Research

Research finds out what is already true, so the design stage does not have to guess and does not have to look things up itself. It answers five questions, for each criterion that needs one: has someone built this already; are there guides or recipes covering it; do reputable sources recommend a way to do it; is an assumption the task leans on still true; and it hands the answers to the design stage in a form design can act on without looking again.

Research does not decide. It reports what exists and what is true of it. Choosing between two usable options is design's work. Research does not read to the bottom of anything: a subject is never exhausted, so the criteria are what stop this stage, never the subject.

Every finding names where it came from and when it was looked at. A finding with no source is not a finding, and the model's own recall is never the answer, only a lead worth confirming with one search.