How to learn AI

I have spent the past three years building and working with AI. In this time, I have worked on 14 apps in my free time, built AI systems at work, and even started to teach people how to use AI. Two of the questions I often get are, “How are you so good at AI?” and “How do you learn and keep up to date with the latest in AI?” At first, it was a mystery to me as well. But in retracing my steps, I realized something fundamental to learning. This is how you can learn anything, including AI, and it's very simple: find what you love, then use it as the vehicle to learn what you want to learn.

That’s the whole thing. You can literally stop reading now.

I've always loved creating and building things. When I was young, I recreated the Monopoly game with Nigerian cities using cardboard and paper. I remember obsessing over getting the consistency right for the board, the Chance and Community Chest cards, and paper bills. When I set out to do this, my intention wasn’t to learn anything. I just wanted to make the game available to my friends in a way they could relate to. However, in doing so, I picked up some things about craft and business. I didn’t enjoy learning business in the classroom but I had fun with pricing and negotiations as part of the Monopoly work. At this time, I just enjoyed the process of building and that happened to come with some learning for me. I found that there is a term for this kind of learning in psychology - incidental learning.

The same thing happened with AI. Software was one medium of creation that I hadn’t explored as much. So imagine my joy when I realized that AI could unlock this for me.  Starting in 2023, I threw myself in and started building. When I built my first app, which was a fortune teller app, it felt awfully similar to the experience in my Monopoly days. I just wanted to build something with AI, but I ended up learning about prompting. I was very new to using AI at this point. ChatGPT launched a few months earlier and Google had just launched Gemini. Through building the app, I understood the specifics of how you give AI the right set of instructions to elicit the right outcome. At this point, I started to feel the learning in the moment and realized that I could actually continue to do this to stay on the edge when it comes to AI.

For my next project, I decided to do something more complex - a conversational starter app. This included more technical concepts like authentication. I remember spending a couple of days just to figure out how to enable “log in” for folks who want to use the app. It also allowed me to start exploring the use of multi-step agents. Instead of just input and output, I learned how you could actually plan and have agents go on for a bit longer. For the next project after that, I went more complex with a text-to-speech app. This gave me an understanding of agent skills - repeatable instructions that are codified to allow more predictable results from agents.

Increasing the complexity of my projects and the AI tools I used for them over time was intentional. In doing this, I had engineered the learning process. While it’s still informal in a way, it was no longer incidental. I was deliberately upskilling by applying AI to what I love - building.

I believe this principle can be applied to other areas beyond software. Say you love cooking and want to learn AI. You could co-create great recipes with an AI agent or take on different cooking projects with the help of AI. And with every cooking project, you deploy the AI tools that you haven’t tried yet. For instance, when personal agents like OpenClaw became a thing, you could use one of the available frameworks to create a digital sous-chef. With this, you are learning more with every new project you work on while doing what you already love doing. And you can take these learnings to other areas of your life.

I am not a software engineer by trade. However, the fundamental concepts when working with AI are the same across different fields. I have been able to bring some of the things I learned to my work. One thing I spent the past couple of months on is my personal agent using Nanoclaw. I learned how to manage memory, context, and build customizations to extend the capabilities of my agent. Recently, I used these learnings to create team agents for my team to help us with shared administrative tasks.

With AI, it's still early days, and there is so much more to be learned and to be done. Pick something you love and use it as a vehicle to learn.