What you'll get in this lesson: See the whole course in four stages, so you know where you are and what comes next.
Prerequisites: No software installation required. Think of one thing you'd like AI to do for you every day.
In this lesson I want to talk about two things: what this course actually teaches, and why it's worth building an AI Agent of your own.
In the very first lesson, L1-01, we use Pi to build an Agent that is yours. Every stage after that adds more abilities to that same Agent. By the end, it becomes your second brain. You can treat it as your twin, your partner, or your staff. To me, it's a company.
Why build it yourself?
Think about it. When you ask GPT something today, it's smart, like a PhD. But it's a stranger. It isn't your employee, and it isn't your partner.
So building our own Agent isn't about replacing the big AI companies. Of course not. We still rely on them. I just want the control to stay with me.
AI models will keep changing, and that's normal. When a new model comes out, you can switch to it right away, so you're less likely to be left behind. Outside tools and resources can be plugged in when you need them, and the strongest AIs can come and use your Agent to get work done for you.
So there's really just one thing to take care of: your own Agent. What it remembers, what it has learned, and your data all belong to you.
The course in four stages
Read it from the bottom up:
- L1 | Build it: get your Agent up and running.
- L2 | Lay the foundation: help it remember things, and give the AI that builds it a fixed way of working. Skip this, and things get messy fast.
- L3 | Start working: give it tools so it can start working for you.
- L4 | Connect outward: plug in tools other people have built, and open your Agent to other AIs.
I've laid out the whole path. Every step uses what you built in the step before, so just follow the order, one step at a time. No rush.
L1: Build an Agent of your own
The first step has one goal: build an Agent that is truly yours. It runs on your own computer, and you can talk to it from the messaging app you already use.
After L1, it has a name, it can read and write files, it can write code, you can chat with it on your phone, and it still remembers you after a restart. It isn't just a chatbot. It does things for you.
So who writes the code? Not you. I can't write code myself. A coding AI writes it based on your ideas. Your job is to use it and talk to it. The whole course is designed around a subscription of about US$20 a month, so you don't need to spend a lot.
L1 is "do first, understand later." Follow along, and learn as you build and use it. Once you've built one with your own hands, you'll understand what an AI Agent really is.
L2: Lay the foundation
Once the Agent can chat, a lot of people start piling on features. Hold off for now.
Think back: every time you start a new chat, do you have to explain who you are all over again? We're all tired of that. I don't want to babysit an AI either. So the first thing L2 does is help it remember who it is and who you are, and give the AI that builds it a fixed way of working.
A few parts of L2 I especially want to mention:
- Memory: it remembers you, and at set times it tidies up its own memory, a bit like dreaming, so the important things carry over to the next day.
- Skills: think of your Agent as a new employee. A Skill is how you teach it to do something. Whether it's a workflow or a way of working, you write it down and hand it over, it learns it, and next time it does things that way.
- Emergency stop: before it blows something up, you need a brake.
- Multiple Agents: in our early survey, a lot of people said they wanted multiple Agents. That means more than one AI: several roles helping you together. Each role has its own personality and its own memory. They can chat together in the same group, and you can have them hold a meeting and hand you a summary. Once you've built it yourself, you'll feel what it's like when several Agents work together.
Why lay the foundation first? When I ran that survey, someone wrote that their first version always went smoothly, but as soon as they started adding features they got stuck, and stayed stuck for two weeks. Honestly, that used to be me too. AI does a great first version, then you add features, and either it breaks something by itself, or fixing one new thing breaks an old one.
After L2, every new feature has a set place to go, and the AI tests its work before handing it to you, so things are much less likely to get messy.
L3: Give your AI employee its tools
By now, you basically have an AI employee. What L2 does is teach it how to behave, in other words, how to train this employee. Now it's finally ready to work.
But think about it: if you hired a new employee and gave them no tools, of course they couldn't do the job, right? So the most important thing in L3 is giving it tools so it can do good work: planning work, finding information, and getting things done.
So what do we want it to do for us after L3?
First, automation. When the time comes each day, it starts on its own. For example, it checks today's news, saves those stories into your Agent, and sorts them out. Then, through an automation workflow such as n8n, or code built straight into the Agent, it makes images, writes posts, and publishes them. This kind of work is where an Agent shines.
Here's a recent example of mine. I had to go out for an event one day, but there were two open-source models I really wanted to dig into. In the past, I would have had to sit at my computer and spend the time myself.
But now, I've already given my tools to my Agent, and we've tried this a few times before, so there's a fixed way of doing it. While I was out, I told it what to do from my phone. It ran the tests and sent me the results, and I told it what to test next. In the end it handed me a test report, recorded the whole process on its own, and even remembered how to use the model from then on.
That's what we're after: letting the Agent do more for us.
L4: Connect to the outside world
The last step is connecting it to the outside.
One direction is bringing things in: your calendar, your ledger, your photos and files. You can also just talk to it out loud, or use your own web chat page.
Then there are APIs and MCP. Think of them as a door that outside tools open for you. Any tool, platform, or website that has opened this door can be connected straight to your AI Agent. How do you connect it? Just ask your AI. That's the fastest way.
The other direction is opening up: you open your Agent to others, only the parts you want to share, so other programs and other AIs can connect and use your memory to work for you. For example, with a browser extension you can call on it right beside the page you're reading. Other AIs can even manage projects together with your Agent.
We no longer have to write one version for this AI and another for that one. You just maintain your own Agent. Of course, the more you connect, the more you need to keep it in check: when it wants to change itself, it has to tell you first. Small changes follow the rules you set, big ones wait for your approval, and anything that goes wrong can be rolled back.
In the last lesson, I'll show you how to learn from code other people have shared publicly. You don't even need to read the code. If you have an idea, you can use their work as a reference and turn someone else's success into a success you've tested yourself. In the end, I hope you won't need me anymore, and you'll be able to build things on your own.
After all four stages, you'll have a second brain
On the left, your life, other people's tools, and even new AI models are all things you plug in and swap whenever you like. On the right, the strongest AIs, other programs, your browser, and your phone can all come to your Agent to get things done for you.
The one in the middle is the only thing you need to take care of. Rather than becoming an expert in some AI tool, I'd rather you become an expert in your own Agent.
Next step: go build it in L1
The hardest step in learning AI is the first one. But once you take it, things start moving on their own.
So after reading this, head straight to L1. Just log in and start. Please actually build it, okay?
👉 Start L1-01: build your first Agent →
L1 is free after you sign in. Start with L1-01, follow it step by step, and you'll have an Agent of your own.
Not ready to start yet? That's fine. Scroll to the bottom and tap "Next lesson →" to keep reading the free articles, then come back to L1-01 when you're ready.
Once you've built something, you're welcome to share it in our showcase area. If you bought the co-learning plan, you can also share it in the members-only group, with me and with everyone. Really, I'd love to see what you make.