The Problem with Vibe Coding
AI-Assisted Development Is About Flow, Not Feeling
The tech world has a new darling: “vibe coding.” Coined by AI researcher Andrej Karpathy in early 2025, the term describes a programming approach where developers describe what they want in natural language and let AI generate the code, essentially coding by “vibe” rather than deliberation. Within weeks, the concept exploded across social media, tech blogs, and even mainstream publications. Yet beneath the playful terminology lies a fundamental mischaracterization of what’s actually happening when developers collaborate with AI. The term “vibe coding” suggests a directionless, almost mystical process. But this couldn’t be further from the truth.
The Misdirection of “Vibes”
When Karpathy introduced vibe coding, he described it as “fully giving in to the vibes, embracing exponentials, and forgetting that the code even exists.” His tongue-in-cheek description involved accepting whatever code suggestions the AI gave without review, asking for “the dumbest things,” and hitting “Accept All” on changes. If errors arose, he’d simply paste them back to the AI rather than debugging traditionally. While amusing, this description fundamentally distorts the cognitive mechanisms involved.
The word “vibe” implies something ephemeral, directionless, and based on feeling rather than thought. It suggests that developers are simply channeling some mysterious creative energy, letting randomness guide their work. This portrayal, however, fails to capture the highly directed nature of the cognitive process involved. When developers work with AI coding assistants, they aren’t abandoning structure or purpose. Instead, they’re engaging in a sophisticated form of directed search through possibility spaces.
Understanding AI-Assisted Development as Directed Search
What actually happens when a developer collaborates with an AI? Far from “vibing,” they’re engaged in an iterative process of specification, evaluation, and refinement. Each prompt to the AI represents a search query in the vast space of coding solutions. The developer begins with a clear mental model of the desired outcome and uses natural language to guide the AI toward that specific goal.
Consider how this actually works in practice. A developer might start with: “Create a React component for a user authentication form.” This isn’t a vibe, it’s a precise specification. When the AI returns code, the developer evaluates it against their mental model, then provides further direction: “Add email validation” or “Include a forgot password link.” Each interaction narrows the search space, moving systematically toward the envisioned solution.
This process mirrors how experienced developers have always worked, just with a different interface. Traditional coding involves searching through documentation, Stack Overflow answers, and one’s own memory to find the right syntax and patterns. AI-assisted development simply changes the search mechanism, not the directed nature of the process.
The Flow State of AI Collaboration
Rather than “vibing,” what developers experience when working effectively with AI is closer to what psychologist Mihály Csíkszentmihályi termed “flow”, a state of complete immersion in an activity where actions and awareness merge. In this state, the developer maintains clear goals while experiencing a seamless interaction between thought and execution.
Flow requires several conditions that are perfectly met by AI-assisted development: clear goals (the desired functionality), immediate feedback (the AI’s code generation), and a balance between challenge and skill. The developer must possess enough expertise to evaluate the AI’s output and guide it effectively, while the AI handles the mechanical aspects of syntax and implementation.
This flow state explains why AI-assisted development can feel almost magical. The frustrations of programming—forgetting syntax, hunting for documentation, and fixing typos—fade, leaving only the challenge of the problem itself. But this isn’t abandoning direction or purpose; it’s achieving a more direct connection between intention and implementation.
The Importance of Intentionality
The danger in the “vibe coding” narrative is that it encourages a passive, undirected approach to development. If newcomers believe that AI coding is about surrendering control and “going with the flow” in a directionless sense, they’ll produce exactly the kind of problematic code critics worry about: untested, ununderstood, and potentially dangerous.
Effective AI-assisted development requires more intentionality, not less. Developers must maintain a clear vision of their goals, understand the principles behind the code being generated, and carefully evaluate each iteration. The AI is a powerful tool for navigating possibility space, but the developer must be the navigator, not a passive passenger.
This intentionality extends beyond individual coding sessions. AI developers need to prioritize architecture, maintainability, and handling edge cases. They must understand not just what the code does, but why it’s structured that way and how it fits into larger systems. The AI can speed up the journey, but the developer must know the destination.
Reframing the Conversation
As AI becomes increasingly central to software development, we need more precise language to describe these new workflows. “Vibe coding” may have served its purpose in capturing initial attention and excitement, but it’s time to move beyond casual terminology that obscures the real cognitive processes at work.
Perhaps we should speak of “flow-assisted development” or “directed AI collaboration.” These terms better capture the active, intentional nature of the process while acknowledging the unique psychological state that effective AI collaboration can produce. They emphasize the developer remains in control, using AI as a sophisticated tool rather than surrendering agency to it.
The educational implications are significant. If we frame AI-assisted development as “vibing,” we risk training a generation of developers who don’t understand their tools or their craft. But if we frame it as achieving flow through directed search, we encourage developers to maintain high standards while leveraging AI’s capabilities.
Beyond the Buzzword
The rapid adoption of AI coding assistants represents a genuine paradigm shift in software development. But paradigm shifts require not just new tools, but new mental models and vocabulary. The “vibe coding” meme has served its purpose in marking this transition, but it’s time to develop more sophisticated frameworks for understanding AI-assisted development.
As educators, developers, and technologists, we have a responsibility to characterize these new workflows accurately. The future of software development won’t run on vibes, it will run on the thoughtful collaboration between human intentionality and AI capability. By recognizing AI-assisted coding as a form of directed search that can produce flow states, we can harness its power while maintaining the rigor and purposefulness that quality software demands.
The next time someone mentions “vibe coding,” perhaps we should gently correct them: “You mean flow-assisted development?” The importance of disciplined thinking in excellent software development shouldn’t be overlooked as we adopt powerful new tools; it’s more fundamental than simply correcting word choices. After all, even in an age of AI, there’s no substitute for knowing where you’re going and why you’re going there.