They simply need to change the way relevancy is measured. They need to implement some mechanism that can evaluate the quality of the page. The algorithm should penalize sites that have content very similar to other sites (like those that scrape github or stackoverflow), low effort sites, or sites that are infested with too many ads.
And since so much quality information is in youtube videos, and they already generate transcripts, why can't you search through those?
Thanks. The only one left for me is YouTube now. On a WAN show Linus asked Luke what product released less than 10 years ago by google he was using and they couldn't think of one. It was the same thing for me. I've been asking friends and colleagues ever since, the answers are interesting.