Live data from Hacker News

Show HN: Wikipedia as a doomscrollable social media feed

xikipedia.org

51–60 of 151 posts

Re: Show HN: Wikipedia as a doomscrollable social media feed

#52

Earlier quoted context omitted.

I believe in privacy but generally people are fine with rec algorithms running on a server if it's transparent enough/self hostable. Mastodon/DuckDuckGo/HN/etc all don't need to download a huge blob locally. (If you do want it to run locally, hosting the blob on a CDN or packaging this as an app and letting someone else host it would probably improve the experience a lot)

Mastodon/HN do not have a personalized weighted algorithm. On HN you see what everyone else sees, and on Mastodon the feed is chronological. DuckDuckGo offers some privacy, but still sends your search queries to Bing. Also, all three of the examples are projects that have years of dev effort and hosting infrastructure behind them - Xikipedia is a project I threw together in less than a day for fun, I don't want to pu…

I really like how you have done things. Didn’t mind the waiting time.

Thank you for making my day a little brighter.

Re: Show HN: Wikipedia as a doomscrollable social media feed

#54

Earlier quoted context omitted.

> Why are you serving so much data personally instead of just reformatting theirs? Because then you only need to download 40MB of data and do minimal processing. If you were to take the dumps off of Wikimedia, you would need to download 400MB of data and do processing on that data that would take minutes of time. And also it's kind of rude to hotlink a half a gig of data on someone else's site. > What am I missing? 4…

no but...why are you passing 40mb from your server to my device in a lump like that? All I'm getting from your serve is a title, a sentence, and an image. Why not give me say the first 20 and start loading the next 20 when I reach the 10th? That way you're not getting hit with 40mb for every single click but only a couple of mb per click and a couple more per scroll for users that are actually using the service? Look…

Who made you do anything? It's a fun website. If you don't like it, move along or make one yourself. I could understand if you were paying for something, but this is free.

Re: Show HN: Wikipedia as a doomscrollable social media feed

#55

Please fix the loading issue and I’ll return! I think you don’t need to pull all the data at initialization, you could lazily grab a couple from each category and just keep doing it as people scroll.

The loading issue is just a hug of death, the site's currently getting multiple visitors per second, and that requires more than a gigabit of bandwidth to handle. I sort of need to pull all the data at the initialization because I need to map out how every post affects every other - the links between posts are what take up majority of the storage, not the text inside the posts. It's also kind of the only way to prese…

I think I'm missing something, but does every user get the same 40MB? If so, can you just dump the file on a CDN?

Re: Show HN: Wikipedia as a doomscrollable social media feed

#57
post #18

This is a very neat idea. I am not sure why the page needs to load 40mb of data and make me wait 5 mins before the first view. I'd probably also add some ranking criteria to surface good quality articles that maximize the "I learnt something new today" factor. Overall kudos to the developer for original thinking.

Presumably the 40mb of data is not from Wikipedia, but the Javascript tracking code bundle needed to turn it into a doomscrollable social media feed. ;) By those standards, I think it’s pretty lightweight! For comparison, the Instagram iOS app is 468.9mb, more than ten times the size…

Re: Show HN: Wikipedia as a doomscrollable social media feed

#60

How does it actually work? Can you add an "about" page that goes into the algo? Or can you add more info on the readme on github? I'd love to learn more.

I might add a proper explanation at some point, but for now you can view-source the page and read the code, there really isn't that much of it.

Impressive! We're a university lab and published recommendation algorithms. Never knew that doomscrolling could be this addictive this fast, thnx!

Please consider taking an hour and push this to a Github with quick readme. Scientists and developers would get it. We have been building a torrent-based alternative to Youtube for a few years. Not many knowledge out there around operational frontpage algorithm.

Edit: https://github.com/rebane2001/xikipedia

Post reply on HN