Claire Vo interviews Tomasz Tunguz about how he uses AI to manage and process information from 36 different podcasts. Tunguz explains his system, the "Parakeet Podcast Processor," which downloads, transcribes, and summarizes podcasts, extracting key topics, quotes, and potential investment theses. He uses FFmpeg for audio conversion, Gemma3 for transcript cleanup, and a local DuckDB database to manage the processed files. Tunguz also discusses his AI-driven blog post generation process, which includes grading the AI's writing using an AP English teacher model for critique and improvement, and touches on the challenges of replicating a personal writing style with AI.
Sign in to continue reading, translating and more.
Continue