A paper writer that searches the library before it types

What it does paper-ai is a web-based writing environment built on Next.js that pairs a Quill rich-text editor with an LLM. You can ask the AI to generate text or search for literature; it queries Semantic Scholar, arXiv, or PubMed for actual papers and weaves the results—and their citations—into your draft. Supabase handles user authentication behind the scenes.

The interesting bit The README is refreshingly honest about the glue-code nature of the project, spending as much space on bug fixes as on features. The author had to fight React’s async state to stop the Quill cursor from jumping backward, and built an invisible AuthRefresher component to force a server round-trip after OAuth because Next.js App Router’s server components can miss fresh cookies on redirect. It’s a candid look at the friction of marrying a classic WYSIWYG editor to a modern RSC framework.

Key highlights

• Searches real academic databases (Semantic Scholar, arXiv, PubMed) rather than inventing references.
• Built-in Quill editor for direct formatting and editing of AI-generated text.
• Supports custom AI endpoints and API keys via environment variables.
• Runs on Vercel or via a provided Docker image.
• Internationalized routing ([lng] dynamic segments).

Caveats

• The README notes the app currently relies almost entirely on SSR with little static generation, which may affect performance.
• The project is essentially integration code between an LLM API, academic search APIs, and a Quill frontend; the value is in the wiring, not novel algorithms.

Verdict Useful for students or researchers who need a quick drafting tool with legitimate citations. Skip it if you’re looking for a reference manager or a serious LaTeX workflow; this is a lightweight web editor, not an IDE for science.