yingaowang-casia/shushu-novelty-finder

Shushu Novelty Finder is a tool for researchers who want to figure out what's genuinely new in their field before committing to a project. Instead of just brainstorming ideas that sound good, it walks you through the existing literature in a specific research area, identifies what's already been done, and then proposes novel ideas that actually fill a gap. It's designed as a "skill" for OpenAI's Codex, meaning you install it and then trigger it by typing a command in your Codex chat. The tool works in two main modes. The first is Literature Lineage mode, where you give it a research direction and it maps out the landscape: grouping similar papers, explaining what each key paper actually contributed, and flagging which areas are already saturated versus which gaps still have room for new work. The second is Idea Generation mode, where you provide either a direction or a "seed paper" and it outputs ranked research ideas. These aren't just one-liners, each idea comes with the closest existing work it would compete against, a proposed minimum experiment, suggested baselines, anticipated reviewer objections, and clear criteria for what results would kill or support the idea. The people who'd use this are graduate students, academic researchers, or anyone trying to write a paper and wondering if their idea is already taken. For example, if you're working on combining retrieval-augmented generation with large language models for chemistry, you could ask the tool to first map out who's done what in that space, then generate three to five concrete ideas that haven't been explored yet, each with a reality check on whether a reviewer would buy it. What's notable is the emphasis on intellectual honesty. The tool explicitly avoids saying "nobody has done this" without evidence, tags unverified claims, and includes a built-in checklist to make sure its output actually holds up. It pushes you to think about kill criteria, what would make you abandon an idea, before you sink time into experiments.