Open sourcing

I have been working intensively on open source contributions for the past three weeks.

For a long time, I was simply publishing my own code and waiting for others to contribute to one of my repositories.

I realized that if I wasn’t even making contributions myself, who would care to contribute to my projects?

So I examined the most active public repository of a tool I use every day: Raycast.

The public repository hosts all the extensions available and built by the community. I started with the Spotify extension.

I had a persistent issue with it.

When generating a playlist using AI, there were problems such as:

Always receiving the same recommendations (the model wasn’t performing web searches)

Inability to access my last generated playlist

Inability to customize a playlist after generation (required re-generating with a more tailored prompt)

So I dove in, cloned the repository, and began tinkering. While AI agents are excellent for navigating new codebases and understanding their structure, I found myself enjoying opening files and tracing cascades of function declarations.

I used AI agents to research topics, how to address the issue with the most effective approach, etc. It helped significantly!

An example is the Download Manager extension.

I discovered that Quick Look on macOS provides a command-line interface to generate images of the Quick Look preview, which we could then save as a base64 string to render the image in Raycast’s markdown preview.