Kestin and colleagues (Scientific Reports, June 3, 2025) compared home AI tutoring with classroom peer problem-solving and teacher feedback in Harvard introductory physics in 2023. Existing peer groups of two or three were assigned by chance, then switched conditions for the second weekly lesson. Content and worksheets were identical; sequenced AI guidance, recorded introductions, setting and pacing differed together. Tutor post-test scores were higher, but the inspected article and supplement do not specify test-time AI access. It reports 194 eligible students; combined score counts do not confirm that each supplied results under both methods. This positive immediate package comparison neither isolates one component nor tests delayed retention.