The Podcast Pipeline Nobody Sees: Behind the Scenes with AI Agents and AI gen
summary
A fully automated research pipeline collects gigabytes from social, papers, videos and GitHub, transcribes audio with Whisper, scrapes with Selenium fallbacks, and stores embeddings in Chroma DB plus relations in Neo4j; it generates 55–80 candidate questions, clusters them with a three-model consensus, outputs structured JSON with source quotes, and narrows to final prompts via n8n workflows and local LMs (Quen). It cuts hours to minutes but requires ~20% human polish for style, bias checks, and hallucination detection.
AI-written summary · about
chapters (11)
- 0:00🚀 Build Stuff Conference – Get Your Tickets!
- 0:30Introduction: The Art of Asking Great Questions
- 3:10About the Podcast & Audience
- 6:00The Evolution of the Pipeline (Pre-AI → Agentic)
- 9:40Data Gathering: Sources, Scraping & Context Building
- 14:20From Data to Questions: Generation & Scaling
- 18:10Clustering with Multiple Models & Consensus
- 22:00Enrichment: Merging, Improving & Structuring Questions
- 25:00Agentic Workflow & Tech Stack (n8n, LLMs, Vector DBs)
- 27:30Limitations: Bias, Style & Human Review
- 29:00Final Thoughts: AI as a Co-Creator, Not Replacement
related talks
24mAI and Web Development: Hype or Reality
12mCodeless Generative AI Pipelines: Prompt Automation
01h 01mLessons Learned Building the Ultimate AI Bug Reporter
26mWelcome to the AI Jungle! Now What?
22mAI in User Research: Tools, Use Cases and Best Practices
30mBuilding Reliable AI Agents for Publishing: A DSPy-Based Quality Assurance Framework
01h 02mGenerative AI for the .NET Developer
56mAI Agents That Don’t Suck: How to Build Ones That Actually Drive Business Value