Vibe Coding Glossary

A standardized way for software systems to communicate with one another.

Whenever an application gets weather data, sends an email, translates text, processes a payment, or talks to an AI model, there is usually an API involved. Most modern software is assembled from many APIs rather than built entirely from scratch.

API keys are often used as passwords for software systems, so they should not be pasted publicly.

A software program designed to perform a specific task.

An app can be a mobile app, desktop app, or web app. Many "apps" today are actually websites that behave like applications.

The word "app" used to mostly mean phone apps, but now it is used very broadly.

Code that can start a task and continue doing other things while waiting for it to finish.

Fetching data from an API, reading a file, or sending an email can all take time. Async code lets the program keep running instead of freezing while it waits.

In JavaScript, async code often uses the keywords async and await, which AI assistants use frequently.

The process of proving who someone is.

Authentication is what happens when a user logs in with an email, password, magic link, fingerprint, or token. It answers the question: "Are you really you?"

Authentication is often shortened to "auth," which can also casually include authorization.

The process of deciding what someone is allowed to do.

After authentication proves who a user is, authorization decides what they can access. For example, an admin may see settings that a regular user cannot.

Authentication asks, "Who are you?" Authorization asks, "What are you allowed to touch?"

The hidden part of an application that handles data, accounts, rules, and processing.

The backend often manages databases, user logins, permissions, payments, emails, and business logic. Users rarely see it directly, but it is where much of the serious work happens.

Many projects look like design projects at first, but become backend projects once users, payments, or saved data are involved.

The application used to access and display websites and web apps.

Chrome, Safari, Firefox, and Edge are browsers. They translate HTML, CSS, and JavaScript into visible, interactive pages.

Each browser has built-in developer tools — including a console, network inspector, and element viewer — that AI assistants frequently ask people to open.

A separate line of development inside a version-control system.

A branch lets people experiment without changing the main version of a project. When the experiment works, it can be merged back in.

Branches are one of the safest habits in AI-assisted coding.

An error or unexpected behavior in software.

Bugs range from tiny visual annoyances to serious failures. Much of software development is bug-finding rather than feature-building.

The term became famous after Grace Hopper's team found an actual moth trapped in a relay of the Harvard Mark II computer on September 9, 1947. The logbook entry reads: "First actual case of bug being found." The moth is preserved at the Smithsonian.

The process of preparing source code so it can run as an application or website.

A build may compress files, check for errors, optimize assets, and create the final version used for deployment.

"Build failed" often means the computer found a problem before the app went live, which is annoying but useful.

Temporary storage used to make things load faster.

Browsers, servers, and apps often save copies of files or data so they do not need to fetch everything again. Cache problems can make a fixed website still look broken.

Pronounced "cash." Phil Karlton, a developer at Netscape, once said: "There are only two hard things in computer science: cache invalidation and naming things." This quote has been floating around the internet ever since.

A network of servers that delivers website files from locations closer to users.

CDNs make websites faster by serving images, scripts, styles, and other files from nearby servers instead of one faraway place.

CDNs are one reason modern websites can feel fast even when visitors are in different countries.

An automated pipeline that tests and deploys code changes whenever they are pushed.

CI runs automated checks every time new code arrives. CD then deploys the changes automatically if they pass. Together they reduce manual steps and catch problems early.

GitHub Actions is a common CI/CD tool. "The CI is failing" means automated checks found a problem with the latest code.

A text-based way to control a computer or software tool.

Instead of clicking buttons, people type commands. Many AI coding instructions use CLI commands because they are precise and repeatable.

CLI and terminal are closely related, but not exactly the same: the terminal is where commands are typed, and the CLI is the command-based interface itself.

A saved snapshot of changes in a Git project.

Commits create a history of the project, making it possible to see what changed and when. Good commit messages explain the reason for the change.

Small, clear commits are easier to undo than one enormous commit.

A reusable piece of a user interface or application.

Buttons, cards, menus, forms, and headers can all be components. Modern frameworks often build entire apps from components.

Components help keep design consistent across a site or app.

A place where software messages, warnings, and errors are displayed.

In a browser, the console is part of developer tools. AI assistants often ask users to check it when a website looks fine but something underneath is broken.

The browser console is different from the terminal, but both can show useful clues.

A small piece of data a website stores in a browser to remember information.

Cookies can keep a user logged in, remember preferences, track sessions, or store shopping cart contents. AI assistants sometimes say "clear your cookies" when diagnosing login or loading problems.

Not all cookies are tracking cookies. Session cookies, authentication cookies, and preference cookies are all common and often harmless.

The four basic actions most applications perform on data.

Create adds something, read views it, update changes it, and delete removes it. Many apps are essentially polished CRUD systems.

If an app stores users, lessons, posts, products, or messages, CRUD is probably involved.

The language used to style web pages.

HTML provides structure; CSS controls appearance. Colors, fonts, spacing, borders, layout, and responsive design are usually handled with CSS.

The "cascading" part means that style rules can flow downward and override one another.

A structured place where information is stored and retrieved.

Databases store users, products, messages, payments, settings, and anything else that must be remembered. They are often more valuable than the visible app itself.

A spreadsheet can feel like a database, but a real database is designed for larger, safer, more structured operations.

To find and fix problems in software.

Debugging is detective work. The visible problem is often only a symptom of something hidden elsewhere.

Good debugging starts with reproducing the problem: making it happen again on purpose.

To publish software so it can be used in its intended environment.

Deployment usually means moving code from a local machine or test environment to a live website, app store, or server.

Deployment is often where hidden environment or configuration problems appear.

A piece of software that a project needs in order to work.

Dependencies are often libraries, packages, or tools installed from places like npm. If a dependency is missing or outdated, the project may fail.

AI assistants often say "install the dependencies" before running a project.

Another word for a folder in a file system.

Directories organize files into a structure. Terminal commands often ask people to move into a particular directory before running something.

"Root directory" usually means the main top-level folder of a project.

A tool for packaging software with everything it needs to run.

Docker helps software behave consistently across different computers and servers. It reduces the "it worked on my machine" problem.

Docker's logo is a whale carrying shipping containers.

A human-readable website address.

A domain points people to a website, app, or online service. It is easier to remember than a numerical server address.

Domains are rented, not permanently owned, and must be renewed.

A specific address where an API can be reached.

Different endpoints usually perform different tasks, such as fetching users, sending messages, or creating payments.

An API is the system; an endpoint is one specific door into that system.

A specific setup where software runs.

Common environments include local, development, staging, and production. Each may have different settings, data, and permissions.

Many bugs happen because two environments are almost, but not quite, the same.

A setting stored outside the main code that changes how software behaves.

Environment variables often hold API keys, database addresses, secret tokens, or configuration values. They let the same code work differently in local, test, and production environments.

Secret environment variables should not be pasted into public code or screenshots.

A larger structure or toolkit used to build software in an organized way.

A framework gives developers patterns, rules, and ready-made tools. React, Vue, Angular, Django, and Laravel are examples people may encounter.

A library is something code calls when needed; a framework often shapes how the whole project is built.

The visible part of a website or application that users interact with.

The frontend includes screens, buttons, forms, menus, animations, and page layouts. It talks to the backend when it needs data or needs something saved.

Frontend work is where design and code meet most visibly.

A reusable block of code that performs a specific task.

Functions can calculate values, validate forms, send emails, update screens, or process data. They help keep code organized and reusable.

A good function usually does one clear thing.

A version-control system for tracking changes to files over time.

Git allows people to restore previous versions, compare changes, collaborate, and experiment safely. It is one of the foundations of modern software work.

Git and GitHub are not the same thing. Git is the system; GitHub is a popular website that uses it. Also: the name "git" is British slang for an unpleasant person. Linus Torvalds, who created it, has confirmed he named it after himself.

A web platform for storing, sharing, and collaborating on Git repositories.

GitHub is where many code projects live. It also handles issues, pull requests, reviews, documentation, and automation.

Many AI coding tools can connect directly to GitHub repositories.

A command-line tool for searching text inside files.

Grep is useful when a project has many files and someone needs to find a word, phrase, function name, or error message quickly.

The name comes from an old editor command meaning "global regular expression print."

The standard language used to define the structure and content of a web page.

HTML creates headings, paragraphs, buttons, images, forms, and links. It is usually the first technology people encounter when building websites.

HTML is not a programming language. It is a markup language, meaning it describes what things are rather than what they do.

The protocol browsers use to request and receive information from websites.

When someone visits a webpage, the browser sends HTTP requests and receives responses. APIs often use HTTP too.

HTTP is one of the basic languages of the web.

The secure version of HTTP.

HTTPS encrypts information traveling between a browser and a website. It is especially important for logins, payments, forms, and private data.

The little lock icon in a browser usually means HTTPS is being used.

An official HTTP status code that means: "I'm a teapot."

HTTP status codes are the short responses servers send back to browsers. 200 means OK. 404 means Not Found. 418 means the server is a teapot and refuses to brew coffee. It was added as an April Fools' joke in 1998 and is still in the official spec.

Despite being a joke, 418 is a real, registered HTTP status code. Multiple real servers and APIs have implemented it. There was once a serious proposal to remove it, which was met with enough developer outrage that it was kept.

The traditional first program a developer writes in any new language or environment.

It does exactly one thing: display the text "Hello, World!" on screen. By convention, if you can make that work, the environment is set up correctly and you are ready to begin.

The tradition traces back to a 1974 Bell Labs memo by Brian Kernighan. It has since appeared in almost every programming language ever created. When a developer starts something new, Hello World is the handshake.

A software application used for writing, editing, testing, and managing code.

An IDE often includes a code editor, file browser, search tools, debugging tools, terminal access, and extensions.

Visual Studio Code is one of the most commonly used IDE-like editors.

To bring code, data, or functionality from one place into another.

In programming, importing often means using a library, component, or function written elsewhere. This avoids rewriting everything from scratch.

If an import fails, the project may be missing a package or using the wrong file path.

A programming language used to make websites and applications interactive.

JavaScript can respond to clicks, update pages, validate forms, talk to APIs, and run entire applications. It is one of the core languages of the web, alongside HTML and CSS.

JavaScript is not the same as Java. The similar names have confused people for decades.

A lightweight format for storing and exchanging structured data.

JSON is common in APIs, settings files, and application data. It uses pairs of names and values, which makes it fairly readable once the pattern is familiar.

Pronounced "Jay-son." Despite the name, it is used far beyond JavaScript.

A collection of pre-written code that can be reused.

Libraries save time by providing ready-made features such as charts, calendars, buttons, animations, or date formatting.

A project may depend on many libraries, which is powerful but can also create maintenance issues.

Running on your own computer rather than on the internet or a remote server.

Developers often test things locally before putting them online. Local work is safer because real users are not affected.

"Localhost" usually means your own computer pretending to be a web server.

A record of events, actions, warnings, or errors produced by software.

Logs help people understand what happened before, during, and after a problem. They are essential for debugging.

Logs can contain sensitive data, so they should be handled carefully.

To combine changes from one branch or source into another.

Merging is how experimental work becomes part of the main project. Sometimes changes conflict and must be resolved manually.

A "merge conflict" does not mean disaster; it means two changes touched the same area and a human must choose how to combine them.

A structured change to a database.

Migrations may add tables, rename fields, change relationships, or update stored data. They help database changes happen safely and repeatably.

Migrations should usually be reviewed carefully before running on a live project.

A runtime that allows JavaScript to run outside the browser.

Node.js is widely used for backend services, development tools, build systems, and command-line scripts.

Node.js made JavaScript useful for much more than webpage interactions.

A package manager commonly used with Node.js projects.

npm downloads and manages libraries, tools, and dependencies for JavaScript projects. AI assistants often suggest npm commands.

Many projects depend on hundreds or thousands of npm packages, most of them installed automatically. In 2016, a developer unpublished an 11-line package called left-pad. Within hours, thousands of projects worldwide stopped building. The internet had a moment.

Software whose source code is publicly available for anyone to view, use, and contribute to.

Many popular tools — including VS Code, Linux, React, and many AI frameworks — are open source. This means bugs can be fixed by the community, not just the original creators.

Open source does not always mean free to use commercially. Licenses vary and matter.

A bundle of code distributed for reuse.

Packages can provide small utilities or large frameworks. They are usually installed through package managers such as npm.

Packages can save enormous time, but they also become part of the project's dependency chain.

A small change or fix to software.

Patches usually correct bugs, security problems, or minor issues without redesigning the whole system.

"Patch version" often refers to the smallest type of software version update.

A number that identifies which program on a computer should receive network traffic.

When a dev server starts at "localhost:3000," the 3000 is the port. Different programs listen on different ports to avoid collisions.

Common development ports include 3000, 4000, 5173, and 8080. If a port is "already in use," another program is running there and needs to stop first.

A request to review and merge code changes into a project.

Pull requests are common on GitHub and similar platforms. They let others inspect, discuss, test, and approve changes.

Often shortened to PR.

A debugging technique where you explain your code, line by line, to an inanimate object.

The act of articulating a problem out loud — even to a rubber duck — forces the brain to slow down and think differently. Developers often find the bug mid-explanation, before the duck has said anything at all.

The technique comes from the book The Pragmatic Programmer (1999). Many developers keep an actual rubber duck on their desk. It works. AI assistants are essentially very chatty rubber ducks.

To restructure or improve existing code without changing what it does.

Refactoring makes code easier to read, maintain, or extend. AI assistants often suggest refactoring when code works but has grown messy or repetitive.

Good refactoring does not add features or fix bugs — it only improves the structure. Tests help confirm nothing accidentally broke.

A JavaScript library for building user interfaces from reusable components.

React is commonly used for dashboards, apps, forms, portals, and interactive websites. AI coding assistants often generate React code because it is widely used and component-based.

React is a library, but people often casually talk about it like a framework.

A document that explains what a project is and how to use it.

The README is usually the first file people read when opening a code project. It may include setup steps, commands, examples, and notes.

The name literally means "read me." Whether developers actually do is a matter of ongoing debate.

A storage space for a software project and its history.

A repository contains code, documentation, configuration files, assets, and Git history. "Repo" is the casual short form.

When an AI says "the repo," it usually means the whole project folder as tracked by Git.

To return software, data, or a deployment to an earlier state.

Rollbacks are used when a new change causes problems. They are a safety mechanism, not a failure.

Good systems make rollbacks fast and calm.

A URL path that maps to a specific page or function in an application.

In a web app, routes determine what happens when someone visits a particular address. "/dashboard," "/login," and "/settings" are all routes.

AI assistants often say "I'll add a route for that" when building a new page or API endpoint.

Software delivered over the internet on a subscription basis rather than installed on a computer.

Google Docs, Notion, Canva, Slack, and most modern web apps are SaaS products. The provider handles infrastructure, updates, and maintenance — users just open a browser and sign in.

When someone says they are "building a SaaS," they usually mean a subscription-based web app. ChatGPT is also a SaaS product.

A small program or set of instructions used to perform a task.

Scripts can automate repetitive work such as renaming files, updating data, running tests, or preparing deployments.

In software, "script" does not mean dialogue. It means instructions the computer can run.

A set of tools that helps developers work with a platform or service.

An SDK may include libraries, examples, documentation, and helpers for connecting to an API. It makes integration easier than starting from scratch.

An API is the doorway; an SDK is a toolkit for using the doorway more easily.

Running software on your own server or infrastructure rather than using a managed cloud service.

Self-hosting gives more control over data, privacy, and cost, but requires managing your own setup, updates, and backups. Many AI tools offer a self-hosted version for users who want to keep data private.

Self-hosting is not the same as running something locally. It usually means running it on a server you control — not just on your laptop.

A computer or program that provides services to other computers.

Servers can host websites, send files, manage databases, process API requests, or handle authentication.

A server is not always a physical machine in a room; it may be a cloud service.

A period of active, authenticated access for a user.

When someone logs in, a session begins. When they log out or the session expires, access ends. Sessions are usually tracked with cookies or tokens behind the scenes.

"Session expired" means the login period has ended and the user must sign in again. "Invalid session" usually means the system can no longer verify the user's identity.

A language used to work with relational databases.

SQL can create tables, search records, update information, delete data, and connect related pieces of information. AI assistants often produce SQL when changing database structure or fetching data.

SQL is usually pronounced either "sequel" or "ess-cue-ell." Both are common.

A technical report showing the chain of function calls that led to an error.

Stack traces look intimidating, but they often point directly to the file and line where something broke.

The most useful line is not always the first one. The clue may be lower down.

The current data or condition of an application at a given moment.

State includes things like whether a menu is open, what a user has typed, which items are in a cart, or whether a request is loading. Modern frameworks like React are built around managing state.

"State management" becomes important when an app grows large enough that keeping track of changing data gets complex.

A cloud platform that provides databases, authentication, storage, and APIs.

Supabase is popular for AI-assisted app building because it combines several backend needs in one service.

It is often described as an open-source alternative to Firebase.

The formal grammar rules of a programming language.

Syntax controls where commas, brackets, quotation marks, and keywords must go. A tiny syntax error can stop a program from running.

Syntax errors are often easier to fix than logic errors because the computer usually knows where it got confused.

A CSS framework that styles pages using small utility classes.

Instead of writing separate CSS rules for everything, Tailwind lets developers apply styles directly in the HTML or component markup. AI assistants often use it because it is fast and predictable.

Tailwind class names can look strange at first, but they are usually just tiny styling instructions.

A window where people type commands to interact with a computer.

Terminals are used to install packages, run projects, search files, deploy apps, and inspect errors. They look severe but are often just a direct communication channel.

The terminal is powerful because commands can be copied, repeated, and automated.

An automated check that verifies a piece of code behaves as expected.

Unit tests usually check one small function or feature at a time. When a change breaks something, tests catch it before real users do.

AI assistants often generate tests alongside code, or suggest "adding tests" when building something new. "The tests are failing" means automated checks found a problem with the latest change.

A piece of information used to represent identity or access permission.

In software security, a token proves that a user or service has permission. It acts like a temporary pass that the system can verify.

Authentication tokens and AI tokens share the name but serve entirely different purposes.

A version of JavaScript with added type checking.

TypeScript helps catch mistakes before code runs by making developers describe what kind of data should be used. It is common in larger or more serious JavaScript projects.

TypeScript eventually becomes JavaScript before it runs in the browser or server.

The visible and interactive parts of a digital product.

UI includes buttons, menus, screens, forms, icons, colors, typography, and layout. It is what users see and touch.

UI is closely related to UX, but UI is more about the interface itself.

The address of a resource on the web.

Webpages, images, API endpoints, downloads, and files can all have URLs. A URL tells the browser where to go and what to request.

A domain is only part of a URL.

The overall experience a person has while using a product or service.

UX includes clarity, speed, ease, emotion, trust, accessibility, and whether the thing helps people do what they came to do.

A beautiful UI can still create bad UX if users get lost.

A named container for storing information in a program.

Variables can hold numbers, text, dates, lists, settings, or other values. Programs use variables to remember and manipulate information.

Good variable names make code easier for humans and AI assistants to understand.

A system for tracking changes to files over time.

Version control makes it possible to compare changes, restore earlier versions, collaborate, and experiment safely.

Git is the most common version-control system in modern software work.

An automatic message sent from one system to another when something happens.

For example, a payment system might send a webhook when someone pays, or a form service might send one when a form is submitted.

An API is often something your app asks for; a webhook is often something another app sends to you.

The sequence of steps used to complete a task or process.

In AI-assisted coding, a workflow might include prompting, editing, testing, committing, and deploying. A clear workflow prevents chaos.

Good workflows are boring in the best possible way.

A human-readable format often used for configuration files.

YAML appears in deployment settings, automation workflows, Docker files, and cloud tools. It is sensitive to spacing, which can make it surprisingly fussy.

The acronym is recursive — YAML includes YAML inside its own name. A programmer joke.

An AI system that can take actions or perform multi-step tasks, often using tools.

A basic chatbot answers a question. An agent may search files, edit code, run tests, and report back.

"Agent" is used loosely, so always check what the system can actually do.

A workflow where an AI system performs a sequence of actions toward a goal.

Instead of answering one prompt, the AI may plan, act, inspect results, revise, and continue. This is common in newer coding assistants.

Agentic systems are useful, but they still need boundaries, review, and human judgment.

The information currently available to an AI system in a conversation or task.

Context includes the conversation history, any files or code provided, instructions, and any other background the AI can see. When someone says "add that to the context," they mean share it with the AI so it can use it.

The more relevant context you give, the better the AI's response tends to be. The Context Window determines the limit on how much context can fit at once.

The amount of text, code, or information an AI model can consider at one time.

If the context window is too small, the AI may lose track of earlier details. Larger context windows can handle longer documents, codebases, and conversations.

A bigger context window helps, but it does not guarantee perfect attention to every detail.

Numerical representations of meaning used by AI systems.

Embeddings help computers compare ideas, not just exact words. They are often used in search systems, recommendations, and document retrieval.

Embeddings let a system know that "car," "vehicle," and "automobile" are related even if the words are different.

Training an existing AI model further on specific examples or tasks.

Fine-tuning can make a model better at a specialized style, format, domain, or behavior. It is not the same as simply giving instructions in a prompt.

Many use cases do not need fine-tuning; better prompts or better retrieval may be enough.

Connecting an AI's answer to specific information, sources, files, or data.

Grounding helps reduce made-up answers by giving the AI something concrete to refer to. It is especially important when accuracy matters.

RAG is one common way to ground an AI system in external information.

Rules or limits that guide what an AI system should and should not do.

Guardrails can help with safety, tone, accuracy, formatting, privacy, or preventing unwanted behavior.

Guardrails can be technical, legal, ethical, editorial, or all of the above.

When an AI system produces information that sounds confident but is false, unsupported, or invented.

Hallucinations are especially risky with names, dates, legal rules, citations, code details, and niche facts. Verification matters.

A hallucination is not usually a lie. The system is generating plausible text, not checking truth the way a human researcher would.

The process of using an AI model to generate an answer or prediction.

Training is how a model learns. Inference is what happens when someone actually asks it to do something.

Most people using AI tools are using inference, not training the model.

An AI model trained to process and generate language.

LLMs power tools that can answer questions, draft text, summarize documents, write code, translate, and reason through instructions.

LLMs do not "know" things in the human sense. They generate responses based on patterns learned from training and context.

A standard that helps AI systems connect to external tools, data, and services.

MCP can let an AI assistant work with files, databases, calendars, code repositories, or other systems in a structured way.

It is part of a broader movement toward AI assistants that can use tools rather than only produce text.

An AI system's ability to retain and recall information across conversations or sessions.

Basic AI tools start fresh every conversation. AI tools with memory can remember preferences, past decisions, project details, and user context over time.

"Persistent memory," "chat memory," and "agent memory" are all variations of this idea. Memory is different from the context window, which only lasts for a single session.

The AI system that receives input and produces output.

In AI tools, the model is the underlying engine that generates text, code, images, summaries, or decisions.

Changing the model can change the quality, style, reasoning, and reliability of the answer.

A system that can work with more than one kind of input or output.

A multimodal AI might understand text, images, audio, video, code, or documents. This is why some AI tools can analyze screenshots or describe pictures.

"Mode" here means type of information, not mood.

The instruction or input given to an AI system.

A prompt may be a question, command, role description, example, document, or detailed set of requirements. Better prompts usually produce better results.

Prompts are not magic spells, but precise context and constraints help a lot.

The practice of designing prompts to get better results from AI systems.

This can include giving examples, defining tone, setting constraints, asking for step-by-step output, or specifying the format of the answer.

The phrase sounds grand, but much of it is just clear communication.

An AI approach where the system retrieves relevant information before generating an answer.

RAG is useful when an AI needs to answer from documents, knowledge bases, company files, or updated information rather than relying only on its training.

RAG can reduce hallucinations, but only if the retrieved sources are relevant and reliable.

A high-priority instruction that tells an AI how to behave.

A system prompt may define the AI's role, tone, rules, limitations, and available tools.

When an AI refuses, formats things a certain way, or follows special rules, the system prompt may be part of the reason.

A setting that influences how predictable or varied an AI's output is.

Lower temperature usually produces more consistent answers. Higher temperature can produce more variety, creativity, and occasional weirdness.

For code, lower temperature is usually better. For brainstorming, a little more randomness can be useful.

A small chunk of text that an AI model processes.

Tokens can be words, parts of words, punctuation, or spaces. AI systems often measure input, output, memory, and cost in tokens.

Long words may be split into multiple tokens, and short common words may be one token.

The ability of an AI model to use external tools or trigger predefined actions.

Tool calling lets an AI search files, call APIs, run calculations, update records, or perform other structured tasks instead of only generating text.

This is one of the building blocks behind agentic AI systems.

A database designed to store and search embeddings.

Vector databases help AI systems find information by meaning rather than exact keyword match. They are common in document search and RAG systems.

This is how a system can find "cancel subscription" even if the document says "terminate membership."

A style of software development where builders use AI assistants to create, iterate, and ship software through natural-language prompts rather than writing code from scratch.

The term captures the intuitive, feel-your-way approach that AI tools make possible — describing what you want in plain language and collaborating with the AI to make it real. This glossary exists because of it.

Vibe coding dramatically lowers the barrier to building software. Understanding the vocabulary — even at a basic level — helps enormously when things go wrong.