Research Radar

Overview

The Research Radar agent automatically monitors arXiv, Hacker News, and Papers With Code daily to surface breakthrough AI research papers before they become mainstream. Designed for developers and founders building AI products, this agent filters papers by practical relevance to modern AI stacks—including agents, RAG, MCP, embeddings, and LLM inference—scores them by scientific novelty and architectural impact, and delivers curated daily digests. Over time, it builds a living research library in Notion that becomes an invaluable reference for staying ahead of the AI innovation curve.

How It Works

1. Search arXiv (Exa) — Perform semantic web searches across arXiv's AI, machine learning, and NLP categories to identify papers submitted in the last 24 hours.
2. Search Papers With Code (Exa) — Scan for trending papers and new state-of-the-art results in LLM, agent, and RAG categories using AI-native neural search.
3. Search Hacker News (Exa) — Query for 'new' and 'Show HN' posts about AI papers that have gained traction (>50 points) in the past 24 hours.
4. Find ML Community Discussion (Exa) — Discover Twitter/X threads and discussions where papers are going viral among the ML research community.
5. Filter and Score Papers (Claude Opus) — Use advanced reasoning to evaluate each paper on two dimensions: scientific novelty (1–5) and practical relevance to agent platforms, RAG, MCP, or LLM inference (1–5). Retain only papers scoring ≥3 on practical relevance.
6. Write Plain-English Summaries (Claude Sonnet) — Generate concise 4-sentence plain-English summaries for each retained paper, covering what was done, why it matters, key results, and concrete architectural implications.
7. Log Papers to Research Library (Notion) — Create a database entry for each paper with title, authors, arXiv ID, topic tags, novelty score, relevance score, and the summary.
8. Post Daily Digest (Slack) — Send a curated message to your research team channel featuring the top 2–3 papers of the day with their summaries.

Key Features

• Automated Multi-Source Monitoring — Scans arXiv, Papers With Code, Hacker News, and Twitter simultaneously in a single daily run, eliminating manual research overhead.
• Intelligent Relevance Filtering — Uses Claude's reasoning to score papers on both scientific novelty and practical applicability, ensuring your team sees only papers with real architectural value.
• Living Research Database — Builds a searchable Notion library of papers with metadata, scores, and summaries—a searchable knowledge base for team decisions and architecture discussions.
• Plain-English Summaries — Converts dense academic papers into 4-sentence actionable summaries, so non-researchers can quickly grasp implications.
• Daily Slack Digest — Delivers top papers directly to your team's Slack, keeping everyone informed without requiring individual subscriptions or manual aggregation.

Use Cases

• AI Startup R&D Teams — Founders and engineers building agent platforms or RAG systems can identify breakthrough papers in their domain 6–12 months before they influence the broader tooling ecosystem.
• ML Engineering Leaders — Stay informed of emerging techniques (new embedding methods, LLM inference optimizations, MCP protocols) before competitors adopt them, enabling faster architectural decisions.
• Research-Driven Product Teams — Use the Notion library during sprint planning to surface relevant papers that inform feature roadmaps and technical decisions.
• Competitive Intelligence — Track which papers competitors and industry leaders are discussing on Twitter/X in real time, revealing which research areas they're investing in next.