https://youtu.be/QEJpZjg8GuA

I'd rather they acknowledge Alec as the inspiration/source for this post at the beginning and explicitly, rather than just mentioning the video in passing midway through, but at least they do link to it!

With all the new stuff coming out in the LLM field, I've taken a cynically mechanistic view to this:

We're basically being conditioned by (the currently popular crop of) social media to work in very short context windows, which aren't sufficient for advanced reasoning.

So yes, totally. Turn it off and go read a book.

What we really need is "responsible" recommendation systems (that allow the joy of discovery while aggressively damping rage bait and extreme view points). They'd need to be trained with some kind of socially beneficial reward function rather than pure engagement or advertiser dollars.

Could such a recommendation systems operate on top of existing social graphs?

Not very healthy - it's like a never ending feed of "someone is wrong on the Internet".

For the record: "right" here is roughly equivalent with the political position of the US Democratic party.

Unfortunately these subreddits are not very balanced, so when I do take a break, I see that the other side "wins" to a noticeably larger degree. Again, small country.

I feel so much freer!

Getting away from the algos is untenable if you use the mainstream internet in any capacity. The trick is to be more intentional about gaming them to your advantage. My feeds usually surface things I value because I am deliberate about what signals I reinforce. I don't engage with content that outrages or upsets me, so it doesn't show me it. Some people are rage addicts and want to get into a doom loop because it fulfills their psychological sense of certainty that things will only get worse. Other people are ignorant of the algorithms and how they work so they never realize it is presenting a distorted picture.

Having varied pipelines for information intake is important too. Forums like this that are non-algorithmic, doing your own searches, visiting websites you like directly, all of this lends itself to that end. No need to go luddite if you like your internet things. Just be conscious of consumption.

I think there is an obvious answer though: taking control of the algorithm via AI. I don't think we're there yet, but it's gotta be a matter of time until somebody makes a local AI agent that browses all these feeds and then filters them to your satisfaction (x% about politics, y% upbeat, z% violence, z% about video games).

1. Use social media at only one particular time of the day. Inside a strict time window. That's it. Even if you are sitting idly on a car, traveling, or even standing in a queue for something, don't open the scroller app. Just be. Even if you are sick and in bed, don't open the scroller. Make a conversation, read a book, watch a movie, listen to a talk, but never open the scroller. This I learned from Cal Newport, and at the risk of sounding hyperbolic, I will say this has made my life better. And if you do this for some months, you will like the changes in your brain.

2. Don't consume any content without premeditated intent. Don't aimlessly scroll, ever. This point is there in the OP. I scroll a particular FB group only for ~10 ninutes a day. I scroll my very narrow CS/Math/Programming YouTube account for ~15 minutes and add videos to Watch Later. That's it.

LinkedIn already has this feature and it's significantly reduced the amount of rage-inducing influencer hot takes that show up on my feed. You can also turn off your watch history to get far fewer recommendations on Youtube.

I still personally find LinkedIn and Youtube to be a net-negative on my mental health, but these settings have helped a lot.

His more important point is that recommendation algorithms combined with replies create terrible conversations. That helps explain why it's so hard to discuss anything complicated on social media.

You also need filtering to remove the manufactured pop-culture dross like the Kardashians that is the new opium of the people.

i am really against this notion. why do we have to surrender ourselves to this forced reality. i really don't find any other beneficial use of social media except of advertise and making profit. (personally i see it immoral to feed this evil machine and run away). so the only real solution is to delete social media altogether or don't tap on user's backs and guide them to more moderate use because it doesn't makes a difference.

Its especially interesting recently as Youtube encourages you to search for something before giving you recommendations, so you get to "seed" your session with topics you like. If nothing comes to mind I'll just start with Practical Engineering and go from there.

The only downside is that I can't "like" content to help the creators, since I'm not logged in

I no longer see Facebook, Instagram or X and I’m okay with that.

I can't remove the apps because I might need them to check something important or write someone, so I forced to use my willpower to skip these videos everyday.

>Taste is an abstract, ineffable, unstable thing. A listener to music or reader of a book cannot truly tell if they will enjoy something before they experience it; pleasure in a piece of art is never guaranteed. So when encountering an artwork, we immediately evaluate it by some set of mental principles, and, hopefully, find the beauty in it, feel affirmed, even if we can't quite describe what that beauty is or how exactly we determined it in the first place. Taste is supposed to be ambiguous. As the Italian philosopher Giorgio Agamben summarized in his 1979 monograph on taste, "Taste enjoys beauty, without being able to explain it."

and then

>As taste requires surprise, it also thrives on challenge and risk, treading too far in a particular direction. Safety may avoid embarrassment, but it's also boring.

and so

>The bounds of aesthetic acceptability become tighter and tighter until all that's left is a column in the middle. While popular styles shift, like moving targets, the centralization and normalization persist.

[0] https://freetubeapp.io/

[1] https://blog.eutopian.io/building-a-better-linkedin/

My solution to the horrendous algorithmic recommendations that YouTube tees up these days is DF Tube (Distraction Free YouTube). It's a plugin available on the Firefox & Chrome app stores, perhaps others. As with all browser plugins, decide whether you're willing to trust the developer.

DF Tube goes the full mile by attacking the YouTube DOM and eliminating (with extreme prejudice) anything other than a search bar, a video, and core menu functions. That means the following are gone:

- Notification Bell - Feed - Autoplay - Trending - Comments/Live comments - Related videos

Here's a link to chrome: https://chromewebstore.google.com/detail/df-tube-distraction...

Here's a link to firefox: https://addons.mozilla.org/en-US/firefox/addon/df-youtube/

https://feedme.digital

Reverse chronological is sacrosanct and it will always remain ad free to keep incentives aligned with it being a place I like to spend time. Every tag also has its own RSS feed.

It's still invite-only, but anyone reading this page is obviously a great fit so here's an invite link: https://lynkmi.com/accounts/signup/?invite_code=333ee833-e3d...

If you took away the algorithms, I don't think you'd necessarily have a relaxing social media feed. You'd still have people sharing so-called "engaging" material and you'd still have to deal with "context collapse" and disagreeable discussions. I think the anonymous nature of online connections inherently make them more impersonal compared with actual face to face ones. And being constantly connected to other human beings digitally (even strangers) is incredibly addictive.

As a space fan, I necessarily see a lot of people on Twitter who are blindly pro- or anti-Elon. Both types get blocked, not because I disagree with them but because I don't need that sort of rabid content in my feed.

Quick edit to add: block all parody accounts on Twitter as well. They almost never are actually worthwhile.

Also, did you know you can block advertisers on Twitter? It's very catarthic.

I understand why this never happened (algorithms are, after all, optimized to the benefit of the company, not the user), but still it's a shame that this was never explored (at least not to my knowledge).

So far that doesn't include any algorithms I've run into.

Turned off "Discover" on my Android phone. Was weird first. I felt like I might miss out on something, some important bit of information. "Sometimes it does show me interested things" I thought. And, true, sometimes I get shown a scientific article that would missed otherwise.

But just like when I deleted my Facebook and Twitter accounts years back, I did not miss any important event.

https://fosdem.org/2025/schedule/event/fosdem-2025-5601-buil...

Only on Mastodon!

https://scrolldaddy.app

If you don’t have the ability to alter your feeds, taking a break from them is the next best thing.

Social media is a cancer. It only benefits those who have the money to power the algorithms.

I haven't been on any kind of Facebook/Instagram/Twitter social network for several years now, I also encourage people to try giving it up for 2-3 months, and I wager that it will feel distinctly odd and uncomfortable if/when you return. From a naive person's perspective, social media is flat out strange. It's got its own culture, lingo, and social constructs that feel a little odd. It's easy to see how it might be manipulative in a lot of subtle ways. But I do admit I lost touch with some people whom I miss. Each person should find their own balance with algorithmically determined content feeds, but I am a firm believer it should be, on balance, less.

Also, if you ever want to revisit a video, just use chromes history, but you'll find also this rarely happens if ever.

I read a lot of online content, from all kinds of sources. Different types of content: short-form, long-form, memes, WaPo editorials, sports, politics, tech, stuff, weird stuff, off-beat, serious, rants, opinions, facts.

The most delightful experiences I've had is when something totally random pops in from someplace. It could come from anywhere, but I've noticed that the best surprises come from places like longreads.com, which collect good writing across a diverse set topics and sources. Pretty much all social platforms do a horrible job at this, and recommend content that is so similar to content I've already consumed that the additional value of that content is extremely low, often even negative.

I think the ideal algorithm for me would be randomly suggestions after filtering out the garbage. No ads disguised as journalism, no influencer content, no clickbait, no spam, no AI slop, etc. I would jump on a platform that does this immediately. Even better, if the platform allowed me to control the knobs on what I consider to be garbage and not garbage.

The solution is very simple, just limit the like count of posts in your feed.

Any day now. Any day now.

Too much disinfo: community notes and grok are IMO just running cover for the disinfo firehose.

Saw the highest profile figure on the platform (yes him) retweet the most knee jerk takes that could be easily fact checked, but weren’t.

Instead of getting upset or trying to fight it, I yanked out the algo slop cable and am back in the real world. It’s great.

Edit: I didn’t really use it before 2024, so I cannot comment on what it was like under the last management.

I also tend to seek out conflicting views to my own when reading books, so it’s not that I’m just raging at ‘the other side’ either.

It sounds too invasive. To violating. Too extreme. Too much power in the hands of the button holder.

But this is effectively what we’re doing with our phones and watches in particular. It’s one of the reasons I’ve disabled notifications on almost everything.

The electrodes aren’t necessary.

They’re more crude but it’s the same idea. You tell a company you like a general topic by choosing a channel, magazine, or section of the newspaper. They give you content that is curated. This curation is automatic and sophisticated in social media apps but is more manual in legacy media, but the idea is generally the same. At some point someone or something is taking a guess at what kind of content you want to see.

All of these forms of passive content involve advertiser influence just the same as social media algorithms.

Maybe algorithmic media isn’t always so alarmingly bad? Maybe it’s okay to have some downtime and just allow someone else to fill that time? It was acceptable for the comics section of the newspaper to waste some time in your day but it’s so harmful for for some silly meme content to do the same just because technology was involved in the latter.

Socrates was even complaining about this, and it’s arguably far worse what is happening today than what he was seeing.

There will still be communities and fringe opinion and that is healthy. You won't have content generated just to push advertisements alone which is not.