I mapped 8.5M research papers into an interactive atlas

Takes 8.5 million papers from arXiv and maps them into a navigable visual atlas, co-locating each paper with its associated dataset, code, peer reviews, and supplementary material. The core frustration it solves is real: reading a paper normally means eight tabs, three search queries, and a lost afternoon. The atlas makes adjacency visible — you can see how clusters of papers relate spatially, which is a genuinely different way to survey a field than keyword search or citation graphs. The scope is what distinguishes it from the handful of research-graph tools that have launched in this space. 8.5 million papers is not a demo dataset. The honest reservation: it starts with arXiv coverage, so anything published only in closed venues is missing, and the UI has the rough edges of a solo-built research project. But the underlying idea — unified context per paper, spatially organized — is one that research-tooling builders should study. -> Best for: AI engineer, ML researcher, or technical PM building research or knowledge workflows