Writing Articles in English with AI as a Spanish Speaker: An Experiment

I was born in Buenos Aires, Argentina, and although I started speaking English from a relatively young age (around 6 to 8 years old), I never lived in an English-speaking country. The story might have been different had I spent at least a few years of my childhood in a country that forced me to speak English in public, but I never felt that my fluency came close to that of my mother tongue. While my English level is estimated at C1, both my ability to articulate complex ideas and concepts and the breadth of my vocabulary are severely limited when compared to their counterparts in my native language.

Although I never devoted much of my time to writing, I always liked reading, and I always liked the idea of writing. As a programmer, I've written several technical articles in English over the years, but I always felt that "something was missing" when I compared my writing to that of the native-speaker leading voices of the industry. I would spend huge amounts of time perfecting short texts, no more than 500–800 words, yet without ever finding the secret to writing long, coherent pieces. This changed somewhat when artificial intelligence agents entered the game, letting me lay out some of my ideas with a lyricism and prose previously forbidden to my natural ability to write in a language that was never entirely "mine." But the advent of AI has turned us all into copies of the same collective author: the same word choices, the same style, the same way of writing. Yes, we've gained a tremendous capacity to produce words in industrial quantities, and it's become far easier for someone like me to generate long, coherent articles, but all of this comes at a very steep cost: authenticity.

A possible solution

In a somewhat different context, I recently came across a piece of advice for interacting with artificial intelligence agents more efficiently: write your prompts in your mother tongue, so that you can express yourself with greater precision. While I'm not sure whether this holds true as far as the effectiveness of a prompt in English over a prompt in Spanish goes, especially for programming, it struck me as a curious idea to carry this over to writing itself: if I could express my ideas in Spanish, and then work with an AI agent to translate them into English, perhaps the final result would be more faithful to what the article wanted to be and to express when it existed only in my head.

This article is the first experiment in a string of texts I intend to write the same way: zero use of AI, man-hours devoted to refining the concepts I'm trying to express, completely natural and organic word choice. Once the piece is finished in Spanish, I'll work with Codex or Claude to translate it so that every concept, every paragraph, and every word is properly translated, in a deliberate way—and not as some poorly supervised vomit of words that wants to say something but never quite says it.

Will it work? I really don't know. But at this very moment, we're all experimenting with our agents to achieve what still seems impossible: honest work, refined with the help of agents, but for and by humans above all else.