"Vibe coding" is a term for building software by describing what you want in plain language and letting an AI tool write and run the actual code, while you stay focused on whether the result feels and works right, not on the syntax underneath it.
You do not need to know what a "function" or a "database schema" is to start. You need to know what you are trying to make, who it is for, and how to tell when it is wrong.
A tiny example
Here is the kind of prompt a beginner might write in their first week, and what is actually happening behind it.
Make a page that shows a countdown to my birthday, July 1st. Make it feel celebratory, not corporate.
Notice what that prompt does well: it states the function (a countdown), the specific data (July 1st), and the feeling you want (celebratory, not corporate). That last part matters more than people expect. AI tools are good at tone if you actually give them one to aim for.
A second example, a bit more involved
Build a simple page listing my favorite local hiking trails. Each one should show a name, difficulty, and a one-line description. It is for friends asking me for recommendations, so keep it casual, like a text message, not a travel brochure.
Same shape, same three ingredients, just applied to something with a bit more structure. The size of the idea changes. The underlying approach does not.
What you are responsible for
- The goal. No tool can guess what "done" looks like for your idea.
- The judgment. Does this look right? Feel right? Is this the actual problem you have?
- The iteration. First results are drafts. Vibe coding is a conversation, not a vending machine.
What the AI is responsible for
- Translating your description into working code
- Handling the technical plumbing, syntax, structure, and most bugs
- Explaining what it did, if you ask
Three myths worth clearing up now
"It only works for simple stuff." This free course sticks to small projects on purpose because they are the fastest way to prove the idea to yourself in one sitting. But the same loop, describe, test, refine, ship, scales up to real ongoing projects too. That is what the modules after this course are for. Not a different approach, the same one, with more capable tools around it.
"You have to get the prompt perfect the first time." No. Most real results take a few rounds of plain-language feedback. A rough first prompt that gets refined beats a perfect-sounding prompt you spent twenty minutes overthinking.
"If you cannot read the code, you are not really in control." Reading code and being in control are not the same skill. A film director does not operate the camera. What keeps you in control is testing the result against what you actually meant, and describing what is wrong when it is not right.
Assignment 1.1 - Name the gap, twice
You're done when: you have two honest, specific sentences describing things you actually want built, not generic examples.
0 of 5 items checked
Your notes save to your account and this browser.
Self-check
In vibe coding, whose job is it to decide what "done" looks like?
Yours. The AI can produce working code, but only you know what the actual goal is and whether the result genuinely matches it.
Which of these prompts gives an AI tool the most useful context?
Answer: C. It states the function, the specific data, and the feeling. All three matter.
Why does not needing to read code conflict with being "in control" of a vibe-coded project?
Control comes from testing the result against your actual intent and describing what is wrong in plain language, not from operating the technical machinery directly.