Class of one

Inspiration

Most schools don't have a computer science teacher, and even fewer show students that fields like cybersecurity, networking, or quantum computing are even careers you could go into. You can't choose a path you never knew existed.

Also, AI chatbots hand you finished answers, which feels helpful but usually means you learn nothing. A real teacher makes you work for it. I wanted to build that teacher.

What it does

Class of One is an AI computer science teacher for students whose school doesn't have one. It won't just give you the answer, it asks what you tried and nudges you instead. It covers nine fields, not just coding: cybersecurity, networking, AI, algorithms, databases, game dev, quantum computing, and more. It runs real Python right in the browser (no installs needed, works on a school Chromebook), remembers who you are, can talk out loud, and gives you a roadmap of projects, next topics, and free courses once you finish a field. No login required, works on any device.

How I built it

Plain HTML/CSS/JavaScript, no framework, so it loads fast on weak devices. A small serverless function on Vercel talks to Google Gemini and keeps the API key private. Python in the browser runs through Pyodide. I built it in stages during the week, starting with one coding path and expanding from there based on what a self-taught student would actually be missing.

Challenges we ran into A CSS rule meant for a big diagram accidentally sized a tiny icon to 300x300 pixels. One character fix, but took a while to track down. My "is the AI connected" status check got stuck forever because it sent an empty message to Gemini, which rejected it, and I never handled that case. Google retired the Gemini model I was using in the middle of the build. Had to catch it from the API's own error message and swap models. The sidebar just disappeared on mobile with no way to reopen it. Had to rebuild it as a slide-out drawer. Accomplishments that we're proud of

Everything in it actually works, not just the parts that make a good screenshot. Live AI, 46 real lessons, a working Python sandbox, voice, memory, and it still loads with no login on a locked-down Chromebook, which was the whole point.

What we learned

The hardest part wasn't the AI, it was deciding what the AI should refuse to do. And accessibility isn't one feature, it's a lot of small choices (no login, fast load, works on old hardware) that all have to hold at once.

What's next for Class of one

More free resources, a couple more interactive widgets, maybe a way for students to see what others built. Same goal either way: no student locked out of computer science because their school never opened the door.