{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/f02d818fa4ff4a43bfe97ec3686fe4ae\" frameborder=\"0\" width=\"1728\" height=\"1296\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1296,"width":1728,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1296,"thumbnail_width":1728,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/f02d818fa4ff4a43bfe97ec3686fe4ae-9036515f078fbae3.gif","duration":169.25866699999997,"title":"Deterministic AI for Legal Research and Drafting","description":"This Loom presents Bucket, an AI workspace designed for India’s 1.5 million advocates working in under-resourced districts and lower courts. It addresses the risk of generic legal AI fabricating authorities, referencing a February AI petition by India’s Chief Justice that used a completely fabricated case as binding authority. Bucket claims to auto-generate case arguments and predict likely bench questions with deterministic, traceable outputs grounded in a 20 million judgment library. Its architecture uses a five-layer pipeline including deterministic scoring, retrieval from top five precedents via Pinecone, iterative relevance grading, and a final generation step that traces every claim to specific sourced clauses."}