# Example: Podcast Production Workspace

This example tests Domain Context Modeling in a creative, collaborative domain rather than a permitting or compliance-heavy domain.

## Product Concept

An AI-assisted workspace that helps small podcast teams plan episodes, manage guests, collect assets, generate transcripts, create clips, review edits, and prepare publishing packages.

## Prompt A: Conventional

```text
Help me design an AI-assisted podcast production workspace.

Create a front-end prototype showing the dashboard, episode planning, production tasks, assets, review workflow, and AI assistant behavior.

Use the domain context to shape the interface and AI behavior. Do not surface framework labels like "Concept Map," "Sources of Truth," or "Agent Boundaries" unless those labels would make sense to podcast producers. Prefer task language, progressive disclosure, source ranges, review states, blockers, and exception handling.
```

## Prompt B: Domain Context

```text
Help me design an AI-assisted podcast production workspace.

Use this domain context:

Domain:
Podcast teams plan, record, edit, review, package, publish, and repurpose audio episodes. The product helps producers coordinate creative work, manage production assets, track reviews, and use AI to draft supporting materials. It does not replace human editorial judgment, guest consent, or publishing approval.

Domain Affection:
This domain is worth caring about because podcast teams are protecting creative trust: the guest's words, the producer's judgment, the show's voice, and the audience's attention. Skilled producers notice pacing, consent, source fidelity, editorial intent, and whether a clip represents the conversation fairly. A careless product would treat generated quotes, summaries, and clips as content inventory instead of editorial material with context, rights, and reputation attached. The product should help teams feel creatively in motion, editorially careful, and confident that AI is assisting rather than overriding judgment.

Domain Opportunity:
The domain opens opportunities beyond task management: collaborative episode rooms, source-linked clip workbenches, guest approval portals, editorial style memory, show voice libraries, recurring segment templates, reviewer lanes, rights-aware publishing packages, and channel-specific repurposing. Informal producer behaviors such as saving standout moments, tracking guest sensitivities, and reusing proven episode structures could become intentional product capabilities. The product should explore ways to make creative momentum, source fidelity, and approval readiness visible without turning production into a compliance dashboard.

Domain Outcomes:
User outcomes include faster episode preparation, fewer missed review items, fairer guest representation, stronger source traceability, and more confident publishing. Business outcomes include retained teams, more published episodes, higher reuse of approved assets, and clearer value for multi-show production teams. Trust and creative outcomes matter: the product should protect editorial judgment, consent, show voice, and audience trust even when adding automation or repurposing features.

Metaphor And Medium Translation:
Podcast teams understand studios, edit bays, episode boards, review passes, release checklists, and clip reels. Those metaphors can help structure work, but the product should not copy physical studio constraints literally. Digital production can make every quote traceable to source ranges, every clip linked to consent and rights status, every review item resolvable, and every channel package adapted from approved material. AI-native features should extend the production model through draft assistance, source-aware suggestions, and consistency checks rather than replacing editorial authority.

Core Concepts:
Show, Season, Episode, Segment, Guest, Recording, Transcript, Asset, Clip, Review, Edit Request, Publishing Package, Channel, Release.

Concept Definitions:
A Show is the ongoing podcast brand or program. It is not one individual episode.
A Season is a grouped set of Episodes when the Show uses seasonal structure.
An Episode is a planned or produced installment of a Show.
A Segment is a meaningful section within an Episode, such as intro, interview, ad read, listener question, or closing.
A Guest is a person featured in an Episode. A Guest is not the same as a collaborator or reviewer.
A Recording is source audio or video captured for an Episode.
A Transcript is text derived from a Recording. It may contain errors until reviewed.
An Asset is a file used in production or publishing, such as audio, artwork, music, show notes, images, or legal releases.
A Clip is a short excerpt created from an Episode or Recording for promotion or reuse.
A Review is feedback on an Episode, Segment, Clip, Transcript, or Publishing Package.
An Edit Request is a specific requested change from a Review.
A Publishing Package is the assembled metadata and files needed to publish an Episode to one or more Channels.
A Channel is a destination such as Apple Podcasts, Spotify, YouTube, newsletter, or social platform.
A Release is the act or record of publishing a Publishing Package to a Channel.

Relationships:
A Show contains Episodes and may contain Seasons.
A Season contains Episodes.
An Episode contains Segments, Recordings, Transcripts, Assets, Reviews, Clips, and one Publishing Package.
A Guest may appear in many Episodes.
A Recording may generate one or more Transcripts.
A Transcript belongs to a Recording and may support Clips, show notes, and summaries.
A Clip must trace back to a source Recording, Episode, or Transcript range.
A Review may create Edit Requests.
An Edit Request belongs to one reviewed object and must have a resolution state.
A Publishing Package belongs to one Episode and targets one or more Channels.
A Release records publication of a Publishing Package to a Channel.

States:
Episode: idea, planned, recording scheduled, recorded, editing, review, approved, packaged, published, archived.
Recording: expected, uploaded, processing, transcribed, approved, rejected.
Transcript: generated, needs review, corrected, approved.
Clip: proposed, drafted, reviewed, approved, scheduled, published, rejected.
Review: requested, in progress, changes requested, approved, closed.
Edit Request: open, in progress, resolved, rejected.
Publishing Package: draft, missing assets, ready for approval, approved, scheduled, published.
Release: scheduled, published, failed, withdrawn.

Actions:
Producers can create Episodes, invite Guests, upload Recordings, assign Reviews, approve Transcripts, approve Clips, assemble Publishing Packages, schedule Releases, and resolve Edit Requests.
Reviewers can comment, request edits, approve items, and reject proposed Clips.
AI can summarize Recordings, draft show notes, extract quotes, propose Clips, identify unresolved Edit Requests, generate title options, draft social copy, and map Transcript ranges to Clip ideas.
AI cannot approve final editorial content, publish Releases, invent Guest consent, use restricted Assets, or mark Transcript text as accurate without review.

Rules:
Every Episode must belong to one Show.
Every Clip must trace back to a source Recording, Episode, or Transcript range.
A Transcript cannot be treated as accurate until reviewed or approved.
A Publishing Package cannot be marked ready if required Assets are missing or restricted.
A Release cannot be published by AI without producer approval.
Guest consent must be recorded before publishing Guest-sensitive clips or quotes.
Restricted Assets cannot be used in a Publishing Package or Clip without explicit clearance.
Open Edit Requests block final approval unless a producer overrides with a reason.

Language:
Use "Episode" for the installment being produced.
Use "Recording" for raw captured media.
Use "Transcript" for derived text.
Use "Clip" for excerpts made from source media.
Use "Publishing Package" for the assembled files and metadata before release.
Do not call an AI-proposed Clip "approved" until a human approves it.
Do not confuse "scheduled" with "published."
Do not use "final" while Edit Requests are open.

Sources of Truth:
For source media, trust uploaded Recordings.
For transcript text, trust approved Transcript versions.
For guest permissions, trust signed releases or producer-recorded consent.
For asset restrictions, trust asset metadata and rights notes.
For publishing status, trust Channel responses or producer-recorded status.
If AI summaries conflict with Transcript or reviewer notes, flag the conflict.

Agent Boundaries:
AI may summarize, draft, classify, suggest, compare, and prepare.
AI must cite source ranges when proposing Clips or quotes.
AI must ask before marking Transcript text approved.
AI must not publish, approve final editorial content, claim guest consent, or use restricted Assets without clearance.
AI should escalate when reviewer feedback conflicts, consent is missing, source ranges are unclear, or publishing would expose unapproved content.

Failure Modes:
Confusing Show with Episode.
Treating raw Transcript text as accurate.
Publishing a Clip without guest consent.
Using restricted music or artwork in a Publishing Package.
Treating scheduled Release as published.
Ignoring open Edit Requests.
Losing traceability from Clip to source Recording or Transcript range.
Inventing quotes or guest claims.
```

Create a front-end prototype showing the dashboard, episode planning, production tasks, assets, review workflow, and AI assistant behavior.
```