In this article
The day "vibe coding" got a name
On February 2, 2025, the AI researcher Andrej Karpathy posted a short note that named something a lot of developers had quietly started doing. "There's a new kind of coding I call vibe coding," he wrote, "where you fully give in to the vibes, embrace exponentials, and forget that the code even exists." You describe what you want in plain language, the AI writes the code, and you steer by feel. He later called it a throwaway tweet. Within a year it was an industry term.
The same shift is now happening with words. Call it vibe writing: you bring the intent and the meaning, the AI helps produce and shape the English, and you steer the draft instead of typing every sentence from scratch. This post is about what that actually means, why it fits non-native English writers better than almost anyone, and how to do it without ending up with generic text that is not really yours.
So what is vibe writing?
Vibe writing is the writing analog of vibe coding. You supply the what — the idea, the argument, the thing you are trying to say — and the AI helps with the how — fluent wording, structure, grammar, tone. You are still the author. You just stop spending all your energy on the mechanical layer and spend it on judgment instead.
It is an emerging term, not a fixed one. Writers have started using it to describe drafting "by energy, tone, and rhythm" rather than grinding out each sentence. The useful core, borrowed straight from coding, is the division of labor: the human owns intent and judgment; the machine accelerates execution.
Why this matters more for non-native writers
Here is the part that rarely gets said. For native English speakers, vibe writing is a convenience. For non-native writers, it removes a bottleneck that was never about thinking.
If English is your second or third language, your ideas were never the problem. You can reason, argue, and analyze perfectly well in your first language. What slows you down is execution: the article you dropped, the collocation that came out wrong, the sentence that is grammatically fine but reads as translated. Vibe writing separates the idea from the English, so the language stops gatekeeping the thought.
The data backs this up. A 2025 arXiv preprint analyzing over two million biomedical papers found that AI-assisted writing grew far faster among non-native English speakers than native ones after ChatGPT arrived — roughly a 400 percent rise in non-English-speaking countries against about 180 percent in English-speaking ones — and that it is measurably narrowing the publication gap between native and non-native English-speaking scientists. The tool helps most exactly where the language barrier was highest.
The bottleneck was never your ideas
This is worth sitting with, because it reframes how you should feel about using AI to write. Using help with English is not cheating your way around a thinking problem. The thinking is yours. Grammar, idiom, and register are execution — the layer a tool can genuinely take off your plate without taking anything that matters.
That reframing also tells you where to keep control. The meaning, the argument, the specific point only you can make — those stay with you. The moment the AI starts supplying the ideas instead of the English, you have stopped vibe writing and started outsourcing.
The right way: you steer, AI drafts, you verify
Good vibe coding is not blind acceptance. As one engineer put it, if an AI wrote every line but you "reviewed, tested, and understood it all, that's not vibe coding — that's using an LLM as a typing assistant." The same standard applies to writing. The version that works is collaborative: you guide and review rather than accept everything that appears.
In practice that means a loop. State your intent. Let the AI draft. Compare a couple of options instead of grabbing the first. Read the result against what you actually meant, and fix anything that drifted. AI is fluent, but fluent is not the same as faithful — it will happily write a clean sentence that says something slightly different from what you intended.
"Just use ChatGPT" is not vibe writing
There is a difference between vibe writing and pasting a prompt into a chatbot and shipping whatever comes back. The second one has a name now too: slop. It is fluent text you cannot defend, did not shape, and did not learn anything from.
The gap is review and ownership. A blank chatbot will give you a confident paragraph about your topic. Vibe writing gives you your paragraph, in better English, that you understood and approved on the way through. One leaves you with text. The other leaves you with text plus a little more command of the language.
The risks, named honestly
Three real risks come with this, and pretending otherwise would be its own kind of slop.
- Slop. AI writing tends toward the generic — the safe, average phrasing that fits any essay and belongs to no one. Left unchecked, your writing flattens into the same beige register as everyone else using the same tools.
- Losing your voice. Your odd, specific way of putting things is an asset, not an error. If you accept every "smoother" suggestion, you can sand off exactly what made the writing yours.
- Not learning. This is the big one for language learners. If you outsource every sentence and never look at why a fix is a fix, your English does not improve. You become dependent on the tool instead of better at the thing.
All three have the same antidote: stay involved. Review, choose, verify, and take one thing away each time.
A vibe-writing workflow you can run today
You do not need a system. You need a loop you actually repeat.
- Write the point first in whatever language you think in. Get the idea down before you worry about English.
- Let the AI draft the English version, and treat it as a first draft to react to.
- Ask for options and pick the one that matches your meaning and tone, rather than the first thing offered.
- Check it against your intent sentence by sentence, and correct any drift.
- Keep one phrase you did not know before. That is the compounding part.
If your draft is correct but still feels off, that is usually the translated-from-your-L1 layer — and it has its own fixes in why your English sounds translated and how to make your English sound natural, not like an AI.
Keep your voice, keep learning — where Diglot fits
This is exactly the loop Diglot is built for. It is a bilingual editor: you write in the language you think in, get fluent English back, and the tooling is designed to teach as it helps, not just hand you an answer.
- Start from your own meaning. The AI writing assistant works with your intent rather than replacing it, so the ideas stay yours.
- Shape, do not just accept. The paraphrasing tool offers versions to choose between and explains the change, so you keep your voice and learn the pattern.
- Stay the author. For work where it matters, the Authorship Certificate records how the document was actually written — proof that the thinking was yours.
- One bilingual workspace. See the full ESL writing tool built for writers who think in one language and publish in English.
Vibe coding worked because the machine got good enough to handle the execution while the human kept the judgment. Vibe writing works the same way — and for non-native writers, it finally puts the language where it belongs: as a tool you steer, not a wall you climb. Bring the idea. Keep the judgment. Let the English catch up.

