I wrote a bit about my techniques a few years ago:
I wrote a bit about my techniques a few years ago:
https://enkiv2.medium.com/against-trendism-how-to-defang-the-social-media-disinformation-complex-81a8e2635956https://enkiv2.medium.com/trendism-cognitive-stagnation-21c8e003df83
My favorite techniques for circumventing trendism (my term for the style of algorithmic recommendation that suggests what "people like you" think is "hot right now") involve randomness. While reverse-chronological feeds are not hyper-targeted to *you*, they are still subject to waves of cyclic interest that splash across social networks, and anything curated is going to operate the same way (for instance, whenever I hear about some obscure anime from the mid-70s on twitter, I see five or six new videos about it about a month later, because the people making those videos are following the same people on twitter as I am -- so even though the subject isn't "new", one person bringing it to light creates a micro-bubble of interest that is strictly less information-rich because of how it's located in time). Public library collections are the same way: there is a person in charge of deciding what books are available, and they make that decision based on an attempt to have something relatively recent about a large number of subjects that they aren't very familiar with. Likewise, if you select search terms, they are limited by what you happen to have at the top of your mind. But if you use one of the many third party programs intended to, say, play a randomly selected youtube video or a randomly selected song on spotify, you get exposed to something that, odds are, nobody within four hops of you in your social network has been exposed to at all -- something that literally would never appear to you under normal circumsances. You can also randomize search terms programmatically (for instance, searching for pairs of random dictionary words). This kind of thing *also* affects the recommender algorithms, injecting a certain amount of noise and play into a system that tends to narrow its model of you over time.
I also like to specifically omit material that's already highly visible, if metrics are available. Showing only tweets from your TL with no likes or responses, for instance, will scrub the activity that you're already flooded with and unearth the rare stuff. Often, randomizers will allow you to filter out popular material, and this is a great combination.
While libraries and new book stores (and, to a certain degree, used book stores) are selective about material in a way that can prevent you from being exposed to the really weird stuff, library sales (especially the big yearly ones) are more fruitful. There are a lot of reasons why books might be donated to a library & end up on sale -- someone might have bought a book and not liked it, or someone might have inherited lots of books from a dead relative or a sudden breakup. Popular books are somewhat underrepresented at library sales because people tend to *like* them in a totally uncomplicated way, and they also get bought up quickly, so the second day of a multi-day sale will have fewer. Books that are valuable also get nabbed up by nomadic bands of professional resellers who go from book sale to book sale, phones strapped to their arms, checking the price of everything with a barcode -- so stuff that has a reputation gets nabbed up, and stuff that has no barcode (either because it wasn't made via normal publishing channels or it was made before bar codes became common) stays available (and often deeply discounted) after the first day. I periodically go to book sales and spent $60, coming home with a hundred or more books I've never heard of; I try to read all the way through all of them, on the off chance they have something interesting, & most go right back to resale after.
Another thing I like to do is organize my reading based on arbitrary constraints. I have this giant hoard of obscure books -- so I can select a book to read at random using dice, or I can choose to read only authors whose names end in A, or only books with red covers. This year (aside from book clubs) I have been reading only books written by women (with extremely varied genres -- recently read a book on past life regression hypnosis and I'm working through one on designing taxonomies for business software).
A sometimes-neglected source of interesting ideas is juxtaposing your own past ideas in a randomized way. While reading, I keep lists of unusual or striking phrases. I have a script that recombines the words in this list in a random way, & I periodically look at the results of that, in case something interesting comes up. The phrases I write down are generally from books, so they are loaded with half-forgotten context. Scrambling that context, and joining it with some other context, can be very fruitful.