The Billionaire Who Said He'd Made Mistakes

It was a podcast that did it. I heard the founder of Alpha School being interviewed, and he genuinely, obviously cares about education. Sure, he had an ego, the sort you can imagine a billionaire would have. But he was also willing to say he's made mistakes and will make mistakes in the future.

That stopped me. Not the two-hour learning claims, not the growth data, not the vision of AI tutors. The admission of error.

Education reform has a graveyard, and it's full of confident people. Every decade or so someone arrives with total certainty, a new structure, a new acronym, a new silver bullet, and the certainty is precisely the problem. When you're certain, you stop measuring. You stop listening to the kids in front of you. You defend the model instead of the students. Teachers know this pattern intimately because we've lived through wave after wave of it.

What I heard on that podcast was different. It didn't seem like a system based on ego or some billionaire's pipe dream. It sounded like someone treating a school the way a good engineer treats a prototype. Try it, measure it, be honest about what broke, fix it, try again. Say out loud that you'll get things wrong. That's not weakness. In teaching we have a name for it: reflective 

practice. It's what we ask of every graduate teacher, and it's almost never what we ask of the people redesigning entire systems.

And the Alpha model deserves scrutiny. It's expensive, it's young, and education authorities in the US have fairly described the approach as untested. Which is exactly why the humility matters so much. An untested model run by someone who admits mistakes can improve. An untested model run by someone who can't is just an expensive experiment on other people's children.

I think this aspect of both boldness and empathy is key to what is needed at the moment, and that is what I'm trying to achieve with PrimarAI. It's why I published the whole model openly, every assumption, every structure, under a licence that lets anyone pull it apart. I'd rather invite investigation than seek approval. I've made mistakes building it and I will make more. The trick is to build a system that expects that, the same way we teach kids that getting it wrong is how you learn.