthetic pleasures came as a happy byproduct.

I’m also not arguing that these experiences can’t be explicitly designed, or scaled to many people. Y Combinator, the startup accelerator program, uses a huge amount of authored structure to convey a set of ideas and values: talks, dinners, office hours, rituals, feedback, deadlines, and so on. This program isn’t a game; these people aren’t “playing founder”. But for most participants, these structured activities are incredibly engaging and rewarding. I claim that’s because YC’s structure is about engaging more deeply and capably in something that matters enormously to these founders.

There are many important lessons to take from games as we try to transcend the Primer. It’s possible to design interactive environments which help people understand complex ideas through simulation and realtime feedback. It’s right to care enormously about motivation and aesthetic experience. Scaffolding and dynamic support can help people reach further. Designed environments can support social connection and community participation. And so on. But I believe the central task here is to draw on all these lessons as we construct a new medium, one centered on supporting meaningful action, rather than on producing aesthetic experiences.

Discovery learning is all you need

In some of the Primer’s most vivid passages, Nell learns not by being instructed, but through inquiry and experimentation in a highly structured environment. That’s how she learned to start a fire in a passage I quoted earlier. When many technologists talk about “building the Primer”, they paint a world in which we learn primarily or exclusively through that kind of minimally guided exploration.

I understand the appeal. Inquiry learning implies hands-on immersion, rather than dull lectures. And those moments of discovery can be quite thrilling. It’s also true that you’re more likely to remember insights you produced yourself. Perhaps you’ll even understand the idea more thoroughly, since you had to construct all the pieces which produced it. Still: while I’m quite fond of discovery learning, we shouldn’t conclude that it’s all we need.

Outside The Diamond Age, this issue has been debated at great length in the educational psychology literature (see e.g. Lee and Anderson’s 2013 review or Kirschner et al’s infamous 2006 critique). I’ll quickly summarize a few of the key issues.

Discovery learning methods often ignore learners’ extremely limited working memory capacity. In many experiments, we’ve found that when people solve a problem by straining their cognitive resources to their limits, they aren’t left with enough capacity to actually learn from the problem.

Likewise, these methods often fail to consider the realities of long-term memory. We forget new knowledge in unfamiliar topics quite quickly unless it’s reinforced through subsequent retrieval. And if we’d like learners to apply new skills or to recognize new patterns fluently, they’ll need practice, or some other intervention which will let them consolidate those memories for automatic use.

Discovery learning emphasizes the concrete. That’s great for intuition-building and engagement. But much domain knowledge relies on abstraction, and it’s often not clear how to introduce that kind of understanding without explanation. When material is introduced without abstraction, we often find that students struggle to transfer what they’ve learned from one concrete case to another. Relatedly, concreteness tends to emphasize informality and imprecision. That’s often helpful initially, but the precision and power of formal notations and concepts are important to meaningful action in most fields.

It’s worth noting that the Primer isn’t pure discovery learning! When Nell is young, the book and its characters provide guided instruction through storytelling. When Nell grows old enough to read on her own, she gains access to books within the Primer, which we’re told she studies avidly. There’s an important pattern to the books. Nell usually encounters a subject first through a discovery-oriented activity (like the chains in Turing’s castle), then once she’s built intuition and meaning through concrete interaction, she finds a book which details the topic more formally and in greater depth. It’s hard to imagine her developing her skills without those books. Of course, the books don’t suffice to make Nell’s mastery totally plausible. Outside of martial arts and survival skills, there’s no mention of explicit practice, and few instances of elaboration over time. But even in the wildly optimistic world of the Primer, discovery learning alone is not enough.

Constructing a new vision

I want a new vision to chase, something which subsumes what enchants me about the Primer: hands-on immersion, vivid responsiveness, a sense of total assurance, and devotion to the emotional heart of learning and creation.

Even as I leave the Primer behind, I still see dynamic media as central to each of those properties. Interactive representations and simulations make immersion possible for more topics. Realtime feedback enables vivid responsiveness. Dynamic scaffolding and adaptive practice can help give me assurance that I can learn what I want. Adaptive metacognitive support and vibrant audiovisuals support my emotional experience.

But we do need to leave the Primer behind. It works its magic through manipulation, through isolation from people or projects with real meaning, and through a misguided reliance on aesthetic pleasures. I see no way to incrementally rescue its structure from these flaws.

Instead, I think we should invert the Primer’s structure. Our system should be designed to help us engage in the projects and interests we find meaningful, rather than to “achieve learning goals”. Our system should be woven into the world around us, where the things we care about are happening, instead of in a fantasy world confined to a screen. Our system can support engagement by helping people participate more fully in their interests, and by amplifying the intrinsic satisfaction which results—rather than by aiming for entertainment, and hoping for meaning as a side effect, as the Primer does.

Such a system would no longer be a “primer”, a sequestered lesson to complete before real participation, useful only to novices. It would be more like a general-purpose enabling environment, a tool for thought which would increase the likelihood and speed of learning-dependent action.

What form might this system take? If it’s woven into our involvement with the world, as I think it must be, it’s not quite a book anymore. It seems to be a medium in the sense that air is a medium for sound: it’s an ever-present conduit, providing the support, structure, and representations we need to do things we care about. This probably means a kind of ubiquitous computing. My sense of that shape is still quite hazy, but you can see a first sketch of what a general enabling environment might look like in my recent presentation, “How might we learn?”. Clearly, there is much more to do. I hope that this essay can help free some of my colleagues who have been likewise captivated by the Primer—and that they can help me design the system which will supersede it.

The ideas in this essay owe much to years of exchanges with Alec Resnick, Bret Victor, May-Li Khoe, Taylor Rogalski—and, especially, Michael Nielsen, whose conversation helped incubate so many of these views. My thanks to all.

My work is made possible by a crowd-funded research grant from my Patreon community. If you find it interesting, you can become a member to help make more happen. You’ll get more essays like this one, previews of prototypes, and events like seminars and unconferences.

Finally, a special thanks to my sponsor-level patrons as of July 2024: Adam Marblestone, Adam Wiggins, Andrew Sutherland, Andy Schriner, Ben Springwater, Bert Muthalaly, Boris Verbitsky, Calvin French-Owen, Dan Romero, David Wilkinson, fnnch, Greg Vardy, Heptabase, James Hill-Khurana, James Archer, James Lindenbaum, Jesse Andrews, Kevin Lynagh, Kinnu, Lambda AI Hardware, Ludwig Petersson, Maksim Stepanenko, Matt Knox, Michael Slade, Mickey McManus, Mintter, Peter Hartree, Ross Boucher, Russel Simmons, Salem Al-MansooriSana Labs, Thomas Honeyman, Todor Markov, Tooz Wu, William Clausen, William Laitinen, Yaniv Tal