Proposed comparison: strong AI alone, AI with shared documents, or AI using the separately reviewed 1Cademy topic-and-link workflow. Hold materials unchanged; match sources, background coverage, model access and time. Fictional cases contrast a material required to work across temperatures but tested at one with a one-temperature goal where that test suffices. Score justified changes, warranted checks and needless objections. Pair safely paused work with equipment damage before human assessment; score appropriate review or automatic protection. Judge explanations without AI; scorers should not know which aid was used. Record response time and total preparation, review, upkeep and first- and later-reader effort. This targets goal/evaluator adequacy and response allocation, not retention or randomized-advice-lean effects. 1Cademy (undated; retrieved 2026-11-09) and Bainbridge (1983-11) motivate, not validate it; shared blind spots remain possible.