Vibe Coding

The term “vibe coding” was coined and popularized by Andrej Karpathy, a prominent computer scientist, AI researcher, former AI leader at Tesla, and former OpenAI co-founder. He introduced the term in February 2025 to describe an AI-assisted software development technique where a developer uses large language models to generate code through natural language prompts without reviewing or editing the code but focusing on guiding, testing, and iterating with the AI. Karpathy described vibe coding as “fully giving in to the vibes, embracing exponentials, and forgetting that the code even exists”.wikipedia+1. The original post is accessible through the following link.

-center

The term was popularized especially with Karpathy’s use of AI tools to prototype software by providing goals and feedback, letting the AI handle the coding details, marking a shift from traditional coding to an intuitive, AI-driven process. This approach prioritizes rapid experimentation and creativity over precision or fully understanding the code.dev+1

Thus, Andrej Karpathy is credited as the originator of the term “vibe coding.”

The Problems of Vibe Coding

Reinventing The Tire: Systems generates the code from the beginning. If we define a good software, we can say that depends on how functionally easy to update the system with new features. On vibe coding systems, the agents always rewrites the code independent from the requirement. For instance; if you want to add a new condition in order to route the input request, the agent does not edit the specified part of the code. Rather, the agent rewrites the code from beginning, even there is no change on other parts of the code.

🧠 Teaching

Content List

Vibe Coding

The Problems of Vibe Coding

References

Graph View

Table of Contents