evanhu1 / talk2arxiv

What it does

• Clean HTML reader. arXiv papers load from arXiv's HTML5 version (with ar5iv as a fallback), with native MathML equations. bioRxiv papers load from bioRxiv's full-text pages. Both get a table of contents and a reading-progress pill.
• PDF fallback. Papers with no HTML version yet (new bioRxiv preprints, some arXiv papers) open as a PDF, rendered with EmbedPDF (PDFium in WebAssembly). Highlighting and chat work there too; the model reads the PDF itself, figures included.
• Highlight and ask. Select any passage and click Ask AI or Explain. The passage goes to the chat with its equations as LaTeX.
• Citation previews. Click a bibliography citation in an HTML paper to preview it in a card, or a bottom sheet on mobile. Ask about this paper adds the cited paper to chat alongside the original. arXiv and bioRxiv references use full text or
• Mobile reading. A section tracker at the top, a floating glass composer, and a swipe-dismissable chat sheet keep the paper in place. Drafts survive opening and closing chat.