Split guest, episode, research cards, and questions. The host should open evidence beside each question and see confirmed vs still TBD at a glance.
(Diagram: four-table prep layout.)
Suggested structure
Guest — name, bio, verification status.
Episode — air date, link to guest, outline status (draft / host-reviewed / final).
Research cards — source URL, excerpt, topic tags, confirmed / TBD.
Questions — wording, episode link, card(s) this relies on, host-reviewed flag.
Let the agent draft cards and questions; the host confirms facts and marks questions reviewed—optionally via Inbox change requests.
(Local demo: review agent-prepared edits before they become the episode’s source of truth.)
Host checklist
- Episode question view: every question links to at least one card.
- TBD cards stay visibly unverified—don’t read them as settled facts.
- After review, mark questions reviewed and move outline status to host-reviewed.
Sample prompt for your agent
For episode N / guest X, collect public sources into research cards (URL + excerpt), mark confirmed vs TBD, then draft 8–12 follow-ups each linked to supporting cards. Do not alter already reviewed questions.
Boundaries
- One long doc mixes evidence and questions so verification is slow.
- Questions without sources can’t be checked quickly on recording day.
- Mixing TBD and confirmed in one status invites treating rumors as facts.
Structured prep makes the host’s job “question + evidence,” not “untangle a blob.”
