Building Apps with AILesson 5 of 87

Contents

Building Apps with AI

Lesson 5 of 87. Sign in to keep your place.

Module 1 · The machine, and what the model can see of it

Getting a machine that can actually build

9 min read

The honest hardware question

You need less than people say and more than the marketing implies.

A laptop with 8 GB of RAM and 20 GB of free disk will run everything in this course. 4 GB will struggle once a build and a browser are open together. A node_modules folder for one ordinary project is 300 to 600 MB, and people run out of disk before they run out of memory.

You do not need a GPU. Nothing here trains a model. The AI runs on somebody else’s machine and arrives over the network.

If you are on a phone, or a Chromebook, or a shared machine you cannot install software on, skip to the hosted section below. That route is real, and it is how a large number of people are building.

The four things to install

A terminal. Already there. On Windows, install WSL (wsl --install in PowerShell, then reboot) and use the Ubuntu terminal it gives you. This is not optional advice: nearly every instruction you will find online assumes Linux or Mac, and WSL makes them work instead of half-work.

Node.js. Install it through a version manager rather than directly, because within a month some project will demand a different version. nvm on Mac and Linux, nvm-windows or fnm on Windows. Then nvm install --lts.

Git. git --version will tell you if it is already there. If not, brew install git, sudo apt install git, or the installer from git-scm.com.

An editor. VS Code is free and is what most tooling assumes. If you would rather not run Microsoft’s build, VSCodium is the same editor compiled without the telemetry, and everything in this course works in it.

Then check the whole chain in one go:

Bash
node --version   # v20.x or later
npm --version
git --version

If any of those says command not found, the install did not finish or your terminal is still using its old settings. Close it and open a new one before you believe anything.

The AI tool itself

Three shapes, all with a real free tier as of early 2026:

  • In the editor. Cursor and Windsurf are VS Code forks with an agent built in. GitHub Copilot is an extension for the editor you already have. Continue is open source and can point at a free model.
  • In the terminal. Claude Code, Codex CLI, Aider. Aider is free and open source, and works against whichever model you can access.
  • In the browser. Replit, Lovable, Bolt, v0. Nothing to install.

Price moves constantly. What does not move: the free tiers are limited by volume, they are enough to learn on, and the difference between tools matters far less than the habits in this course.

If you cannot install anything

This is a completely workable path, not a consolation.

GitHub Codespaces gives every free account a monthly allowance of hours on a full Linux machine with VS Code in the browser. It is the closest thing to a real development environment that runs on a phone.

Replit runs entirely in the browser and has a mobile app that is usable for small edits. StackBlitz runs Node inside the browser tab itself, with nothing on a server.

The honest limits: typing code on a phone keyboard is slow and you will not enjoy long sessions, free compute hours run out, and browser environments sleep when you close the tab. For reading code, reviewing a diff, fixing a typo and redeploying, a phone is fine. For a three-hour build, borrow a keyboard.

The first five minutes of any project

Bash
git clone <url>      # or start a new project
cd <folder>
npm install          # reads package.json, downloads dependencies
npm run dev          # starts the development server

Then open the address it prints, usually http://localhost:3000.

When this fails, it is almost always one of four things: you are in the wrong folder (pwd tells you), Node is the wrong version (the error will say so), port 3000 is already taken by something you started yesterday, or npm install did not actually finish. Read the last five lines of the output before doing anything else.

Set one thing up that everybody skips

Make a folder for projects and keep them all in it. ~/code or ~/projects, one folder per app.

It sounds trivial. It is the difference between finding last month’s work and rebuilding it, and it takes ten seconds now.

Every lesson here is open. Your progress is recorded once you pass the Building with AI exam.

Check your understanding

npm run dev fails on a project that worked yesterday, with an error naming an unsupported Node version. What is the appropriate response?

Pick the one you would defend. Your answer stays private.

Next lesson · 8 minThe terminal, for people who have been avoiding it

Stuck on something?Ask Beatriz, your AI tutor

How to cite: Addaly (2026) ‘Getting a machine that can actually build’, in Building Apps with AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/building-apps-with-ai/a-machine-that-can-build (Accessed: 13 October 2026).