I stopped reading Docker tutorials and finally learned it by building apps

I was initially hesitant to try building my own self-hosted applications because Docker felt daunting. Using containers, images, volumes, and Docker Compose all seemed a little advanced to the uninitiated. However, with an AI chatbot acting as a troubleshooter, I dove headfirst into using it to install apps I actually wanted, and in the space of a few days, projects like Navidrome, Audiobookshelf, Kavita, Sub/Wave, and Immich taught me all I needed to know.

I started with a real app instead of a demo

My first container taught me more than many tutorials could

I had been making the classic beginner’s mistake of believing I needed to fully understand Docker before launching a project. Like many people, I am a hands-on learner, and no matter how many tutorials I read, they don’t solve any real problem; therefore, I don’t end up learning anything of substance.

My first project was to build Sub/Wave, a self-hosted AI DJ that plays my own music library. I was very keen to see this in action, so every command felt like it had real purpose; I had a vested interest in why my music library needed a mounted volume and why port 4533 mattered. This motivation was a game-changer for understanding every step of the build.

All the Docker fundamentals quickly fell into place. The idea of a Docker image (in this context, a lightweight, standalone software package) was no longer abstract. I found I could easily delete and recreate containers without losing my entire music library. I started to understand how Volumes work to protect my data. All of these concepts made much more sense when applied to real-world usage.

My confidence built alongside each app

From Navidrome to Immich, I noticed the progress I made with every step

Each application presented a new challenge. Building Audiobookshelf, the self-hosted Audible alternative, taught me the importance of organizing multiple media folders and mapping them correctly. Kavita, the self-hosted e-reader, was configured differently and made me understand persistent storage.

Immich, the self-hosted Google Photos alternative, was the biggest leap. This relied upon multiple services working together, which was intimidating at first, but it made me realize how Docker actually made things easier by enabling us to manage complex applications within a single Compose file. And by the time I arrived here, I was confident enough with how Docker components work that I didn’t feel too overwhelmed.

The key here was repetition. The workflow stayed the same; only the variables changed. Every project required pulling images, defining volumes, exposing ports, adding environment variables, and using a select few essential commands to achieve the same results. Once I started recognizing the common structure of each self-hosted project, Docker started to make much more sense.

AI was my learning companion throughout

ChatGPT helped me quickly and efficiently debug and understand errors

As I went into Docker completely green, there were inevitable stoppers along the way, especially at first. Whereas not so long ago I’d have to Google issues and hope for the best, or sift through several Stack Overflow threads for answers, here I could paste the error log into ChatGPT and ask it not only for a solution but also to describe the issue in detail.

Believe me, there were many occasions when I simply couldn’t have continued without this valuable assistance. And because the explanations were conversational, I began to see the reasoning behind the fix, which reinforced my understanding of how Docker works.

However, there were several occasions when copying commands didn’t work because of discrepancies in my folder structure, hardware, or operating system. It was therefore important not to think of AI as a magic solution to every issue I faced, but more as a mentor that could patiently and repeatedly explain concepts and troubleshoot issues until I understood exactly what was going on.

Learning by doing beats following a tutorial

Tutorials work in a perfect world; real projects teach by repetition

In my experience, Docker tutorials assume ideal conditions and success at every turn. In the real world, projects are much messier, and along the way, I encountered containers that refused to start, databases that needed rebuilding, ports that were already occupied, and configuration files with tiny mistakes that caused surprisingly big problems.

While these issues were frustrating at the time, this is where real learning occurred. My extensive troubleshooting taught me how Docker behaves when things go wrong and how to fix them, which is arguably more helpful than watching someone demonstrate how things work perfectly.

It also really helps if you are genuinely invested in a project, and I can recommend trying any of Immich, Sub/Wave, Audiobookshelf, Kavita, or Navidrome. Start with something that aligns with your hobbies and interests, and you won’t mind spending time fixing an error, since the end goal is something cool and relatable.

Build something you’ll actually use, and Docker is far more fun

To paraphrase Benjamin Franklin, “Teach me, and I forget…involve me, and I learn,” and this is certainly true of my experience with Docker. Building real applications gave every command, volume, and Compose file a practical purpose. And asking AI for assistance didn’t feel like cheating, especially since I asked questions that offered explanations along the way.

By combining genuine enthusiasm with a way to overcome the inevitable hurdles, I not only ended up with five usable self-hosted apps, but also a deep understanding of Docker and the confidence to try my next project entirely unassisted.