About this site

How the data is produced

Nothing here is typed in by hand, and nothing a machine produces is published without a person seeing it. Each episode goes through four steps:

1. The episode is read from the podcast's public RSS feed: number, title, publication date, description and the links in the show notes.
2. The audio is transcribed by AssemblyAI, which returns word-level timings and separates the speakers.
3. The transcript is passed to a language model, which returns a structured list of what was discussed — media, running bits, segments — with the verbatim quote for each. The model never emits a timestamp: it returns the quote, and the timestamp comes from matching that quote back to the timed transcript.
4. Everything lands in a review queue before it is published. Low-confidence extractions are looked at one at a time; the rest are approved in batches.

So parts of this site are machine-written and human-reviewed, never machine-published. When something is wrong, it is because a person let it through, and a correction is the fix.

About the timestamps

Every timestamp on this site is measured against the ad-free program audio — what the hosts actually recorded — and never against any particular listener's download.

That distinction matters more than it sounds. The podcast is ad-stitched on delivery: each download is assembled when it is requested, with a different set of adverts at different lengths. Two people can sit nearly a minute apart in the same episode, and the gap changes at every ad break rather than staying put.

So the player works out your copy's own offsets when it loads and converts as it goes. When you are hearing an advert it knows that too, and waits.

Writing about real people

The hosts are real people, and this is a catalogue of a published podcast rather than a file on anybody. Every entry about a host is anchored to a specific episode, a specific moment in it, and the quote itself.

Nothing is stored about anybody's children — no names, no ages, nothing about what they have watched. Where an age rating exists it is a rating of the media, contributed by listeners, citing the episode where the subject came up.

There is no mood tracking, no sentiment scoring and no live status of any kind. "Last reported on the pod" is a dated citation, not a claim about how somebody is today.

Corrections

Corrections are welcome and are the main reason this site has an inbox. If a timestamp is off, a short is misattributed, or something should not be here at all, write to [email protected]