Show HN: HackerNews but for research papers

Original Article ↗ Hacker News Discussion ↗

An experimental site that presents research papers in a Hacker News–style feed is drawing strong interest as a potential way to surface and discuss new work across fields. Early users praise the concept but report significant usability issues, especially on mobile, along with missing features such as tagging, search, RSS, better contrast, and support for math, journals, and non-CS disciplines. Beyond UI feedback, commenters question whether an algorithmic feed suits how scholars actually read papers, suggesting narrower subject focus, AI-generated summaries, and careful community building and moderation as critical to making the platform genuinely useful.

Overall reaction to the concept

  • Many commenters like the idea of a Hacker News–style site for research papers and say it’s “a thing that should exist.”
  • Several report they’ve been wanting something similar or had previously built similar projects that struggled to gain traction.
  • Some see it as especially useful for non-academics or ex-academics who want visibility into current research without living on arXiv or Twitter.
  • Others are skeptical that researchers want an algorithmic “feed” of papers; they typically search with specific questions in mind.

Scope, focus, and community-building

  • Multiple commenters argue the site is too broad; suggest starting with a narrow niche (e.g., AI/ML, computer vision, FPGAs, PL, or a single field) and building a small expert community first.
  • There’s tension between serving career scientists (who have little time and need depth) vs. a broader audience (who may only want high-level summaries).
  • Some propose curated “most important papers” lists per field and expert-driven sections.
  • Concerns raised that most papers are niche and low-ROI to read; recommendation to lean on AI-generated summaries plus discussion of those.

UX, UI, and accessibility

  • Strong, repeated complaints about mobile: layout breaking, horizontal overflow, comment view issues, spacing, sort dropdown click bugs.
  • Text contrast and readability widely criticized; titles and nav links are too light and hard to skim.
  • Requests for a simpler, HN-like aesthetic (even “just HN with a different color”) and better information density preservation.
  • Several users note the site requires JavaScript, unlike HN, and ask for at least basic no-JS functionality.

Features and data sources

  • Common requests:
    • Tags, topics, and sub-categories (including CS, EE, bio, chem, health, history, law, etc.).
    • Search (by title, author, DOI, venue, keywords) and the ability to filter/subscribe to topics.
    • RSS feeds, “top rated” default sort, lists like HN’s, and sorting by newest.
    • Linking to arXiv abs pages (not direct PDFs), Google Scholar, and showing journal/venue info.
    • AI-generated ~100-word summaries, with community-editable corrections.
    • LaTeX support in titles and comments; PubPeer integration; pulling signals from GitHub, Hugging Face, other hype trackers.

Technical and infrastructure issues

  • Many sign-up/login problems: email rate limits, confirmation errors, 400s, browser-specific failures.
  • Site currently scrapes arXiv; pagination is inefficient (fetching all items client-side).
  • Some corporate and campus networks block .xyz domains; several suggest moving to .com.

Meta: moderation and scholarly discourse

  • Users stress that moderation quality will determine whether it stays high-signal like HN or degrades.
  • Debate over whether linear, tree-based comments are suitable for scientific discussion; ideas include DAG-like structures and richer referencing between comments.
  • Some doubt such forums can work for the vast majority of “mundane” papers, but see value for high-profile or cross-disciplinary work.