Kyle Young, Upper School French Teacher, Hathaway Brown School

I used to have a discussion prompt for my French students: what would the perfect new app be? Recently, I asked myself: What if instead of just talking about it, they could build it? What if they could craft a working prototype in a single class period without touching a single line of code—all while working entirely in the target language? That was exactly what my students were able to do in a classroom activity for our unit on technology. The activity was surprisingly fun and communicatively rich, and I am excited to share the idea with my fellow language educators.

The activity is built around a concept that has gained traction in the tech world over the past few years: “vibe coding.” The term refers to the practice of building software by describing what you want in natural language and letting an AI translate that description into working code. Those with no technical background can now build functional applications simply by communicating clearly with an AI. I have experimented with vibe coding in some of my own projects. Once I realized that the core skill involved is the ability to express ideas clearly and precisely in language, I immediately saw the potential for the world language classroom.

The setup was relatively straightforward. I created a custom AI assistant with a specific and simple directive: respond only in French, regardless of what language the student uses. This idea could work on several different platforms, but since my school uses the Google Suite of tools, I created the activity as a Gem—a custom version of Google’s Gemini AI with specific instructions that shape its interaction with the user. In this case, I instructed the AI to help the user to build software but to respond only to commands or queries written in French. I further specified that the agent must act at all times as if it does not even understand English. Students were then asked to use this AI to build a small interactive learning tool of their choice. It could be a vocabulary quiz, a verb conjugation game, a flashcard app or anything else they thought would be useful for studying French. The only rule was that all communication with the AI had to happen in the target language.

Students were immediately motivated because the stakes felt real: if they could not communicate their vision clearly in French, the AI would not build what they wanted. Vague or imprecise language produced results that did not match their intentions, which in turn prompted them to revise, clarify and try again. This iterative cycle of expressing, evaluating, and refining mirrors exactly the kind of negotiation of meaning that is central to developing communicative competence.

By the end of a single class period, students had built working prototypes by using nothing more than their French. We concluded the activity with a brief showcase of what everyone had created and the results allowed the students’ creativity to shine. One student had come up with a trivia game where the goal was to build a pizza: for each question answered correctly, players could add an additional ingredient to the pizza, which was rendered in real time as a 3D model. Another student came up with a French-language version of a children’s word game that she used to play with her father in Turkish. Each prototype was unique to the student who had crafted it and tailored to her interests and vision.

Several features of this activity make it particularly valuable from a pedagogical standpoint. The feedback is immediate and unambiguous: the student’s prompt either results in the desired output, or it does not. This gives students a clear and honest signal about the precision of their language.  In addition, the activity is inherently differentiated. More advanced students can attempt more ambitious projects and push themselves toward more sophisticated vocabulary and syntax, while students at lower levels of proficiency can still accomplish more modest but nevertheless meaningful goals with the language skills they have. Moreover, it accomplishes two things that are very difficult to achieve simultaneously in more traditional classroom activities: students have the opportunity to produce a significant amount of written output while remaining in a genuinely dynamic interaction.

There are, however, some practical considerations worth noting. The activity works best at more advanced levels where students have enough vocabulary and grammatical range to describe a project with some degree of specificity. Using a vibe coding activity with lower-level students would likely require some significant modification, such as providing vocabulary scaffolding or modifying the AI’s instructions to allow it to provide more language support. Also, as with any AI platform used with students, it is worth checking with your school’s IT department to ensure the tool you plan to use meets your institution’s data privacy requirements.

Finally, while this activity does fit nicely into a unit on technology, it is not a coding lesson, and students will not learn any new programming skills. Nevertheless, it does help develop some AI literacy skills by familiarizing students with the current capabilities of the technology and some of the new ways people are using it. More than anything, however, students will emerge with a concrete experience of their language as a functional tool capable of producing real results in the real world. I cannot wait to see what else my students can create with the power of their words.