Joe Liemandt
Alpha School
Joe Liemandt is a technology entrepreneur and the principal of Alpha School, a K-12 educational institution. His work focuses on utilizing artificial intelligence and individualized learning models to accelerate academic mastery and develop practical life skills in students.
22 claims checked on air: 4 contradicted 1 overstated 11 supported 6 unverified 5 flagged
What they said on air - supported
The modern industrialized model of universal compulsory classroom education originated in Prussia before being imported elsewhere.
"the industrialized model started over in Prussia, right? And then was imported." (said at 0:03:33)
Historical scholarship confirms that the standardized, state-mandated model of universal compulsory education originated in 18th- and 19th-century Prussia. International educational reformers, notably Horace Mann in the United States and Konstantin Ushinsky in Russia, extensively studied and imported elements of the Prussian system—such as age-graded classrooms, standardized curricula, compulsory attendance, and state-certified teacher training—which became the foundation for modern public schooling systems.
Benjamin Bloom's 2 Sigma Problem paper showed that students receiving one-to-one mastery tutoring perform two standard deviations (2 sigma, or top 2%) better than students in conventional classrooms.
"There's a famous paper from the '80s, Bloom's 2 Sigma Problem, where he's like if you give an individual tutor to mastery level, kids can learn at 2 sigma, right? 2 sigma is top 2%. We could get everybody—so when you talk about the frontier, he's basically claiming you can get everybody to the top 2% performance today." (said at 0:29:42)
Benjamin Bloom's seminal 1984 paper, "The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring," reported that students who received one-on-one tutoring with mastery learning scored approximately two standard deviations (2 sigma) higher than students in conventional classrooms. In a normal distribution, an effect size of +2 standard deviations places the average tutored student at approximately the 98th percentile (the top 2%) of conventionally taught students.
Project Follow Through in the 1970s was the largest educational study ever conducted to evaluate effective ways to teach children.
"There's this Project Follow Through, a big project in the '70s that was like the largest educational study ever about effective ways to teach kids." (said at 0:35:30)
Project Follow Through was a large-scale, federally sponsored national longitudinal educational project initiated to evaluate different instructional approaches and educational models for teaching economically disadvantaged children. Published scientific literature documents that substantial time, money, and evaluation efforts were invested into the national Follow Through Project to test instructional methods and their effects on disadvantaged children's achievement.
Memorizing information or math facts to cognitive fluency prevents the retrieval of that information from consuming working memory capacity.
"if you memorize something to fluency, when you use it, it doesn't take a working memory slot." (said at 0:53:12)
The claim aligns with foundational principles of cognitive load theory and human cognitive architecture. Working memory has severe capacity limitations when processing novel information, whereas long-term memory stores vast numbers of schemas. When knowledge or facts (such as basic mathematical facts) are learned and automated through extensive practice—achieving cognitive fluency—they can be retrieved from long-term memory with minimal or no working memory load, thereby preserving working memory capacity for other elements of a task.
There are no learning science papers showing that middle school girls are incapable of learning middle school math and science.
"I make them go find me the learning science paper that shows middle school girls can't learn math/science. There are none of those. They can learn this math/science and it's middle school level" (said at 1:14:11)
Large-scale meta-analyses and cognitive science research consistently show that middle school girls are fully capable of learning grade-level mathematics and science, with no evidence in the educational or psychological literature demonstrating an inability to learn these subjects. A 2010 meta-analysis of 242 studies comprising 1,286,350 participants found an overall effect size of d = 0.05, indicating no meaningful gender difference in mathematics performance. Similarly, an earlier meta-analysis covering over 3 million students found no gender differences in problem-solving in elementary or middle school, with girls showing slight advantages in computation.
- supports: New trends in gender and mathematics performance: a meta-analysis. (Psychological bulletin 2010) · cited 919x in the literature
"First, we meta-analyzed data from 242 studies published between 1990 and 2007, representing the testing of 1,286,350 people. Overall, d = 0.05, indicating no gender difference, and variance ratio = 1.08, indicating nearly equal male and female variances." (abstract, results, passage verified)
pubmedfull study (doi) - supports: Gender differences in mathematics performance: a meta-analysis. (Psychological bulletin 1990) · cited 1662x in the literature
"An examination of age trends indicated that girls showed a slight superiority in computation in elementary school and middle school. There were no gender differences in problem solving in elementary or middle school" (abstract, results, passage verified)
pubmedfull study (doi)
State standardized testing under the Texas STAAR (State of Texas Assessments of Academic Readiness) program begins in the third grade.
"And I'm like, "Well, the Texas STAAR starts in the third grade."" (said at 2:05:12)
Under the Texas state assessment system, the State of Texas Assessments of Academic Readiness (STAAR) program begins administration in the third grade, testing elementary students annually in mathematics and reading beginning at Grade 3.
- supports: Differences in Academic Performance between Elementary Charter Schools and Traditional Pub… (Journal of Advances in Education Research 2017) · cited 2x in the literature
"In this study, the degree to which differences were present in reading for Grade 3 students; in reading and writing for Grade 4 students; and in reading and science for Grade 5 students between charter and traditional elementary public schools in Texas was determined. Participants in this study were Grades 3, 4, and 5 students who took the 2015 State of Texas Assessments of Academic Readiness Reading, Writing, and Science tests." (abstract, passage verified)
openalexfull study (doi) - supports: Predicting Reading Performance by Texas Student Demographics Characteristics: A Statewide … (International Journal of Social Learning (IJSL) 2021) · cited 3x in the literature
"Analyzed in this research study was the degree to which demographic characteristics (i.e., economic status, ethnicity/race, English Language Learner status) of Grade 3 students in Texas schools was related to their reading achievement as assessed by the State of Texas Assessment of Academic Readiness (STAAR) Reading test." (abstract, passage verified)
openalexfull study (doi)
A study found that an LLM trained on dating advice outperformed human parents at providing social-emotional advice to teenagers.
"The surprising finding was it crushed the parents at giving social-emotional advice to teens." (said at 2:21:10)
A preregistered, randomized experimental study evaluated dating and relationship advice provided to 404 adolescents (aged 13–17) across realistic conflict scenarios. The study compared advice from human parents against advice from off-the-shelf LLMs (ChatGPT and Gemini) and a fine-tuned LLM specifically trained to provide relationship advice (AskElle). Across primary outcome metrics—including perceived coping support, likelihood of implementation, and trustworthiness—LLM-generated advice outperformed parent-provided advice by approximately 0.3 standard deviations. Furthermore, when adolescents were able to directly compare sources across repeated exposures, they demonstrated a strong preference for the fine-tuned relationship advice LLM over human parents.
- supports: An Experimental and Qualitative Comparison of the Quality of Dating Advice from Parents Ve… (? 2025)
"Here we report a preregistered [osf.io/md2ek], randomized experiment (N=404 adolescents aged 13-17) to compare the quality of relationship advice from parents against that generated by LLMs... The LLM-provided advice was split into two sources: a fine-tuned LLM (AskElle) that was trained to provide relationship advice or an off-the-shelf LLM (either ChatGPT or Gemini)... The LLMs outperformed parents’ advice by approximately 0.3 standard deviations across the primary outcomes, all pr(ATE>0)>.99... Once participants could directly compare all three sources, they overwhelmingly preferred the fine-tuned LLM’s advice, pr(Diff)>.99 over both parents and off-the-shelf LLMs." (abstract, results)
openalexfull study (doi)
Learning science shows that the more prior knowledge a person has, the faster they learn new material.
"The learning science shows that the more you know, the faster you learn, right?" (said at 2:22:35)
Extensive cognitive science and neuroscience research supports the claim that prior knowledge accelerates and enhances the acquisition of new, related material. In cognitive psychology and schema theory, pre-existing knowledge frameworks (schemas) facilitate more rapid encoding, integration, and consolidation of schema-congruent information in memory networks.
High IQ combined with high conscientiousness on the Big Five personality traits are the primary individual predictors of student success in traditional education.
"Then with IQ, the second one, once you get to the next level, it's IQ and Big Five conscientiousness are the two" (said at 5:52:30)
Extensive meta-analytic evidence confirms that general cognitive ability (IQ) and conscientiousness are the primary individual predictors of academic achievement across educational levels. Meta-analyses synthesizing data from hundreds of thousands of students demonstrate that cognitive ability accounts for the largest share of variance in academic performance, while conscientiousness is the strongest and most consistent non-cognitive predictor, offering substantial incremental validity over intelligence alone.
Higher IQ provides more working memory capacity, facilitating problem-solving with difficult material.
"which is IQ, you get more working memory slots. So if you put in material that's too hard and stuff, you can struggle through it easier." (said at 5:52:40)
Psychometric and cognitive research demonstrates a very strong positive correlation between general intelligence (particularly fluid intelligence) and working memory capacity (WMC). WMC reflects the ability to actively maintain, update, and manipulate multiple elements or bindings simultaneously, which serves as a key predictor and facilitator of complex reasoning and problem-solving performance.
There are approximately 55 million school-aged children in the United States.
"There's 55 million kids in America and we need to get it to everybody." (said at 10:11:30)
National statistics and demographic data indicate that there are approximately 55 million school-aged (K-12) children in the United States across public and private elementary and secondary schools.
Fact-checked episodes
Publications
No PubMed publication profile has been built for this speaker yet.