Oh, so it’s mostly a side effect, but they are still primarily being trained to predict the next word.

And the only solution to the dead internet theory is scanning our eyeballs for Worldcoin. There doesn’t seem to be any non-dystopian timelines in our future.

I agree, but that might complicate things. How about something like this?

Quality Engagement Score (QES)

QES = (PCM * AVU) / MAU, where:

• PCM = Posts + Comments per Month
• AVU = Average Votes per User (total monthly upvotes / MAU)
• MAU = Monthly Active Users

PCM measures raw activity, while AVU factors in community approval.

I appreciate your perspective, but my focus is on enhancing our measurement of community activity; if you have a more effective metric in mind, I’d love to hear it instead of just pointing out flaws.

30k communities and 9M posts per day. I find the number of posts per day very hard to believe. Each community would have an average of 300 posts per day, and most communities are abandoned. Maybe it’s the bot communities that repost all the Reddit posts that inflate the number so high.

I don’t like keeping duplicate files, especially in my main drive where I don’t store media. If I didn’t mind duplicate files, it wouldn’t be an issue.

If the files are literally duplicated (exact same bytes in the files, so matching md5sums) then maybe you could just delete the duplicates and maybe replace them with links.

If it was only a handful of ebooks I’d consider using symlinks but with a large collection that seems daunting, unless there is a simple way to automate that?

Automatically sorting books by category isn’t so easy. Is the metadata any good? Are there categories already? ISBN’s? Even titles and authors? It starts to be kind of a project but you could possibly import MARC records (library metadata) which have some of thatinfo in them, if you can match up the books to library records. I expect that the openlibrary.org API still works but I haven’t used it in ages.

If there’s still no simple way to get the metadata based on the file hashes, I’ll just wait until AI becomes intelligent enough to retrieve the metadata. I’m looking for a solution that doesn’t require manual organization or spending too much time. I’m wondering if there’s a way to extract metadata based on file hashes or any other method that doesn’t involve manual work. Most of the files should have title and author metadata, but some won’t. I’m not in a rush to solve this issue, and I can still find most ebooks by their title without any organization after all.

Yes, they are all media but they are not specific to a single type of media. Today I may want to find a book and tomorrow a song with the same program. So the files can be literature, audio, movies, series, etc.