Hacker News new | past | comments | ask | show | jobs | submit login

> The app uses algorithmic predictions, which Kevin Systrom sees as ‘the future of social.’

I see this as the anti-future. All I want from social (e.g. Facebook and Instagram) is a chronologically ordered list of things I chose to follow. That's it.




You know that joke that facebook is an over-engineered birthday reminder app? I would make the point that social media are also over-engineerd news feeds.

The vast majority of people get their news from some kind of feed. Be it Facebook, Instagram, Reddit, Twitter, whatever. Scrolling a feed is what gets news to most people. I can see the purpose of making an app that does just that, instead of having to also live with 3000 social features i don't care about (and all the noise and pointless posts they bring to the feed).

I also don't agree on the chronologically ordered list of things i choose to follow. I tried to do that and as soon as you put any kind of news outlet it's going to flood your feed of things you don't care about. I want to follow TechCrunch, Ars Technica or The Verge, but you can be sure that I don't want to see every article they post. I just want to have them on my radar for that more noteworthy stories. And I also want to discover new things, trending articles or topic that might have nothing to do with what i specifically follow.

That's why i also go on HN or reddit for news, because there's some kind of user curation where only the interesting articles ends up on my feed. I get to disover things that i don't already know and follow, plus i get curated news selection that should filter the noise while letting me know what's the hot topic of the week. But if the articles are selected by upvotes, then we have echo-chambers and both this site, reddit or any other social media are prime examples of this. You really get only a narrow and often biased slice. An app whose sole purpose is to give you news and articles you find interesting might have an incentive to do so without trapping you in a filter bubble. Or maybe not, only time will tell.


This approach is fine if it works for you and I'm sure many people want the same, but it's definitely not what I or the GP want. Facebook's hostility to chronological ordering was one of the big reasons I stopped using it way back when and I haven't used news feeds much over the years for the same reason.

I have a collection of bookmarked news sites that I check regularly, which show their articles without any personalized algorithmic reshuffling or hiding. Knowing that I see the same content as other visitors do is a plus for me. I don't need to worry about missing stories I care about because some AI decided the opposite, and I feel like it helps me avoid filter bubbles and clickbait.

This isn't to say that one approach is right or wrong, just that this is what works for me and some other folks.


I know the dubious security claims turn many off of it, but I really, really, really, really like the way Telegram handles lists of content.

You don't get a big pile of all of your subscriptions in a list, like a twitter feed. You get a list of all of your subscriptions, like a list of IM chats. Selecting one of them gives you a chronological list of posts from that channel.

I think the downside is there seems to be a limit to the number of chats a human can keep track of. Algo feeds let you subscribe to thousands of things without thinking about it too much, that would get overwhelming with how Telegram does things.


I want control over what I see.

Democratic design instead of authoritarian design is the future of social media to me. I am long on Reddit because it allows you to subscribe to specific subreddits. I am long on RSS - which has seen a resurgence - because of the control it gives you over what you see.

If only there was a "reddit for video" where I could see categorized videos instead of the garbage the YouTube algorithm gives me - of which I have almost no control over.


Devil's advocate, I'm going to guess you do not want exactly that and you'd be happier with a quality alg.

There is far, far too much information out there, and the things outputted from the things you follow is a 'decent start' but it's not what you want.

Systrom is exactly right and better algs is what we need.

TikTok is a great social medial platform, it's fun, irreverent, and the alg allows you to find niches in content that would not exist otherwise. I've found this guy who does videos on his sheep herding dogs, it's so fun to watch.

Applying a great filter to all of my youtube/news feeds is one of the most meaningful ways to improve my content feed.

Yes - we want to make sure people have the option to follow exactly what they want and hopefully to adjust the feed parameters ... but I would love to get rid of the fuzz in my feed.

This is decent innovation and it could affect a lot of people, it's fine to be a bit sus but I don't think we should be cynical.

You don't have to use it.


I guess this is my pessimism about an algorithm really ever working well for me. On FB and other social sites, I'm hesitant to engage any one-off content because my feed becomes flooded with it. I guess if I had more fine-grained control to say that I don't like something and WHY I don't like it, I'd have more confidence that this could be useful .


I mean, Instagram has a “not interested” menu item (behind …), with the additional option 'don’t show posts from ${account}' and “this post makes me uncomfortable”.

They also additionally introduced chronologically sorted “following” and “favorites” feeds. I haven’t had a Facebook in a decade, but after hardly using IG for years, I’ve started using it again. These features have honestly improved my opinion of Meta significantly, you can at least utilize it as a tool now.


My issue is that the controls aren't fine-grained enough for me. I'm not a heavy IG user, so that doesn't matter as much to me. (DISCLAIMER: The only reason I even have IG is because I worked at FB and my team made a couple components used in IG)


Agreed on FB but TT is another thing altogether. I guess Quality matters.


Makes sense. They are different use cases as well. On FB, I'm interested in people, not "content", whereas TikTok is about the content, which does require some means of discovery.


Interest based text recommendation is actually a proved idea in Chinese world, e.g. Toutiao as a personalized news aggregation app (https://techcrunch.com/2019/10/21/is-there-room-for-a-u-s-eq...), and Zhihu (a Chinese version of Quora)

Both apps are quite mainstream & huge DAU versus their English counterparts. The delta ingredient is likely alg.


> I'm going to guess you do not want exactly that and you'd be happier with a quality alg.

I'm going to guess that it's possible, if not likely, for me to have higher engagement with an alg while being less happy. A simple way to find out is choice. I like that Twitter has both modes.


You might like Mastodon, then? That's exactly what it offers.

(That it only offers a chronological feed is the main thing I don't like about it, to the point that I wrote my own client to implement rudimentary prioritization)


It depends on how they choose to tune their algorithms. If they tune for "engagement" with an infinite scroll that the user doesn't have any control over, then it's probably going to be another thing that makes the world worse while making them wealthier.

If they tune for quality, have a scroll that bottoms out when the next article no longer meets some high bar, and gives the user some control over the sources and presentation, then it could be decent.


That's what I want as well, and the Fediverse (Mastodon, etc) and RSS seem like the main things that fit that criteria. I don't use RSS for articles as much as I would like but I do use it for podcasts, of course, and the fediverse has been a fun experiment for me.


So true. It's really what killed Facebook for me. The timeline was great. The feed was horrible and still is.

I don't care about discovery unless it happens naturally, eg I go looking for something or a friend recommends it. Same with the HN feed, I basically consider most of your friends so it works really well. An AI that has a purpose of just making sure I stick around in the site as much as possible? No.

I've gone and found news sites that were the exact same: chronological content in text format. Hacker news is like a role model for websites for me.


Why do you want a chronological feed?

(Personally, and maybe this isn't great, but I use through social media when I'm bored and want to see something semi-interesting. Ordering by an estimate of interestingness seems like a no-brainer in this scenario.)


> All I want from social (e.g. Facebook and Instagram) is a chronologically ordered list of things I chose to follow. That's it.

Then you would like Mastodon/ActivityPub.


Humans are for training


I mean... isn't the solution just an RSS reader then? With like a system to recommend feeds if you want to get real fancy.


next the people will be asking to be able to right-click save a random meme


RSS does that pretty well.


That’s cool. The engagement metrics show more usage and ad clicks when the feed is ML-driven, so that’s what they’re gonna give you. Thanks for playing.


Agreed. One of my issues with Mastodon is that I cannot see content from people I don't subscribe to. Even if all of my follows like/favorite a toot, I won't see it. This is a flawed design. This is the main reason Twitter is better than Mastodon, IMO. It makes my Mastodon feed pretty calm, but it also feels a bit dead, which makes me wonder why I opened the app at all.


You do see it if they boost the post.

Also, since 4.0 you can follow hashtags. Currently a restriction on this is that you only see posts that made it to your instance because someone else on your instance is following the author. So it works better if you are on an instance with many users or users with similar interests to yours. It works very well for my use cases, most of the content I see is through hashtag follows.


I love this about Mastodon. I hated that Twitter showed people who follow me what I was liking today, and I hated that Twitter constantly showed me stuff people I followed were liking.

If I want my followers to see something, I have to consciously make a post that says "hey check out this thing". It's great! I open Mastodon, look at what the people I'm following are saying or pointing at, and then I close it and get on with doing something else instead of spending half an hour being distracted by Shit Twitter Thinks I Might Like.


I don't think this is true. I see "Boosted" posts in my feed from people I don't follow. Maybe this is a server specific feature? I am using Hachyderm.io.


This is a standard Mastodon feature. You will see post that people you follow "boost", but not things they only "favorite".

My understanding is that the Mastodon devs are trying to separate "increase the distribution of a post" from "give positive feedback to the post author".


yeah people can boost posts on mastodon (essentially a retweet), but you don't see the asinine algorithmic things like replies to people you don't follow (unless you go look for them)

I do like that there's no reply component to boosts though. You boost it up or not. No sniping.


"When a measure becomes a target, it cease to be a good measure."




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: