When Vibecoding Goes Too Well

A short essay on the adventure of an economist in the world of Vibecoding and Language Learning.

I believe that the greatest impact of AI has not been in the boardrooms of multinationals or in the cubicles of corporate employees. It has been on people like you and me, who quietly think of it as our long-awaited revolutionary moment that will finally unleash our full potential. The vibe coding movement — and tools like Anthropic’s Cowork — have raised those aspirations even higher. Riding on a wave of positive vibes, pun intended, I took up the mantle myself: I decided to tackle my language learning ordeal by building a website grounded in narrative and storytelling. I ended up with a personal tool that only I use.

This is, I think, one of the major underestimated downsides of vibe coding: it works too well. It builds something so thoroughly in your own form that no one else will understand it, let alone find a use for it.

I have always found languages fascinating. When I was young, my parents and I would play a game: finding similar-sounding words between Mandarin and Arabic, languages they are both fluent it. We would laugh at how silly one word or the other sounded, and I always wondered whether it was ever on purpose — whether some funny-sounding word had ended up in a language just to get a chuckle out of foreign listeners. That game developed into a lifelong interest in cultures, histories, and etymologies. It also sparked a long pursuit of trying to learn different languages — one that kept running into the same wall.

My biggest issue was never aptitude. It was sustaining the dedication. I always found myself tired, demotivated, bored, or simply busy. I started to wonder whether the problem was the tools themselves. I tried building my own curriculum, but it turns out that almost every app and website — no matter how innovative, no matter how technology-laden — is built on the foundation of traditional teaching. Duolingo gamifies the classroom. It does not replace it.

So I built Wordpaths.

Wordpaths is a personal web application built on my premise that languages are not isolated systems but a single, centuries-long conversation. Rather than teaching vocabulary through repetition or grammar through drill, it maps the historical routes words have traveled — across empires, trade networks, colonial encounters, and cultural exchange.

The app has three sections. The Root Explorer presents words in their origin language and asks you to sit with them before revealing where they traveled. Sukkar in Arabic before you know it becomes azúcar in Spanish and sugar in English, carried along medieval trade routes. The reveal is unhurried — a deliberate departure from the gamified urgency of conventional language apps. The Language Bridges section organizes connections by historical corridor: eight centuries of Al-Andalus linking Arabic to Spanish, 333 years of Spanish colonial rule embedding itself in Tagalog, the Silk Road carrying Persian vocabulary into Turkish, Classical Chinese spreading governance and philosophy into Korean and Vietnamese. Each bridge is a named historical relationship, not an abstract pairing.

The World Journey is the heart of the app. It traces six stations — Gibraltar, North Africa, Byblos, Malta, Dunhuang, and East Asia — each representing a moment where languages collided and transformed each other. Gibraltar opens with the Arabic name Jabal Tariq embedded in a Spanish rock on British soil. Malta reveals the only Semitic language written in the Latin alphabet, its Arabic core — ħobż for bread, xemx for sun, dar for house — surviving a thousand years of subsequent conquest. Dunhuang preserves manuscripts in a dozen languages, connected by Chinese script. The journey doesn’t teach languages. It teaches how languages happen.

My approach to language learning is history-focused — I am always looking for the origin story. Language is always in flux: new words arrive, others go extinct. I even maintain a catalogue of Arabic-Spanish connections where the Arabic root words have essentially vanished from modern usage, surviving only in their Spanish descendants. Yet understanding that past — the route a word traveled, the empire or trade network that carried it — helps the learner grasp not just vocabulary but the context in which meaning developed. That is what Wordpaths tries to do.

Whether this app is the full manifestation of my self-teaching method, I cannot yet say, nor can I say for sure whether this kind of language teaching would interesting to anyone else. When I gave my wife access to the early version, she said the experience was like having a walk inside my mind, and I am not sure this is something I feel others should experience. But I am certain of this: the AI revolution is being won not in the business world but at the individual level. I use AI more and more to improve my daily life rather than my office productivity. The retail user — the person building a language app for themselves at midnight because no existing tool fits the way their mind works — is the true winner of today’s AI moment. The tool worked too well. It built something only I understand. And I am not entirely sure that is a problem.