A Brief History of How Humanity Has Always Taught Itself

Every few thousand years, humanity figures out a completely new way to learn.

This has nothing to do with a better textbook. It’s not about a new teaching method. Not an updated curriculum framework. It is something more fundamental than any of those.

It’s a shift in the basic infrastructure of how knowledge moves from one human mind to another. A change so deep that it rewrites what learning is, who gets access to it, and what it costs to transmit everything a civilization knows to the generation that will carry it forward.

We are living through the third of these shifts right now (I know, it sounds extreme, but stay with me).

To understand where we are going (and why it matters so much which direction we choose) it helps to understand where we have been. Because each transition in the history of human learning gained something extraordinary. And each one gave something up.

The thing that was given up is what we are now in a position to actually recover.

Learning 1.0: Knowledge lived in people

For the vast majority of human history there was no writing. There were no books, no schools, no curricula, no standardized tests (I know, sounds nice). There was no separation between knowledge and the person who held it.

Knowledge was people.

One of the stories I like to tell about Learning 1.0, is something I saw in National Geographic years ago.

Aboriginal Australians navigated a continent seven million square kilometers wide using what they called songlines. These were routes encoded in song, in which every verse corresponded to a landmark, a direction, a story about what had happened at that place and what it meant.

So, any person who knew the songs could travel hundreds of miles through country they had never physically visited. Researchers working with Aboriginal elders/leaders found that some of these songlines encoded accurate geographical information stretching back more than ten thousand years. These were really connected to landscape features changed by rising sea levels at the end of the last Ice Age (mind…blown).

Ten thousand years of reliable geographic data, and it was all stored in songs. Passed down through relationships, repetition, and the specific focus of attention that one human being can give another.

The oral traditions of Learning 1.0 were not “primative” in the way we might think about it looking back.

They were very sophisticated technologies for encoding, storing, and transmitting knowledge that were refined over millennia. Each shaped by the simple fact that anything which didn't work was forgotten, and anything which survived had to work exceptionally well.

What made them work was a set of features that we have largely stopped noticing, because we gave them up so completely.

Learning 1.0 was personal. The elder didn’t teach the same content in the same sequence to every apprentice. She watched. She calibrated. She understood that this you needed to hear something in story before you could hold it in sequence, and that you needed to walk a path before you could carry its meaning. The instruction was shaped around the learner because there was no other option.

A simple fact that shaped Learning 1.0: When knowledge lives only in people, you cannot afford a system that loses half its students.

Learning 1.0 was also contextual. The Aboriginal elder did not teach the songline in a room. She walked it. The plants along the path were part of the lesson. The sound of the wind at a particular ridge was part of the lesson.

Knowledge and the conditions under which it was learned were inseparable because the brain stores them together. Modern neuroscience has a name for this called embodied cognition.

The oral tradition didn't know the name. But, it had built a practice around the reality for fifty thousand years.

Learning 1.0 was emotional. The great oral traditions are stories structured around conflicts, real-world consequence, various transformations, and of course, meaning. These cultures had discovered, through long experimentation, what neuroscience has since confirmed about emotion. The stories that survived in oral tradition survived because they were designed to be felt and to be remembered, and those were the same design.

And, finally, Learning 1.0 was relational. You did not learn from a text that you did not have a relationship with. You learned from a person who knew you (and who had watched you), who had their own relationship with the knowledge they were transmitting, who gave you that knowledge wrapped in a relationship that was itself part of what made it stick.

The research on what makes learning powerful has been converging, for decades, on this exact set of features. Personal. Contextual. Emotional. Relational. The oral tradition had all four baked structurally into its design.

It also had a fatal limitation.

It could not scale.

Everything powerful about Learning 1.0 depended on presence, attention, and human relationship. These have always been the scarcest resources in existence. The songline that was broken through death or displacement was gone in a way that no book can be gone.

Learning 1.0 was powerful but also extremely vulnerable…

Learning 2.0: Knowledge escaped the body

Around five thousand years ago, in Mesopotamia and Egypt and China and the Indus Valley, human beings began to write things down.

This was a change so fundamental that it rewired the relationship between knowledge and the human mind. For the first time in history, knowledge could leave the body and could be stored in clay and papyrus and paper. It could travel and survive the death of its author.

A book written in Alexandria in the third century BC could teach a student in New York in the twenty-first century AD. This is no small feat, and quite frankly, is the miracle on which the entire modern world that we now know was built upon.

In the fifteenth century the Printing Press scaled this miracle. The scientific revolution, in the sixteenth and seventeenth centuries, systematized it. And the expansion of “compulsory” schooling in the nineteenth century democratized it in ways we never imagined.

By 2020, global literacy rates exceeded 87 percent. In 1800, they were below 15 percent. This is among the most significant expansions of human capability in history, and it happened through the creation of Learning 2.0.

Learning 2.0 brought us the textbook, the trained teacher, the standardized curriculum. the school year, and the grade level.

These were the specific organizational forms that made knowledge transmissible at scale and that allowed the miracle of writing to reach not just the scribes and scholars and wealthy families who had always had access to it, but eventually almost everyone.

Learning 2.0 was a genuine triumph. It also made a trade. One that we can feel as students, educators, and parents (I have five kids, so believe me on this one).

The trade was simple. In order to deliver knowledge at scale, you had to make it generic.

The textbook could reach everyone because it was written for no one specifically. The standardized curriculum could be delivered to millions of students because it was calibrated to the “average” student. The school, as an institution, was organized not around what each learner needed but around what could be delivered efficiently to all learners simultaneously as orderly as possible.

We lost the personal connection. Learning experiences often became transactional. The relationship became a distraction from the serious business of content delivery.

We built, in other words, a system that was extraordinarily good at transmitting knowledge broadly and structurally incapable of seeing the specific child in front of it.

If you are a teacher, you’ve felt this before. Trying to make connections with 120+ students on my roster was one of the most difficult aspects of being a teacher.

The science of learning has known for decades what Learning 2.0 was getting wrong. Contextual embedding. Emotional engagement. Metacognitive development. Spaced practice. Contingent feedback. The research has been pointing in one direction for fifty years.

But the tools available for mass education made this effectively impossible. You cannot be specifically contingent on thirty students simultaneously. You cannot build and hold a detailed model of each learner's understanding while also managing a classroom and delivering a lesson. The science knew what was needed. The Learning 2.0 infrastructure could not provide it.

Until now.

Learning 3.0: The system learns the learner

We are at the beginning of the third shift in human learning.

Like the transition from oral tradition to writing, and from manuscript to print, this one makes things possible that were structurally impossible in the previous era.

For the first time in the history of mass education, a learning system can hold a detailed, accurate model of an individual learner.

It can understand a specific child, with a specific cognitive profile, a specific history of understanding and misunderstanding, and a specific set of contexts in which learning opens up and closes down.

The AI learning systems being developed and deployed right now (with all their limitations, with all the caveats that honest engagement with the evidence requires) can do something that neither the oral elder nor the classroom teacher nor any prior educational technology has been able to do at scale.

These systems can see each learner individually, continuously, with a precision that accumulates over time.

You could always do this on-one-one (even if you missed some things along the way), but whenever we went past that it became impossible.

These systems can notice a four-minute pause. They can recognize that a student's errors cluster around a specific conceptual gap rather than general confusion. They can identify that understanding arrives through story rather than symbol, through concrete before abstract, through context before principle. T

They can even track the specific pattern of a learner's engagement and notice, with statistical sensitivity that no human observer managing thirty students can match, the moment when something has shifted.

And (this is the part that I think matters most, and that most current deployments are not yet getting right) they can give this information back to the learner.

AI systems tell the learner what the system has noticed about how their mind works. Give them the map of themselves that the oral elder gave the apprentice and that the classroom has never had the tools to provide.

This is the restoration that Learning 3.0 makes possible. It is a recovery of what was traded away to achieve that miracle of scale. It brings back the specific, accurate, individualized recognition that every human mind deserves and that the standardized system was just unable to ever provide.

The Trade That Each Era Made

Learning 1.0 had depth but no scale. It could see each learner with extraordinary precision and reach almost no one.

Learning 2.0 had scale but no depth. It could reach almost everyone and see no one specifically.

Learning 3.0 is the first era in which both might be possible simultaneously. It’s the first era in which a learning system could, in theory, reach every child and see every child.

Well, at least in theory...

In practice, whether Learning 3.0 becomes the tool that finally closes the gap between what the science of learning knows and what educational systems do, or whether it becomes a more sophisticated version of the sorting machine that Learning 2.0 built…depends entirely on choices that are being made right now.

Plenty of folks are leaving Learning 2.0 systems and going back to Learning 1.0. Homeschooling is on the rise. Device free schools on the rise. Tutoring on the rise.

But, the masses are stuck in a Learning 2.0 world that has left so many without the instruction and learning experiences they need.

Right now choices are being made on what the AI systems optimize for. There are choices being made about who has access to the most powerful tools, and choices about what role human teachers play in a world where machines can do the things that machines can now do.

These are the oldest questions in education, arriving in a new form.

What is learning for? Who is it for? What do we owe each other as learners?

I have been living with these questions for years. I have read the research, visited the classrooms, talked to the teachers and the technologists and the policymakers and (most importantly) the students.

I have to be honest. I’ve tried to understand not just what Learning 3.0 is, but what it should be. What it must be, if it is going to be worth building. And, it’s not easy.

The answer I have arrived at is more of a learning answer, than it is a technology answer. And it begins with a single principle that the oral tradition understood and that every subsequent educational era has struggled to honor at scale.

The learner is the point.

Not the curriculum. Not the institution. Not the credential or the test score or the metric. The specific human being who is trying to understand something is the point.

That is what learning has always been, at its best.

That is what Learning 3.0, built right, could make available to every child on earth.

The distance between could and will is what the rest of this series, and the book LEARNING 3.0 I’m writing is all about.