{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/a039070a9d02414b987ea6ee3dd04736\" frameborder=\"0\" width=\"1662\" height=\"1246\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1246,"width":1662,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1246,"thumbnail_width":1662,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/a039070a9d02414b987ea6ee3dd04736-d7710e8e0b8bb588.gif","duration":212.282,"title":"Bringing Taste to Science with Atlas","description":"This Loom explains how Atlas is designed to prevent homogenization in AI for science by encoding researcher taste rather than just standardizing literature workflows. It describes Atlas as a paper management system with reading tracking, recommendations, meaning-based filtering via embeddings, and higher-level “maps” of papers by themes, including insight into which labs drive areas. Atlas also uses synthetic data for saved visualizations and enables direct Claude integration so users can query their GeoCorpus database, draft paragraphs using only their corpus, and generate plots in the style of figures saved to a mood board. The main goal is to bring taste to science to counter convergence toward the mean."}