An AI chat window is useful. But once you are building something real, the work is not only happening in the chat. It is happening across your computer: in project folders, files, a code editor, a terminal, a browser, and sometimes a local database or server.

That collection is your workspace.

Understanding the workspace changes how you work with AI. Instead of saying “make me an app” into an empty chat, you can open the real project and say: “Read this folder. Explain what is here. Tell me what you would change before touching anything.” The agent can work from the same evidence you see.

The five parts of a working AI workspace

None of these replaces the others. The editor shows you the files. The terminal runs the project. The browser shows you the result. The AI helps you operate all of it, but it needs a clear workspace to do useful work.

Project folder: files and instructionsEditor: inspect and editTerminal: run commandsAI agent: plan, build, testBrowser: see the running result

A project is more than its code

A beginner often imagines a website as “the code.” In reality, a healthy project also includes the material that tells people and agents what the code is for.

my-project/
  README.md          What this project is and how to run it
  CONTEXT.md         Current goal, decisions, and constraints
  src/               The app's source code
  public/            Images and other public assets
  package.json       Tools the project needs
  .env               Local secrets and settings — never publish this
  HANDOFF.md         What changed and what happens next

The exact folders will vary. The point is not to memorize a template. The point is to recognize that the workspace holds three kinds of truth:

1. The product: code, assets, content, and data. 2. The instructions: the goal, rules, decisions, and open work. 3. The operating record: what was run, what changed, what passed, and what needs attention.

Files & Context taught you how Markdown can carry the second and third kinds of truth. The workspace is where those files meet the actual running product.

The project root: start here, not somewhere inside

The project root is the main folder that contains the project. It is the place you should normally open in VS Code and the place where you should start Codex or Claude Code.

Starting an AI agent in the wrong folder is like dropping a contractor into a random room of a building with no map. It may see one file but miss the instructions, dependencies, and surrounding system.

Before you begin work, check four things:

• What folder am I in?
• Is this the project root?
• Is there a README.md, AGENTS.md, or CONTEXT.md to read first?
• Is the project already running somewhere?

You do not need to answer all of this alone. Ask the agent to orient you. But make it show you the answer before it changes anything.

The normal build loop

Most AI-assisted building is a loop, not one magic prompt.

Open the project
        ↓
Read the context and inspect the files
        ↓
Make one focused change
        ↓
Run the project or tests
        ↓
See the result in the browser
        ↓
Review the changed files
        ↓
Record what changed and choose the next action

The browser is important. A successful command does not always mean a working product. You still need to look at the page, click the behavior, check the mobile layout, and decide whether the result is what you meant.

This is the division of labor that makes AI powerful without putting you on autopilot. The agent handles repetitive implementation and diagnostics. You keep control of the goal, the taste, the decision, and the approval.

Git and GitHub: the project's memory and backup

Once a project matters, its folder needs a history. Git is the version-control system that records changes to the files. A commit is a named checkpoint: a moment you can inspect, compare, share, or return to if a later change goes wrong.

GitHub is a hosted home for a Git repository. It is where the project can be backed up, shared with collaborators, reviewed in a browser, and connected to deployment services. Git works on your computer; GitHub gives that work a shared remote home.

Your project folder
        ↓ Git commit: a checkpoint
Local Git history
        ↓ Git push: send approved work outward
GitHub repository
        ↓ optional deployment
Live website or application

This is why Git belongs in the workspace even when you are working alone. An AI can make a surprisingly large number of changes quickly. A commit before and after a focused task gives you a clear answer to three questions: what changed, did it work, and how do I return to the previous version if needed?

You will learn the actual commands in Class 6 and the full review workflow later in Project Control. For now, remember the role: Git is the project's memory; GitHub is its shared, backed-up home.

What belongs where?

Do not keep the only copy of an important decision in a chat. If a decision affects the product, put it in the workspace where the next session can find it.

Practical exercise: map a real workspace

Choose one project you already have—even a folder of notes is fine. Open it in VS Code. Then ask an AI agent:

Read its answer beside the file tree. Correct anything it misunderstood. Then create or improve a README.md with the project purpose and the next action.

Check your understanding

1. Why is a project folder more useful to an AI agent than a pile of pasted files in chat? 2. What is the difference between the editor, terminal, and browser? 3. Why should you start an AI agent from the project root? 4. Where should a decision live if it matters to the next work session?