Your agent is a translator
The analogy that finally explains vibe coding: a translator fluent in every language, waiting for you to know what to say.
// trace: where this idea came from
- ↳ video diary @ 7:04 (la analogía del traductor)
- ↳ video diary @ 8:01 (por qué el inglés rinde más)
- ↳ Entry 2-2: The AI does what you say, not what you mean (la regla del spec, ahora como definición del oficio)
Honest disclosure: if you sat me down to write a hello world from scratch, Python is the only language where I wouldn’t freeze. And yet I ship applications weekly. The analogy that resolves the paradox:
Vibe coding is having a translator who speaks every language: Chinese, Russian, French, all of them. You just say “tell them this.” Your entire job is knowing what to say. ▸ 7:04
This reframes what the skill is. Nobody praises a diplomat for speaking through an interpreter; they judge whether the message was worth translating. The rule from entry 2-2 becomes a job description: the profession left standing is knowing what to say, precisely, completely, and first of all in your own mother tongue. If you can’t explain the feature to a person, no translator can save you ▸ 7:45.
saber qué decir es el oficio →
Practical notes from the field
Prompt in English when you can. These models trained overwhelmingly on English; even the programming languages themselves are English (print, let, if). Instructions in English simply land better ▸ 8:01. The mirror case proves the mechanism: a language with no written corpus can’t be prompted at all. No text, no translator.
And my brother’s objection, logged honestly: he says what I do with Jules isn’t even vibe coding, it’s managing a junior programmer, task in, task out ▸ 21:40. Maybe. Either way the junior speaks every language ever documented, and the manager’s only tool is clarity…
// continued in
no entry has continued this idea yet: the arc is still open