How I build learning games without a game engine

The learning games I build at work are small web apps, not engine projects. Why that works, and why each game gets its own copy of the code.

Abdullah Abdul Samad, . A 1 minute read.

At work I build learning games for the MissionX AI Textbook, short missions that teach AI ideas to school students up to grade 12. You can play them there. This is about the two decisions that made building a lot of them manageable.

No game engine

None of the games use a game engine. Each one is a small website built with React and TypeScript.

That sounds like the wrong tool until you look at what the games actually do. A student reads a short story, makes choices, drags things into place, watches a robot act on their rules and gets a result. That is buttons, cards, drag and drop and a bit of animation. The web does all of it, loads quickly on a school tablet, and every developer already knows it.

Copy, don’t share

The rule I would argue for hardest is that games never share code. To start a new one, I copy the closest existing game and change the story, the art and the rules.

Shared libraries feel tidier, but a shared component that changes for one grade can quietly break a game for another that nobody has opened in months. With copies, a finished game stays finished, and upgrades happen one game at a time, when they are worth it.

What AI changed

I build these with AI coding tools, and a consistent structure is a big part of why that works. When every game follows the same written rules, an assistant can be pointed at the guide and an existing game and produce something that fits on the first try.

That leaves my time for the parts that need judgement: whether a level actually teaches the idea, and whether a ten-year-old will understand the instructions without help.