# Post Analysis: @thesupermannx on the Harvard AI-tutor RCT (2026-09-28) Back to [[Users/@thesupermannx/thesupermannx-x-profile|@thesupermannx profile]] · Paper note: [[Users/@thesupermannx/Papers/AI Tutoring Outperforms In-Class Active Learning (Sci Rep 2025)|Sci Rep 2025]] · Related pattern: [[Topics/Ribosome Pausing and Cotranslational Folding/Related Trends#1. "AI cracked open X" posts that re-label old science|Ribosome topic: the "AI cracked open X" pattern]] **Verdict:** The study numbers are mostly right, but 'AI has deleted every reason universities exist' contradicts the authors, who say the tutor should complement and not replace classes, in a two-lesson, one-course study. ## The post (verbatim) Posted **Sep 28, 2026, 5:34 AM CT**. Link: [Sci Rep s41598-025-97652-6](https://www.nature.com/articles/s41598-025-97652-6). > Researchers proved AI has deleted every reason universities exist. > > Harvard University ran a controlled experiment pitting a custom AI against their own top-tier classrooms. > > And the results are going to collapse the higher education bubble. > > They took 194 undergraduates and split them up. > > One group learned physics in one of Harvard’s best hands-on, active-learning physical classrooms. Group work. Instructor support. The premium university experience. > > The other group went home and learned the exact same material with an AI tutor. > > The AI didn't just win. It embarrassed the institution. > > Students using the AI learned more than twice as much as the students in the elite Harvard classroom. > > They scored 30% higher on the final assessment. > > And they did it in less time. > > Let that sink in. > > A piece of software sitting on a laptop outperformed a world-class faculty in one of the most elite learning environments on Earth. > > Universities have always justified their exorbitant tuition with two things: access to elite knowledge and the physical classroom experience. > > This study just proved both of those moats are gone. > > When software can teach you complex physics twice as well as a $60,000-a-year institution, the math of higher education breaks permanently. > > The AI didn't just give the students answers. It used strict pedagogical guardrails. It guided. It questioned. It forced the students to do the cognitive work. > > It offered perfect, one-to-one tutoring, personalized to the exact moment a student misunderstood a concept. > > That level of attention is mathematically impossible to scale in a physical lecture hall. > > For a thousand years, the university was the only place to get a premium education. > > Now, it’s the bottleneck. > > If AI can double your learning speed for a fraction of the cost, what exactly are students taking on decades of debt to pay for? ## Media The image is a screenshot of the real paper's Scientific Reports page: title, the five authors, "Published: 03 June 2025", and the start of the abstract. ✅ It is the correct paper. ## Engagement (Oct 24, 2026) | Impressions | Likes | Reposts | Quotes | Replies | Bookmarks | |---:|---:|---:|---:|---:|---:| | 1,273,398 | 14,352 | 3,050 | 709 | 1,083 | 7,891 | ## Claim-by-claim check **Ratings:** ✅ accurate · 🟡 overstated or simplified · 🔀 misattributed · ❓ unsupported by the paper · ❌ contradicted by the paper | # | Claim | Rating | What the sources show | |---|---|---|---| | 1 | "Researchers proved AI has deleted every reason universities exist" | ❌ | The authors say the AI tutor should not replace in-class teaching. The study is two lessons in one course | | 2 | "Harvard University ran a controlled experiment" | ✅ | An RCT in Harvard's PS2 course, Fall 2023 | | 3 | "194 undergraduates and split them up" | 🟡 | 194 analyzed. It was a **crossover**: every student did one lesson each way, so there was no fixed split | | 4 | "One of Harvard's best hands-on, active-learning physical classrooms" | ✅ | The control was research-based active learning | | 5 | "went home and learned … with an AI tutor" | ✅ | The AI lessons were done at home | | 6 | "learned more than twice as much" | ✅ | Median learning gains "over double" | | 7 | "They scored 30% higher on the final assessment" | 🟡 | Median post-test was 4.5 vs 3.5, about 29% higher. It was a short post-lesson test, not a course "final" | | 8 | "in less time" | ✅ | Median 49 minutes vs a 60-minute class. 30% of students took longer than 60 minutes | | 9 | "strict pedagogical guardrails … forced the students to do the cognitive work" | ✅ | The tutor was designed on the same best practices, using instructor-written step-by-step solutions | | 10 | "perfect, one-to-one tutoring" | 🟡 | Students rated the explanations highly (83% as good as or better than instructors'), but the tutor followed instructor-made scaffolds | | 11 | "twice as well as a $60,000-a-year institution"; university "moats are gone" | ❓ | Not studied. The authors suggest the tutor may not win for higher-order synthesis | ## What the post leaves out - **Scope:** two topics (surface tension and fluid flow) in one intro course. - **The tutor's design:** GPT-4 scaffolded by instructors' own solutions. - **Ceiling effects** on the test. - **The paper's date:** June 2025, 15 months before the post. - **The authors' own position:** the tutor should complement in-person teaching. ## Bottom line - **What holds up:** the numbers. - **What does not:** the conclusion. The authors explicitly reject it.