5 comments

[ 2.6 ms ] story [ 14.6 ms ] thread
To be honest, it's jq and feedparser that are doing the heavy lifting, plus bat and pandoc for formatting.
Why have the LLM prompt in a github gist rather than in the script or a file beside it?
Is this rage bait for shell script people?
The framing as "bash pipes" is slightly misleading since jq and an LLM do the heavy lifting. But I don't think that's the point.

What I like is the philosophy: use an LLM as a personal filter based on a plain-text description of your interests. No training a recommendation algorithm on engagement signals. No black-box "for you" feed. Just "here's what I care about, show me relevant stuff." The pattern of "LLM as personal curator" seems underexplored.