Uploaded June 2026 | Updated November 2026, 2 days ago
"Hallucinations are not a bug in my classroom. They are the assignment." That's Giorgio Lagna's framing for one of the 13 AI tutors he's built into his biology course at Santa Clara University. In this episode of Educast 3000, the UCSF research scientist and college lecturer joins Ryan Lufkin and Zach Pendleton to walk through what happens when you design AI not to answer questions but to refuse them, not to be right but to argue back. The conversation covers his "Crucible" course design, the DISCO framework research he published this year showing measurable equity outcomes, and why he believes higher ed is making a strategic mistake by treating AI as an integrity problem instead of an assessment problem.
In this episode:
- Toll booth gems, boss battle gems, and certification gems explained
- Why deliberate AI errors teach better AI literacy than warnings about hallucinations
- The cost math on AI-powered oral exams (NYU's run came to 42 cents per student)
- How community college students are now running CRISPR research with AI co-investigators
- What Pope Leo XIV's first encyclical said about AI and education the day before this recording
About Educast 3000: Melissa Loble and Ryan Lufkin host fellow educators and leaders, delivering sharp commentary and trend analysis on the systemic shifts across the entire education ecosystem. From the high-stakes ethics of AI to the science of how we actually learn, the show tackles the innovations, ideas, and challenges defining education today.
New episodes drop regularly. Find the full archive, show notes, and more at instructure.com/resources/podcast or subscribe wherever you get your podcasts.
"Hallucinations are not a bug in my classroom. They are the assignment." That's Giorgio Lagna's framing for one of the 13 AI tutors he's built into his biology course at Santa Clara University. In this episode of Educast 3000, the UCSF research scientist and college lecturer joins Ryan Lufkin and Zach Pendleton to walk through what happens when you design AI not to answer questions but to refuse them, not to be right but to argue back. The conversation covers his "Crucible" course design, the DISCO framework research he published this year showing measurable equity outcomes, and why he believes higher ed is making a strategic mistake by treating AI as an integrity problem instead of an assessment problem.
In this episode:
- Toll booth gems, boss battle gems, and certification gems explained
- Why deliberate AI errors teach better AI literacy than warnings about hallucinations
- The cost math on AI-powered oral exams (NYU's run came to 42 cents per student)
- How community college students are now running CRISPR research with AI co-investigators
- What Pope Leo XIV's first encyclical said about AI and education the day before this recording
About Educast 3000: Melissa Loble and Ryan Lufkin host fellow educators and leaders, delivering sharp commentary and trend analysis on the systemic shifts across the entire education ecosystem. From the high-stakes ethics of AI to the science of how we actually learn, the show tackles the innovations, ideas, and challenges defining education today.
New episodes drop regularly. Find the full archive, show notes, and more at instructure.com/resources/podcast or subscribe wherever you get your podcasts.






![The AI-List: A Clips Episode on AI Across the Classroom
Ryan Lufkin and Melissa Loble pull together the sharpest AI moments from recent Educast 3000 conversations, featuring five guests who come from very different corners of education but keep running into the same questions: What should we actually teach? How do you assess real learning when AI can do the work for students? Whats the educators role now? And what does it mean to prepare students to be responsible citizens, not just capable ones?
From CS classrooms to liberal arts seminars, neuroscience-informed frameworks to institutional policy, the perspectives dont always agree. Thats what makes it worth watching.
Featured guests:
- Dr. Joseph Youngblood II, Chancellor of Kean Global
- Kelly Shiohira, Director [CONFIRM TITLE], Global Science of Learning Education Network
- Joe Potvin, Adjunct History Professor at Simmons University and Senior Portfolio Manager at Cengage
- Matt Winters, AI Education Specialist at the Utah State Board of Education
- Sanjay Srivastava, CEO of Vocareum
About Educast 3000: Melissa Loble and Ryan Lufkin host fellow educators and leaders, delivering sharp commentary and trend analysis on the systemic shifts across the entire education ecosystem. From the high-stakes ethics of AI to the science of how we actually learn, the show tackles the innovations, ideas, and challenges defining education today.
New episodes drop regularly. Find the full archive, show notes, and more at instructure.com/resources/podcast or subscribe wherever you get your podcasts. The AI-List: A Clips Episode on AI Across the Classroom](https://i.ytimg.com/vi/32Plz-KYd2A/mqdefault.jpg)



