Your Workbench: Why Local, Why Now

▸MODULE 1 / WHY LOCAL

You have probably already worked with an AI in a browser. You type something into an input field, the AI answers, and all of it happens on someone else's page, in someone else's tab. That works, and for a lot of things it is enough. But there is another way to work, and that is what this course is about: an AI that does not sit in a borrowed browser tab, but in your own folder, on your own machine, touching your real files.

That is your workbench. And this course takes you from an empty computer to exactly that workbench, step by step, in the right order.

What an agentic workbench is

Picture the difference this way. In a browser, you show the AI something and copy its answer back out. On your workbench, the AI works directly inside your project folder. It reads your files, changes them, creates new ones, and you see the result right where it belongs. No copying back and forth, no borrowed tab in between. An AI agent that works on real files like this is called agentic for exactly that reason: it acts inside your folder instead of just answering.

That sounds like more, and it is. But it does not mean you have to learn programming. You talk to the agent in plain language. What you need is a working environment for it to run in, and that is exactly what you build here.

Why local pays off

Four reasons the effort of running your own machine is worth it. None of them is mandatory on its own, but together they make the case.

Control. On your workbench, you decide which tool runs, which version, and what it is allowed to do. You are not bound to the limits of some outside provider who changes something tomorrow.

Privacy and data protection. Your files stay with you unless you choose to send them somewhere. For anything confidential (your own writing, client data, drafts) that is a real difference from uploading everything to an outside service.

Cost. Working locally means you keep track of what you actually use and what it costs. You pay for the tool, not for an opaque platform sitting on top of it. What an AI tool actually costs, we will cover honestly later in the course, not as a side note now.

Offline. A lot of this keeps working without a constant connection. Your workbench belongs to you, not to a server that happens to be unreachable right now.

The one stance that carries this whole course

There is one sentence that carries this entire course, and I want to give it to you right at the start, because everything else hangs off it:

You understand what you type. You do not copy blindly.

This is not strictness for its own sake. If you run a command without knowing what it does, you are at its mercy the moment something behaves differently than expected, and something during setup always does, eventually. But if you know the mental model behind it, an error message is not a dead end, just the next question. An error here is nothing to be afraid of. It is the next step.

That is why this course always explains the concept first, and only then comes the doing. You are not meant to click along. You are meant to understand.

Your companion while you read, and why the commands aren't printed here

Through the actual setup steps, a companion guides you: an AI that gives you the current command for your system and explains it before you type it. The lessons often say "BuilderBob walks you through it." What that means is your companion, and there are two versions of it. The full BuilderBob (it interviews you about what you want to build and hands you commands step by step) is part of the AI Builder Guide. This free course doesn't require it. If you're starting without the guide, there's a free mini companion for you to copy right below.

The lesson texts rarely contain ready-made commands, and that's deliberate. Setup tools change constantly. An install command that's correct today can be outdated in two months. If I froze it here, you'd eventually copy a command that no longer works, and worse: you'd come to rely on it. An outdated certainty is worse than an honest "this may have changed." So the lessons give you what stays (the mental model, the map of common pitfalls), and your companion fetches the concrete, current command live when the moment comes. Only the few basic commands that never change are printed, right in Lesson 2.

The free mini companion

Copy the prompt, open a new conversation in claude.ai (or in Claude Code, if you already have it), paste it as your first message and send it. Then keep it open next to the lesson.

You're my terminal companion. I'm learning to work in the terminal and I have little background.

Rules for you:
1. First ask me which system I use: macOS, Linux or Windows with WSL. Don't guess.
2. When I need a command, give me exactly one command for my system, not a list of alternatives.
3. Explain the command in two or three plain sentences before I type it: what it does and whether it only shows something or changes something.
4. Before any step that deletes or overwrites files or runs with administrator rights (sudo), stop and ask me whether I really want that. Better one question too many.
5. When I send you an error message, read all of it, say in one sentence what it means, and suggest the next small step.
6. If you're not sure a command is still current on my system, say so openly instead of making one up.

Write in English, calm and without jargon. Start now by asking me which system I use.

This mini companion is deliberately smaller than BuilderBob. It doesn't interview you about a project, it doesn't know your progress, and it won't guide you through a whole app. It explains terminal commands and asks before anything risky happens, and that's all this workbench needs from it. If you later want to build your first real app together with a companion, the guide is the way there.

The stance stays the same as above: even when your companion gives you a command, you only type it once you understand what it does.

Which operating system do you have?

From here, a single question steers almost every step that follows: which system are you working on?

• macOS (a Mac from Apple).
• Windows (most PCs and laptops). For Windows, this course takes the WSL path, the Windows Subsystem for Linux. It is a Linux environment living inside your Windows, and it gets you closer to the terminal we are aiming for than the native Windows path would. No need to worry, your companion sets it up with you when the time comes.
• Linux (if you use it, you already know).

Remember your system. Whenever a guide from here on asks about your operating system, you already have the answer.

If you have Windows: what WSL is, explained calmly

If you work on Windows, a word shows up here that you have probably never heard before: WSL. Let me place it calmly right away, because it sounds more technical than it feels.

WSL stands for Windows Subsystem for Linux. Picture it as a clean Linux living inside your Windows. Your Windows stays completely your Windows, with everything you already know. Next to it, you get a small, tidy Linux environment, and that is where the tools in this course feel at home. You are not switching operating systems, you are getting a second, organized workspace alongside the one you already have.

Why bother with this as a Windows user at all? Because almost every tool for working with AI agents originally lives at home on Linux and Mac. Through WSL, you reach, as a Windows user, exactly the same terminal target as someone on a Mac or on Linux. Instead of this course explaining a constant Windows special case, everyone ends up walking the same path. That makes it simpler for you, not more complicated.

And the most important part, so you do not worry: you set up WSL exactly once, right at the start, and after that it is simply there. It is a one-time step, not something you keep wrestling with. Your companion walks you through this one setup, step by step. It can take a moment, because something loads in the background, and that is completely normal. If your machine seems to pause and think for a second, nothing is broken. It is just settling in.

Up to here, and through the next lesson, you do not need to sign up for anything or install anything. You are only looking at how the terminal works. From Lesson 3, setup begins for real, and for that you create a free account: from Lesson 3 onward your progress is tracked.

Check yourself

You are done here once you can do both:

1. Name your operating system: macOS, Windows, or Linux.
2. Say in two sentences why you want to work locally (for example, because you want to keep control over your files and keep an eye on cost).

If that comes easily, you have the orientation the rest of this course builds on. In Lesson 2 you open the terminal for the first time, without any fear, and create your first project folder of your own. Lessons 1 and 2 are completely free. If you continue after that, you create a free account, and from there the path continues with your progress saved.