Android’s sideloading changes forced me to create an APK installer

I have no problem using ADB with the command line every time. But between remembering where I put the executable, opening a terminal window, connecting the device, and pasting the APK file into the path, it’s a bit of work. It would be faster to download a small program on my computer to handle most of the work.

Yes, options such as ADB AppControl, EasyADB or QuickADB are already available in the market. However, I wanted to build something that focused more on the APK installation workflow – wireless debugging pairing, file/directory browsing and unpacking – than a full ADB management suite.

Although I’m a moderately competent programmer, between work and two kids, I don’t have time to write a program completely from scratch. So I thought I’d give AI-powered “vibe coding” a try. This had the added benefit of allowing me to try out a new programming language I was unfamiliar with: Get.

Why Go? It’s a simple language that’s perfectly suited for such a compact project (no need for Rust), but it easily compiles into a small, cross-platform executable (sorry, Python). Plus I wanted to build a cool looking terminal app using pretty Bubble Tea TUI frame which I recently came across. With the design in mind, I set out to find an AI agent to help me.

I entered AI tools Just know that asking someone to build an app for you is a gamble. Expensive models from Anthropic and OpenAI can handle simple tasks once, but when you leave a complex project to an agent, you often end up in a tedious back-and-forth to resolve errors. I decided I wanted a more hands-on approach to interacting with ADB and building a terminal-based UI.

I wasn’t going to spend hundreds of dollars on Opus 4.6 for this little project. Even with careful planning, there’s always the risk that models will get stuck in an expensive construction loop or run into an unavoidable problem. Instead of asking Claude or Codex to handle the task themselves, I integrated my regular IDE with OpenCode and a very affordable $10 Go subscription. This gave me access to the MiniMax M2.7 and the very impressive GLM 5 to use alongside my slow human brain and keyboard-obsessed fingers.

I myself used the bare bones of the structure – folders, a few placeholder functions, .gitignore, etc. – posted and wrote a quick AGENTS.md to describe the project’s requirements, goals, and coding features. Then, rather than asking the AI ​​to finish everything, I slowly went through the steps I knew I needed: planning, building, reviewing, and adjusting each step along the way. I highly recommend using version control like Git so you can undo major AI mistakes and avoid the inevitable rabbit holes.

First, I got a basic ADB wrapper running on Go, which allows me to run commands like adb devices to find and then use connected phones adb install Pushing APKs to them. Next, I expanded the work commands to the main terminal interface to walk through the process. It was pretty simple, but not the most user-friendly experience.

Finally – and arguably the hardest part of the back and forth – was building the UI flow. It didn’t help that I was learning a new language and framework at the same time, but I was able to make manual fixes myself by having the AI ​​agent expose the theme via a configuration file. This saved time (and credits) by avoiding the need to burn resources on every little detail.

In the end, the combination of human planning, AI creation, and tag team tweaking produced a solid little program in just three days of off-duty work. I certainly couldn’t write ~3000 lines of code back then. Not bad for $10 and I have plenty of credit left for another idea.

This is a compact 7MB executable – no installation required. If you don’t have it, it can automatically download ADB from Google or use your existing PATH setup. The app starts ADB when needed and closes it on exit. You can pair with debuggable Android devices via USB or Wi-Fi, browse files using the built-in file explorer, and install APKs directly. It also supports opening and installing software packages including APKM, XAPK and APKS files.

The good news is that I got a program that works and it’s really helpful. If there’s one drawback, it’s that I didn’t learn Go as much as I initially expected. By asking AI to solve problems for me, I missed out on learning the language’s syntax and patterns in depth. I gained some understanding from reviewing the code and following the logic, but I wouldn’t call myself proficient in Go even though I wasn’t into problem solving.

And that’s the real risk of vibe coding: not fully understanding how the code works. I’ve found a few harmless bugs and even a zip slip vulnerability while reviewing AI-generated changes, but I can only imagine the risks when building anything related to encryption, API keys, or sensitive credentials. My advice to anyone trying Vibe coding is to make an effort to plan and understand, not just how your app works should do work, but also learn what tools you need, how to manage a project, and what usually goes wrong. That way, you’re in a better position to check AI output when it inevitably misses the mark.

If you want to try my app, you can find the releases and source code here Small APK Installer GitHub repo. I haven’t had a chance to try the Mac or Linux builds yet, so I believe Go compiles everything correctly. If you spot any bugs, feel free to open an issue — and maybe curse the AI ​​for making it so easy to build (and break) things.