The AI Learning Hub Journal
◆ Getting Oriented

What Vibecoding Actually Means

The whole loop: you describe, the AI writes, you judgeyou are directing a very fast, very literal helper — it has read a huge amount of code, but knows nothing about you1 · YOU DESCRIBE IT"a page where someoneenters their email andI get notified"2 · THE AI WRITES ITit turns your words intothe exact instructions acomputer needs3 · YOU JUDGE ITtry it, look at it, andsay exactly what iswrong or missingsaying what is wrong starts the loop again — each round gets closer to what you meantWHAT YOU BRINGthe idea, the judgement and the taste —you decide what it is for, and whetherwhat came back is actually rightWHAT THE AI BRINGSspeed, patience, and the exact wordinga computer needs — but it knows nothingabout your business until you tell itIT DOES EXACTLY WHAT YOU SAID — WHICH IS NOT ALWAYS WHAT YOU MEANTclear thinking about what you want is now the hard part — not typing speedYOU ARE NOT PRETENDING TO BE AN ENGINEER — AND THIS IS NOT MAGICthe skill is explaining clearly what you want to exist, then looking hard at what comes back
You describe, the AI writes the exact instructions, and you judge what comes back — saying what is wrong is how it gets better.

Describing, Not Typing

Vibecoding means building software by describing what you want in ordinary language. An AI writes the code. You say "a page where someone enters their email and I get notified", and something roughly like that comes back. You look at it, say what is wrong, and it changes. That is the whole loop. The word is a bit flippant, but it points at something real. The skill has moved from typing the instructions a computer understands to explaining clearly what you want to exist. You are not pretending to be an engineer, and you are not being handed magic. You are directing a very fast, very literal collaborator. It has read an enormous amount of code. It has no idea what your business is until you tell it.

  • You supply the intent, the judgement and the taste; the AI supplies the exact wording a computer needs
  • The loop is describe, look, react — not write code, run it, hunt bugs
  • Clear thinking about what you want is now the bottleneck, not typing speed
  • It does exactly what you said, which is not always what you meant

The Three Shapes of Tool

Tools in this space come in three broad shapes. Knowing which one you are using explains most of your experience. Chat-based assistants are a conversation window: you ask, it gives you code or instructions, you copy them somewhere. In-editor assistants sit inside the place the code actually lives. They can see all your files and change several at once. Prompt-to-app builders take a description and hand you a running thing on the internet. Hosting and a database come attached — a database being the organised store where an app keeps its information. Specific products change every few months, so learn the categories rather than the brand names. As a beginner, a builder that gives you something running is usually the fastest start. Move to the others when you hit its ceiling.

  • Chat assistants: flexible, but you are responsible for where the code goes
  • In-editor assistants: see your whole project, better for changing existing work
  • Prompt-to-app builders: fastest to something live, least control underneath
  • Products churn constantly — pick by category and by what you need next

What This Genuinely Changes

Here is the honest version of the claim. The cost of finding out whether an idea is any good has collapsed. Work that used to mean months of someone else's time and a real budget can now be a weekend and a subscription. That matters most for the ideas that were never worth commissioning. The internal tool that saves your team four hours a week. The calculator your customers keep asking for. The rough version you need before anyone will fund the real one. What has not changed is this: software that handles other people's money, health or personal data still needs someone who can read what was written. Getting to a working thing is now easy. Knowing whether it is safe to hand to strangers is the part this course spends the most time on.

  • The cost of testing an idea has fallen further than the cost of running one
  • Ideas too small to justify an engineer are now the sweet spot
  • Getting something working and knowing it is safe are separate problems
  • Speed is real; the caveats later in this course are also real

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.