Intention Is All You Need: Vibe Coding as a Hermeneutic Practice

I was recently at an AI meetup where I first heard the phrase “Intent is all you need.” The idea is that in the era of “vibe coding” — or more accurately, AI-assisted development — the code itself is generated by the machine. The developer just needs to communicate their intention, and the AI handles the heavy lifting.

It struck me as fascinating that “intention” has become the central currency of modern coding. It’s not just about eliminating the need for deep technical syntax; it’s about this somewhat “fluffy” concept of intent. To put it bluntly, I would have understood if they said “requirements are all you need.” You feed in the specs, and the AI builds the product. But intention? I asked around at the meetup, but nobody could quite pin down what that meant.

When I got home, I dug into the research and found some fascinating perspectives. The article that popularized this notion — Advait Sarkar’s “Intention Is All You Need” — was actually quite critical of it. He argues against “vibe coding” approaches that give too much agency to the AI, warning that it could have a homogenizing effect on software. He claims it creates an illusion of free will, where the programmer feels liberal but is actually being steered by the model. Sarkar suggests we need a “reflective layer” — leaning into existentialist philosophy to see programming as an evolutionary process of “becoming.” Practically speaking, he advocates for prompting practices that challenge the AI, complicating the conversation rather than just accepting the first output.

I think we can take this even further by looking at the world of hermeneutics — the ancient study of interpretation. Historically, hermeneutics was used to decode religious and legal texts through what’s called the “hermeneutic circle.” The idea is that understanding isn’t linear; it’s a constant dialogue between the parts and the whole. You read a sentence to understand the book, but you need the context of the book to truly understand the sentence. In the Romantic era, this was applied to literature with the assumption that an author encodes an “intention” into a text, and the reader spends their time approaching that intention, even if they never fully reach it.

When we code with AI, we are essentially entering into this kind of hermeneutic conversation. The AI agent isn’t a neutral tool; like the reader in the hermeneutic process, it is an interpretive instance that transforms our intention. The big difference here is that, unlike a dead author who can’t talk back, the programmer and the AI are in a live, interactive loop. The intention is sent, interpreted by the agent, and then reviewed and reflected upon by the programmer, leading to a refined intention.

In this light, AI programming feels less like a cold technological discipline and more like a practice of “conversational hermeneutics.” It’s a dynamic process of making meaning together — something that becomes obvious the moment you start chatting with an AI agent.