What's interesting for ten years is likely really interesting. I often find a better selection of links on the third or fourth page of HN thanks to their algorithms, of stories that have held position for several days. Perhaps these are more evergreen topics than faster rising and falling front page submissions.
Essentially you are filtering hacker news content based on the type. Why not simply allow the user to choose what kind of articles they want to see? For example I'm only interested in the latest trends so I should be able to see just that and no other content at all. This could increase the number of users interested in your app
I like the idea and would be happy to see it expanded upon. I know that HN already downranks posts that get a lot more comments than upvotes because these tend to be highly controversial, political etc. I sort of wish those posts were not in my default filter at all, it's easy to get sucked in but it's rare that I learn anything valuable from them.
I'm pretty sure some topics are just outright muted to prevent abuse (not that it's a bad thing).
For example, if you were on /new over the past 6 months, it was extremely common to see a "Twitter"-titled article get 20+ upvotes and never see the frontpage. People were outright obsessing over Twitter, posting everything they could find, no matter how benign or uninteresting it's content was. Locking the topic until verifiably interesting things happen was a great solution. It probably works similarly for other keywords, too.
This is old code, and the mods make tweaks here and there without warning, so the details may have changed. (I'm not sure if gravity changed from 1.8 to 1.6 (???).)
I didn't test it personally, but it feels like HN is using something very similar. And there are a few black box analysis of the sorting of the front page that got similar results. Also, minimaxir is the kind of person that is probably running an script to use the HN API to verify the claims.
I don't truly see value in someone else hand-picking a few stories and assuming those are the ones I want to read. HN has a simple, easy UI to pick a few stories to read. If there are topics I don't want to read... I just don't.
Basically, if you are trying to solve the problem of "I don't want to read news", that is a different problem than "I want a hand-picked list of stories".
HN is a bit of an echo chamber at times, but I still like it. I think with collaborative filtering though HN would end up a filter bubble and the users would be worse off.
One kind of collaborative filtering algorithm is "Personalized PageRank" (PPR). Like the regular PageRank you do a bunch of random walks on the graph of users and items where the upvotes are the edges of the graph. What makes it "Personalized" is that you start each walk from the user you are making recommendations for.
If you limit the walk to just 3 steps then it becomes "users who upvoted what you upvoted also upvoted this".
What I like about it is that each recommendation is supported by concrete shared liked items which makes each recommendation easily explainable to the user.
I implemented this algorithm for my project (https://linklonk.com/item/3292763817660940288) that includes RSS feeds as sources of information and it's been working well as a prioritizing RSS reader. It would be great to see this algorithm tried on a bigger website like HN where there are so many upvotes/items/users to work with.
In that case every user would be a curator of content that they have chosen to upvote. If you happen to like someone's upvotes (ie, you upvote the same content as they do) then you would see their other upvoted items ranked higher for you.
The OP is just one curator that has decided to make their liked content public, but PPR would make every user of HN a potential curator of content for you.
Filter can't control the source, only what to come through. "Garbage in, garbage out" is harshly true. I wish for a lot of Show HN and its traction gains.
Perhaps you don't see the value because you don't value your time in the same way that others do. Skimming through a bunch of titles I'm not interested in does have a cost. The cost is exacerbated if viewing the titles via the New view which includes lots of spam. Also every news site displays hand-picked articles so nothing new there. I've often wished HN would implement a basic story categorization feature so that users could specify their preferences which would apply a basic filter to the homepage story list. Anyway kudos to the author of this new tool!
I question anyone willing to browse New also values their time in wading through all the spam just to be one of the first commenters in hopes of getting internet points no one cares about.
To add to that, there is also a lot of clickbait that, for many people, if they see it they are going to jump in, read the discussion, and comment, even if it's not particularly healthy. "Hide" can work for this, having a way to never see it in the first place could be better, for some
Without analyzing how useful it is, I clicked in and found a few links I missed on HN before but very interested in. I'd say it's useful at lease for me today, which is a pure plus
I don’t mean to be rude but if you don’t see the value, just don’t use it? You’re perfectly fine not reading stories on HN that you don’t care about, so just add this site to that list and move on. Not everything has to solve YOUR problem.
The '10 years ago' is interesting. Could be fun if you add also 3/5/7/10 years tab. It's always fun reading content from The 'About' and 'Privacy' buttons are not working.
Consider adding 'archive' section where we could quickly see older newsletter issues
HN really is a wonderful resource if you look beyond the current trends. How are you curating the posts -- do you have a guideline on what makes it on the page?
I built something similar a while back too, indexing all posts with (YYYY) in the title [0]. Let's do more projects like this!
It's not me mining the gold, time is. It's the best filter of all :)
Good luck with your project! I imagine you have to click through quite a few links to curate the site every day, hopefully it saves the effort for other people.
Ask HN: Can use the #1 supercomputer for any project I want. What should it be?
hn-miw-i on Jan 12, 2013: Bitcoin. Pay off your student loans one Merkle tree at a time.
hn-miw-i on Jan 12, 2013: Ok well blocks generate at an average of 1 / 15 minutes, so about 96 blocks a day. Each block generate 25 BTC reward. At ~14.00 USD/BTC on MtGox right now that about $34k.
I know I’m speaking from hindsight, but if you’d told me the largest BTC exchange is based on Magic: The Gathering, I would’ve known nothing good would come of this.
I think it would be useful to give information on the selection criteria and process (automated/manual) - I am assuming by the subtitle this is a manually curated selection.
I'm thinking of curating a list of curated lists of interesting technical articles. But I'm worried someone else already did this, because then we'd need a list of lists of lists and we all know what would come next.
* I don't really see the value-add of a curated list of what is already a curated list. Down the rabbit hole...
* The site's definition of what is included and what is excluded is extremely vague. "general news", "unique views", "interesting stuff". I mean come'on...
* At least HN has a relatively excellent definition of what counts as acceptable content (relative to other social media platform's definitions, which are deliberately vague to allow complete hegemonic content control).
* If I had a buck every time someone said "Hacker News needs to be more hacker and less news", only to realize that their definition of "hacker" and "news" was very unique to them and so supposing that there should be some filter to "objectively make HN better" was actually very subjective and of little use as a publicly available tool.
IMHO, this should have been a private tool of theirs. They could have probably achieved their subjective goal of "more hacker less news" with a little script to filter out HN posts for specific domains and keywords.
It's concerning that you have such a low bar for using such extreme language, which seems a bit tone-deaf given the point you are trying to make. Couldn't you have phrased that in a less flamebait way?
To elaborate and address your valid point, however: OP implies a confusion between the subjective and objective. The site makes statements in an objective way as if they are facts about what is "news", "interesting", and what is bad content. Being a publicly available tool reinforces that implication further.
My final takeaway point, as explained by all the previous points before it, was that this is incorrect because the OP's objectively described views are in fact very subjective. This is one in a long line (dozens?) of "HN but more X less X"es, and most make this mistake, which is why they never gain any ground.
I could have been less discouraging, however, so I'll try and take that onboard.
Hackers building out their own private versions of the site, an invisible weave atop the common view: events for folks they follow to chat with, profiles in an iframe with custom links--personal pet features.
Remember dark matter developers? Who knows how much of our echoes reverberate in the space of a holographic universe, streams consumed in a wrapped-app above.
So the social piece is the architecture to enable it, but not published, to make a kind of ownership stickiness...
96 comments
[ 2.5 ms ] story [ 336 ms ] threadEDIT: There is a typo in the word "business".
See https://en.wikipedia.org/wiki/Lindy_effect
But I don't know if this is based on evidence.
For example, if you were on /new over the past 6 months, it was extremely common to see a "Twitter"-titled article get 20+ upvotes and never see the frontpage. People were outright obsessing over Twitter, posting everything they could find, no matter how benign or uninteresting it's content was. Locking the topic until verifiably interesting things happen was a great solution. It probably works similarly for other keywords, too.
This is old code, and the mods make tweaks here and there without warning, so the details may have changed. (I'm not sure if gravity changed from 1.8 to 1.6 (???).)
I didn't test it personally, but it feels like HN is using something very similar. And there are a few black box analysis of the sorting of the front page that got similar results. Also, minimaxir is the kind of person that is probably running an script to use the HN API to verify the claims.
Basically, if you are trying to solve the problem of "I don't want to read news", that is a different problem than "I want a hand-picked list of stories".
https://en.wikipedia.org/wiki/Collaborative_filtering
It would be nice to define "local overrides," though. Much like how we have /etc/ configs, then dotfile, env vars, and then options on the CLI.
If you limit the walk to just 3 steps then it becomes "users who upvoted what you upvoted also upvoted this".
What I like about it is that each recommendation is supported by concrete shared liked items which makes each recommendation easily explainable to the user.
I implemented this algorithm for my project (https://linklonk.com/item/3292763817660940288) that includes RSS feeds as sources of information and it's been working well as a prioritizing RSS reader. It would be great to see this algorithm tried on a bigger website like HN where there are so many upvotes/items/users to work with.
In that case every user would be a curator of content that they have chosen to upvote. If you happen to like someone's upvotes (ie, you upvote the same content as they do) then you would see their other upvoted items ranked higher for you.
The OP is just one curator that has decided to make their liked content public, but PPR would make every user of HN a potential curator of content for you.
OTOH since it's hand-picked I have no expectation it'll last very long.
The '10 years ago' is interesting. Could be fun if you add also 3/5/7/10 years tab. It's always fun reading content from The 'About' and 'Privacy' buttons are not working.
Consider adding 'archive' section where we could quickly see older newsletter issues
HN really is a wonderful resource if you look beyond the current trends. How are you curating the posts -- do you have a guideline on what makes it on the page?
I built something similar a while back too, indexing all posts with (YYYY) in the title [0]. Let's do more projects like this!
[0] https://hn.lindylearn.io
About the curation: it's just whatever fits my taste.
Good luck with your project! I imagine you have to click through quite a few links to curate the site every day, hopefully it saves the effort for other people.
Ask HN: Can use the #1 supercomputer for any project I want. What should it be?
hn-miw-i on Jan 12, 2013: Bitcoin. Pay off your student loans one Merkle tree at a time.
hn-miw-i on Jan 12, 2013: Ok well blocks generate at an average of 1 / 15 minutes, so about 96 blocks a day. Each block generate 25 BTC reward. At ~14.00 USD/BTC on MtGox right now that about $34k.
I clicked the About link, but it didn't activate.
* I don't really see the value-add of a curated list of what is already a curated list. Down the rabbit hole...
* The site's definition of what is included and what is excluded is extremely vague. "general news", "unique views", "interesting stuff". I mean come'on...
* At least HN has a relatively excellent definition of what counts as acceptable content (relative to other social media platform's definitions, which are deliberately vague to allow complete hegemonic content control).
* If I had a buck every time someone said "Hacker News needs to be more hacker and less news", only to realize that their definition of "hacker" and "news" was very unique to them and so supposing that there should be some filter to "objectively make HN better" was actually very subjective and of little use as a publicly available tool.
IMHO, this should have been a private tool of theirs. They could have probably achieved their subjective goal of "more hacker less news" with a little script to filter out HN posts for specific domains and keywords.
This is the most depressing toxic response masked as constructive criticism that I've read on this site in, well, at least a day.
OP should ignore this and continue to put your ideas out there even if they may not be perfect or match someone else's.
It's concerning that you have such a low bar for using such extreme language, which seems a bit tone-deaf given the point you are trying to make. Couldn't you have phrased that in a less flamebait way?
To elaborate and address your valid point, however: OP implies a confusion between the subjective and objective. The site makes statements in an objective way as if they are facts about what is "news", "interesting", and what is bad content. Being a publicly available tool reinforces that implication further.
My final takeaway point, as explained by all the previous points before it, was that this is incorrect because the OP's objectively described views are in fact very subjective. This is one in a long line (dozens?) of "HN but more X less X"es, and most make this mistake, which is why they never gain any ground.
I could have been less discouraging, however, so I'll try and take that onboard.
Hackers building out their own private versions of the site, an invisible weave atop the common view: events for folks they follow to chat with, profiles in an iframe with custom links--personal pet features.
Remember dark matter developers? Who knows how much of our echoes reverberate in the space of a holographic universe, streams consumed in a wrapped-app above.
So the social piece is the architecture to enable it, but not published, to make a kind of ownership stickiness...
The first rule is...
“I hate the news”
https://news.ycombinator.com/item?id=5051902