Course contents
What vibe coding is
Explain what vibe coding is, where the term came from, and which parts of the work stay yours when an AI agent writes the code.
Vibe coding means building software by describing what you want in plain language and letting an AI coding agent write, run and fix the code. You say what the app should do and how it should feel. The agent works out how: which files to create, what goes in them, and which commands to run to check the result.
The term comes from Andrej Karpathy, an AI researcher and OpenAI co-founder. In a post on February 2, 2025, he described a new kind of coding where you “fully give in to the vibes” and “forget that the code even exists.” The phrase spread so fast that Collins Dictionary named vibe coding its Word of the Year for 2025.
What changes, and what doesn't
In traditional programming, you turn an idea into instructions a computer can follow, one line at a time, in a language such as JavaScript or Python. With an agent, you skip the typing. You still make every decision that matters: what to build, what counts as finished, whether the result is correct, and whether it's safe to put online.
Picture yourself as the client and the site manager on a building job. The agent is a fast, tireless builder who has read every manual but has never seen your site. It builds what you describe, fills gaps with guesses, and tells you when it thinks it's done. Your job is to describe well, look at what it built, and catch the guesses that are wrong.
Your agent reports: “Done! The contact form works.” What should you do next?
The loop you'll repeat all course
Every build in this course follows the same loop. You describe a change. The agent reads your project, plans, edits files and runs commands. You look at the result in a browser and either accept it or describe what's still wrong. Each pass takes a few minutes, and a real app is dozens of passes, so getting fast at the loop matters more than any single prompt.
Put one pass of the vibe coding loop in order.
- The agent edits files and runs commands
- The agent reads the project and proposes a plan
- Save a checkpoint, or describe what's still wrong and go again
- You check the result in the browser, not only the agent's summary
- Describe the change you want and how you'll know it's done
A spectrum, not a switch
People use the term loosely. At one end, you never read the code and accept whatever looks right on screen. That's fine for a throwaway prototype or a birthday page. At the other end, professional developers use the same agents but read every change, run automated tests and own the result. This course starts you near the first end and moves you toward the second, because the moment real people use your app, “it works on my screen” stops being enough.
| Pure vibes | Guided vibe coding (this course) | AI-assisted engineering | |
|---|---|---|---|
| Reads the code? | Never | Skims each change and asks questions | Reads every line |
| Checks results by… | Clicking around | A short checklist per feature | Automated tests and code review |
| Good for… | Throwaway experiments | Personal and small-business sites | Products with many users or sensitive data |
| Main risk | Hidden bugs and data leaks | Missing a subtle flaw | Slower to get started |
Where is pure vibe coding fine, and where do you need to slow down and check carefully?
A countdown page for a friend's birthday
What stays your job
- The idea and the details. The agent can't read your mind. Who is the site for, and what must it show?
- Checking. The agent's “done” is a claim. You confirm it in the browser.
- Safety. Agents make apps work; they don't automatically make them safe. Module 4 teaches the basic rules.
- What goes online. Publishing is a decision you make, not a side effect of building.
Plain-English definitions from the AI Coding Dictionary.