# Introduction

Hi everyone. I'm really excited that you decided to join this course. I can't wait to show you what is possible with these agents and using AI for personal and work. I'm really grateful that you're going to be spending the next few hours with me.

About five months ago I started using AI agents at work — shifting from chat with ChatGPT to a whole new world that was open to developers. I started using those tools that were mainly built for developers for business use cases. I spent around 500 hours with these agents, using Cursor and Claude Code, both in the context of my personal life and at work.

What makes my experience unique is that I go through a lot of content — there's a lot of noise in the market and a lot of YouTube videos, everyone is talking about it. I cut to the core of what is important, remove the noise, and experiment with those capabilities, insights, and tips in a work environment. I've experienced the problems that happen in a work setting. In that way, I can bring some unique insights and enable others to build practical tools that help them in an actual environment — not just theories and personal projects with limited audience.

Just to be straight up from the beginning: I'm not a developer, and this course is not intended for engineers and developers or people who just casually use ChatGPT. If you are using AI regularly, if you're doing similar tasks over and over and find yourself spending a lot of time on things that you think should take much less time, and if you wish you could do more and are willing to spend a few hours learning — these tools have a bit of a learning curve. Claude Code is not something you'll get in one hour, but if you spend one week with it, you'll learn it to a very good degree and you'll be able to get your time back in the first month. It's definitely an investment worth making.

If the answer to all of those questions is yes, this is the right course for you. If not, and if you're just casually using ChatGPT, then staying with that or using Claude chat is good enough for you.

To give you a mental model: what's the difference between Claude chat and Claude Code? Think about Claude chat as a smart friend you're texting — you get advice from it. Claude Code actually takes actions for you. It's like hiring a smart assistant. You might be thinking: it has the word "code" in the name, so it's not for me. Rest assured, there's nothing scary about code these days. It's just a language, and you don't have to speak that language — you have a translator, which is natural language. So you just talk to it.

Claude Code runs in three different environments. I'll go through all of them with you through the course. I have a preferred interface, and then there's a more advanced one in terminal. For you, terminal is just going to be another chat window — you talk into it and it takes actions for you. I have the terminal interface explained in the advanced section of the course, but we'll start with an easier, nicer UI for using Claude Code as the core of the course.

In this course you'll also learn how these agents work and how you can build systems around them. This is not about just telling you tools or hacks because tools change. I'll give you the set I'm using right now and walk you through it. But this space is rapidly changing, new tools come out, and I'll try to keep you updated throughout the course and produce more content.

The first principle is that you need to understand how they work and the principles behind how they operate. Then you can build systems around them. Through doing and working with them, you're going to learn how to master them in no time.

Take this course step by step. Complete it with me. I'm going to do everything from scratch — install apps, setups, integrations — all of that with you. Some things will be new for me too, and I'll show you how I solve them with the tool along the way. Don't rush it. If you don't understand some concepts at the beginning, they're all going to connect together.

We'll have a bigger project that connects everything I go through at the beginning of the course into one cohesive capstone so we can get practical experience there.

The mental framework we'll cover has six stages:

1. **Context** — what context is, the importance of it, how you gather it, manage it, and keep it useful and optimized for agents.
2. **Organization** — how to organize files so agents can navigate your information and documents.
3. **Skills** — how to teach skills to agents. When you're talking to ChatGPT or Claude chat, they're quite generalist. With skills you can make agents very specialized in whatever you want to do. There have been very few cases where I couldn't do something with skills.
4. **Integrations** — how to connect agents to pretty much all the tools you're using through MCPs, which I'll explain later.
5. **Planning** — how you can plan what you want to do and have agents execute it for you.

The shift in mindset is moving from chat mode with agents — just responding like a friend — to delegating work to agents, training them, and having oversight on what they do so they execute and you approve.

I've seen a lot of videos and podcasts where people say these agents run autonomously and do all the work for them. I don't really believe that — at least I haven't experienced being able to fully let something autonomously run without looking over its shoulder and understanding what it's done. Even when they do work, you have to go back and review it. It's important for a human to be in the loop. I'll show you how to do that with a good harness and oversight so agents don't take actions you don't want them to take. Through iterations you can also improve the quality of the work — there's a huge difference between setting up agents properly and not, and the output quality is day and night. Human involvement to review is quite important.

Here's the course structure at a glance:

We'll start with **foundations** — installing the tools and setting them up. Then I'll cover **configuration**, where I open the box of how Claude Code works in the environments we're setting up, what capabilities it has, and how to fine-tune it from the beginning to get good quality output. Then I'll **connect it to a variety of tools** that are useful for everybody. I've picked tools that are general enough that everyone would have a use case for them, since many of you are coming from different backgrounds. Once you know how to connect to these general tools, you should be able to connect to your organization's tools on your own.

From there we'll do a **capstone project**, which I'll explain in detail. Then there's an **advanced section** covering terminal (the most powerful way to use Claude Code, with all features available), and Git for versioning — useful if you work with multiple people or on multiple devices. You can skip that if it's not relevant to you.

For **bonus content**, I'll show you how I've built my knowledge base using Obsidian — a tool that displays markdown files in a really nice way and connects them together. And image generation using Nano Banana, which you can use to create images from Claude Code for social media or whatever you want. This course won't go deep into image and video generation — I'll just touch on it because it could be useful and interesting for some. There's a whole world of visual arts generation through AI that this course definitely won't cover.

The core of the course is teaching you the tool, how it works, connecting it to your existing toolset, and doing a project. After that, you should be able to start using it for your work and personal life and build little systems and tools to achieve pretty much anything you want.

**Sections one and two:** we'll install Cursor, which is where we'll run Claude Code. Cursor is a tool developers use, but I'll show you how to set it up — it's quite easy and straightforward. Then we'll set up your personal configuration. I'll tell you what CLAUDE.md is, skills, memory, and much more.

Then we'll **connect the tools** — your email and calendar (most people use Google Workspace), how to gather context and content from websites and YouTube, how to convert any document to text that agents and LLMs can read and understand, even with tables, images, and diagrams. I'll also do a demo of how I connect it to Notion, my personal knowledge base, and pull in content from there.

For the **capstone project**, you might say: why planning a trip to Japan? I could do that with ChatGPT. True. But I chose this project because I didn't want the topic itself to be complicated or irrelevant to people from different roles. From the conversations I had with people interested in this course, they come from very different backgrounds — sales, product, hiring — and the use cases shared across all of them are almost non-existent in a business context.

The second reason is that the real power of these agents shows when you connect the full knowledge of your organization — your context, product, business model, customers. That's where the power is. I can't share information from the company I work at, and I can't create sample company data. So I decided to start with something that doesn't have existing context, where I can build that context from scratch, but it's relatable to anyone. I'm actually taking a trip to Japan and I did it exactly with agents — I can show you the output at the end. We'll do it together using everything we learned prior to the capstone so you can see how those capabilities work in action.

**Advanced section:** terminal and Git — full Claude Code power, version control, Obsidian. I foresee many of you using Obsidian with Claude Code. It's an easy setup, more user-friendly than Cursor, and applies more to personal use cases. For business scenarios, Cursor is a better environment to load and run Claude Code. For personal work, Obsidian is a better choice. We'll also do some fun image generation from Claude Code.

With that, I'm closing the introduction and really excited to go to the next chapter with you and get you started with setting up Claude Code and Cursor.

Thank you.
