I Built an Esperanto Game with AI—Then the Vocabulary Became Harder Than the Code

 A few months ago, I wondered if I could build an Esperanto vocabulary game with AI.

I wasn't a programmer. I was a first-year AI student in my 50s—but I had been learning Esperanto for more than ten years.

So I combined something I knew well with something I had only just begun learning.

Today, that small experiment has grown into an English–Esperanto vocabulary game I can share with learners around the world.

Wait... I Made a Web Game?

The project started with Gemini Canvas.

I asked AI to help me create a simple vocabulary game, and much faster than I expected, I had something working on the screen.

My reaction was basically:

Wait... I made a web game?

Then I actually started using it.

Buttons needed fixing. Mobile usability needed work. Sounds and interactions had to be tested.

My “simple game” wasn't so simple anymore.

The Real Problem: 2,300+ Words

I thought coding would be the hard part.

I was wrong.

The vocabulary became harder than the code.

Languages don't match perfectly word for word. One word can have several meanings, and in a quiz, two answers can sometimes both look correct.

The code can work perfectly.

The question can still be bad.

Eventually, I was working with more than 2,300 vocabulary entries.

AI helped me organize, compare and convert the data, but I still had to review the results as an Esperanto learner.

That became the biggest lesson of the project:

AI helped me build the prototype quickly. Making the content trustworthy required much more human attention.

From Korean to a Global Version

As the game grew, so did the AI tools I was learning.

Gemini Canvas helped with the first version. I later experimented with Google Antigravity while improving parts of the interface and structure, and used ChatGPT extensively while reorganizing and converting vocabulary data.

Then I asked myself:

Esperanto is an international language. Why keep the game only in Korean?

So I began building an English–Esperanto version.

It wasn't just translation. I had to check meanings again, deal with ambiguous pairs and duplicates, and reorganize the vocabulary.

I also created a 547-word beginner vocabulary set for the global version, rather than simply moving the entire 2,300+ Korean collection into English. I wanted new learners to have a more manageable place to start.

Somewhere along the way, my little coding experiment had become a real learning tool.

Why I'm Sharing It

I've studied Esperanto for more than ten years, and many people have helped me along the way.

Now I'm a beginner again—this time in AI.

I still don't consider myself a programmer. But this project taught me something I didn't expect:

I didn't have to learn everything first and then build something. Building something I cared about became the way I learned.

And now the game that started as my AI experiment can be shared beyond Korea.

For me, that's the best part.

I started building the game to learn AI.
Somewhere along the way, I built something I could give away.

🔗 And this game is only one part of how Esperanto has become my personal laboratory for learning, building, and creating with AI.