In the adviser-preference comparison, less movement toward the adviser’s assigned preference could coexist with influence from the common teaching package. Hypothesis: selecting alternatives, evidence emphasis and presentation could steer choices under either AI lean. Inspect omissions and editorial choices first; if feasible, later compare defensible framings of the same evidence separately from adviser-lean variation. Check choices, sources of reasons and explanations of why a competent person could choose otherwise. Evidence may warrant persuasion; neither zero influence nor felt ownership alone establishes good judgment. 1Cademy raises this concern (undated proposal; captured 2026-11-09). Anthropic’s own report (2026-10-27; one task/model, amplified leans) and the debate preprint (public draft v1, submitted 2026-10-04; selected fictional debates) do not demonstrate package steering.