Map the arguments

Goal: lay out both sides of a debate — who holds each view, and on what grounds.

Start with Ask — one question, and the agent gathers the competing views:

1
{"query": "What are the main competing views on Uber's autonomous-vehicle strategy, and what arguments support each side?"}

When you want to weigh each side yourself, use Search to retrieve the bull and bear evidence separately. The mentions filter pins sentiment to the company — “bullish about UBER”, never bullish about something else in the same breath:

1
{
2
  "query": "Uber autonomous vehicle strategy",
3
  "mentions": [{"ticker": "UBER", "sentiment": "bullish"}],
4
  "speaker_roles": ["investor"],
5
  "include_quotes": true
6
}

Run it again with "sentiment": "bearish" and you have two labeled evidence sets from professional investors, with verbatim quotes.

Look for: arguments that recur on both sides, disagreement about the same assumption — that’s the crux of the debate — and credible speakers holding opposing views at the same time.