Love this idea. I agree with the other comment that its not filtering well, but i hope you can get it dialed in.

Sometimes hackernews feels like the drive into SF. Just bombarded with Ai garbage billboards.

nice, yeah, I will keep working on it

I wanted to make this general purpose enough that it could handle large datasets (e.g. i want to do thiings like https://sprinklz.io/demo/headlineNews), so I was hesitant to do things like make 1-LLM-call-per-article, but I'll keep exploring ways to make it better