Show HN: HackerNews but for research papers (papertalk.xyz)

319 points by sleno ↗ HN
Hey guys, I love HN! I wanted to extend the same aesthetic and community towards things beyond tech-related news.

I thought it would be cool to get the same quality of community gathered around the latest and greatest research coming out.

Let me know what you guys think of what I have so far. It's still early so there are probably bugs and other quality issues.

If there's any features missing that you'd want let me know.

ALSO, if any of you are familiar with the map of the territory of any particular field, please let me know! Would love to pick your brain and to come up with a 'most important papers' section for each field.

Thank you!!

-stefan

208 comments

[ 5.0 ms ] story [ 239 ms ] thread
I love the idea and simplicity, thank you for sharing!

Small bit of feedback: would it be possible to make the UI a bit more mobile friendly? Or, alternatively, is there an API that others could use to build different interfaces?

Again, this is a Thing That Should Exist, so thank you for bringing it into the world.

I'll work on that, sorry yes i didn't spend much time on making it mobile friendly yet!
I second this. Mobile is how I primarily read news. IMO it should be the first platform you support.
noted, thank you!
fixed some mobile issues
Tags would be great
Will implement, thanks for the feedback!!
If you had watched me use this over my shoulder for a minute, this is what you would see:

I went to the page, scanned the list to find something that has comments, but everything is listed as "0 Comments". I clicked on "Comments" for the AGI paper, but nothing happens -- it's not a link. There's no "Discuss" link. On HN clicking on the paper title opens the upstream link, not comments, but on your site I discovered that it opens the discussion page. Once there at https://www.papertalk.xyz/papers/2404.10731v1, I noticed that there are in fact 3 comments, but the comments counter at the top still says "0". The comments are also just random gibberish.

I hope you iterate on this project and that it gets massive traction!

Really appreciate the feedback! I'm gonna ship this up, yes there are some gibberish comments i left from testing. Thanks for letting me know about the bugs, and other UI issues, will fix ASAP.
Fyi: the UI is completely broken on mobile. I cant read anything at the above link because it extends off the edge of the page in both directions and I can't horizontally scroll.
fixed some of the major issues
> but everything is listed as "0 Comments"

Also, everything is listed as "0 points" which I found it weird, how would then the ranking work? HN gives 1 point to every story submitted.

OT: I used to see other HN-inspired forums for other topics. I remember once a data mining topic that showed up here. Can anyone here reading this share those forums as well.
Cool! Have you considered tagging/sub-categorizing? That would be helpful to organize topics.

I’ve often seen papers have a list of keywords somewhere on the first page that could be helpful for indexing

It's on the todo-list. Are you looking specifically for like sub-categories of a particular field? Like in computer science AI, Cryptography, etc.?
I work at an FPGA company so anything FPGA-related would be cool :)

To capture the most popular topics, you can grab the topics of the top N most attended academic conferences as a starting point

Amazing idea, thank you!!
.xyz ltds are blocked on some corporations for security proposes. I suggest a .com domain.
Thanks for letting me know, forgot about that. Will try to find a domain that uses .com
really? so corporations are doing ltds blocking? Why?
Because corporate security is... the positive spin is probably "heuristics" or "statistical", the negative spin is "cargo cult". Lots of scams use(d) .xyz domains so just block the whole TLD. Whether this is a good cost/benefit is left as an exercise to the reader.
Sign in isn't working for me after confirming my email on both Chrome and Safari.
Sorry about that, I'm gonna have to do some testing. Works for me on Brave. I'll fix ASAP
Thanks! Seems to be working now. Next issue I see is that the formatting of a paper discussion page seems to be rendering content out of bounds of the view on mobile Safari.
Oh hell yeah!

It can be hard to sift through the all the random papers on Arxiv or something, but I do try my best to keep up with current research. Obviously I don't have the time or ability to read every single paper (much as I wish I could!), but I do try and at least read the abstract, introduction, and conclusion to CS papers that are relevant to me, though the "relevant to me" part can be hard to figure out a lot of the time.

Having an "HN-like" experience for research papers could really help with that, I love the idea!

Out of curiosity, what did you build this in? Since I do think this has potential to get to HN-levels of success in the academic world, have you thought about scaling?

I'm using Astro with React and Supabase. haha in regards to scaling I know what I built isn't ready but I figured I'd worry about that bridge after I run into problems.
Looks amazing!

UI Request: Make the font darker. That light font on a light background is extremely hard to read.

Will do, thanks for the feedback!
Mobile comment view is messed up on Safari.

The top level view seems to leave a bit too much margins on mobile.

Thanks for the feedback, i need to give mobile some more attention. Will fix ASAP
Huh. You know, I think HackerNews folks must approach literature searches and reading academic papers differently from the scholarly world at large. I can't think of any of my colleagues who would be looking for an algorithmically driven feed of papers to peruse. Usually we have specific questions we are trying to answer in research.
A lot of my colleagues either:

- read arXiv (the algorithm there is pretty simple, but it's a feed of sorts), or

- follow twitter accounts to learn about new results

So, like it or not, a lot of people are getting their feed from algorithms.

I think you missed my point that the idea of a "feed" being something scholars need is not really clear. It is more that a feed is something that website developers know how to make
Yeah sorry I was being overly oblique above: I know a lot of academics who learn about a lot of results via twitter. Personally I don't do it, but a lot of people in academia do.

I'd consider twitter a feed, and one driven by a proprietary algorithm at that. I'm not sure if it's better or worse to have the interest of these academics guided by a company like twitter, but maybe having some other options is good.

I love the idea! But there is one little thing. I would love to see an easy and clean interface like HackerNews but using a Blue color instead. I think that would be awesome!
Is there an RSS feed I can subscribe to?
I like the concept. My feedback is to increase the color contrast. It's very difficult to read with the light gray text on light background. I'm using Firefox on Ubuntu.
Noted, thank you!
I agree. I honestly think the UX would be much better if it just looked exactly like HN but with a new color other than orange.
I am glad to see more projects in the vein of helping people find research papers. I am delighted by the discourse of papers on HN. Indeed, the comments on research papers here can help me better understand something that's outside my field of expertise. But HN is not specifically for research. Thus, it feels like there is room for another space for people to come.

I like the idea of subcategories. Some notes on the experience:

  - When I load the page, it takes 1-2 seconds for papers to load. This doesn't feel great.
  - The grey colours in the list make the page hard to skim. I recommend a darker colour for the text.
  - The colour of the navigation links is too light. It is hard to see the links.
  - The information density is spot on. Amazing work!
This is a community project. Try to get a few people on one niche who know what they are talking about commenting on your site actively (This HN post may help you find those people!). If your site became good at computer vision, the field in which I work professionally, I'd be there every day.
Really appreciate the detailed feedback. Community is the hard part for sure, will do my best to make it as good as HN
Chech and compare layouts on a phone (y vs yours), and fix it so linebreaking matches
Looks interesting.

FYI. I tried signing up. Got: "Email rate limit exceeded"

Sorry, didn't expect to get as much traction as this has been getting. Need to do some scaling haha
Would be nice to have the journal the paper is published in on the post.
How about Bio and Chem as well?
Relatively easy to add, glad to know people are interested in these fields as well.
Seconding this, I love HN because it gives a wide gamut of exposure, it's a glorious nexus of information, however as a MolB I'd love to have something more narrow when it comes to narrowing things down to academic work.
Looks quite similar, never heard of it but will take a look
Took a look. Emergent mind looks really polished, however it's tailored towards only AI it looks like.
Hey all, Matt here, Emergent Mind's founder.

Yes, Emergent Mind is 100% focused on AI/ML papers from arXiv. I think it makes more sense to focus on a niche because you can tailor everything to that niche, vs creating a general research paper site which won't wind up speaking to any audience well.

For anyone curious about Emergent Mind: it surfaces trending AI/ML papers by monitoring social media (HackerNews, Reddit, X, YouTube, and GitHub) for discussions about papers, then ranks them based on the amount of engagement they're getting (similar to how HackerNews uses upvotes). Then, for all trending papers, it automatically summarizes them using GPT-4o and links to relevant discussions so you can learn more.

We're working on a bunch of new capabilities that we'll announce soon too.

Feedback welcome: matt@emergentmind.com

Did you consider just running a Lemmy instance with perhaps different time decay parameters?
Can you explain? Don't know what a Lemmy instance is. I did think about using time decay to give a more HN-like experience but haven't gotten around to implementing it yet.
I think the ideal future for science publication would be something like:

- some database like arXiv for most all the results, obviously mirrored a few places

- multiple frontends like this!

- no more journals: replace journal review with an endorsement from a similar organization, on the arXiv (or similar) page

Thanks for taking one of the required steps to making this possible!

Appreciate it man, thank you
The quality standard of peer review, especially within computer science, is declining. I've read lots of papers only to discover obvious method flaws or dishonest statistics. With a simple comment section, this would be trivial to filter out.

And don't get me started on the price for publication and for readers that supposedly fund this "detailed peer review process" that misses so many obvious flaws.

The peer review process is still needed, but I would love to see it be more public. I've been wanting this exact thing for years. A simple comment section of the average readers and their thoughts.

>With a simple comment section, this would be trivial to filter out.

Pubpeer.com + browser plugin

I agree.

Journals have been publishing "responses" for ages, so there always has been a "comment section" of sorts. But it could be so much easier!

Personally I haven't really seen or been involved in a response. It would not surprise me if that tradition is dying because it's so much easier to comment somewhere on the internet where some people will see it. But of course without an official endorsement from the journal / preprint server / whatever, it's hard to know which forum you should direct your response to.

A critical original response should be accepted as true and be advertised at the top of the paper if not responded to within 14 days. The response should involve corrections or a timely planning for it.

I don't like how software has to be actively maintained to be usable or taken serious but the results are hard to question.

An average paper from arXiv has 10 people in the world who can appreciate every detail of the text, half of them are the authors, their advisors, and paper's referee, the other half will be mentioned in the acknowledgements. This is especially true e.g. for mathematics. I want to point this out because papertalk.xyz features many texts like that.

Moderating some kind of "official" internet forum for papers will require a lot of time, and most researchers won't bother to follow it, because messages will be of low value, from crackpots and occasional ill-spirited anonymous commenters.

In 2016 arXiv run a survey about new features. Comment section, votes on submissions etc. were not popular. Many people were against this stuff, so it was not implemented.

There have been projects like PubPeer that tried to explore this territory and never really took off.

I see a value in a website like HN with links to papers of bigger interest and discussions for the general public.

E.g. someone posts a link to a new publication/preprint about cancer research, a supposed superconductor discovery, massive high energy physics experiments, etc. and users are free to discuss the meaning, beauty, impact, and credibility of the results.

But I'm not sure if this can work for mundane papers that are >99% of published research. I also think it will be much harder to market to the researchers. As someone who read, wrote, reviewed papers.

> And don't get me started on the price for publication and for readers that supposedly fund this "detailed peer review process" that misses so many obvious flaws.

Readers don't fund the peer review process. People on the editorial board and peer reviewers do their job as a community service and have nothing to do with the inflated subscription prices.

Looks neat, is this a self-made thing or a project one could use to deploy a similar forum? I've been noodling on the idea of a HN style board for a non-tech audience that is older. I think the more primitive UI will resonate with the target demographic.

Yours looks too nice for my use case honestly :) Weird complement but I really do like how it looks, you did a good job.

Thanks man, yeah self-made. Obviously took a lot of inspiration from HN haha