It this point, I would probably invest some time into automating it even more. Download all articles, and save them into PDF. Create full text search over them. Use LLM for sentiment analysis...
I spend 3-10 minutes scrolling over at least 1/3 of 5000 headlines. Today I open 12 in tabs doing that, I've looked at 3 of those.
I got rid of most things, I only keep [pubDate, url, title] and only if the title doesn't contain any badwords and the pubDate isn't older than the oldest item.
Each item comes with a hours and minutes ago a translate to English button, the domain name, and a blacklist button.
While it does take opml most of my feed lists are just txt files with one url per line.
It loads one of those files then pulls x feeds per second and parses y per second where x and y are adjusted to the pending requests and pending parse jobs.
While you could certainly do many interesting things most of my experiments didn't survive my desire to keep it simple. I for example use to dump the results on a web page. I couldn't convince myself it was useful to the goal.
It this point, I would probably invest some time into automating it even more. Download all articles, and save them into PDF. Create full text search over them. Use LLM for sentiment analysis...
We did something like that at bitcoin boom.