Curator

Curator is a just-for-me project that helps me find interesting links. It generates a list of recently-published content on the web that is similar and relevant to the articles, links, and quotes I post here.

Use case

Sometimes I fall behind on reading my feed subscriptions, and I want the best stuff to bubble up. Curator grabs the last X days of articles, compares them to all the posts from this website, and generates a Markdown digest of content related to my interests. At least, in theory.

Right now Curator is in an experimental form, and I’m still refining its inputs. This early iteration reliably turns out a couple solid links per digest, but it hasn’t surfaced anything unexpected. My theory is that the more feeds and posts on this site I have to add to it, the more likely it is to find something I might not on my own.

Screenshots

I run Curator at the command line and it generates a Markdown digest I open in Obsidian.

How I built it

Curator is a Python script that started as a ChatGPT brainstorming session on how to use ChatGPT to find good links for a blog, like any good snake eating its own tail.

It is a work-in-progress, and I’m still fine-tuning the prompt to help it understand more deeply what constitutes a link being interesting.

This is the first time I’ve used the ChatGPT API to compare large sets of text and it was very easy if not immediately effective. Also, I had to optimize this script to cache aggressively make as few API calls as possible so as not to blow through my limits too quickly. I’m going to keep playing with this one and see if I can get it to the point where it turns up several post-worthy links.

Back to Projects