Fourteen kids trained racing AIs in one afternoon

On a July afternoon at Wright State University, fifteen teenagers sat down at lab machines expecting a lecture about artificial intelligence. They did not get one. Instead, each of them got a small, untrained AI and one instruction: teach it.

The seat swap

Kids this age are drowning in AI talk. It writes their captions, finishes their sentences, and hovers over everything like weather. What almost none of them have felt is the other side of the glass: being the one the machine learns from. That was the whole objective of the camp. Not AI as a magic trick to watch, but AI as a student to coach.

Nobody in that room watched a demo. The AI watched them.

Three games, one lesson

The camp ran as an arcade with three missions, and each one is the same idea wearing a different costume. In Doodle Dojo, you draw and the machine learns to recognize your drawings: teaching by example. In Inbox Defender, you sort real-looking mail into scam, safe, and sale, then your filter sits a twenty-message exam alone: teaching by judgment. And in the AI Grand Prix, you drive laps around a 3D circuit while a rookie AI studies your driving, then trains generation after generation to beat you: teaching by demonstration.

The third seat was the hardest to build. Before the Grand Prix earned its place, I built and killed three other versions of it: a virtual pet you trained with treats, and two racing games that worked perfectly and bored me anyway. Boring is a bug. A kid who is bored learns exactly one thing, which is to stop paying attention.

Race day

The finale put every car on the same track at once. Fourteen students raced the field that day, each car carrying its trainer's name and its trainer's habits, because that is what training data is: your habits, made executable. The projector showed the whole swarm at helicopter height. The room did not sound like a classroom.

The podium came down to fifteen points out of a thousand: 992, 989, 987. Three cars, three drivers who had never written a line of code, separated by the quality of their teaching.

The camp never closed

The workshop ended when the afternoon ran out. The Grand Prix did not. The fourteen cars from camp day still hold the track as the resident legends, and the arcade is now in open play: anyone can walk in, train a car, and join the field for a week. The leaderboard has never been reset. Some of those kids are still on it.

That was the experiment: skip the lecture, hand over the chalk, and let a room full of teenagers find out that a model is just a story about what comes next, and that they are the ones telling it.