Every episode below is broken down into who said what, when, with the receipt — attributed claims, dated falsifiable predictions, verifiable facts, and notable quotes, each carrying a timestamp and a verbatim transcript quote. Browse it here, or point your agent at the API / MCP.
Recent-episode coverage: 2026-08-23 → 2026-08-23. Older archive episodes are not indexed yet — AiPod says “not covered” rather than guessing. Weekly ingest moves the cutoff forward.
Claim volume across the indexed Lenny's Podcast episodes. Each is a compiled page your agent can pull with /api/entity/<name>.
Enterprise sales · 21 Information edge · 8 Enterprise pricing · 7 Product demos · 5 Enterprise pilots · 5 Sales champion · 4 Enterprise procurement · 3 Executive sponsor · 2 Founder-led sales · 2 CRM sales stages · 2 Professional services · 2 Pincer model · 1 Forward-deployed engineers · 1 AI sales collateral · 1
Newest first. Each page is generated from the structured record — summary, segments, claims by speaker, predictions, quotes.
This is what your back catalog looks like as structured, citable knowledge instead of an audio archive: every claim attributed, every prediction on a scoreboard, every quote sourced to the second. Linkable per episode, queryable by any agent.
Don't scrape these pages — query the structure directly. Start at /llms.txt, then GET /api/context?q=… or connect over MCP. Every result is cited; empty means not covered.