Evidence for podcast and story naming

Bounded review, 4 October 2026. Two proposed meanings, not implementation approval or a complete pipeline audit.

History and question scope

The original content-leftovers.json L-02 asks whether Podcast Video should remain, become Podcast Episode or become Interview. It proposes renaming every selected record while describing its own example as a clip. The new review corrects that premise. The original card remains historical.

The content-naming.json N-10 slash-name table proposes Brand Story and Origin Story as aliases. The approved no-slash rule is not approval of that alias. The existing queue already separates this pair as N-10/brand-story; Interview/Q&A and Course/Lesson have their own recorded answers. The ratified content-type list includes the earlier Brand Story / Origin Story wording. The new recommendation amends the boundary without pretending the type never had approval.

The August 16 final-amendments dictation asks how Brand Film, Campaign Film and Brand Story differ. Campaign membership is now settled, while broad video terminology remains pending. This packet must not decide that pending question through a new story label.

The July 10 decision-impact matrix lists both podcast_video and podcast_episode as keepers and podcast as retired. Older governance text instead describes podcast mapping to podcast_video. The selected live alias maps podcast to podcast_episode. These are implementation-history differences to reconcile, not evidence of a new user answer.

The broad A-10 definition proposal remains open for item-level reconciliation. The current approved process already requires individual boundaries and evidence. It does not ratify every draft definition by accepting a standing queue.

Example inspection

• The original TikTok record's publisher caption identifies the speakers and points to episode 9. Browser playback showed a 2:48 total, including a sampled frame near 2:42 with a seated speaker, microphone and captions. Earlier frames near the opening were also inspected. This is sampled visual evidence, not a complete audiovisual review or a verified time alignment to the parent.
• The publisher episode page identifies episode 9 and lists chapters through 1:13:11. Its metadata was read; the whole recording was not reviewed. The combination supports the excerpt counterexample without relying on a duration threshold alone.
• Read the beginnings section of Patagonia Australia's company history. It supplies a concrete origin narrative, while the page continues into later history. Do not label the complete page solely from that section.
• Read the main Patagonia ownership letter. It contains brief founding context but centers on a later ownership decision. This is an authored classification counterexample, not a claim that the publisher supplied our taxonomy label.

Selected live checks

The prior selected schema, sample and registry responses are retained in docs/reference/2026-10/2026-10-04/names-definition-preflight.json. That receipt predates the playback inspection above. Its earlier media-not-viewed limitation is preserved as history.

The selected sample remains labeled podcast_video. The registry selection returned active podcast_episode and podcast_video content types and an active origin_story concept. The selected alias was podcast to podcast_episode. These observations do not establish global absence of other story labels, consumer compatibility or classification accuracy.

The additional routing probe and schema are saved verbatim in names-definition-routing-evidence.json. The selected live rows are podcast and podcast_video, both active with moments_policy skip. No podcast_episode row was returned by that exact selection. No policy was changed.

Local consumer trace and later contract

Read backend/services/moments_eligibility.py: active routing rows supply an exact normalized type lookup; missing types skip with an unknown-type reason. Read the caller in backend/routers/pipeline_scene_classify.py: it makes this routing decision before downloading and checks again after probing an unknown duration. This establishes a local dependency on type spelling, not deployed-code equivalence.

The local analysis router's video type list includes podcast_video and supplies the type prompt. A later rename must trace that list, the actual composed prompt, validation, every scalar type writer, routing configuration, read views, filters and tracker rules together. The rich-classification alias resolver is not proof that the scalar content-type writer uses that alias. The local concept vocabulary includes origin_story; its physical storage is an implementation fact, not the reason to recommend its meaning.

Before implementation, verify show → episode membership, excerpt → source relationships, ordered parts, publication/capture identity and evidence for time alignment. Determine multiplicity and deletion behavior from the actual schema and consumers. Preserve human corrections and uncertain links. Check media capability, duration, cost, reanalysis and scheduler behavior independently of display names. An unknown-type skip must remain visible in observability.

For story search, retain subject links and distinguish whole-piece evidence from a passage or moment. Queries, saved trackers and cited answers must preserve that distinction and permissions. No schema, prompt, processing or automation changes follow from this review.

Choice mapping

• L-02: one complete-episode name across media, with excerpts preserved; no bulk relabeling.
• N-10/brand-story: Brand Story versus origin-focused narrative; the rest of N-10 remains partial.
• A-10: no blanket approval requested; individual definition review remains unfinished.