Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: I made Hacker News but for research papers (paperlist.io)
287 points by kashnote on Feb 13, 2023 | hide | past | favorite | 89 comments



What you'll discover is that the moment this falls off the HN front page, your traffic goes to zero. You'll need to manually recruit users, every day, typically by going to wherever they are.

Twitter is your friend here. DM everyone you can, and keep checking in with them to see if they're getting value out of it.

My personal experience was that I showed up, was delighted to see so many relevant papers, but then lost interest due to zero discussion.

That doesn't mean that it's not interesting. It is. But when you're making a social network, you're fighting a massive amount of inertia. The default case is for no one to come.

Give us a reason to keep coming back, and we will.

FWIW I speak from experience, having spent a few years running an HN clone that eventually tapered off.

EDIT: Also, your upvote algorithm needs a bit of work. The most upvoted papers are currently hidden on the second page. And in fact, at this stage, pagination doesn't make a lot of sense -- just show all the submissions.


All good points. To be honest, I hate Twitter so I don't know if I'll be trying to share it on there, but I agree that promoting discussions is the way to go.

PS Fixed the voting algorithm (was using ASC instead of DESC). Probably needs more work but thanks for pointing it out!


Yeah, the reason people keep coming back to HN is that there are different stories every day. That's the heart of its success. If you want this to succeed, you can't rely on simply sorting by asc, nor can you rely on community upvotes to do the work for you. You'll need to manually place papers on the front page, just like HN does (in its semi-automated-but-human-driven way).

And get ready for it being a slog. It's seriously soul-grinding to try to curate a collection of new things each day and see only a handful of people come. But you have to do it if you want to succeed long term; there's no other way. I was able to keep up with HN for about four months, before basically collapsing of exhaustion. Then traffic growth stopped, and never resumed.


+1 to @sillysaurusx's comment. I'd like to add on a recommendation to read The Cold Start Problem by Andrew Chen. It goes into the trials and tribulations of creating a product dependent on network effects.

More than just a problem of no longer having an "artificial" influx of users via HN, you will need to figure out how to build what Chen calls an "Atomic Network" of users who are self-sustaining on the platform. The first step to that end will be figuring out how to attract the "hard side" of the network (i.e. people who find good papers and post them). What keeps them coming back?

I'd highly recommend considering what the smallest network of users would be that could sustain something like Paperlist and what features they might want.

Some questions that might be helpful to get started: * What incentive does a given user have to return to the site? * How would a user be drawn back to the site, e.g. RSS, notifications, etc.? * What incentive does a user have to contribute to Paperlist over submitting to HN, reddit, etc.? * How do you attract a new user the to site? Put another way: how do you increase the "virality" of Paperlist? E.g. why would someone link to Paperlist today instead of Arxiv?

I hope this is helpful.


An RSS feed would be helpful. That's how I follow Hacker News (news.ycombinator.com/rss).


I think it might be worth partnering with someone who specializes in community building, then.

The easy part is what you’ve done already. It truly looks awesome, but unfortunately it really is just the beginning. The much, much harder part is getting people to keep coming back and contributing themselves.

There’s nothing technologically impressive about HN, the reason we’re here is because everyone’s already here, and the reason everyone’s already here YC and Paul Graham. Without huge names like that behind your project, you’re going to be growth hacking and bootstrapping.


> I think it might be worth partnering with someone who specializes in community building, then.

Curious if anyone has any concrete experience on what this step looks like / how thats done.

And for whatever it's worth I agree on some level that building the site is the easy part but I still remember when building a functioning well-designed web app seemed daunting / impossibly complex and it still is to most people. A beautiful working site like this is always an achievement.

[speaking to myself mostly] I think anyone that can build social platform alternatives like this should as often as possible to create a richer ecosystem less in danger of being dominated by abusive / destructive monopolies.

Were still in the car brand phase of social platform options when it seems like we could be at a artisan coffee roasters level.

I want to find my next social media app at a farmers market.


loll if you're not willing to do whatever it takes to promote your site and get engagement it'll die. I also dislike Twitter, but it is the perfect place to promote this.


I wonder if bots that post about the trending papers on Twitter, Mastodon etc. can drive more consistent engagement. I will forget to visit this website, bit would definitely subscriber to the bot and likely end up browsing more of the side when I visit after having followed the social media post.


Did your platform permit LaTeX in comments?

To get profound discussions, the discussions need to be able to delve towards the fundamentals, i.e. towards both mathematics on one hand and the basic sciences on the other (physics, chemistry).

I yearn for a HN for scientific papers, and while there are often comments by actual physicists, mathematicians, chemists,... on HN, there is a lot of "science fan boy" noise.

If actual mathematicians and scientists can not use the notation systems they are familiar with, the platform is doomed to word salad.

Growing up I once saw a piece of graffiti in the city's dead center, it read either:

"Every system creates

the bullshit it deserves"

or:

"Every system deserves

the bullshit it creates"

I can't remember which exactly.


100%, my experience getting to front page of HN twice with https://sqwok.im

I found Twitter to not be a great place to recruit new users, granted it was probably partly because I was using the official account for the site, not a personal account with my face, and the nature of "real-time" means w/o coordinated events where people show up at the same time, the bounce rate is high.

What was the hn clone you built? Still working on community site(s)?


What you'll discover is that the moment this falls off the HN front page, your traffic goes to zero. You'll need to manually recruit users, every day, typically by going to wherever they are.

Yeah front page on HN = high bounce rate. You need more to keep 'em coming back.

I don't think this is that good of an idea, imho. arXiv and SSRN already do this. There are no shortage of paper syndication and aggregation services.


> I don't think this is that good of an idea, imho

I don't think it necessarily is a bad idea either. The problem is that you can't just make a simple looking website and call it a "HN for X". HN is nothing without the users and comments. So if the idea is to bring the research community to the site itself, the idea is.. good? At least in theory. It might still be impossible in practice.


A narrower, niche focus helps, because easier to get a higher concentration on fewer things. Then you want a natural way to expand the focus. It was easier for reddit and HN, without reddit and HN to compete with.


That's not entirely true. Reddit when was launched was the underdog to a very popular Digg, and maybe to a less extent Slashdot. Reddit gained popularity when Digg updated their interface to something that most users disliked strongly. And HN had reddit as a competitor right from the start - maybe not the same reddit of today, but still notable.


A bit crass these days, but this is probably where Reddit-style fake accounts will probably shine. Maybe a mix of manual work + judiciously crafted chatGPT prompts?


I've thought about fine tuning a GPT using stylometry to sound like specific HN users. You'd have to do that if you want it to succeed, since otherwise it'll all sound same-y.


I know these are all arxiv.org so far, but if you want to include research papers beyond that plaform, can I encourage you to use DOIs for links? It'll make links persistent and easy to cross-reference with citations and other systems. I did a spot check of the 14 links you have so far all have DOIs. I'm happy to give advice!

(I work at Crossref.org)


I came here as well to recommend the DOI. So many nice things can be had by using this rather than your own IDs, it means that site now becomes a tool that composes with other tools.


I'm always interested in hearing about less-than-mainstream uses of DOIs.


The DOI is a natural foreign key, look at semantic scholar and the integrations that arxiv is doing in the bottom 1/3. You are generating your own IDs, so you won't compose with the rest of the ecosystem.

Your keys mean nothing and given a handle to a paper, they can't reference your site.


Hey @afandian, can I interview you about the tech stack you all used to build search at Crossref.org?


Always happy to share more details about how we build open scholarly infrastructure! Can do here, or reply with contact details.


Related, but I created Scholars

https://www.scholars.io

Where researchers and others interested can create "reading rooms" of papers and discuss. I have also been thinking about something similar to this where a list of papers that people are currently discussing is shown.

This looks great too!


Is there a way to see what other people are reading and if I could join/collaborate with them?


This is cool! I suggest adding some more hints to indicate how important/relevant different papers are. Maybe some tags to classify papers in different fields, and number of citations or amount of discussions online as a metric of how big a paper is? (I know that's not a great metric but I think it's still better than seeing a bunch of papers and not knowing which ones are worth reading). You can also spark discussions. Seeing "0 comments" discourages anyone to look at the comments, but even having an AI-generated summary as a comment can be encouraging and spark discussions. Just a thought.


A few years ago I created upvote.pub (source code: https://github.com/danielecook/upvote.pub)

It never really caught on - but it was a fun project. I do hope something like this takes off. It was able to import publications using DOIs, PMIDs, PMCids, Arxiv, and Biorxiv.


The problem is, there are way too much papers, even if you focus already to some domain, like deep learning. But this seems to be completely open to any topic?

For machine learning, https://www.reddit.com/r/MachineLearning/ is currently a good place, or Twitter and following some of the authors you are interested in, or just set some Google Scholar alerts.


Is that a problem though? HN covers a very broad spectrum yet it still functions.

Keeping it broad by allowing any topic but raising the bar to require a peer reviewed paper might even decrease the velocity of conversation, allowing better quality conversation. That is, if the lower velocity can allow a community to form still


It works on HN for at least the reason that HN post titles are often relatable. The same can't be said about paper titles even if you're in the field.


Hard disagree. If a user submits a paper, chances are it's because they're interested in it. I'm interested in this precisely because I might find myself reading about something in a topic I know nothing about that's interesting to somebody.


I actually enjoyed that the topics are broad. I studied physics, but worked as a software developer my whole life.

The topics on the site are stuff that I was interested in at some point in my life: python, GRBs, code style, whatever GPT, thermodynamics and communication.

I used to read papers but I don't anymore, and this site lets me keep discovering topics even if I'm not a scientist anymore. I don't need 25 posts a day on gamma ray bursts, I want one post a week and a fun discussion (which isn't there yet).


It doesn't have to be every paper, just papers people find interesting. HN is even larger, it can be any subjects including papers but still survives.


This looks awesome! FYI You may have a bug where you treat unicode as ascii. Notice the " “post-symbolicâ€" here [1]

[1] https://paperlist.io/paper/724614047


Ah interesting. I'm using LaXeXML to translate LaTeX to HTML. Might be something up with that. Will take a look at it!


This is really nice - the simplicity of the interface and ability to read papers without sacrificing the readability of math equations is awesome. Congrats on publishing it. Do you mind sharing how you convert from pdf to html (if at all)? I'm building a tool for reading, curating, and visualizing personal knowledge [1]. Apps like yours and others mentioned in this thread are nice ways to discover papers and address blind spots during literature reviews.

[1] https://www.knowledgegarden.io


No search ? :-(

Sorry, I won't use it: search is the only way to let your users discover the content that THEY ARE LOOKING FOR (and maybe to engage) instead of JUST CONSUMING content pushed by others or algorithm...


You're on a site with no search


HN has a search bar at the bottom of the page


That's cleverly worded. HN does not have search, however.


woah I stand corrected, my apologies


Personally I didn’t find your tone really warrants an apology, and anyway the search feature is pretty unobtrusive :)

Also for what it’s worth I kind of disagree with the top level poster, search can come later after there’s enough discussion to actually index.


Well, I'll be happy to use it when some kind of search will be implemented (even google search!)

The thing is: as mentionned in other comments, research papers may spread really different fields... so some kind of either search or category may help to find good content wrt specific knowledges.


I know someone else has said there's not too much whitespace, but I disagree. Comparing HN to this, I'm seeing 9 links per page compared to 26 for HN. Even Reddit fits 10 per page, and that's with thumbnails. It goes up to 18 with compact view. It needs much more content. Love the concept, but the UI leaves a lot to be desired in my eyes.


Maybe it was intended to surf the website only from vertical ultrawides:

https://imgur.com/a/o613uYC


that huge amount of whitespace is a dealbreaker for me too. needs to be more compact at least as an option for me to use it.


Cool idea! One of the great things about HN is the high quality discussion. Maybe you can think of some ways to bootstrap that? Like ask submitters to leave a micro review saying why the result is cool or something they learned or some way they applied the work? Ideally not just repeating the abstract


I got out of academia, but I'd love to keep reading the new hot and influential papers to keep up with the research trends in the research communities I am interested in. So I like this idea!

However, there are so many junk papers that the paper title itself is rarely useful. Too much context is lacking: where was it published? How strong is the conference / journal? When was it published? How many citations so far? What sub-sub-sub-field is it in?


I wonder if there might be a way to incorporate metrics like h-index, which I would think that many researchers themselves would be interested in. Perhaps something like that might detract from the posts standing for themselves, but I just thought I'd throw that out there if you find it might be useful. Great work though, seems like a great idea to me. I hope you can draw in a community around it successfully.


This website is promising. Here are some suggestions/questions.

For some reason, I can't use the trackpad to navigate forward and backward. Did you disable this somehow? Just a minor annoyance.

In your html-rendering, it would be nice to see an outline similar to the PDF renderer in Chrome. It is a very nice feature to quickly see the sections of a PDF or paper.

Suggestion: Allow people to comment on a specific line in the paper, and include a link to that line.

Good work!


I like the site, but because you only support arxiv, you are issuing out on lots of great research. You should consider supporting doi, i tried to load one of my papers (not on arxiv) and its rejected - now im sad!


Looks great. I think if there's one thing this could benefit from it would be Genius-style comments and annotations. https://genius.com/Genius-how-to-annotate-and-edit-on-genius...



Please allow me to share my favorite paper: bitcoin.org/bitcoin.pdf

Most of great papers I have read are located not on arXiv. And maybe consider to do the paperlist.io a torrent tracker because of availability issues of some papers over time like there is happening with links on Wikipedia pages.


I really like the UI, very visually striking without too much whitespace, and without getting in the way.


Appreciate it!


Is this meant for only CS/ML research?

It’s a great idea either way! I look forward to keeping up with it.


This is so cool. Simple concept, clean interface. Nice job. I hope it takes off.


I personally am a fan of using lobste.rs and filtering for pdf only [0], it's not *just* research papers but that is the lion's share of it. Plus they've already got a community so (to the people complaining about lack of discussion there) maybe check it out.

[0]: https://lobste.rs/t/pdf


You need to fake users to get the network effects to start.

Use a LLM to generate comments maybe finetuned on HN

When real people start showing up they will see comments and assume its not a wasteland. As real people come reduce the bot comments.


Need some users and your set!


The hardest part for any new site really.



Suggestion: if/when this expands to include life science papers add an ‘in mice’ tag.

It’s frustrating to read amazing research paper titles only to find out it was done with mice.


Why some papers have HTML representation and some not? Would expect the way it worked was if a link points to arxiv it'll automatically generate one.


How can I follow the feed? Rss or email news?


I went to sign up, but it doesn't work without an email. That's an important part of the HN experience.


HN has a set restricted topic: “Anything that good hackers would find interesting.” (from the guidelines). But, for PaperList, are there people interested in research papers – i.e. any and all research papers? Or would it be appropriate to narrow the focus on some subfield, or perhaps on “groundbreaking papers”, “practical papers” or even “controversial papers”? That last one ought to get some traffic, at least.


a site I would be interested in is: “papers a hacker would be interested in” (aka papers you could nerdsnipe someone with) which is unfortunately not what this website seems to be.


This is actually what I'm going for, which is why I described it as "Hacker New but for research papers." I think it'll definitely be possible with proper moderation


Thoughts:

* Allow logged in users to add tags to submissions (maybe after reaching some karma threshold). Tags could either be a private feature, or a shared feature.

* Allow filtering of all submissions by one or more tags

* "Favorites" feature for logged in users


Nice. I dream one day someoe would build HackerNews with tags so I could easily filter the news by subjects.


Would like to see a tag showing the subject of the paper. Many of the titles are not very descriptive.


Pull down to refresh is not working on mobile for some reason.


Looks super clean! I'll definitely be trying to use this.


How do you intend to combat spam?

My own immediate response was "great, I'd shove all my shittier papers on there", and I wouldn't even particularly consider myself a spammer for doing that.

Let alone proper spam (i.e. non-papers)


Any plans to add RSS?


Same here. That’s how I follow HN. Would definitely follow paperlist if there was a RSS feed.


Great idea but you should include every field of Science


Are you planning to implement search and filter tags?


am i alone in thinking that the people sifting through research papers are least likely to spend time sharing it with the crowd?


Sweet! I signed up :) Hope to see it do well


There is horizontal scroll on mobile


Filter tags would be very helpful.


I wanted to signup but then saw that email textbox.


why would you create separate site and not subreddit where you can get way more traction from existing users?..


An HN clone, not reddit.

But one huge perk of not having a subreddit is you don't have the mass of users come from other subreddits where the standards of comments tend to be more copypasta troll-like than HN.


> users come from other subreddits where the standards of comments tend to be more copypasta troll-like than HN.

I somehow didn't notice this on relevant subs (machinelearning, singularity, etc)


There's also the risk of not getting that halo effect and passing footfall.

There was something similar a few years back, upvote.pub: https://news.ycombinator.com/item?id=16273171

I don't know the reasons for it folding, but I imagine it was difficult to sustain.




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

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

Search: