Vibe Coding - Creating Code with Natural Language
1. Vibe Coding
- The term “Vibe Coding” first appeared in a post by Andrej Karpathy on X (Twitter) in February 2025.
- He is a co-founder of OpenAI and previously served as the Head of AI at Tesla.
"There's a new kind of coding I call vibe coding, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists."
[Source]
X - Andrej Karpathy
- The core idea of “Vibe Coding” is a development approach where instead of manually writing code, developers describe requirements in natural language and let AI generate the code.
- Just like creating music by following rhythm, coding is performed by riding the “vibe,” progressing based on flow and atmosphere.
In summary, Vibe Coding is an AI-driven development paradigm where you build by requesting tasks from AI, focusing more on outcomes and flow rather than the code itself.
2. Code Assistant
- A Code Assistant is an AI tool that generates or modifies code based on natural language instructions.
- Examples include GitHub Copilot, Claude Code, OpenAI Codex, and Gemini Code Assist.
- These tools support Vibe Coding by helping developers focus more on “what” to build rather than “how” to implement it.
| Code Assist | Provider | Base Model |
|---|---|---|
|
|
- GitHub - Microsoft |
Supports multiple models including GPT (GPT-mini, GPTCodex), Claude (Sonnet, Haiku), Gemini, etc. |
|
|
- Anthropic | Claude Opus, Claude Sonnet, Claude Haiku |
|
|
- OpenAI | Based on OpenAI GPT, GPT-mini, GPTCodex models |
|
|
- Google - Google DeepMind |
Gemini family models (Flash, Pro) |
3. Prompt
- A prompt is a sentence used to instruct an AI Code Assistant on what task to perform.
- There is no strict format, but in practice, structured instructions using Markdown are commonly used.
- The more clearly you specify what to build, how to build it, and under what conditions, the better results you can achieve.
- This is because prompts are the key element that allows AI Code Assistants to accurately understand user intent and generate appropriate outputs.
- The methodology and experience of crafting effective prompts become valuable know-how in real-world development.
- Programming languages have evolved to become more human-friendly over time.
- Initially, there was machine code, which was difficult for humans to understand. This led to the emergence of Assembly language, which represents instructions in a readable text form.
- Later, higher-level languages such as C and Java, which are closer to human language, became widely used.
- Now, we have entered an era where simply describing desired functionality in a prompt allows AI to generate code.
- Regardless of the programming method used, all code is ultimately translated into machine code and executed by the CPU.