Product Expert Insights

KEY FEATURES

I built this to make 300+ hours of product leadership wisdom instantly searchable. Instead of scrubbing through episodes, users ask a question and get cited answers in seconds—powered by RAG architecture.

AI-powered answers from 300+ Lenny's Podcast interviews using RAG architecture
Expert insights from leaders at Google, Meta, Stripe, Airbnb, Figma, Netflix, and more
Direct links to source episodes for deeper exploration
Semantic search across 15,685 transcript chunks using vector embeddings
No hallucination - AI only uses information from actual podcast transcripts
Clean, minimal UI focused on readability

Try it now!

HOW I BUILT THIS

Tech Stack

Frontend: Vanilla HTML, CSS, JavaScript
Backend: Python (Flask)
AI: OpenAI (embeddings), Claude API (generation)
Vector Database: Pinecone
Deployment: Render
Code Editor: Cursor (AI-assisted IDE)

My Process:

Building on the "Claude CTO" approach I learned from Zevi Arnovitz (Meta PM) on Lenny's Podcast, I had Claude act as my technical advisor throughout the project. Key architecture decisions included chunking strategy (500 tokens with overlap), prompt engineering for structured responses, and a summary-first UI pattern inspired by Rotten Tomatoes.

Challenges Solved

Data pipeline: Parsed 303 markdown transcripts with YAML front matter, chunked into searchable segments
Prompt engineering: Designed prompts to generate consistent, citation-rich responses without hallucination
Package compatibility: Navigated Python dependency issues during deployment (pinecone-client → pinecone rename)

V2 Ideas

Filter by topic, company, or guest (e.g., "growth advice from Stripe PMs only")
Save favorite answers for later reference
"Related questions" suggestions based on query