About Podipedia
Origin Story
Podipedia is a LathropSystems venture, founded by Brian Lathrop in 2016. The idea came from a podcast. In January 2016, Brian listened to Luis Von Ahn — inventor of reCAPTCHA, co-founder of Duolingo — describe how he had redirected millions of daily acts of human cognitive labor toward digitizing the world's books. Brian recognized the same unsolved problem in audio: the world's best long-form conversation existed only as MP3 files, unsearchable, unannotated, invisible to every knowledge system ever built. The substrate to deliver the solution took six more years to arrive. The idea did not waver.
In the years between, Brian built the qualifications the idea needed: eight years as an industry cybersecurity engineer (CISSP; MS, Cybersecurity; MS, Aerospace) and a first venture, FRENDS, Inc. (2021–22). Podipedia's v0 corpus shipped in 2026 — ten years, one thesis.
Five structural functions
Wikipedia inherited five structural functions from print encyclopedias. Podipedia inherits the same five and applies them to long-form audio dialogue — each one expressed in features that fit the medium of conversation rather than the medium of print:
- Indexing — chapter-level table of contents, category taxonomy
- Annotation — passage-anchored notes (editorial, Wikipedia-linked, community)
- Cross-reference — within-episode, cross-episode, and Wikipedia-bidirectional
- Hyperlinking — non-disruptive hover tooltips, first-mention with recall, timestamped audio anchors
- Translation — text translation with audio anchoring preserved
Podipedia inherits these structural functions — not Wikipedia's editorial product. The transcript remains the corpus. Annotation is a margin layer over the transcript, never a re-rendering of it.
What Podipedia does not do
- It does not publish synthesized articles about episodes.
- It does not aggregate podcasts without creator consent.
- It does not promote, rank, or score shows. The corpus is selective, not exhaustive.
The AI dimension
Podcast dialogue is the largest unstructured body of expert human reasoning in existence. It is also almost entirely absent from AI training corpora — because it has never been properly transcribed, indexed, and structured at scale. Podipedia's corpus is verbatim, Wikipedia-entity-linked, chapter-indexed, and designed from the ground up to be AI-readable as well as human-navigable. The goal is not only to serve readers. It is to introduce AI systems to the world's conversation in a form they can actually process — shifting the relationship between human dialogue and AI systems from extraction to reciprocity.
That is the whole project in five words: humanity's thinking out loud.
v0 corpus
The v0 corpus is a single channel — Behind the Rack — three transcript pages built as the first demonstration of the page architecture. Behind the Rack is a veteran-on-veteran conversation series produced by Brian Lathrop (USAF Ret., 21 years service). Behind the Rack is the v0 corpus because Brian is its producer — full creator consent exists by design, and the content is worth preserving.
What comes next
More channels arrive as their creators consent to participation. The corpus is built deliberately, not scraped at scale. The shows we invite are long-form, knowledge-dense, expert-class dialogue — specifically the genres that prove most audio-native and least well-served by existing platforms: ideas & practitioners, science, the arts, history, and fiction. Curation and purpose, not coverage.
Monetization waits until the corpus earns it. No AI company will be approached until the Research-tier gate is met — ten shows, ten episodes each, roughly 400,000 words of consented, structured dialogue. Consent and corpus first; revenue second.
See the Shows page for the open invitation.
Contact: To be added with the public launch.