Open-source tool filters papers before spending on deep analysis
Research Radar is an tool that finds the few new arXiv papers relevant to a person's research. It collects every new paper in selected categories through RSS and an API, removes duplicates, and compares each abstract with a markdown file describing the research interests. A cheaper AI model scores the abstracts from 1 to 10 in batches.
Only the highest-scoring papers receive the expensive treatment: a stronger model reads the extracted PDF text and produces a summary, key findings, limitations, and links to the person's own work. The results arrive as a morning HTML digest, with an optional Telegram alert for essential papers. Changing one interests file adapts the same process to , physics, biology, economics, or another field.
AI is used only for the two scoring and analysis passes; ordinary Python code handles collection, duplicate removal, text , and presentation. Its model-independent design can also run through Claude Code or s using an existing .
Key points
- Collect new arXiv papers through RSS and an API, then remove duplicates.
- Score abstracts from 1 to 10 with a cheaper model working in batches.
- Use a stronger model only for top papers and extract findings, limitations, and relevance.
- Change one markdown interests file to reuse the process in another research field.
- Limit AI calls to filtering and deep analysis to reduce unnecessary token use.