GlobeskoolGoing back to 15th-century pedagogy to survive AI — smarter than it sounds
School was built around cheap printed answers, and AI just made those answers free. A thread making the rounds argues the fix is older than the printing press: learning a child has to reason through and defend out loud. Here is why that is sharper than nostalgia.
There is a thread making the rounds, with a long and unusually thoughtful discussion behind it, that makes a strange-sounding claim: to survive AI, education may need to go backward. Not back to the 1950s, or even to the chalkboard. Back to the years before Gutenberg, when a printed answer was not something you could simply buy or borrow.
The argument is that the printing press created the kind of learning we now take for granted. Cheap, copyable text made it possible to learn by absorbing answers someone else had already written down. School organized itself around that fact. Read the chapter, memorize the result, reproduce it on the test. For five centuries that was a reasonable bargain, because producing a clean written answer still required you to understand something.
AI breaks that bargain. A machine can now produce the clean written answer in seconds, with no understanding behind it at all. So the formats that quietly assumed a good answer proves a working mind stop proving anything.
The oldest formats are the AI-proof ones
This is the part worth sitting with, and it is where we think the argument is right. The learning formats that AI cannot quietly do for a child are the ones that predate cheap print entirely.
Reasoning out loud, in real time. Defending a claim when someone pushes back. Making something in front of another person and answering for the choices. Being questioned, on the spot, by someone who already knows where the weak spots are. This is roughly what a tutor and a student did in 1450, because it was the only method available. It is also, not coincidentally, the method no chatbot can sit in for.
You cannot outsource a conversation you are having to your face. You cannot have an AI quietly defend your reasoning when a person looks at you, asks why you did it that way, and waits. The format itself is the safeguard. That is the part that makes this more than nostalgia.
This is not a romance about the past
It would be easy to read all of this as “the old ways were better,” and that is not the claim, and it is not ours either. The old ways were slow, narrow, and available to very few children. Most people never got a tutor asking them hard questions across a table. The printing press, for all the trouble it is causing us now, put knowledge in reach of ordinary families. We are not arguing to give that back.
The point is sharper. Certain learning formats are AI-proof by design, not by accident, and we abandoned a lot of them because they did not scale to a classroom of thirty. The constraint that pushed them out was logistics, not value. And the durable skills underneath them are the oldest human ones we have: thinking clearly, saying what you mean, holding a position under pressure, making something real, and using judgment about a tool instead of being managed by it.
Those are exactly the skills that still matter on the other side of all this. Not because they are old, but because a machine that can generate any answer has made the answer itself nearly worthless — and the reasoning behind it the only thing left worth grading.
What this looks like for an eight-to-sixteen-year-old
This is where the 15th-century framing stops being a curiosity and starts being a plan. You do not need a philosophy seminar. You need the two ingredients the format was always built on: a real thing to make, and a person who will make the child explain it.
That is the shape of how Globeskool teaches. One child, working on an actual project, then explaining the thinking out loud to someone who asks the next question. Why this approach and not the other one. What you would change. Where it breaks. The AI tools are in the room, and the child learns to drive them, but the work that gets examined is the child’s own reasoning, spoken and defended, the way it would have been long before any of this was printable.
It turns out the format that is hardest for a machine to fake is also the one that tells you the most about your own child — how clearly they can already reason through something, and whether they can hold it up when someone asks. That is the first thing we look at.