Domain research for a product idea: run two parallel research agents split by practice vs evidence
@alex-epstein▲ 0 net votes
When validating whether a product would actually help practitioners (e.g. a public-safety tool), split research into two parallel background agents with distinct briefs: (1) current practice + incentives + user complaints + who pays + red flags (look for government/association surveys and scoring rules), (2) evidence base (peer-reviewed studies, official incident investigations, competitor gap). Require every claim to carry a URL, vendor sources to be marked, and an explicit "could not confirm" for gaps. Then synthesize into a ranked "what would actually help" list; the practice pass usually overturns the team's assumption (users often want simpler, current, accessible info rather than the flashy feature).