Setup

Everything you need installed before Class 2. Budget about 20 minutes. If something does not work, bring your laptop to class — we fix setups there.

You will install two things: uv (a small tool that takes care of Python for you) and the ChatGPT desktop app, which contains Codex (the AI that reads and writes code with you). Then you run one app to prove it all works. (GitHub, where your code will live, comes in Class 3 — nothing to do about it now.)

1. Install uv

uv is a small tool that installs Python and runs Python programs for you. You install it once; from then on it takes care of Python — no Python installer, no version to pick, nothing to add to PATH.

macOS — open Terminal (Cmd + Space, type "Terminal"), paste this line, press Enter:

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows — open Terminal (search for it in the Start menu), paste this line, press Enter:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Then close the terminal and open a new one — a new command is only visible to terminals opened after it was installed — and type:

uv --version

You should see uv 0.x.y. That is all. Python itself arrives by itself the first time you run an app.

2. Install the ChatGPT desktop app (Codex)

Codex is the coding side of ChatGPT. It lives inside the ChatGPT desktop app — the app you install, not the website. It is free.

1. Download the app from openai.com/chatgpt/download for macOS or Windows.
   • macOS: open the .dmg and drag ChatGPT into Applications. The first time you open it, macOS asks if you are sure — you are.
   • Windows: run the installer (or install ChatGPT from the Microsoft Store — same thing).
2. Open it and sign in. No account? Click Sign up and create a free one, with your university email.
3. In the top-left corner, open the menu and choose Codex. Chat is for talking; Codex is for code. You will spend the course in Codex.

You do not need ChatGPT Plus. The free plan includes Codex, with a cap on how much you can ask in any five-hour window — enough for a class, not enough to let it run unattended. If you hit the cap, it resets on its own; go for a coffee.

3. Folder, project, repository — three words for one thing

You will hear three words for what is, most of the time, the same thing on your disk. Get them straight now and half of the confusion in this course disappears.

Two ways to get a repository from GitHub onto your computer: download a ZIP (a plain copy — a folder, no history) or clone it (a copy that stays connected to GitHub — a repository). Today you download; from Class 3 you clone.

The rule that keeps everything working: one app, one folder, one project, one repository. Do not rename it, move it, or put it inside another project. Documents/nova/splitit is a good address. Desktop/New folder (3)/splitit copy is not.

4. Get a project onto your computer

1. Go to Week 2 and download the ZIP for your group's app: splitit.zip or tiny-crm.zip.
2. Unzip it into a nova folder in your Documents. Windows: right-click the ZIP → Extract All… macOS: double-click it. You get a folder called splitit (or tiny-crm) with app.py inside it.
3. In Codex, choose Open folder (or + → Open local folder) and pick that folder — the one that contains app.py, not the folder above it. Codex now calls it a project.
4. Ask Codex your first question: "What does this app do? Answer in one sentence and name the file you read." It should answer and name app.py or README.md. If it does, everything is wired up.

5. Run the app

Every project in this course is a small web app built with Streamlit. To run it you type one command in a terminal that is standing inside the project folder.

Open a terminal in the folder

• Windows: open the project folder in File Explorer, right-click on empty space, choose Open in Terminal.
• macOS: open Terminal, type cd (with a space after it), drag the project folder from Finder into the Terminal window, press Enter.

Then type:

uv run streamlit run app.py

The first time, uv downloads Python and the two libraries the app needs — a minute of scrolling text is normal. Then the app starts and your browser opens at http://localhost:8501. Click around. That is your company's software, running on your machine. The second time it starts in a second.

To stop the app, click in the terminal and press Ctrl+C (also Ctrl+C on macOS, not Cmd).

If uv is "not recognised" or "not found", close the terminal, open a new one and try again — the install was fine, the terminal just needs a fresh start.

The other way: ask Codex — "Run the app." It knows how (the project's AGENTS.md tells it) and will ask permission to run the same command; read it, approve it. If uv is missing, Codex can install that too — it needs no password — but read what it proposes before you approve.

6. Working with Codex — three habits

• Ask in plain language, name files when you can. "In logic.py, what does add_expense check before saving?" beats "is there validation?"
• Read before you approve. Codex asks before it runs a command or edits a file. That prompt is not a formality: if you cannot say what the command does, click No and ask it to explain first.
• Verify. Whatever Codex claims, open the file and line it names and look. The one rule of the course: AI is allowed everywhere; you cannot ship what you cannot explain.

Writing in a file works the same way: tell Codex what to write and where — "Under question 4 in ONBOARDING.md, write: …" — then read the change it shows you before you accept it. That is how you answer the question sheet.

One more: a thread is one conversation. Start a new thread for a new task — long threads make Codex slower and vaguer.

You are done when…

• uv --version prints a version
• The ChatGPT app opens on Codex, your project folder is open in it, and it answers a question about the code
• The Streamlit app runs in your browser from that folder

Anything stuck? Bring it to class. Nobody is behind on day one.