Ten Years Ago We Flipped Education. In the Age of AI, What Do Children Still Need to Climb?

TL;DR — Ten years ago we talked about flipping education. Now the world is talking about AI literacy: UNESCO has released a student AI competency framework, and countries are packaging AI literacy into courses delivered by the hour. That is a good thing, and it has limits. An education experiment from a decade ago has a question for it: Is there a syllabus for collaborative literacy? Can systems thinking be scheduled? These capacities only grow inside real tasks. Flipping opened the door. What children still need to climb in the AI era is figuring out how to actually grow that literacy, in a world where answers have gotten cheap.

Ten years ago, the hottest word in education was “flip.” Turn the teacher’s lecture into student-directed learning. Turn the textbook into a project. Turn the classroom into the world. That was also when I started guiding children along a less conventional educational path.

Ten years later, the hottest phrase is “AI literacy.” And this time, the world is moving faster.

Countries Are Scheduling AI Literacy

In 2024, UNESCO released a student AI competency framework, breaking AI literacy into 12 competencies across four dimensions: human-centered mindset, AI ethics, AI technology and applications, and AI systems design. The framework puts critical engagement at the center, not just tool operation.

Policy followed quickly. Some countries folded AI general education into primary and secondary school, mandating minimum hours per academic year from elementary through high school. Others set national targets — AI education systems spanning every level of schooling, built out by 2030. Behind all of these moves is a shared assumption: literacy can be curricularized, standardized, and delivered by the hour.

That direction makes sense. Curriculum-based delivery is the only path that scales. It keeps AI education from becoming a privilege of well-resourced schools and extends reach to rural and disadvantaged communities. When a country needs millions of children to have access to AI education, there is almost no alternative.

Still, the decade-long education experiment I ran left me with a different intuition about this.

Is There a Syllabus for Collaboration? Can Systems Thinking Be Scheduled?

My experience is this: some capacities cannot be put on a timetable.

Judgment, collaboration, systems thinking — the very things named in every future-of-work report — do not come from completing a course or passing a test. They require context, friction, and a shared goal that simply has to be met. In “Four Kids, One Summer, One Website That Had to Launch”, I wrote about students who genuinely developed systems thinking only after spending two weeks stuck in front of a blank page, pushed by a real business problem until something clicked. That clarity couldn’t be handed to them in a lesson.

Collaboration is the same. You can teach “five principles of teamwork” in a classroom, and students will absorb them. But a student actually learns to collaborate inside a real project — after arguing with someone who sees things differently, redividing the work, watching someone nearly quit, and still finishing together. That capacity is not taught into existence. It is pressed into existence.

So when I see designs like “8 hours of AI literacy per school year,” I respect the ambition — and I worry. A classroom can deliver knowledge. The rarer, harder-won practical capacities are something a classroom cannot deliver.

This Is the Old Problem of Scale vs. Depth, Wearing New Clothes

Task-based approaches have their own limits.

Real tasks depend heavily on adult time and family resources. In “The Hidden Ledger of Education Innovation”, I calculated that a single summer’s real-task project required more than 140 hours of adult involvement behind the scenes. That density is not something every family or school can sustain. If the argument stops at task-based learning without acknowledging curriculum-based delivery, good education becomes a privilege for the few.

Curriculum-based and task-based approaches are two sides of the same enduring problem: scale or depth? Curricula serve scale; real tasks serve depth. The hardest challenge in the AI era is not choosing a side. It is finding ways to connect the scale that curricula can reach with the depth that real tasks produce — so that more children who gain access to AI education also have a genuine chance to grow their capacities through something real.

Flipping Opened the Door. Climbing Is the Work of the AI Era.

Back to the two words that anchor this. Ten years ago I talked about flipping. Now I find myself thinking more about climbing.

Flipping is about opening a door. Children are no longer defined solely by a standard curriculum, and AI has made knowledge and answers unprecedentedly cheap. That is a genuinely good thing.

But climbing is what happens after the door is open. When answers are everywhere, what becomes scarce shifts: clarifying a fuzzy problem, collaborating with someone who thinks differently, integrating across uncertainty and disorder, taking responsibility for your own choices. AI cannot replace these yet. They are also exactly what real tasks grow.

So what do children need to climb in the AI era? More than learning to use tools — tools will keep changing. What they need, in a world where answers are cheap, is to grow the capacities that AI will not devalue: judgment, collaboration, accountability, and the will to turn an idea into something real.

In the overview essay “From Flip to Climb”, I argued that the core of education is bringing children into contact with the real world. AI has not changed that. Literacy still only grows slowly through practice. Techniques go out of date, but literacy that has genuinely formed becomes character — something that stays with a child for a long time, and can shape an entire life.