"Vibe Prompting": the new way to talk to your AI

For a few weeks now, a new concept has been making its way through the developer world: “Vibe Coding.” But this approach can just as easily apply to any kind of AI prompting work.

Popularized by OpenAI co-founder Andrej Karpathy, vibe coding is an approach where developers use natural-language instructions to generate code with artificial intelligence, focusing more on concepts than on technique or syntax. Karpathy describes the method as a way to “fully give in to the vibes” and “forget that the code even exists.”

Its rise coincides with the arrival of non-coder creators who build applications in a matter of days with the help of AI tools like Cursor or Windsurf, coding on instinct — in short… “vibe coding.”

“Vibe Prompting” enters the chat

If you work in communications, marketing, or content creation, the derivative concept known as “Vibe Prompting” will speak to you.

Forget ultra-technical briefs — instead, you talk to the AI the way you’d talk to a colleague or an art director: “Write me something that feels like spring and renewal,” “Generate an image that’s a mashup of Wes Anderson and a 90s music video,” “I want this to breathe, to feel light, airy,” and so on.

That’s exactly what vibe prompting is. Except this time, it’s an AI that has to understand. The result? More exploration, more nuance, and often, better creative surprises. You build something through iteration, like an artist working a canvas.

But it also means the human doesn’t adapt to the machine — it’s the other way around.

With “Vibe Prompting,” we’re moving closer to a more natural interaction, closer to real art direction. We’re no longer just describing an object or an action — we’re conveying an intention, an emotion. The AI becomes an assistant that grasps our overall vision rather than a mere executor of mechanical instructions.

Grok and “Unhinged” mode: vibes in action

A notable example of this vibing trend is xAI’s introduction of different personalities for Grok 3, including the “Unhinged” mode. This mode lets the AI adopt a freer, less conventional tone, delivering bolder — and above all, unpredictable — responses. This kind of tone and style customization shows how vibes can be built right into AI interactions, letting users pick the mood that best fits their needs or preferences.

GPT-4.5 and evaluating “vibes”

The idea of AI vibes is gaining ground. OpenAI has even introduced a criterion called “vibes,” which analyzes how well a model perceives and reproduces the tone a user intended. In blind tests, its new LLM, GPT-4.5, reportedly showed a better grasp of vibes and adapted more effectively to shifts in tone and intent. It also demonstrates stronger aesthetic intuition and greater creativity, excelling at helping with writing and design.

Sam Altman, CEO of OpenAI, said ahead of GPT-4.5’s launch that the model is “the first to feel like talking to a thoughtful person,” highlighting its ability to deliver emotionally intelligent responses.

This means AI is becoming increasingly able to understand and generate content based on the desired feel, strengthening the effectiveness of vibe prompting and paving the way for even more intuitive, personalized interactions.

Vibe prompting has real implications for our work:

• We can now guide the AI with the emotion we want. The result: more spontaneity and a considerable time saving.
• A brand is an atmosphere, an attitude. Vibe prompting makes it possible to train an AI to pick up on that DNA and generate content that stays true to it.
• By playing with vibes, we explore unexpected variations instead of boxing ourselves into an overly rigid brief.

Our interaction with AI is changing

We talk a lot about AI and creativity, often as if the two were opposed. But vibe prompting shows that we can use AI without losing the soul of the creative process.

We’re moving from a model where AI executes tasks to one where it understands a brief the way a human would. We’re not telling it what to do — we’re telling it what emotion to convey.

And if we look further ahead, we can imagine that tomorrow’s AIs will become even more attuned to these nuances, to the point of picking up on our cultural references, our inspirations, and even our hesitations.

Personally, I prefer this dynamic, where the AI adapts to our vibe, over one where we have to talk to it in a technical way, like a machine.