Evidence

How it works

Every claim in the product — a theme, a signal, a recommendation — traces down to evidence you can open. An evidence item carries the quote, who said it, which study and theme it belongs to, when it was captured, and its kind:

• Supports (evidence) — the quote backs the belief.
• Nuance (counterEvidence) — the quote complicates it. Nuance is a reading on the belief, not a rebuttal.

Evidence comes only from analyzed, high-quality interviews with active themes. Low-quality material never enters the pool, so a citation is trustworthy by construction.

What you can do

Pull an evidence packet for any object — a signal, a study, specific themes, or specific interviews — with nuance included when confidence matters. Or search semantically: the index covers every evidence quote, so "did anyone mention pricing?" is one query, not a read-through.

When it's quiet

No hits for a topic does not mean nobody cares. The index contains evidence quotes only — not transcripts or notes — so it may just mean nobody was asked. That's what research gaps are for.

From your agent

• search_evidence — semantic search over all evidence; supporting quotes by default, pass kind: counterEvidence to search nuance instead.
• get_evidence — the audit trail behind a signal, study, theme, or interview.