BuildC3 (@buildc3) on ThreadsWhat happens when you point Claude Code at 50 YouTube videos and ask it to think? You get production-ready insights. Zero hallucination. Almost zero tokens.HomeSearchCreateNotificationsProfilePinMoreMoreBackMoreLikeCommentRepostShare
Published on
BuildC3 shares a cheat sheet for a free AI research pipeline combining Claude Code with Google’s NotebookLM. The workflow points Claude Code at up to 50 YouTube videos, uses NotebookLM as a pre-built RAG system running on Google’s servers, and produces production-ready insights with near-zero token consumption.
The 4-step process: search YouTube for relevant videos, feed the URLs into NotebookLM (which handles the heavy analysis), then use Claude Code to orchestrate the output into deliverables. NotebookLM accepts up to 50 sources per notebook and can produce a first insight in about 5 minutes.
The claim is that this replaces a $500/month research stack for free — leveraging Google’s infrastructure for the compute-heavy RAG work while Claude Code handles orchestration and final output formatting.