// the find
bojone/papers.cool
Cool Papers - Immersive Paper Discovery
Cool Papers is a web service for browsing arXiv preprints with an inline PDF viewer and Kimi (a Chinese LLM) summarization. The repo itself is mostly a changelog and a Chrome extension — the actual backend and frontend live elsewhere. Primarily aimed at Chinese-speaking ML/AI researchers who want a faster way to triage daily arXiv drops.
The category union/difference syntax (cs.AI+cs.LG-cs.CY) for building custom paper feeds is a genuinely useful idea that arXiv itself doesn't offer. Historical data integration from the Kaggle arXiv dataset means most older papers don't require a live scrape. The Atom feed support for per-category and per-venue subscriptions is a clean touch. The Chrome extension that redirects arXiv/OpenReview/ACL pages to the Cool Papers view is the kind of quality-of-life feature that actually changes daily workflow.
The repo contains essentially no source code — it's a changelog, a Chrome extension stub, and a Zotero plugin. You can't self-host, audit, or contribute to the actual service. Kimi integration is a hard dependency on a single Chinese commercial LLM, so the summarization feature is useless outside China without a VPN or if Moonshot changes their API. The bioRxiv data source was already dropped because Cloudflare blocked the scraper, which signals fragility in any non-arXiv expansion. Reading history and preferences are stored locally in the browser with no account system, so syncing between devices requires manually exporting and sharing a URL.