Skip to main content
Globeskool
← The Journal

Karpathy is building AI+Education — what parents should actually care about

Andrej Karpathy is building in AI and education, and it set the conversation alight. The real question for parents is not whether AI is coming to learning, but what pedagogy wraps around it.

A while back, one of the most respected names in artificial intelligence put out a short announcement: he was starting a company in AI and education. Andrej Karpathy is not a celebrity in the usual sense. He helped lead AI work at Tesla, he was part of the early team at OpenAI, and the people who build these systems tend to listen when he talks. So when he said he was launching an AI-education company, the tech world paid attention, and the comment threads filled up with people arguing about whether the thing would actually work.

We think most parents reacted to the headline and missed the part worth caring about.

The headline most people heard

The easy version of this story is “AI is coming to school.” That version is not wrong, but it is not useful either. AI has been creeping into classrooms for a while now, mostly in the form of chatbots that write the essay, solve the problem set, and hand a child the answer with none of the thinking attached. If that is what AI in education means, parents are right to be wary. A tool that does the homework does not teach anyone anything.

But that is not what Karpathy seemed to be describing. As we read his announcement, the idea was closer to a teacher-plus-AI partnership: a human expert designs the real substance of a course, and the AI works as a patient assistant that walks alongside one learner, at that learner’s pace. The first thing he set out to build was reportedly not a homework machine but a course where students build a working AI system themselves, from the ground up. The learner does the hard part. The AI keeps them company while they do it.

We want to be careful here. We are reading from a public announcement, not from inside the company, and plans change. Karpathy has since taken on other work and has described the education project as something he intends to return to rather than abandon. So treat the specifics lightly. The direction is what matters.

The question that actually counts

For parents, the useful question was never “is AI coming to education.” It plainly is. The useful question is: what kind of teaching wraps around it.

Drop a capable AI into a school built on worksheets and a bell schedule, and you mostly get faster worksheets. The machinery of the old model stays in charge, and the AI becomes one more thing that keeps thirty children moving through the same material at the same speed, whether it fits them or not. That model was designed for a world that needed large numbers of people to do roughly uniform work. That world is fading, and the school built for it has not noticed.

The more interesting use of AI is almost the opposite. Not one lesson for thirty children, but a different path for each child. AI is genuinely good at something a human teacher with a full classroom cannot do: it can notice exactly where one specific child is stuck, hold their pace, and circle back to the gap that was quietly left behind two years ago. As an adaptive engine, that is real. As a replacement for thinking, it is a trap.

What we have been building toward

This is the direction Globeskool has been working in, so we have a stake in saying it plainly. We teach one child at a time. We use AI as the adaptive engine that meets a child where they actually are, and we keep real projects as the substance, because building something is where critical thinking, problem solving, communication, creativity, and judgment about technology are actually formed. The skills that hold their value are the ones a machine cannot simply hand you.

There is a quieter test underneath all of this. In a good setup, the child is the one in charge of the tool, asking it, directing it, deciding when its answer is good enough. In a bad one, the tool is in charge of the child, feeding them just enough to keep them clicking. Same technology, opposite outcome. The difference lives almost entirely in how it is taught.

If you are trying to work out which side of that line your own child is on, that is a reasonable place to begin.

Find out where your child stands today →