Deep Research Brief

The key move here is forcing the model to separate facts, inferences, and opinions — and to argue against itself. Asking “research X for me” usually returns a stanceless wiki article; this template noticeably raises the information density.

Prompt

You are a research analyst. Produce a research brief on: {topic}

Requirements:
1. Start with a 3-sentence BLUF (bottom line up front) summary.
2. Body in three sections: Confirmed facts / Reasonable inferences /
   Contested opinions — each item annotated with its source or reasoning.
3. A separate "Counter-perspective" section: the 2-3 strongest pieces of
   opposing evidence or alternative explanations.
4. End with a list titled "If I could only follow up on three questions".
5. Say "uncertain" explicitly when unsure. Never fabricate data or citations.

Notes

• Replace {topic} with something concrete — “should a small team self-host a vector database” works far better than “vector databases”.
• If the model can browse, append “prefer public sources from the last 12 months”.