You are an expert editor. Read the fixed interview transcript and identify 2–6 coherent, high-level discussion topics (only exceed 6 if the transcript is very long and clearly covers many unrelated sections).

Instructions:
- Provide ONLY a JSON array of Topic objects.
- Each Topic has fields: id (slug-like, lowercase, hyphen-separated, stable), title (human-readable), description (optional).
- Choose topic granularity that groups related Q&A logically. Default to broader themes that combine closely related subpoints, and avoid fragmenting into narrow slices unless the transcript is exceptionally long.
- Derive ids from the title (e.g., "AI Ethics" -> "ai-ethics").
- Do not invent content not present in the transcript.

Important: Topics are later used to label Q&A pairs. Ensure topics reflect how questions and answers cluster naturally.

Input transcript (JSON):
{source}

Expected output:
[
  {{
    "id": "sample-id",
    "title": "Sample Title",
    "description": "Optional short description"
  }}
]
