AI changes how we work — especially when the work involves structuring content. Best practices are still emerging. To expedite the process, I’m doing more work “with the garage door open” — i.e., sharing work in progress.
With this in mind, I’m starting a new YouTube series called Over the Shoulder where I’ll share how I’m using AI to tweak jarango.com. (One of the advantages of building a self-hosted web presence over time: fodder for experimentation!)
This first video is longer than the rest, since it introduces the project.
Here’s the rub: I’m migrating my Book Notes posts to a dedicated content type with its own metadata structures. The site is currently managed in Jekyll, so this is all happening in plain text — perfect for working with AI.
After this intro, the video switches to the first use case: tweaking the new content type’s Liquid template. I started here because this scenario can be easily replicated with other tools: basically, copying-pasting code from a chatbot.
The key difference here: I’m interacting with the LLM within Emacs using gptel. Nerdy stuff. (Although I don’t get into too many specifics of this setup in the video. It was getting too long as it was!)
Anyway, there are other videos in the works. And of course, this is a real change that’s coming to the site. I’d love feedback — please leave comments in the YouTube video.