Vibe Coding is not ‘let the AI write whatever and hope’. It is a disciplined practice: you stay the architect and reviewer, while an AI coding assistant (Claude, Cursor, ChatGPT, Microsoft Copilot or Gemini) does the typing. You describe intent, the AI proposes code, you read every line, you accept or correct. The goal of this course is concrete: take an idea from architecture to a live, paying product. In this lesson you set up the workbench.

Why a clean setup matters

AI assistants are only as good as the context they can see. A messy repo with no README, no structure and no dependency file produces messy suggestions. A clean, well-described repo (مستودع منظّم) lets the model infer your conventions and stay consistent.

Step-by-step

  1. Create a new folder naseej/ and run git init. This is your product repo for the whole course.
  2. Install Python 3.11+ and create a virtual environment: python -m venv .venv then activate it.
  3. Add three files by hand so the AI has anchors: README.md (one paragraph on what Naseej does), requirements.txt (start with streamlit and fastapi), and .gitignore (ignore .venv and .env).
  4. Open the folder in Cursor (or VS Code with Microsoft Copilot, or Claude in your browser/desktop). Confirm the assistant can read the whole folder.
  5. Make your first commit: git add . && git commit -m "chore: scaffold Naseej repo".

Example prompts

You are my pair-programmer on a Python SaaS called Naseej.
Read my README.md and requirements.txt.
Propose a minimal folder structure for a Streamlit front end and a
FastAPI back end in the SAME repo. Explain each folder in one line.
Do NOT write feature code yet — structure only.
Generate a .gitignore for a Python project that uses a .venv
virtual environment, a .env secrets file, and Streamlit cache files.

Notice the second prompt is narrow and verifiable — that is the habit you build. Vague prompts (‘make me an app’) get vague code; scoped prompts get reviewable code.

Try it

Scaffold your own naseej/ repo following the five steps above. Write your README paragraph in your own words — in English, but feel free to name the Arabic-market problem you are solving (e.g. an invoice helper for Egyptian freelancers). Commit. Then ask your AI assistant to review the structure it proposed and list one risk. You now have a clean workbench and a reviewing partner.

Next