Methodology

Podcast Atlas uses an AI-assisted publishing pipeline to organize publicly available podcast material into reader-facing guides and a connected Wiki. Automation makes the corpus navigable at scale; it does not turn a podcast claim into an independently verified fact.

How the atlas is built

1. Establish the source. Each published episode page records the show’s metadata and, when available, links to the original audio.
2. Create an Episode Guide. Available source material is organized into an overview and navigable sections so a reader can understand the conversation before or after listening.
3. Preserve provenance. A corresponding Source Note connects the Wiki back to the canonical episode page and original recording.
4. Connect knowledge. Concepts and Entities link recurring ideas, people, organizations, places, and works across source notes.
5. Update the living layer. Current Synthesis is rebuilt as the corpus changes; Open Questions keeps unresolved issues separate from the current account; Update History records change over time.
6. Validate before publishing. Automated checks reject missing Wiki targets, ambiguous filename and route collisions, broken Episode-to-Source relationships, malformed canonical URLs, and invalid deployment artifacts.

Evidence and provenance

The original audio is the source of record. Episode Guides and Source Notes are derived navigation and synthesis layers, and links between them are preserved so readers can return to context.

Where a guide uses labels such as [Fact] / [事实] and [Inference] / [推测], they distinguish what a speaker directly stated from an interpretation made while structuring the episode. A fact label describes attribution within the source; it is not independent fact-checking of the speaker’s claim.

Concept and Entity pages may combine material from more than one episode. Their source relationships provide the path back to the contributing podcast material. Current Synthesis represents the atlas’s present organization of that evidence, not a final or universal conclusion.

Limitations and corrections

• Podcast coverage is selective. Absence from the atlas does not imply that a show, episode, person, or idea is unimportant or nonexistent.
• Source audio, descriptions, and transcripts can be incomplete, unavailable, edited, or ambiguous.
• Automated transcription and AI-assisted synthesis can mishear names, flatten disagreement, omit context, or make an incorrect connection.
• Podcast participants may be mistaken. The atlas preserves attribution but does not independently verify every external claim.
• Pages evolve as new episodes arrive. A later synthesis may revise an earlier framing while Update History preserves the chronology of that change.

Readers should verify consequential claims against the original audio and other primary evidence. Any correction review should identify the public page URL, the disputed passage, and the supporting source or timestamp.