First, I use regular bookmarks in Safari for pages I might visit "often," i.e. more than once. This makes them immediately searchable in Alfred and really fast to open.
Secondly, I use a multipronged Emacs system which allows me to easily file and sort links that I browse on a given day into an org-roam journal file. First, anything from Elfeed (rss reader) that I open or engage with gets logged. Second, I can save the currently open browser tab's URL with a keyboard shortcut. I've also mapped a 4 finger trackpad press to this which makes it super automatic and easy to do. I try doing this for everything I read and/or engage with, since I often find myself thinking "what was that one thing," and this has helped me find it on multiple occasions now. Third, there's a dedicated way to capture the URL with a PROCESS todo state tag. All of these show up in my org agenda buffer which means I see them all the time and go through them every once in a while. Fourth, a dedicated shortcut (and trackpad 4 finger force press) both logs the page and sends it to Wallabag, which is a sort of self-hosted pocket alternative. This is synced with my Kobo ereader (with the Koreader operating system) and I tend to read through the interesting articles at a pace only slightly slower than they accumulate, so it tends to be manageable.
For videos, I also have a quick docker service that allows me to yt-dlp a video, save it as mp3, and add it to an RSS feed for a podcast I host for myself, so I can listen to them on the go.
In general I'm trying to separate the act of seeing a piece of media to engage with and consuming said piece of media. One notable benefit is that something that might have seemed interesting later turns out not to be. On the other hand, logging most things I engage with comes in handy a lot when I'm talking to someone and something I barely paid any mind to in the past comes up!
In the future, I intend to expand the system with logging of the content as well as the URL, and adding embedding support with a vector database to allow for both full-text and semantic search. Retrievability is the biggest pain point right now (though ripgrep is already perfect for 80% of the usecases).
Since it's all integrated with org-roam, I can also quickly "promote" links to full nodes, link them to other concepts, etc. This comes in super handy with quickly going through a lot of literature by following arXiv feeds via RSS. Having it all in one system is super handy.