Skip to main content

Common AI Coding Terms

This glossary explains terms you will see when using AI coding tools.

AI Assistant​

A tool such as ChatGPT, Claude, Cursor, GitHub Copilot, or Codex that helps answer questions, write code, debug problems, or work inside a project.

LLM​

LLM means large language model. It is the model behind many AI assistants. It predicts useful text based on your prompt and context.

Prompt​

The instruction you give the AI. A good prompt explains the goal, constraints, examples, and desired output.

Context​

The background information the AI can use. This might include files, error messages, project rules, previous conversation, or sample data.

Token​

A small unit of text processed by the model. Long files and conversations use more tokens.

Hallucination​

A confident but incorrect answer. AI may invent functions, packages, facts, or links. Always verify important output.

API​

An application programming interface. In practical terms, it is a structured way for one program to talk to another program or service.

SDK​

A software development kit. It is usually a package that makes an API easier to use from a specific language.

CLI​

Command-line interface. A CLI tool is used from the terminal, such as git, pnpm, node, or python.

Repository​

A project folder tracked by Git. It usually contains code, documentation, configuration, and history.

Branch​

A separate line of Git work. Branches let you make changes without immediately changing the main version.

Commit​

A saved Git snapshot with a message describing the change.

Diff​

The set of changes between two versions of files.

Test​

Code or a command that checks whether behavior is correct. Tests help you verify AI-generated changes.

Refactor​

Changing code structure without intentionally changing behavior. Refactoring can make code easier to read, test, or extend.

Agent​

An AI workflow that can plan steps, inspect files, run tools, and make changes with some autonomy. Agents still need clear goals and review.

RAG​

Retrieval-augmented generation. The AI searches a knowledge base or document set and uses the retrieved text to answer more accurately.

Model​

The AI system used to generate answers. Different models have different strengths, speeds, and costs.

Temperature​

A setting that affects randomness. Lower temperature usually gives more consistent output; higher temperature can be more creative.

Markdown​

A lightweight text format used for documentation. Files often end in .md or .mdx.

JSON​

A structured data format used by APIs and configuration files.

CSV​

A spreadsheet-like text format where values are separated by commas.