Vibecoding Glossary

Your quick-reference guide to vibecoding and AI development terminology. Bookmark this page — you’ll come back to it.

Terms are listed alphabetically. Click any term to jump to its definition.

A

Agentic Coding

A mode of AI-assisted development where the AI autonomously performs multi-step tasks — writing code, running it, reading error output, and fixing issues — with minimal human intervention between steps. Tools like Claude Code and Cursor Composer support agentic workflows. See Choosing Your AI Tools.

API (Application Programming Interface)

A set of rules that allows different software programs to communicate with each other. In vibecoding, you’ll frequently build or consume APIs — AI excels at generating API code from descriptions of desired endpoints and data shapes.

Autocomplete (AI)

The most basic form of AI coding assistance, where the tool predicts and suggests the next few characters, lines, or blocks of code as you type. GitHub Copilot popularized this approach. Distinct from chat-based or agentic AI coding.

B

Boilerplate

Repetitive, standardized code required to set up a project or feature but not unique to your application’s logic. Vibecoding eliminates most boilerplate — AI handles it instantly. See Project Scaffolding.

C

Chain of Thought (CoT)

A prompting technique where you ask the AI to reason step-by-step before producing output. Produces more accurate results for complex tasks. Example: “Think through this step by step before writing the code.” See Advanced Prompting.

Context Window

The maximum amount of text (measured in tokens) that an AI model can process in a single interaction. Larger context windows allow the AI to consider more of your codebase at once. As of early 2026, context windows range from 32K to 200K+ tokens depending on the model.

Cursor

A popular AI-native code editor built on VS Code. Features include Tab completion, Chat, and Composer (agentic mode). One of the most popular tools for vibecoding. See Choosing Your AI Tools.

D

Debugging (AI-Assisted)

Using AI to identify and fix bugs by sharing error messages, stack traces, or unexpected behavior descriptions. AI is particularly effective at diagnosing common errors and suggesting fixes. See Debugging with AI.

E

Embedding

A numerical representation of text (or code) that captures its meaning in a format AI models can process. Used in RAG systems to find relevant code or documentation. You rarely interact with embeddings directly in vibecoding, but they power many tools you use.

F

Few-Shot Prompting

Providing the AI with a few examples of the desired input/output pattern before asking it to perform a task. Useful for establishing formatting conventions or coding style. Example: showing 2-3 function signatures before asking AI to generate more. See Advanced Prompting.

Fine-Tuning

Training an existing AI model on additional, domain-specific data to improve its performance for a particular use case. Most vibecoders use general-purpose models rather than fine-tuned ones, but some organizations fine-tune for their specific codebase.

G

Grounding

Connecting AI output to verifiable facts or existing code. Grounded responses reference actual documentation, existing functions, or real data rather than generating plausible-sounding but incorrect information. Reduces hallucinations.

H

Hallucination

When an AI generates output that is plausible-sounding but factually incorrect — inventing API methods that don’t exist, referencing nonexistent libraries, or generating code with subtle logical errors. The primary reason why human review (the Vibe Loop) is essential. See Testing AI-Generated Code.

I

IDE (Integrated Development Environment)

A software application for writing code, typically including a text editor, debugger, and build tools. VS Code, Cursor, and JetBrains IDEs are popular choices for vibecoding.

Inference

The process of an AI model generating output from a given input (your prompt). When you ask Claude to write code, the generation process is “inference.” Inference costs time and money — more complex prompts with larger context windows cost more.

Iteration

The practice of repeatedly refining AI-generated code through cycles of prompting, reviewing, and adjusting. The core of the Vibe Loop. Good vibecoders iterate fast and often.

L

LLM (Large Language Model)

The type of AI model that powers vibecoding tools. Trained on vast amounts of text and code, LLMs predict the most likely next tokens given an input. GPT-4, Claude, Gemini, and Llama are all LLMs.

M

Model

The specific AI system generating your code. Different models have different strengths — some are better at reasoning, others at speed, others at following instructions. Most vibecoding tools let you choose which model to use. See Choosing Your AI Tools.

Multi-Turn Conversation

An interaction with AI that spans multiple back-and-forth exchanges, where the AI remembers the full conversation history. Essential for complex vibecoding tasks where you iteratively build and refine code across many prompts.

N

Natural Language Programming

Writing software by describing desired behavior in everyday language rather than formal programming syntax. Vibecoding is a practical implementation of natural language programming.

O

One-Shot Prompting

Providing the AI with a single example before asking it to perform a task. A lighter version of few-shot prompting.

P

Pair Programming with AI

Using an AI coding assistant as your “pair partner” — you provide direction and review while AI handles implementation. The collaborative dynamic at the heart of vibecoding. Unlike human pair programming, AI never gets tired and brings knowledge of millions of codebases.

Prompt

The text input you give to an AI model. In vibecoding, your prompt includes the description of what you want built, relevant context, constraints, and any examples. Prompt quality is the single biggest factor in output quality. See Prompting Fundamentals.

Prompt Engineering

The skill of crafting effective prompts to get the best possible output from AI models. Includes techniques like providing context, specifying format, using examples, and chain-of-thought reasoning. A core vibecoding skill. See Prompting Fundamentals and Advanced Prompting.

R

RAG (Retrieval-Augmented Generation)

A technique where relevant information (documentation, code, data) is retrieved from a knowledge base and included in the AI’s context before generating a response. Many AI coding tools use RAG to include relevant parts of your codebase in each prompt automatically.

Refactoring

Restructuring existing code without changing its behavior. AI excels at common refactoring patterns — extracting functions, renaming variables, converting between patterns. See Refactoring with AI.

Role Prompting

Telling the AI to adopt a specific persona or expertise (“You are a senior React developer” or “Act as a security auditor”) to shape the style and focus of its output. See Advanced Prompting.

S

Scaffolding

Generating the initial project structure — directories, configuration files, boilerplate code — using AI. One of the highest-ROI uses of vibecoding, saving hours of setup. See Project Scaffolding.

System Prompt

Hidden instructions provided to an AI model before the user’s conversation begins. System prompts configure the model’s behavior, persona, and constraints. In tools like Claude Code, the system prompt is set by the tool; in custom applications, you control it.

T

Temperature

A parameter that controls the randomness of AI output. Lower temperature (0-0.3) produces more predictable, deterministic output — good for code generation. Higher temperature (0.7-1.0) produces more creative, varied output — useful for brainstorming or writing.

Token

The basic unit of text that AI models process. Roughly 3/4 of a word in English, though code tokenizes differently. Tokens matter because they determine context window limits and API costs. A 100-line function might be 300-500 tokens.

V

Vibe Loop

The core iterative framework for effective vibecoding: Prompt → Generate → Review → Refine → Repeat. Named for the improvisational, intuitive nature of the process. The Vibe Loop ensures AI-generated code meets quality standards through structured iteration. See Iterative Development: The Vibe Loop.

Vibecoding

AI-assisted software development where humans provide direction, context, and judgment in natural language while AI provides code implementation, with quality ensured through structured iteration. Coined by Andrej Karpathy in February 2025. See What is Vibecoding? for the full story.

Z

Zero-Shot Prompting

Asking an AI to perform a task without providing any examples — just a direct instruction. The simplest form of prompting. Works well for straightforward, well-defined tasks. Example: “Write a Python function that validates email addresses.”

Missing a term? This glossary is regularly updated as the vibecoding ecosystem evolves. Start building with our Getting Started guide.