Backgrounder
This and the next section are crafted by Claude with some light editing.
In February 2025, a two-word term coined by AI researcher Andrej Karpathy sent shockwaves through the software development world: “vibe coding.” What started as a post on X about building software by “fully giving in to the vibes” has evolved into a movement that’s fundamentally challenging how we think about software creation, who can be a developer, and what the future of SaaS looks like.
Vibe coding represents a radical departure from traditional software development. At its core, it’s an AI-assisted approach where developers—or more accurately, “vibers”—describe what they want in natural language, and AI generates the entire codebase. The revolutionary aspect? The human never looks at, edits, or even understands the code. They simply test the output, describe adjustments, and iterate through conversation.
This isn’t your typical AI pair programming where GitHub Copilot suggests completions while you type. This is surrendering the entire coding process to AI while maintaining creative control through natural language direction. As Karpathy described it: “I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.”
Origins and Evolution
The concept emerged from the convergence of several technological advances. Large Language Models (LLMs) like GPT-4, Claude, and Gemini reached a sophistication level where they could understand complex software requirements and generate functional code. Simultaneously, the rise of browser-based development environments eliminated setup friction, making it possible to go from idea to deployed application without installing a single tool.
What makes vibe coding different from traditional AI-assisted development is the complete abstraction of code. Traditional AI coding tools are productivity enhancers for developers who still need to understand programming concepts, review generated code, and make manual adjustments. Vibe coding removes these requirements entirely, making software creation accessible to anyone who can articulate what they want.
The Psychology Behind “Vibing”
The term “vibe” is deliberately chosen—it captures the intuitive, almost meditative state of creation where technical barriers dissolve. Instead of context-switching between creative vision and technical implementation, vibers maintain a continuous flow state, describing and refining their vision while AI handles all technical translation.
This psychological shift is profound. For decades, software development required a dual mindset: creative problem-solving paired with meticulous attention to syntactic detail. Vibe coding splits these concerns, freeing humans to focus entirely on the “what” and “why” while AI manages the “how.”
The Explosive Growth Phase
The numbers tell a compelling story of rapid adoption. Y Combinator’s Winter 2025 batch revealed that 25% of startups had codebases that were 95% AI-generated. Google disclosed that AI now writes 25% of their new code. The coding AI agents market has exploded. This isn’t limited to Silicon Valley startups.
The key insight emerging from both successes and failures is that vibe coding works best for specific use cases: rapid prototyping, proof-of-concepts, internal tools, and “software for one”—personalised applications serving individual needs. For production-grade, mission-critical systems, a hybrid approach is emerging where vibe coding accelerates initial development, but human expertise ensures quality, security, and maintainability.
