A random unexpected open source contribution

This morning (yes, I happen to wake up at 5 AM), I wanted to create an AI agent as a proof of concept for an AI engineering intern role with a company I got referred to. But I had never seen the codebase for an AI agent or knew how they were built, and this made me reluctant to just use whatever Claude generated for me after a simple "Build me an AI agent" prompt.

Striking Gold in Open Source

So, I went on GitHub to see a few examples of how an AI agent is created and came across the OpenHands repo. I started reading the codebase (it was Python, so I was able to navigate my way through it) and then I saw a TO-DO comment in the chunk localizer file.

Finding an actual unsolved TO-DO in an open source repo is like striking gold. I went deep down the rabbit hole:

• I used Gemini 3.5 Flash to understand more about the work I needed to do.
• I used Claude Sonnet 4.6 to write the functions needed.
• I added two cases in the unit test, and everything worked!

OpenHands Pull Request #14699

How You Get Started

Open source can seem intimidating from the outside, but this is exactly how you get started. You don't need to wait for a curated "good first issue." Just start reading unfamiliar codebases, learn the patterns, spot a gap, and dive in.