Tech & Learning Conversations Podcast

The episode contains several substantive ideas about using survey data and AI for student identification, evidence-based decision-making, and the generalizability problem in education research. However, there is considerable filler including extended throat-clearing about getting started in the role, rehashed points about teacher cognitive load, and vague aspirational statements about chronic absenteeism without concrete actionable insights. The Mathematica study (52 additional students placed) and the Learning Collider work on 800,000 students are the concrete substantive moments.

“almost two full classrooms of kids who would have otherwise been overlooked”

“they started to see these relationships between the predictive nature and whether a kid feels connected to a trusting adult in school, uh, and whether there will be chronically absent their grades”

The core insight - that schools miss talented students and that survey data plus traditional metrics can identify them - is known in the education world and was already EOS's prior work. The positioning of AI for predictive analytics on soft factors like 'trusted adult' is somewhat novel but presented as exploratory rather than proven. The Maryland $40M ladder-funding framework is genuinely interesting but briefly mentioned without depth. Overall, the thinking is competent but not distinctly contrarian or first-principles.

“we were able to increase the average number of students placed in advanced coursework by 52 students per school”

“what I think is the next frontier of learning is if you use AI and predictive analytics on this type of information”

AJ Gutierrez has legitimate operational credibility: he co-founded Saga Education, helped build high-impact tutoring into a $7B movement, and is now CEO of Equal Opportunity Schools operating across 900 schools in 35 states. He speaks from direct execution experience rather than theory. However, he is relatively new to the EOS role (started July, still in strategy phase) and much of the discussion is forward-looking rather than deeply reporting on current execution and results.

“Before this role, he co founded Saga Education and helped build the high impact tutoring movement into a national priority, securing more than $7 billion in public investment”

“I started as a CEO, uh, at, uh, Equal Opportunity Schools in July. So I'm still going through the hazing process”

The episode includes some concrete numbers: 52 additional students per school in the Mathematica study, 800,000 students in the Learning Collider dataset, 900 schools across 35 states served by EOS, $40M Maryland initiative, $7B in high-impact tutoring spending, and the 80% miss rate on PSAT qualification. However, these figures are often presented without context or methodology detail. Most AI and chronic absenteeism claims remain abstract and prospective rather than evidence-backed with specifics.

“we were able to increase the average number of students placed in advanced coursework by 52 students per school. Uh, and this was a study around uh, about 400,000 students”

“So the Learning Collider is based out of the University of Texas. Um, that entire initiative is focused on, uh, developing ethical AI tools”

The host Kevin Hogan asks competent setup questions and makes relevant contextual observations (e.g., the three-tiered customer problem in education, the post-COVID ed-tech skepticism), but rarely pushes back or requests deeper specifics. When AJ discusses AI predictive analytics on 'trusted adult' relationships, the host nods along rather than probing: what does 'connected to a trusted adult' actually measure? How is it validated? The LA superintendent example is brought up but not interrogated. The conversation is pleasant and professional but lacks the sharp follow-ups that would distinguish it.

“I think going to your point about comparing education in other industries, one of the things that's interesting I found over the years”

“Yeah, now, as I mentioned earlier, there seems to be a general sentiment”