Why I built Margin.

I’m a high school senior. I run our school’s machine learning club. Roughly sixteen students show up to bi-weekly sessions, voluntarily, because they’re curious. That’s the audience this whole thing is built for.

The courses I learned ML from either skipped the math or skipped the code. The math-heavy ones treated the implementation as obvious; the code-heavy ones treated the math as a black box. I needed both, in the same window. So I’m building both, in the same window.

The shape of the thing

Sixteen sessions across the school year. Each session is a lesson — eight to twelve minutes to read — and a Lab, where you write code that runs in your browser, watch a real model train, and try to break it on purpose. Pyodide for Python, TensorFlow.js for the networks, a small AI tutor that gives you Socratic hints when you’re stuck.

No leaderboards. No XP. No streaks. The research is unambiguous on this — heavy gamification harms students who are already curious. Margin is built around explorable explanations, not arcade mechanics.

What this page will become

A real essay, written over the summer as the platform comes together. For now, this is the stub — the placeholder that says this thing is being made carefully and on purpose.