[PaulHoule]

The hate for algorithms boggles my mind.

Chronological feeds are awful. You'll never see anything from the people who post occasionally because they get drowned out by the people who are posting all the time.

There may be some algorithms that deliberately magnify hate because that's a way to increase engagement, but if you want to create one of those algorithms you can make a training set based on chronological feed + boosting/retweets/reposts.

I'm amazed at how people keep making failing RSS readers that keep failing with the same failing user interfaces that have been failing since 1999; everybody knows RSS has been failing but they never ask why or if we have a choice.

We still see the readers that make you mark things as read, that take their cues from email and newsreaders, that, when you subscribe to N feeds show you N boxes with a list of items, etc.

My RSS reader works like TikTok because I'm not afraid of algorithms.

[openrisk]

> You'll never see anything from the people who post occasionally because they get drowned out by the people who are posting all the time.

If you control the algorithm there is nothing to prevent you from sorting feeds by volume and adjusting their presentation accordingly.

Actually you can imagine countless other UI adaptations depending on preferences, usage patterns etc.

Ideally RSS readers should offer flexible customizations, e.g., with plugins or even some low-code environemnt.

[PaulHoule]

It's one thing to sort a list of things by one criteria, say "a". If you have two criteria, say "a" and "b" you can sort by "a" and break ties by "b" or you can sort by a+b or a-b or something like that... But that's not the same as optimizing both things. Maybe you can say "there are these 5 people who are special to me who I never want to miss their posts" but I think people would struggle to maintain rulesets and might not really be happy with the results they get.

In search rankings for instance you night have a document score (like Pagerank) that tries to identify the quality of a document, and you might have a query-document score that identifies the relevance of a document to a query. It's not trivial at all to find a way to blend those that gives you queries that are both relevant and quality as opposed to just one or the other.

The greatest weakness of my current RSS reader is that it's slow, depending on how much I am using it, articles could be delayed anywhere from a day to a week. For certain kinds of articles [1] recency doesn't matter, but other articles [2] have a definite shelf-life and if you repost them too late you look like a total dope.

I'd like to let articles about sports "cut the line" in front of higher quality articles about other topics but it's really hard to find a balance that's right because I don't want to get flooded with lower quality sports articles. It's one thing to say "let people make up their mind about how to balance these things" but when you really try it you find it's pretty hard. Not only do I have to change the whole way my pipeline works (can't be a batch job anymore) but it's not clear to me how to tune up the selection criteria.

[1] https://www.sciencedirect.com/science/article/pii/S294979062... [2] https://www.usatoday.com/story/sports/nfl/2025/01/19/eagles-...

[openrisk]

If I understand your challenge correctly it could be helped with a reader that gives a taxonomy to classify feeds, as you then get implicitly multiple timelines - one per feed category. Linux desktop readers like akregator and the now abandoned quiterss provide that and it works fine.

[beej71]

I prefer time-based feeds, but that's just me.

I don't think people are objecting to suggestive feeds in general; they're objecting to suggestive feeds whose primary objective is to keep you scrolling for as long as humanly possible to maximize company revenue.

That's simply not what I want to do with my day.

I'd train my own if I wanted to go that route.

[PaulHoule]

The problem with time based feeds is this.

You can read x articles a day; your system ingests y articles a day.

x=y is perfect but requires close-to-perfect balance (if x=0.9y to 1.1y maybe you can adjust your reading habit to your your feed)

if x>y then your system isn't showing you enough, if y<x you are going to miss things you subscribe to based on some arbitrary or random characteristic.

With an algorithmic feed of some kind you choose to read x items a day, your system shows you the best x items a day out of y based on some set of criteria and constraints.

These things are common sense but seemingly nonsensical to a lot of people. For instance our impoverished rights-based discourse (see [1]) about "free speech" presupposes that 100% of people can read 100% of what everybody else posts, realistically platforms can only show people some fraction of what gets posted so one thing is going to get more visibility and other things get less and that's a choice -- it could be random but it's still a choice. (As Rush would put it, "if you choose not to decide, you still have made a choice")

I think the discussion is so impoverished that we never hear that an algorithm could choose to do anything other than maximize profits for a platform, when in fact that is just one thing an algorithm could do out of countless options.

[1] https://www.amazon.com/Rights-Talk-Impoverishment-Political-...

[palata]

> With an algorithmic feed of some kind you choose to read x items a day, your system shows you the best x items a day out of y based on some set of criteria and constraints.

And the whole point of the people who want RSS is that they disagree with the idea that "the algorithm" shows the best x items.

The whole point is that they don't want "the algorithm". And your answer is "Have you heard of the algorithm? I think it's a solution to your problem".

[thuanao]

For most people y > x. You're choosing what posts to see even if the feed is chronological. For a chronological feed, x is the most recent posts you happen to see when checking your feed reader. And x will skew towards people who post at the same time of day you tend check your feed reader. People who post more will also take a larger distribution of x.

Any algorithm would seem to be an improvement on that, right? For example, an algorithm that sampled evenly from all sources. That would at least mean x was distributed across all your sources, so people who post more didn't crowd others out.

[ryukafalz]

> if x>y then your system isn't showing you enough

We've grown used to having a practically infinite amount of things presented to us all the time, but I've come to believe recently that limiting the amount of posts you see can be a fine (and maybe healthier) choice. There's value in being able to say "I'm caught up, I'm done", it can be a natural stopping point.

I'm not as good about limiting myself like this as I would like to be, but it is a goal of mine.

[PaulHoule]

The total amount of content available is y>x.

Some arbitrary process has to reduce the 1,000,000,000 pieces of content produced today to maybe 10 or 100 you can handle.

The very concept of "news" ("fake news" is bad because it is news not because it is fake) is an act of violence against the fabric of reality because a huge number of things happened today, but CNN reported the same 10 of them 100 times. They could maybe cover 1000 different news stories (still a fraction of maybe 100,000 things that happened) but that's a non-starter because someone who watched the network at 3pm would have a totally different impression than someone who watched it at 5pm.

[beej71]

> Some arbitrary process has to reduce the 1,000,000,000 pieces of content produced today to maybe 10 or 100 you can handle.

In my case, I'm the process that does the reduction by subscribing to specific feeds. (Not sure if that's "arbitrary", but I could see it argued either way.)

I admit I miss some content that I would probably like to see. :) But that's just life how it's always been.

As my wife put it, "You don't have enough years in your life to read all the good books, so stop wasting time with bad ones." That's how I think of it. A handful of high-quality, low-volume feeds is great. (But I also follow HN and my curated Lemmy feed, so I get some of the firehose.)

The only "news" I have in my feed is local news. Any interesting national news hits HN or Lemmy.

Again, that's just me. I respect your different usage pattern.

[Eji1700]

> You'll never see anything from the people who post occasionally because they get drowned out by the people who are posting all the time.

I use inoreader and track basically everything I want in a huge list of feeds.

It's pretty trivial to mark stuff as "always flag this", and then leave the rest of the pile as "scan through and manually tag anything else"

Anything low volume goes on the "always" flag list as it takes one review a day of the new content to decide if it's something I care about.

Ultimately what I want, and what most people want, is the ability to just hook up to various data streams and apply rules to it. From there filtering as desired comes pretty easily.

[josephcsible]

> Chronological feeds are awful. You'll never see anything from the people who post occasionally because they get drowned out by the people who are posting all the time.

My preference would be chronological feeds, but following few enough accounts that I can see every single post in it. Then that's not a problem.

[ahalimah]

Which RSS reader do you use in that case?

And is it fair to say there's a middle ground between purely chronological feeds vs algorithms that reward engagement/time spent on the app?

[PaulHoule]

An algorithm doesn't have to reward engagement or time spent on the app. It can reward or punish anything at all -- and that's what it is about algorithms.

My feed reader (that I wrote myself) rewards things that I thumbs up and punishes things that I thumbs down.

[bluebarbet]

Is this mysterious feed reader one that you made yourself?

I agree about the need to have some means of filtering. This is a major weakness of RSS, alongside onboarding and discovery. Algorithms serve the purpose of filtering well. The problem is their opacity.

[PaulHoule]

What people are calling an "algorithm" is really a heuristic.

In the public discourse over "algorithms", the definition seems to be "something that maximizes your (time on site|outrage|clicking)" and not the real definition

  In mathematics and computer science, an algorithm (/ˈælɡərɪðəm/ ⓘ) is a 
  finite sequence of mathematically rigorous instructions, typically used 
  to solve a class of specific problems or to perform a computation. [1]

By that definition any well-defined process that creates your feed (sorting chronologically, alphabetically) is an "algorithm"; in the machine learning age it is easy to make an algorithm that selects for anything at all, except for the capital-T Truth.

[1] https://en.wikipedia.org/wiki/Algorithm

[Havoc]

People aren’t afraid of algorithms…just ones that are Blackbox and controlled by someone else who’s motivations don’t align with yours.

Being concerned about that is downright logical.

[nonrandomstring]

Let it go Paul. It's like the word "hacker". We ain't getting it back.

It's sad the popular perception of a fundamental computer science concept, the great progeny of Al-Khwarizmi Musa, the life work of heroes like Donald Knuth, is sullied and stained with dire B-movie comic-book Bond villainy. Thank the hooligan broligarchs and their manchild ambitions for trashing that bit of computing culture.

Even my mum spits in the dust like a cowboy at the sound of "Algorithm".

[PaulHoule]

It might be easier to convince the people at the CS department to name the "algorithms" class something else.